of breast cancer - pharmaceutical review · pdf filepharmacogenetics in the care of breast...

H8637

Pharmacogenetics of Breast Cancer

Towards the Individualization of Therapy

Edited by

Brian Leyland-Jones

Leyland-Jones

Pharmacogenetics of Breast C

ancerTow

ards the Individualization of T

herapy

Translational Medicine Series 7

about the book…

Pharmacogenetics is becoming increasingly relevant in the diagnosis, treatment, and recovery of cancer patients. A major problem facing oncologists is the out-standing varied efficacy of treatment. Promising advances in pharmacogenetics have allowed the development of effective agents which will enable personalized cancer chemotherapy to become routine for the clinical practice.

Written by experts in the field and combining information that is unable to be found in a single source volume, Pharmocogenetics of Breast Cancer:• combines a complete overview of pharmacogenetics and how it relates to

breast oncology for diagnosis, treatment, and the recovery of patients using an individualized therapy model

• is the first single source reference demonstrating the application of pharmacogenetics in the care of breast cancer patients

• enables physicians a coherent interpretation of the emerging science of pharmacogenetics, aiding them to incorporate individual therapies in their own practice

• gives practical guidance on various forms of specimen collection, tissue selection, and handling procedures

Oncology

about the editor...

BRIAN LEYLAND-JONES is Associate Vice President and Director, Emory Winship Cancer Institute, Emory University, Atlanta, Georgia. Dr. Leyland-Jones’ main research interests are pharmacodynamics, pharmacokinetics, and pharmacogen-etics in oncological clinical trials; translation of preclinical models into the clinic; biomarker endpoints in Phase I/II clinical trials; and screening and mechanistic studies of novel targeted and chemotherapeutic anti-cancer agents. He has authored more than 125 peer-reviewed articles and book contributions, 150 abstracts, and 29 patents.

Printed in the United States of America

[gajendra][D:/Informa_Projects/H5293_Leyland_Jones_112035/z_production/z_3B2_3D_files/978-1-4200-5293-0_CH0000_O.3d] [11/4/08/16:0:41] [1–16]

[gajendra][][D:/informa_Publishing/H5293_Leyland_Jones_112035/z_production/z_3B2_3D_files/978-1-4200-5293-0_CH0000_Series_pageO.3d] [3/4/08/10:56:5][1–2]

TRANSLATIONAL MEDICINE SERIES

1. Prostate Cancer: Translational and Emerging Therapies, edited byNancy A. Dawson and W. Kevin Kelly

2. Breast Cancer: Translational Therapeutic Strategies, edited by GaryLyman and Harold Burstein

3. Lung Cancer: Translational and Emerging Therapies, edited byKishan J. Pandya, Julie R. Brahmer, and Manuel Hidalgo

4. Multiple Myeloma: Translational and Emerging Therapies, edited byKenneth C. Anderson and Irene Ghobrial

5. Cancer Supportive Care: Advances in Therapeutic Strategies,edited by Gary H. Lyman and Jeffrey Crawford

6. Cancer Vaccines: Challenges and Opportunities in Translation,edited by Adrian Bot and Mihail Obrocea

7. Pharmacogenetics of Breast Cancer: Towards the Individualizationof Therapy, edited by Brian Leyland-Jones


Informa Healthcare USA, Inc.

52 Vanderbilt Avenue

New York, NY 10017

# 2008 by Informa Healthcare USA, Inc.

Informa Healthcare is an Informa business

No claim to original U.S. Government works

Printed in the United States of America on acid-free paper

10 9 8 7 6 5 4 3 2 1

International Standard Book Number-10: 1-4200-5293-4 (Hardcover)

International Standard Book Number-13: 978-1-4200-5293-0 (Hardcover)

This book contains information obtained from authentic and highly regarded sources. Reprinted material is

quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable

efforts have been made to publish reliable data and information, but the author and the publisher cannot

assume responsibility for the validity of all materials or for the consequence of their use.

No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,

mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and

recording, or in any information storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright

.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood

Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and

registration for a variety of users. For organizations that have been granted a photocopy license by the CCC,

a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used

only for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Pharmacogenetics of breast cancer : towards the individualization of therapy / edited by

Brian Leyland-Jones.

p. ; cm. — (Translational medicine series ; 7)

Includes bibliographical references and index.

ISBN-13: 978-1-4200-5293-0 (hardcover : alk. paper)

ISBN-10: 1-4200-5293-4 (hardcover : alk. paper)1. Breast—Cancer—Genetic aspects. 2. Breast—Cancer—Chemotherapy.

3. Pharmacogenetics. I. Leyland-Jones, Brian. II. Series.

[DNLM: 1. Breast Neoplasms—genetics. 2. Breast Neoplasms—drug therapy.

WP 870 P536 2008]RC280.B8P43 2008

616.990449—dc22

2008001013

For Corporate Sales and Reprint Permissions call 212-520-2700 or write to: Sales Department,

52 Vanderbilt Avenue, 16th floor, New York, NY 10017.

Visit the Informa Web site at

www.informa.com

and the Informa Healthcare Web site at

www.informahealthcare.com


Preface

I write this preface on Christmas Day, 2007, in Costa Rica. Christmas is always a

special time for me, since my parents were married on Christmas Day sixty

nine years ago. It is almost unimaginable to think how far we have progressed

since 1938. My parents were married in the year before the Second World

War. The nitrogen mustards, from which arose the first chemotherapies, were

developed during the war. I was fortunate during my training at Sloan-Kettering

to work in the laboratories of Fred Phillips, Joe Burchenal, and Jack Fox, which

conducted some of the pivotal work on these compounds. Indeed, the first task

force for childhood leukemia was set up in Washington in the year of my birth.

However, we all have to remember how slowly treatments are disseminated to

the community; I can remember, as a child, my next-door neighbor having a leg

amputated for metastatic lung cancer since no other treatments were accessible

locally!

The rest is history, but again it is difficult to believe that the Nobel Prize

for the discovery of the cellular origin of retroviral oncogenes was awarded to

Bishop and Varmus only in 1989.

Now we face the challenge of too few patients eligible for clinical trials to

adequately test the myriad of targeted therapies available as well as the avail-

ability of high-quality starting material with associated clinical and phenotypic

data. Various platforms are available for the identification of biomarkers: early

technologies have focused on reverse transcriptase polymerase chain reaction

and expression arrays. Now single nucleotide polymorphisms, array comparative

genomic hybridization, methylation signatures, microRNAs, and proteomics are

increasingly included in biomarker profiling; nanotechnology, molecular imag-

ing, and the use of circulating tumor cells are advancing at a tremendous pace.

By the time this book is released, the specimen collection guidelines for breast

cancer, representing a joint effort between the North American cooperative

iii


groups and those comprising the Breast International Group (BIG), will have

been published and be available for open Web access. We have attempted to

cover the breadth of these diverse areas in this book.

I am especially proud of three aspects of the book. The first is the timeliness,

as so many new approaches to individualization of therapy have emerged, and

when two very large randomized trials are aggressively accruing on the basis of

Oncotype and mammoprint. Second, the diversity of the topics covered, ranging

from pathology, tissue handling, DNA-, expression-, and epigenetic-based

approaches, nanotechnology, stem cells, circulating tumor cells, endocrine, and

chemotherapeutic through to pharmacologic aspects and the key pivotal trials.

Third, I am proud that so many of my most distinguished friends and colleagues

from around the world have been so generous with their time and devotion to

make this book outstanding.

Finally, in writing this preface from one of the most beautiful places on

earth, I am deeply reminded of the disparities in access to healthcare. We are all

immensely grateful that my friends and colleagues John Seffrin and Otis

Brawley at the American Cancer Society have been leaders in bringing this

critical issue to the forefront of public consciousness, but we must never fail to

remember that severe disparities in access to care remain, not only in the United

States but globally. Individualized therapy is not only meant to direct therapy to

those who will benefit most but also intended to reduce waste and improve cure

rates dramatically, so that this therapy is available to all who suffer. Let us, the

privileged, never move our focus from the most disadvantaged in our world and

aim at using individualized therapy as a means to make our most sophisticated

treatment available and affordable to all.

Brian Leyland-Jones, M.D., Ph.D.

iv Preface


Acknowledgments

Many thanks are owed on a personal level, especially to all my friends and

coauthors who have been immensely generous in dedicating their nonexistent

spare moments to contributing the various chapters. I am deeply grateful to the

publisher, Informa Healthcare, who not only had the insight to partner with

me on this book but also had the patience to wait for the “nonexistent spare

moments” of the contributing authors to materialize. I must also profoundly

apologize to several colleagues who were approached early on to contribute

chapters but who were unable to because of the original time lines and now must

be frustrated because the time lines were forced to be extended considerably.

Most especially, I must thank my dear, dear colleague, Dr. Brian R. Smith, who

was relentless in organization, follow-up, and advice, a man whose wisdom I

rely upon at so many critical moments and to whose structured approach I aspire.

I am deeply, deeply grateful to Charles Bronfman, whose chair in the memory of

his dear sister, Minda, sustained me over the past 17 years and who now has

inspirationally established his own Institute of Personalized Cancer Therapy in

New York. I am grateful also to the Flanders family who have shown me a depth

of kindness, support, and friendship that is without equal. Finally, I am humbled

by my research and clinical colleagues who instruct and guide me constantly,

and especially by each and every patient who inspires us and teaches, in each and

every encounter, just how personalized breast cancer therapy really is.

v


Contents

Preface . . . . . . . . . . iii

Acknowledgments . . . . v

Contributors . . . . . . . . xi

1. Pharmacogenetics of Breast Cancer: Toward the Individualization

of Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Jenny C. Chang, Susan G. Hilsenbeck, and Suzanne A. W. Fuqua

2. Pharmacology, Pharmacogenetics, and Pharmacoepidemiology:

Three P’s of Individualized Therapy . . . . . . . . . . . . . . . . . . . . 11Shaheenah Dawood

3. Role of Genetic Variability in Breast Cancer Treatment

Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Kandace L. Amend, Ji-Yeob Choi, and Christine B. Ambrosone

4. The Role of Epigenetics in Breast Cancer: Implications for

Diagnosis, Prognosis, and Treatment . . . . . . . . . . . . . . . . . . . . 45Amy M. Dworkin, Tim H.-M. Huang, and Amanda E. Toland

5. Comparative Genomic Hybridization and Copy NumberAbnormalities in Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . 61Nicholas Wang and Joe Gray

vii


6. Gene Expression Profiling as an Emerging Diagnostic Tool

to Personalize Chemotherapy Selection for Early Stage

Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77Cornelia Liedtke and Lajos Pusztai

7. Gene Expression–Based Predictors of Prognosis and Response

to Chemotherapy in Breast Cancer . . . . . . . . . . . . . . . . . . . . . 97Soonmyung Paik

8. Utilization of Genomic Signatures for Personalized

Treatment of Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . 107P. Kelly Marcom, Carey K. Anders, Geoffrey S. Ginsburg,

Joseph R. Nevins, and Anil Potti

9. Personalized Medicine by the Use of Microarray Gene

Expression Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123Stella Mook, Laura J. van ’t Veer, and Fatima Cardoso

10. Predictive Markers for Targeted Breast Cancer Treatment . . . 135Hans Christian B. Pedersen and John M. S. Bartlett

11. Circulating Tumor Cells in Individualizing Breast Cancer

Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151James M. Reuben and Massimo Cristofanilli

12. Individualization of Endocrine Therapy in Breast Cancer . . . . 167Amelia B. Zelnak, Ruth M. O’Regan, and Clodia Osipo

13. TAILORx: Rationale for the Study Design . . . . . . . . . . . . . . . 185Joseph A. Sparano

14. Taking Prognostic Signatures to the Clinic . . . . . . . . . . . . . . . 197Michail Ignatiadis and Christos Sotiriou

15. Tissue is the Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213Adekunle Raji

16. Acquisition and Preservation of Tissue for Microarray

Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231Brian R. Untch and John A. Olson

viii Contents


17. Tapping into the Formalin-Fixed Paraffin-Embedded (FFPE)

Tissue Gold Mine for Individualization of Breast

Cancer Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243Mark Abramovitz and Brian Leyland-Jones

18. Breast Cancer Stem Cells and Their Niche: Lethal Seeds in

Lethal Soil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259Danuta Balicki, Brian Leyland-Jones, and Max S. Wicha

19. Molecular Imaging in Individualized Cancer Management . . . 291David M. Schuster and Diego R. Martin

20. Cancer Nanotechnology for Molecular Profiling andIndividualized Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309May Dongmei Wang, Jonathan W. Simons, and Shuming Nie

Index . . . . . . . 323

Contents ix


Contributors

Mark Abramovitz VM Institute of Research, Montreal, Quebec, Canada

Christine B. Ambrosone Department of Cancer Prevention and Control,

Roswell Park Cancer Institute, Buffalo, New York, U.S.A.

Kandace L. Amend Department of Oncological Sciences, Mount Sinai School

of Medicine, New York, New York, U.S.A.

Carey K. Anders Duke Institute for Genome Sciences & Policy, Department

of Medicine, Duke University Medical Center, Durham, North Carolina, U.S.A.

Danuta Balicki Department of Medicine and Research Centre, Centre

Hospitalier de l’Universite de Montreal; Department of Medicine, Universite de

Montreal; and Breast Cancer Prevention and Genetics and Breast Cancer Risk

Assessment Clinic, Ville Marie Multidisciplinary Breast Cancer Centre,

Montreal, Quebec, Canada

John M. S. Bartlett Endocrine Cancer Group, Cancer Research Centre,

Western General Hospital, Edinburgh, United Kingdom

Fatima Cardoso Department of Medical Oncology, Institute Jules Bordet,

Brussels, Belgium

Jenny C. Chang Breast Center, Dan L. Duncan Cancer Center, Baylor College

of Medicine, Houston, Texas, U.S.A.

Ji-Yeob Choi Department of Cancer Prevention and Control, Roswell Park

Cancer Institute, Buffalo, New York, U.S.A.

xi


Massimo Cristofanilli Department of Breast Medical Oncology, The

University of Texas M. D. Anderson Cancer Center, Houston, Texas, U.S.A.

Shaheenah Dawood Department of Medical Oncology, Dubai Hospital,

Dubai, United Arab Emirates

Amy M. Dworkin Human Cancer Genetics Program, Department of

Molecular Virology, Immunology, and Medical Genetics, The Ohio State

University, Columbus, Ohio, U.S.A.

Suzanne A. W. Fuqua Breast Center, Dan L. Duncan Cancer Center, Baylor

College of Medicine, Houston, Texas, U.S.A.

Geoffrey S. Ginsburg Duke Institute for Genome Sciences & Policy,

Department of Medicine, Duke University Medical Center, Durham,

North Carolina, U.S.A.

Joe Gray Life Sciences Division, Lawrence Berkeley National Laboratory,

Berkeley, California, U.S.A.

Susan G. Hilsenbeck Breast Center, Dan L. Duncan Cancer Center, Baylor

College of Medicine, Houston, Texas, U.S.A.

Tim H.-M. Huang Human Cancer Genetics Program, Department of



Michail Ignatiadis Department of Medical Oncology, Translational Research

Unit, Institut Jules Bordet, Universite Libre de Bruxelles, Brussels, Belgium

Brian Leyland-Jones Winship Cancer Institute, Emory University School of

Medicine, Atlanta, Georgia, U.S.A.

Cornelia Liedtke Department of Breast Medical Oncology, The University of

Texas, Anderson Cancer Center, Houston, Texas, U.S.A.

P. Kelly Marcom Duke Institute for Genome Sciences & Policy, Department

of Medicine, Duke University Medical Center, Durham, North Carolina, U.S.A.

Diego R. Martin Department of Radiology, Emory University, Atlanta,

Georgia, U.S.A.

xii Contributors


Stella Mook Department of Pathology, The Netherlands Cancer Institute,

Amsterdam, The Netherlands

Joseph R. Nevins Duke Institute for Genome Sciences & Policy, Department

of Molecular Genetics and Microbiology, Duke University Medical Center,

Durham, North Carolina, U.S.A.

Shuming Nie Departments of Biomedical Engineering and Chemistry and

the Winship Cancer Institute, Emory University and Georgia Institute of

Technology, Atlanta, Georgia, U.S.A.

Ruth M. O’Regan Emory Winship Cancer Institute, Atlanta, Georgia, U.S.A.

John A. Olson Institute for Genome Sciences and Policy and the Department

of Surgery, Duke University Medical Center, Durham, North Carolina, U.S.A.

Clodia Osipo Loyola University, Maywood, Illinois, U.S.A.

Soonmyung Paik Division of Pathology, NSABP Foundation, Pittsburgh,

Pennsylvania, U.S.A.

Hans Christian B. Pedersen Endocrine Cancer Group, Cancer Research

Centre, Western General Hospital, Edinburgh, United Kingdom

Anil Potti Duke Institute for Genome Sciences & Policy, Department of

Medicine, Duke University Medical Center, Durham, North Carolina, U.S.A.

Lajos Pusztai Department of Breast Medical Oncology, The University of

Texas, Anderson Cancer Center, Houston, Texas, U.S.A.

Adekunle Raji Eastern Cooperative Oncology Group Pathology Coordinating

Office, Northwestern University, Evanston, Illinois, U.S.A.

James M. Reuben Department of Hematopathology, The University of Texas

M. D. Anderson Cancer Center, Houston, Texas, U.S.A.

David M. Schuster Division of Nuclear Medicine and Molecular Imaging,

Department of Radiology, Emory University, Atlanta, Georgia, U.S.A.

Jonathan W. Simons The Prostate Cancer Foundation, Santa Monica,

California, U.S.A.

Contributors xiii


Christos Sotiriou Department of Medical Oncology, Translational Research

Unit, Institut Jules Bordet, Universite Libre de Bruxelles, Brussels, Belgium

Joseph A. Sparano Albert Einstein College of Medicine, Montefiore Medical

Center, Bronx, New York, U.S.A.

Amanda E. Toland Human Cancer Genetics Program, Department of



Brian R. Untch Institute for Genome Sciences and Policy and the Department

of Surgery, Duke University Medical Center, Durham, North Carolina, U.S.A.

Laura J. van ’t Veer Department of Pathology, The Netherlands Cancer

Institute, Amsterdam, The Netherlands

May Dongmei Wang Departments of Biomedical Engineering and Electrical

and Computer Engineering, Georgia Institute of Technology and Emory

University, Atlanta, Georgia, U.S.A.

Nicholas Wang Life Sciences Division, Lawrence Berkeley National

Laboratory, Berkeley, California, U.S.A.

Max S. Wicha Comprehensive Cancer Centre, Department of Internal

Medicine, Division of Hematology-Oncology, University of Michigan, Ann

Arbor, Michigan, U.S.A.

Amelia B. Zelnak Emory Winship Cancer Institute, Atlanta, Georgia, U.S.A.

xiv Contributors

[gajendra][6x9-Standard][D:/informa_Publishing/H5293_Leyland_Jones_112035/z_production/z_3B2_3D_files/978-1-4200-5293-0_CH0001_O.3d] [11/4/08/15:4:48] [1–10]

1

Pharmacogenetics of Breast Cancer:

Toward the Individualization of Therapy

Jenny C. Chang, Susan G. Hilsenbeck, and Suzanne A. W. Fuqua

Breast Center, Dan L. Duncan Cancer Center,Baylor College of Medicine, Houston, Texas, U.S.A.

INTRODUCTION

Mortality from breast cancer results from the ability of some tumors to meta-

stasize to distant sites. Selecting patients with micrometastases at diagnosis is

crucial for clinicians in deciding who should, and who should not, receive toxic

and expensive adjuvant chemotherapy to eradicate these metastatic cells.

Although many individual biomarkers were originally attractive, over the years

most have failed to become clinically useful. In addition, the management of

breast cancer has changed, with the majority of node-negative patients now

undergoing systemic adjuvant therapy because we cannot precisely determine an

individual’s risk of recurrence. A majority of node-negative patients are being

unnecessarily overtreated because if left systemically untreated, only about 25%

of node-negative patients would ever develop recurrence. There is therefore a

critical need to identify patients with sufficiently low risk of breast cancer

recurrence so as to avoid further treatment. In addition, in patients at risk of

recurrence and in need of therapy, optimal therapeutic selection is an increas-

ingly important objective. Recent developments in applying microarray tech-

nologies to breast tumor samples suggest that these new techniques may provide

for the transition of molecular biological discoveries to clinical application and

1


will generate clinically useful genomic profiles that more accurately predict the

long-term outcome of individual breast cancer patients.

BACKGROUND

Until recently, evaluations of prognostic and predictive factors have considered

one factor at a time or have used small panels of markers. However, with the

advent of new genomic technologies such as microarrays, capable of simulta-

neously measuring thousands of genes or gene products, we are beginning to

construct molecular fingerprints of individual tumors so that accurate prognostic

and predictive assessments of each cancer may be made. Clinicians may one day

base clinical management on each woman’s personal prognosis and predict the

best individual therapies according to the genetic fingerprint of each individual

cancer.

Breast cancer is characterized by a very heterogeneous clinical course. A

major goal of recent studies is to determine whether RNA microarray expression

profiling or DNA array gene amplification or gene loss patterns can accurately

predict an individual’s long-term potential for recurrence of breast cancer, so that

appropriate treatment decisions can be made. Microarrays can be used to mea-

sure the mRNA expression of thousands of genes at one time or survey genomic

alterations that may distinguish molecular phenotypes associated with long-term,

recurrence-free survival or clinical response to treatment. These new tech-

nologies have been successfully applied to primary breast cancers and may

eventually outperform currently used clinical parameters in predicting disease

outcome.

Since RNA expression microarray technology provided a method for

monitoring the RNA expression of many thousands of human genes at a time,

there was considerable anticipation that it would quickly and easily revolutionize

our approaches to cancer diagnosis, prognosis, and treatment. The reality

remains extremely promising but is also complex. A potential complication in

the application of microarray technology to primary human breast tumor samples

is the presence of variable numbers of normal cells, such as stroma, blood

vessels, and lymphocytes, in the tumor. Indeed, it has been demonstrated using

gross analysis of human breast cancer specimens compared with breast cancer

cell lines that the tumors expressed sets of genes in common not only with these

cell lines but also with cells of hematopoietic lineage and stromal origin (1,2).

Laser capture microdissection has also been successfully used to isolate pure-cell

populations from primary breast cancers for array profiling (3). Sgroi et al. (3)

utilized laser capture microdissection to isolate morphologically “normal” breast

epithelial cells, invasive breast cancer cells, and metastatic lymph node cancer

cells from one patient and were able to demonstrate the feasibility of using

microdissected samples for array profiling as well as following potential pro-

gression of cancer in this patient. However, with the emerging data supporting

important roles for the surrounding stroma in breast cancer progression and the

2 Chang et al.


labor-intensive and technically challenging nature of laser capture technology

with subsequent amplification of RNA for quantitation, most published inves-

tigations to date have evaluated total gene expression to identify prognostic

profiles, as will be described in the next section.

MOLECULAR CLASSIFICATION OF BREAST CANCER

A study of sporadic breast tumor samples by Perou et al. (2) was the first to show

that breast tumors could be classified into subtypes distinguished by differences

in their expression profiles. Using 40 breast tumors and 20 matched pairs

of samples before and after doxorubicin treatment, an “intrinsic gene set” of

476 genes was selected that was more variably expressed between the 40 sporadic

tumors than between the paired samples. This intrinsic gene set was then used to

cluster and segregate the tumors into four major subgroups: a “luminal cell-like”

group expressing the estrogen receptor (ER); a “basal cell-like” group express-

ing keratins 5 and 17, integrinb4, and laminin, but lacking ER expression; an

“Erb-B2-positive” group; and a “normal” epithelial group (Fig. 1).

In a subsequent study with 38 additional cancers, the investigators found

the same subgroups as before (4), except that the luminal, ER-positive group was

further subdivided into subsets with distinctive gene expression profiles. In

univariate survival analysis, performed on the 49 patients diagnosed with locally

advanced disease but without evidence of distant metastasis, ER positivity was

not a significant prognostic factor on its own, but the luminal-type group enjoyed

a more favorable survival compared with the other groups. Conversely, the

basal-like group had a significantly poorer prognosis. Although small and

exploratory, this study suggests that important differences in outcome can be

ascertained from microarray expression profiling.

An interesting study was reported by Gruvberger et al. (5), who profiled 58

grossly dissected primary invasive breast tumors and used artificial neural network

analysis to predict the ER status of the tumors on the basis of their gene expression

patterns. They then determined which specific genes were themost important for ER

classification. By comparing to Serial Analysis of Gene Expression (SAGE) data

from estradiol-stimulated breast cancer cells, they determined that only a few genes

of the many genes that were associated with ER expression in tumors were indeed

estrogen responsive in cell culture. This observation lent further support to the

hypothesis developed by Perou et al. (1) that basic cell lineages, such as the luminal

ER-positive cell type, can be partly explained by observed genomic gene expression

patterns rather than downstream effectors of only one pathway, such as the ER.

PROGNOSTIC IMPLICATIONS

Microarrays have also been used to predict lymph node status and very short-

term relapse-free survival in two groups (n ¼ 37 and 52, respectively) of het-

erogeneously treated patients (6). Although prediction of nodal status is of

Pharmacogenetics of Breast Cancer 3


limited interest clinically, the study uses innovative statistical methods, rigor-

ously generates estimates of future classifier performance, and further demon-

strates the feasibility of accurate prediction of tumor biology using expression

arrays. In a more focused and somewhat more clinically relevant study, van ’t Veer

et al. (7) used RNA expression microarray analyses to identify a 70-gene

prognostic gene signature (“classifier”) in young, untreated, axillary lymph

node–negative patients using a training set of 44 good (disease free for more than

5 years) and 34 poor (distant relapse in less than 5 years) outcome tumors and

then tested the classifier in a validation set of 19 tumors. The same group (8) has

now extended the study to a total of 295 young (<53 years), stage I–II breast

cancer patients with both node-negative and node-positive disease, using the

Figure 1 Supervised classification of prognosis signatures (van ’t Veer). The 78 tumors

are listed vertically, and the 70 “prognostic” genes horizontally. (A) Study design. (B) Heatmap

of prognostic signature.

4 Chang et al.


70-gene classifier (7). The microarray-based predictions are consistent with, and

perhaps better than, estimates that can be obtained with current prognostic

indices.

GENETIC SUSCEPTIBILITY

A few studies have utilized new genomic approaches for the study of inherited

breast cancer (8). There is accumulating evidence, both epidemiological and

histological, that tumors arising as a result of mutations in the two breast cancer

susceptibility gene families (BRCA1 and BRCA2) are biologically distinct. For

instance, BRCA1 breast cancers are most often ER and PR negative, but BRCA2

cancers more often tend to be positive for these receptors (9). In a seminal paper

published by Hedenfalk et al. (10), seven tumors each from BRCA1 and BRCA2

gene mutation carriers or sporadic breast cancers were compared by expression

microarray analysis. They found that the gene expression profiles of the three

tumor groups differed significantly from each other, underscoring the funda-

mental differences between BRCA1 and BRCA2 mutation-associated tumors. Of

course a potential confounding issue was the differential distribution of ER

between the BRCA1 and BRCA2 tumors. However, even after removal of ER-

or PR-associated genes from the analysis, the two inherited tumor groups were

still discernable. Thus, ER status alone does not fully explain the observed

differences in gene expression profiles. Although this study is obviously very

small, and other confounding issues such as tumor stage, grade, and treatment

cannot be considered, it does set a foundation for larger validation studies to

confirm differential genes, which could then provide important clues to the

etiology of inheritable breast cancer.

Microarrays are also being studied as a way to predict response to systemic

therapy. The neoadjuvant setting is especially attractive for these studies for

several reasons, including early assessment of response to therapy, biopsiable

access to the primary tumor, and considerable reduced sample sizes compared to

those required in the adjuvant setting.

PREDICTIVE IMPLICATIONS

Methods for assessing response in neoadjuvant trials remain problematic. Clin-

ical response to neoadjuvant chemotherapy is a validated surrogate marker for

improved survival (11,12). Women who achieve pathologic complete response

are most likely to have the best clinical outcome, although survival is still

improved in those who clinically respond but do not achieve pathologic complete

response.

In an early study, Buchholz et al. (13) obtained sufficient RNA from core

biopsies of five patients to perform serial microarray expression profiles and

showed that despite differences in therapy, patients with good pathological



responses to neoadjuvant treatment appeared to have gene profiles that clustered

distinctly than those of patients who were poor responders to treatment. More

recently, Chang et al. (14) have shown that gene profiling can be used to

accurately predict response to neoadjuvant docetaxel. The study enrolled

24 subjects, extracted sufficient RNA from all core needle biopsies, and constructed

a 92-gene predictor of response. In a complete cross-validation analysis, which

gives an unbiased estimate of performance on future samples, the classifier

correctly identified 10 of 11 responders and 11 of 13 nonresponders with an

overall accuracy of 88%. In a small validation set, this 92-gene classifier suc-

cessfully predicted response in six patients (Fig. 2). This compares very favor-

ably with the best existing predictive factors for response to specific therapy and

strongly suggests that after appropriately extensive validation, microarray

profiling will be useful for treatment selection. A second neoadjuvant study was

recently published that used cDNA arrays to develop predictors for paclitaxel,

fluorouracil, doxorubicin, and cyclophosphamide, involving 24 samples. A

classifier with 74 markers was developed, with 78% accuracy, suggesting that

transcriptional profiling has the potential to identify a gene expression pattern in

breast cancer that may lead to clinically useful predictors of chemotherapy

response (15).

In vitro drug sensitivity data in NCI60 cell lines sensitive and resistant to

specific chemotherapeutic agents (adriamycin, cyclophosphamide, docetaxel,

etoposide, 5-fluoruracil, paclitaxel, topotecan) were recently published (16).

Gene set enrichment analysis was performed, and gene expression signatures

predictive of sensitivity to individual chemotherapeutic drugs were reported.

Each signature was validated with response data from an independent set of cell

line studies. These signatures were then used to predict clinical response in

individuals treated with these drugs. Notably, signatures developed to predict

response to individual agents, when combined, were found to also predict

response to multidrug regimens (16). These profiles are being validated in

prospective clinical trials.

However, we acknowledge that studies to construct and validate array-

based prognostic and predictive “markers” are complex. These studies must

address all of the concerns associated with ordinary, single-gene markers as well

as a number of considerations unique to array studies. Recommendations for

development of array-based prognostic classifiers have recently been enunciated

by Simon et al. (17). Among the most important points, they recommend that

studies should include (1) adequately large sample sizes in both training and

validation sets, (2) complete iteration of the entire classifier construction process

in estimating cross-validated prediction rates, (3) head-to-head comparison of

alternative classifiers on the same dataset, and (4) the full diversity of cases in

any validation set. In addition, gene expression patterns may be confounded by

several other factors, including ovarian ablation in premenopausal, ER-positive

patients and a different mechanism of action of combination chemotherapies.

6 Chang et al.


Groups of patients with different characteristics, such as menopausal and ER

status or human epidermal growth factor receptor 2 (HER2) overexpression, may

be necessary to definitively determine classifying patterns in these subsets of

patients.

Figure 2 Hierarchical clustering of genes correlated with docetaxel response. Sensitive

tumors (S) are defined as 25% residual disease, and resistant tumors (R) are defined as

greater than 25% residual disease.



CONCLUSIONS

The goal of comprehensive, genomewide approaches is to identify clinically

useful genetic profiles that will accurately predict the outcome to therapy and

prognosis of patients with breast cancer. Despite improvements in technology,

complex mechanisms driving breast cancer evolution continue to present chal-

lenges for the use of genomic approaches to better understand breast and other

cancers. This, combined with our use of different markers, methods, tumors (e.g.,

differing ER and HER2), and measurements of clinical outcomes, impedes the

development of a consensus regarding predictive and prognostic markers for

breast cancer.

As this field matures, genomic studies examining identical breast tumor

sets with multiple complementary technologies (e.g., loss of heterozygosity,

comparative genomic hybridization, gene expression array analysis) will prove

essential to unraveling the genetic heterogeneity characteristic of this disease. A

combined genomic approach is necessary to define the underlying heterogeneous

complexity characteristic of breast cancer. These data should lead to the iden-

tification and characterization of breast cancer subtypes, define the malignant

potential of a given lesion, and predict its sensitivity to specific therapies. These

multidisciplinary approaches should contribute to a better biological under-

standing, and therefore improved clinical management, of breast cancer.

ACKNOWLEDGMENTS

This work was supported by funding provided by National Institutes of Health

(NIH) grants P50 CA58183 and P01 CA30195, the Breast Cancer Research

Foundation, the Emma Jacobs Clinical Breast Cancer Fund, and the U.S. Army

Medical Research and Materiel Command DAMD17-01-0132 and W81XWH-

04-1-0468.

REFERENCES

1. Perou CM, Jeffrey SS, van de Run M, et al. Distinctive gene expression patterns in

human mammary epithelial cells and breast cancers. Proc Natl Acad Sci USA 1999;

96:9212–9217.

2. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.

Nature 2000; 406:747–752.

3. Sgroi DC, Teng S, Robinson G, et al. In vivo gene expression profile analysis of

human breast cancer progression. Cancer Res 1999; 59:5656–5661.

4. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas

distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA 2001;

98:10869–10874.

5. Gruvberger S, Ringner M, Chen Y, et al. Estrogen receptor status in breast cancer is

associated with remarkably distinct gene expression patterns. Cancer Res 2001; 61:

5979–5984.

8 Chang et al.


6. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast cancer

outcomes. Lancet 2003; 361:1590–1596.

7. van ’t Veer LJ, Dai H, van De Vijver MJ, et al. Gene expression profiling predicts

clinical outcome of breast cancer. Nature 2002; 415:530–536.

8. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a

predictor of survival in breast cancer. N Engl J Med 2002; 347:1999–2009.

9. Verhoog LG, Brekelmans CTM, Seynaeve C, et al. Survival in hereditary breast cancer

associated with germline mutations of BRCA2. J Clin Oncol 1999; 17:3396–3402.

10. Hedenfalk I, Duggan D, Chen Y, et al. Gene-expression profiles in hereditary breast

cancer. N Engl J Med 2001; 344:539–548.

11. Fisher B, Bryant J, Wolmark N, et al. Effect of preoperative chemotherapy on the

outcome of women with operable breast cancer. J Clin Oncol 1998; 16:2672–2685.

12. Chang J, Powles TJ, Allred DC, et al. Biologic markers as predictors of clinical

outcome from systemic therapy for primary operable breast cancer. J Clin Oncol

1999; 17:3058–3063.

13. Buchholz TA, Stivers DN, Stec J, et al. Global gene expression changes during

neoadjuvant chemotherapy for human breast cancer. Cancer J 2002; 8:461–468.

14. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling predicts

therapeutic response to docetaxel (TaxotereTM) in breast cancer patients. Lancet

2003; 362:280–287.

15. Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete

pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and

cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22:2284–2293.

16. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of che-

motherapeutics. Nat Med 2006; 12:1294–1300.

17. Simon R, Radmacher MD, Dobbin K, et al. J Natl Cancer Inst 2003; 95:14–18.



2

Pharmacology, Pharmacogenetics,

and Pharmacoepidemiology: Three

P’s of Individualized Therapy

Shaheenah Dawood

Department of Medical Oncology, Dubai Hospital,Dubai, United Arab Emirates

If it were not for the great variability among individuals medicine might as

well be a science and not an art.

Sir William Osler, 1892

INTRODUCTION

Over the last decade significant advances have been made in the management of

breast cancer. This is exemplified by the development of several new agents and

significantly improved survival rates. However, despite our best efforts breast

cancer remains the second most common cause of cancer-related death in the

United States with approximately 40,910 women expected to die of this disease

in 2007 (1). Unfortunately, such trends are not restricted to only breast cancer but

have been observed across a range of solid tumors. A large portion of the

underlying problem is attributed to the fact that the choice of a chemotherapeutic

agent is often empirical based on guidelines that assumes all patients with a

particular malignancy to be homogeneous. The unfortunate outcome of this

11


erroneous belief is that approximately 40% of patients may be receiving the

wrong drug (2).

Numerous studies have consistently observed that the efficacy and toxicity

of chemotherapeutic agents are heterogeneous rather than homogeneous across

populations. Despite a multitiered approach to determining the efficacy and safety

profiles of chemotherapeutic agents, spanning from in vitro assays, animal studies

through to large human population studies, the current practice of administering

the same dose of a given anticancer drug to a population of patients results in

considerable variation in clinical outcome. These effects cover a spectrum of

possibilities ranging from success to failure when considering treatment outcome

as an endpoint, and from no effect to a lethal event when treatment-associated

toxicity is the variable under study (3). Complicating the problem further is the

fact that currently available anticancer drugs show limited efficacy in up to 70% of

patients with adverse drug reactions accounting for as many as 100,000 patient

deaths and two million hospitalizations in the United States every year (4).

The key is to be able to, at the individual level, prospectively identify

patients at risk for severe toxicity and those likely to attain maximum benefit

from a given chemotherapeutic agent. Understanding that an individual does not

live in isolation but is in fact a product of the gene-environment interaction is

also vital to our understanding of the concept of individualized therapy (Fig. 1).

Figure 1 Sources of pharmacologic and pharmacogenetic variability.

12 Dawood


Designing treatment strategies based on individual efficacy and toxicity pre-

dictions will shift clinical practice from simply an “art” to a “science” by

incorporating a combination of physiological variables, genetic characteristics,

and environmental factors (all known to alter the dose-concentration time pro-

file) as well as population epidemiological elements resulting in a practice

paradigm of “personalized medicine.”

THE COMPONENTS OF INDIVIDUALIZED THERAPY: THE THREE P’s

When we administer a drug to a patient, our goal is to deliver a dose that will

achieve maximum benefit with little, if any, associated toxicity. Seldom is this

goal achieved. Individual differences in tumor response and normal tissue tox-

icity have been consistently observed with most anticancer agents. These dif-

ferences are due to a variety of clinical variables (e.g., gender, age, diet, and

drug-drug interactions) as well as variations in drug metabolism and transport.

It is believed that 30% of patients treated with a particular drug will exhibit

beneficial effects from the drug, 30% will not attain any form of beneficial

effect, 10% will experience only side effects from the drug, and 30% are non-

compliant with the drug often as a result of drug-related side effects or perceived

insufficient effect (5). This results in a waste of allocated resources with large

proportions of patients being unnecessarily exposed to adverse side effects. Here

we will describe three components that aim to explain the observed differences

in responses to individual drugs: pharmacology, pharmacoepidemiology, and

pharmacogenetics.

Current dosing of anticancer drugs is based on either a fixed quantity of

the drug or on a dose that is normalized to the individual body surface area.

This method makes the assumption that within a group of individuals there

will be a uniform degree of systemic exposure to the drug. With studies

reporting 2- to 10-fold variations in drug clearance, this assumption is clearly

not valid (6). With drugs used in the field of oncology having narrow thera-

peutic indices, it thus becomes imperative that we understand the mechanisms

behind the observed variability in drug response and toxicity when treating a

patient with cancer. At the clinical level, variability in drug response can be

explained by its pharmacology that describes the pharmacokinetic and phar-

macodynamic profiles of the drug. Pharmacokinetics explain “what the body

does to the drug” by describing the relationship between time and plasma

concentrations of the drug metabolites. This relationship is affected by varia-

bles such as absorption, distribution, metabolism, and excretion of the drug.

The underlying objective of pharmacodynamics is to describe “what the drug

does to the body” by correlating drug concentration to drug effect (both ben-

eficial and adverse effects). Both the pharmacokinetic and pharmacodynamic

components of a drug are not independent, but represent a spectrum of con-

tinuous events starting with the ingestion of the drug and culminating in the

observed clinical effect.

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 13


Variability in any of the pharmacokinetic components will affect the

observed drug effect. This is typically observed in patients with organ dys-

function where the inability to either metabolize or excrete the drug will lead to

exaggerated drug effects. For example, the use of carboplatin in patients with

renal dysfunction exaggerates the thrombocytopenic effect of this drug. Dosing

formulas using creatinine clearance are used to avoid this problem while max-

imizing dose intensity (7). Another example is the use of taxanes, which are

primarily metabolized and excreted by the liver. When administered to patients

with hepatic dysfunction, prolonged bouts of severe neutropenia are observed.

These examples illustrate how clinical variables can affect the pharmacokinetic

profile of a drug leading to disturbances in effect. Good clinical assessment of

the patients is therefore the first and most important step in designing an indi-

vidual treatment regimen that will maximize drug benefit.

Pharmacoepidemiology is a sub discipline of epidemiology that focuses on

understanding why individuals respond differently to drug therapy. It is defined

“as the study of the distribution and determinants of drug-related events in

populations and the application of this study to efficacious drug treatments” (8).

These drug-related events pertain to both beneficial and adverse events. Phar-

macoepidemology, through the use of observational methods, aims to quantify

drug exposure and to describe, as well as predict, the effectiveness and safety of

the drug within defined populations at specific places and time. Factors

important in interpreting the variability in drug outcome include a patient’s

health profile, sex, age, diet, associated comorbidities, disease severity and

prognosis, drug compliance, drug prescribing and dispensing quality, and genetic

profile of the patient and tumor (9,10). Depending on drug effect there are four

main groups of patients: (i) responders, (ii) non-responders, (iii) toxic res-

ponders, and (iv) toxic non-responders (Fig. 2).

Pharmacogenetics, a third and important component of individualized

therapy, focuses on describing the extent to which an individual’s genetic

makeup is responsible for the observed differences in drug efficacy and toxicity

profiles (11,12). This information is then used to make predictions about the

safety, toxicity, and efficacy of drugs in individual patients. Inherited variability

of drug targets, drug metabolizing enzymes, and drug transporters may all have a

major impact on overall drug response, disposition, and associated adverse side

effects by altering the pharmacokinetic and pharmacodynamic properties of

drugs (Table 1).

Genetic variations include nucleotide repeats, insertions, deletions, and

single nucleotide polymorphisms (SNPs). SNPs are the simplest and most

commonly studied DNA polymorphism that occur once every 1900 base pairs in

the 3 billion bases in the human genome (13) and account for more than 90% of

the genetic variation observed. More than 1.4 million SNPs have been identified

in the human genome. Of these more than 60,000 SNPs occur in the coding

region of genes with some SNPs being associated with variations in drug

metabolism and effects (13). When polymorphisms occur in the promoter region

14 Dawood


it can effect the transcription of the gene. When the polymorphism occurs in the

exon (and intron) region or 30UTR of the gene, it can affect translation and RNA

stability, respectively, resulting in either reduced or enhanced activity of the

encoded protein (3).

Two approaches have been used to identify clinically relevant poly-

morphisms. In the reverse genetics approach, a drug is first given to a group

individuals and the variability in response to and disposition of the drug is

observed. Phenotypes of clinical importance are prioritized and then experiments

Table 1 Examples of Genetic Variations and Their Associated Phenotypic Changes

Enzymes Substrate Phenotype

CYP3A4/3A5 D, P, E, T Interindividual variability in

pharmacokinetics

DPD 5-FU Increased toxicity

P-glycoprotein Various anticancer agents Drug resistance

TS 5-FU and antifolates Drug resistance

TPMT 6-MP Increased hematological toxicity

UGT1A1 Irinotecan Increased diarrhea and neutropenia

Abbreviations: D, docetaxel, P, paclitaxel; E, etoposide, T, tamoxifen; 6-MP, mercaptopurine, TPMT,

thiopurine methyltransferases, UGT1A1, UDP-glucuronosyltrasferase 1A1, DPD, dihydropyrimidine

dehydrogenase.

Figure 2 A group of individuals with the same diagnosis do not respond homogeneously

to a single drug. Through the elements of phamacology, pharmacoepidemiology, and

pharmacogenetics, the aim is to identify patients at risk for toxicity or reduced response to

therapy prior to medication initiation.



are conducted to determine if a genetic component is present to explain the

observed variability. Finally, the biochemical and molecular nature of the vari-

ability is determined through extensive resequencing techniques that will link the

polymorphisms with the observed phenotypes. The second and more cost effective

method is the forward genetics approach. This method involves first determining

the presence of genetic variability at specific loci in a group of individuals. A

hypothesis is then formulated regarding these polymorphisms causing phenotypic

variability. Finally, an experiment is performed to investigate the clinical con-

sequences of the genetic variability. Because of the multitude of polymorphisms,

the challenge today is to be able to identify polymorphisms that occur in gene

regulatory or coding regions that have clinical relevance. The field of oncology

adds another level of complexity as issues such as clinically relevant tumor

genome variations not found in the germline and somatic mutations acquired

before or after chemotherapy can affect drug efficacy and toxicity. Moreover,

different polymorphisms of certain genes may exhibit different phenotypes among

different malignancies. Polymorphisms of glutathione S-transferases (GSTs) are a

common example. GSTs facilitate the conjugation of glutathione to xenobiotics,

including anticancer agents (e.g., cyclophosphamide and anthracyclines), envi-

ronmental carcinogens, and reactive oxygen products (14). A variety of func-

tionally important polymorphisms of GSTs have been described. The presence of

the GSTT1 polymorphism has been shown to predispose children treated inten-

sively for acute myelogenous leukemia to greater toxicity and death during

remission (15), while women with breast cancer treated with cyclophosphamide

carrying the GSTP1 polymorphism are more likely to survive (16). This example

illustrates how the phenotypic effect of polymorphisms depends on the underlying

tumor as well as the intensity of regimen used to treat the malignancy. However,

as we begin to unravel and accurately identify the mechanism of action of many

anticancer drugs, polymorphisms in candidate genes likely to influence drug

efficacy and associated adverse events can be identified. Indeed the field of

pharmacogenetics has transitioned from one that determined the genetic basis of

observed inter patient phenotypic variability to one that determines the phenotypic

consequence of an observed genotypic variability.

Linking the information obtained from the pharmacology and pharma-

coepidemiology of a drug and the pharmacogenetics that describes the genetic

makeup of the individual being treated with the drug is thus vitally important in

developing an optimal approach to individualized therapy. The use of isoniazid

in the treatment of tuberculosis is a good example. Epidemiological evidence

shows that polymorphisms differ in frequency among ethnic and racial groups.

This phenomenon was shown early on when interindividual and interethnic

variations of efficacy and toxicity associated with isoniazid treatment was

observed in a cohort of individuals with tuberculosis. The enzyme responsible

for the metabolism of isoniazid is N-acetyltransferase, a phase II conjugating

liver enzyme. Early studies of isoniazid revealed a bimodal distribution in its

plasma concentration among individuals. High plasma concentrations were

16 Dawood


observed among individuals with the slow acetylator phenotype often resulting

in peripheral nerve damage, while low plasma concentrations of isoniazid were

observed among individuals with fast acetylator phenotype and were not affected

(17). Studies have shown large variations in the incidence of the slow acetylator

phenotype (inherited as an autosomal recessive trait) among different ethnic

groups (40–70% of Caucasians and African Americans, 10–20% of Japanese, and

more than 80% of Egyptians) (18,19). Amonafide (20), a topoisomerase II

inhibitor, is another substrate of N-acetyltransferase whose clinical development

has been hampered by highly variable and unpredictable toxicity caused, at least,

in part, by interindividual differences in N-acetylation with fast acetylators

experiencing greater myelosuppression than slow acetylators. This example

illustrates both the importance of pharmacoepidemiological evidence in identi-

fying and characterizing phenotypic variance to individual drugs and how the

pharmacological effects of these drugs depend on pharmacogenetic determinants

that define sensitivity to the drug, and ultimately describes the drug’s behavior.

THE LEUKEMIA STORY: WHAT DID WE LEARN?

Pharmacological and pharmacogenetic studies are increasingly being used to

optimize existing drug therapy with the goal of reducing toxicity while max-

imizing efficacy, leading to improved survival. By utilizing this method

remarkable progress has been observed in the treatment of acute lymphoblastic

leukemia (ALL) with cure rates increasing from a mere 10% (four decades ago)

to nearly 80% in children and 40% in adults (21). This has been achieved

through optimization of use of a variety of commonly used antileukemic drugs

such as mercaptopurine and methotrexate.

Thiopurine methyltransferase (TPMT) is an enzyme that converts thio-

purine prodrugs, such as mercaptopurine, into inactive methylated metabolites.

Prior to inactivation, mercaptopurine is converted into a variety of active

nucleotide metabolites that are incorporated into DNA and RNA, thereby

inducing an antileukemic effect. Patients with an inherited TPMT deficiency

suffer severe, potentially fatal hematopoietic toxicity when exposed to standard

doses of mercaptopurine (and other thiopurine prodrugs such as azathioprine and

thioguanine). More than 95% of the clinically relevant TPMT mutations are

accounted for by three nonsynonymous SNPs (TPMT*2, TPMT*3A, and

TPMT*3C) (22). TPMT activity is inherited as an autosomal codominant trait

with approximately 95% of the population homozygous for the wild-type allele

TPMT*1 (that have full enzyme activity), 10% heterozygous for the poly-

morphism (that have intermediate levels of enzyme activity), and 1 in 300

individuals will carry two mutant TPMT alleles (that do not express functional

TPMT) (23). Using a pharmacogenetic test, developed at St. Jude Hospital,

patients are often classified according to their levels of TPMT activity into

normal, intermediate, and deficient. This test has a concordance rate of 100%

between genotype and phenotype with studies revealing no long- term outcome



changes related to dosage individualization based on the TPMT status of a

patient (24) (Fig. 3). This illustrates how the clinical application of pharmaco-

genetics can be used to minimize toxicity without compromising efficacy.

Methyltetrahydrofolate reductase (MTHFR), a crucial enzyme in the

metabolism of folic acid, catalyses the reduction of 5,10-methylenetetrahy-

drofolate to 5-methyltetrahydrofolate (the predominant circulatory form of folate

and the carbon donor for the demethylation of homocysteine to methionine). As

methotrexate increases serum levels of homocysteine and the active poly-

glutamate metabolites of methotrexate inhibit MTHFR function, patients with a

low level of MTHFR function are at an increased risk of methotrexate-induced

toxicities such as oral mucosities. Two polymorphisms associated with reduced

MTHFR activity have been described (25): (i) 677C > T polymorphism that is

homozygous in 10% of the population where it encodes the enzyme with 30% of

the wild-type enzyme activity and heterozygous in 40% of the population where

it encodes an enzyme that has 60% of the wild type-enzyme activity and (ii)

1298A > C (less common than 677C > T). Recent findings have also suggested

that MTHFR polymorphisms may protect against the development of both adult

and pediatric ALL (26).

5-FLUOROURACIL

5-Fuorouracil (5-FU) is a uracil analog widely used to treat solid tumors such as

breast and colorectal cancer (27). 5-FU is inactivated by dihydropyrimidine

dehydrogenase (DPD), an enzyme that exhibits up to 20-fold variation in activity

among individuals (28). Mutations that result in low DPD activity predispose to

potentially life-threatening gastrointestinal, hematopoietic, and neurological

toxicities (29). Up to 20 mutations (30) have been identified that cause inacti-

vation of DPD. In the general population it has been estimated that 3% to 5% of

individuals are heterozygous and 0.1% are homozygous carriers for the mutation

that inactivates DPD, with DPYD*2A being the most inactivating allele (31).

Although polymorphisms associated with DPD inactivation have been identified,

a number of patients still exhibit severe 5-FU toxicity with no detectable

Figure 3 Individualization of dosing based on TPMT status does not affect cumulative

incidence of relapse. Source: From Ref. 24.

18 Dawood


mutations in the coding region of the DPD gene explaining the difficulty in

implementation of DPD pharmacogenetics in the prospective identification of

high-risk patients for severe 5-FU toxicity.

One of the primary mechanisms of action of 5-FU is inhibition, via fluo-

rodeoxyuridine monophosphate (an active metabolite of 5-FU), of thymidylate

synthase (TS), a critical enzyme required for de novo thymidylate synthesis,

which is required for DNA synthesis and repair (32). Polymorphisms in the gene

encoding TS are one of the few examples of a polymorphism that is associated

with the genetic makeup of the tumor with clinical studies showing an inverse

relationship between TS levels and antitumor response (33). The TSER poly-

morphism of TS is characterized by tandem repeats in the TS promoter enhancer

region of the gene. Higher levels of TS expression and enzyme activity have been

observed with increasing copies of tandem repeats, which result in lower tumor

response rates to 5-FU (34). In a study by Marsh et al (35), it was observed that in

a cohort of patients with metastatic colorectal tumors receiving 5-FU-based che-

motherapy 2 tandem repeats was nearly twice as common in responders than in

non-responders with longer median survival observed in patients with 2 tandem

repeats compared to those with 3 tandem repeats (16 months vs. 12 months) (35).

Polymorphisms of TS that result in 5-FU variability of its pharmacody-

namic and pharmacokinetic properties illustrate the important fact that cancer

being a polygenic disease will need a polygenic solution to guide therapy. To

accurately select patients that are most likely to tolerate and respond to 5-FU

therapy, one needs to use combined genotyping of DPYD and TSER functional

variants as well as taking into account nongenetic factors.

TAXANES

Paclitaxel and Docetaxel are the two most commonly used taxanes in the

treatment of both early and advanced stage breast cancer (36). They exert their

cytotoxic effects by binding to B-tubulin that subsequently results in the stabi-

lization of microtubules and disruption of mitotic spindle formation during cell

division ultimately resulting in cell death. Genes involved in drug transport,

metabolism, and drug target all play an important role in overall taxane efficacy,

with their polymorphisms partly explaining the variability observed in toxicity

and response to these drugs.

The multidrug-resistant transporter, P-glycoprotein, encoded by ABCB1

(MDR1) and expressed in multiple cell types, has been shown to mediate both

paclitaxel and docetaxel influx (37,38). Three polymorphisms in ABCB1 have

been commonly studied (3435C>T, 1236C>Y, and 2677G>T/A) (39), with

most studies showing inconsistent results when trying to establish an association

between these polymorphisms and drug efficacy/toxicity. Studies focusing on

looking at ABCB1 polymorphism in combination may yield more useful clinical

information. Some of the recent studies have reported an association of the

various ABCB1 polymorphisms with tumor response to paclitaxel (ABCB1



2677G>T/A) (40), risk of neutropenia with docetaxel (ABCB1 3435C>T) (41),

and risk of neuropathy with paclitaxel (ABCB1 3435C>T) (42). Oxidative

taxane metabolism occurs via the CYP450 pathway, with CYP2C8 and CYP3A4

primarily responsible for paclitaxel metabolism while docetaxel is mainly

metabolized by CYP3A4 (43,44). The functional significance of the poly-

morphisms of these enzymes has not been established with associations seen in

in vivo studies (45) not observed in studies of cancer patients (46).

TAMOXIFEN

Tamoxifen is a selective estrogen receptor modulator (SERM) that is widely

used in the treatment and prevention of breast cancer. When given in the

adjuvant setting, it reduces the annual risk of recurrence of breast cancer by half

and the breast cancer mortality rate by one-third (47). The pharmacology and

metabolism of tamoxifen is complex and undergoes extensive phase I and II

metabolism with the production of endoxifen and to a lesser extent 4-hydrox-

ytamoxifen (48). These metabolites play an important role in the anticancer

effect of tamoxifen through their high-affinity binding to estrogen receptors and

suppression of estradiol-stimulated cell proliferation (49).

The clinical effects of tamoxifen are highly variable (50) due to the vari-

ability observed in the plasma levels of the more potent metabolites of tamoxifen.

Tamoxifen is converted to endoxifen (4-hydroxy-N-desmethyltamoxifen) pri-

marily by CYP2D6-mediated oxidation of N-desmethyl tamoxifen, while the

conversion of tamoxifen to 4-hydroxy tamoxifen is catalyzed by multiple enzymes

(51). Plasma concentrations of endoxifen is thus sensitive to variants of the

CYP2D6 genotype with the most common variant making the enzyme inactive

(common allele associated with CYP2D6 poor metabolizer phenotype is

CYP2D6*4). One prospective trial revealed that women treated with tamoxifen

who either carried the genetic variants associated with low or absent CYP2D6

activity or were treated concomitantly with drugs that inhibited the activity of

CYP2D6 had lower levels of endoxifen (52). Subsequently, using tissue sample

from a cohort of early stage breast cancer patients treated with tamoxifen, it was

shown that women with a homozygous CYP2D6 phenotype tended to have a

higher risk of disease relapse and a lower incidence of hot flashes (53).

Polymorphisms associated with the activity of the metabolites of tamox-

ifen teach us two important lessons. First, the interindividual variability in the

response to tamoxifen is explained both by genetic variation in CYP2D6 and

the coadministration of drugs that inhibit the activity of CYP2D6. Second, the

toxicity associated with drug can help identify responders.

CONCLUSION

At present time, dosages of anticancer drugs are administered based on the

assumption of homogeneity among individuals and are rarely based on indi-

vidual characteristics. As a result we encounter a situation where by a large

20 Dawood


number of individuals experience unwanted adverse side effects after treatment

without attaining maximum drug efficacy. Such adverse drug reactions impose a

huge burden not only on the patient but also on the health care system. A better

understanding of the pharmacoepidemiological, pharmacological, and pharma-

cogenetic determinants of a chemotherapeutic agent is needed to optimize drug

dosing regimens to obtain maximal efficacy with minimal toxicity. Advance-

ments in the field of pharmacogenetics hold promise in helping to identify which

polymorphisms are likely to be important in clinical oncology thereby enabling

the development of effective agents that will make their way to the clinic in a

timely manner. Realizing the importance of each of the components of indi-

vidualized therapy and incorporating each one into the infrastructure of clinical

trials will allow us to further explore and define the relationships between dif-

ferent polymorphisms and observed phenotypes. Considering the fact that anti-

cancer drugs have narrow a therapeutic range oncology is a field that will

certainly benefit from the application of pharmacogenetic and pharmacological

principles to dosage individualization. The idea of the “one model fits all” should

be accepted as grossly flawed and, as we progress in the field of cancer thera-

peutics, will be replaced in the future with the application of “personalized

cancer chemotherapy” into routine clinical practice.

REFERENCES

1. Jemal A, Siegal R, Ward E, et al. Cancer statistics 2007. CA Cancer J Clin 2007;

57:43–66.

2. Bordet R, Gautier S, Le Louet H, et al. Analysis of the direct cost of adverse drug

reactions in hospitalized patients. Eur J Clin Pharmacol 2001; 56:935–941.

3. Evans WE, Reilling MV. Pharmacogenomics: Translating functional genomics into

rational therapeutics. Science 1999; 286:487–491.

4. Jeffrey S, Ross MD, David P, et al. Pharmacogenomics. Adv Anat Pathol 2004;

11:211–220.

5. Maitland-van der Zee AH, de Boer A, Leufkens HG. The interface between phar-

macoepidemiology and pharmacogenetics. Eur J Pharmacol 2000; 410:121–130.

6. Gaplin AJ, Evans WE. Therapeutic drug monitoring in cancer management. Clin

Chem 1993; 39:2419–2430.

7. Calvert A, Newell D, Gumbrell L, et al. Carboplatin dosage: prospective evaluation

of a simple formula based on renal function. J Clin Oncol 1989; 7:1748.

8. Last JM. A Dictionary of Epidemiology. Oxford: Oxford University Press, 2000.

9. Sander C. Genomic medicine and future of health care. Science 2000; 287:1977–

1978.

10. Vesell E.S. Therapeutic lessons from pharmacogenetics. Ann Intern Med 1997;

126:653–655.

11. Evans WE, Relling MV. Pharmacogenomics: Translating functional genomics into

rational therapeutics. Science 1999; 286:487–491.

12. Relling MV, Dervieux T. Pharmacogenetics and cancer therapy. Nat Rev Cancer

2001; 1:99–108.



13. Sachidanadam R, Weissman D, Schmidt SC, et al. A map of human genome

sequence variation containing 1.42 million single nucleotide polymorphisms. Nature

2001; 409:928–933.

14. Ketterer B. Protective role of glutathione and glutathione transferases in mutagenesis

and carcinogenesis. Mutat Res 1988; 202:343–361.

15. Davies SM, Robison LL, Buckley JD, et al. Glutathione S-transferase poly-

morphisms and outcome of chemotherapy in childhood acute myeloid leukemia.

J Clin Oncol 2001; 19:1279–1287.

16. Sweeney C, McClure GY, Fares MY, et al. Association between survival after

treatment for breast cancer and glutathione S-transferase P1 Ile105Val poly-

morphism. Cancer Res 2000; 60:5621–5624.

17. Evans DA, Manley KA, Mckusch VA, et al. Genetic control of isoniazid metabolism

in man. Br Med J 1960; 2:485–491.

18. Knight RA, Selin MJ, Harris HW, et al. Genetic factors influencing isoniazid blood

levels in humans. Trans Conf Chemother Tuberc 1959; 8:52–56.

19. Evans Da. Survey of human acetylator polymorphism in spontaneous disorders.

J Med Genet 1984; 21:243–253.

20. Ratain MJ, Mick R, Berezin F, et al. Paradoxical relationship between acetylator

phenotype and amonafide toxicity. Clin Pharmacol Ther 1991; 50:573–579.

21. Pui C-H, EvansWE. Acute lymphoblastic leukemia. N Engl J Med 1998; 339: 605–615.

22. McLeod HL, Coulthard S, Thomas, AE et al. Analysis of thiopurine methyl-

transferase variant alleles in childhood acute lymphoblastic leukemia. Br J Hematol

1999; 105:696–700.

23. Krynetski EY, Tai HL, Yates CR, et al. Genetic polymorphisms of thiopurine

s-methyltransferase: clinical importance and molecular mechanisms. Pharmacoge-

netics 1996; 6:279–290.

24. Relling MV, Pui C-H, Cheng C, et al. Thiopurine methyltransferase in acute lym-

phoblastic leukemia. Blood 2006; 107:843–844.

25. Molloy AM, Daly S, Mills JL, et al. Thermolabile variant of 5,10-methyenete-

trahydrofolate reductase associated with low red-cell folates: implications for folate

intake recommendations. Lancet 1997; 349:1591–1593.

26. Franco RF, Simoes BP, Tone LG, et al. The methylenetetrahydrofolate reductase

C677T gene polymorphism decreases the risk of childhood acute lymphocytic leu-

kemia. Br J Haematol 2001; 115:616–618.

27. Heidelberger C, Chaudhuri NK, Danneberg P, et al. Fluorinated pyrimidines, a new

class of tumor – inhibitory compounds. Nature 1957; 179:663–666.

28. Etienne MC, Lagrange JL, Dassonville O, et al. Population study of dihydropyr-

imidine dehydrogenase in cancer patients. J Clin Oncol 1994; 12:2248–2253.

29. Raida M, Schwabe W, Hausler P, et al. Prevalence of a common point mutation in

the dihydropyrimidine dehydrogenase (DPD) gene within the 5’-splice donor site of

intron 14 in patients with severe 5-fluorouracil (5-FU)-related toxicity compared

with controls. Clin Cancer Res 2001; 7:2832–2839.

30. McLeod HL, Collie-Duguid ES, Vreken P, et al. Nomenclature for human DPYD

alleles. Pharmacogenetics 1998; 8:455–459.

31. Lu A, Zhang R, Diasio RB. Dihydropyrimadine dehydrogenase activity in human

peripheral blood mononuclear cells and liver: population characteristics, newly

identified deficient patients, and clinical implication in 5-fluorouracil chemotherapy.

Cancer Res 1993; 53:5433–5438.

22 Dawood


32. Grem JL. 5-Fluorouracil: forty-plus and still ticking. A review of its preclinical and

clinical development. Invest New Drugs 2000; 18:299–313.

33. Leichman CG, Lenz HJ, Leichman L, et al. Quantitation of intratumoral thymidylate

synthase expression predicts for disseminated colorectal cancer response and resis-

tance to protracted-infusion fluorouracil and weekly leucovorin. J Clin Oncol 1997;

15:3223–3229.

34. Horie N, Aiba H, Oguro K, et al. Functional analysis and DNA polymorphism of the

tandemly repeated sequences in the 5’-terminal regulatory region of the human gene

for thymidylate synthase. Cell Struct Funct 1995; 20:191–197.

35. Marsh S, Mckay JA, Cassidy J, et al. Polymorphism in the thymidylate synthase

promoter enhancer region in colorectal cancer. Int J Oncol 2001; 19:383–386.

36. Nabholtz JM, Gligorov J. The role of taxanes in the treatment of breast cancer.

Expert Opin Pharmacother 2005; 6:1073–1094.

37. Shirakawa K, Takara K, Tanigawara Y, et al. Interaction of docetaxel (‘Taxotere’)

with human P-glycoprotein. Jpn J Cancer Res 1999; 90:1380–1386.

38. Sparreboom A, Van Asperen J, Mayer U, et al. Limited oral bioavailability and

active epithelial excretion of paclitaxel (Taxol) caused by P-glycoprotein in the

intestine. Proc Natl Acad Sci U S A 1997; 94:2031–2035.

39. Lepper ER, Nooter K, Verweij J, et al. Mechanisms of resistance to anticancer drugs:

the role of the polymorphic ABC transporters ABCB1 and ABCG2. Pharmacoge-

nomics 2005; 6:115–138.

40. Green H, Soderkvist P, Rosenberg P, et al. mdr-1 Single nucleotide polymorphisms

in ovarian cancer tissue: G2677T/A correlates with response to paclitaxel chemo-

therapy. Clin Cancer Res 2006; 12:854–859.

41. Tran A, Jullien V, Alexandre J, et al. Pharmacokinetics and toxicity of docetaxel:

role of CYP3A, MDR1, and GST polymorphisms. Clin Pharmacol Ther 2006;

79:570–580.

42. Sissung TM, Mross K, Steinberg SM, et al. Association of ABCB1 genotypes with

paclitaxel-mediated peripheral neuropathy and neutropenia. Eur J Cancer 2006;

42:2893–2896.

43. Taniguchi R, Kumai T, Matsumoto N, et al. Utilization of human liver microsomes

to explain individual differences in paclitaxel metabolism by CYP2C8 and CYP3A4.

J Pharmacol Sci 2005; 97:83–90.

44. Leyseng-Williamson KA, Fenton C. Docetaxel a review of its use in metastatic

breast cancer. Drugs 2005; 65:2513–2531.

45. Bahadur N, Leathart JB, Mutch E, et al. CYP2C8 polymorphisms in Caucasions and

their relationship with paclitaxel 6 a- hydroxylase activity in human liver micor-

somes. Biochem Pharmacol 2002; 64:1579–1589.

46. Henningsson A, Marsh S, Loos WJ, et al. Association of CYP2C8, CYP3A4,

CYP3A5, and ABCB1 polymorphisms with pharmacokinetics of paclitaxel. Clin

Cancer Res 2005; 11:8097–8104.

47. Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and

15-year survival: An overview of the randomized trials. Lancet. 2005; 365:1687–1717.

48. Lonning PE, Lien EA, Lundgren S, et al. Clinical pharmacokinetics of endocrine

agents used in advanced breast cancer. Clin Pharmacokinet 1992; 22:327–358.

49. Stearns V, Johnson MD, Rae JM, et al. Active tamoxifen metabolite plasma con-

centrations after co administration of tamoxifen and the selective serotonin reuptake

inhibitor paroxetine. J Natl Cancer Inst 2003; 95:1758–1764.



50. Osborne CK. Tamoxifen in the treatment of breast cancer. N Engl J Med 1998;

339:1609–18.

51. Desta Z, Ward BA, Soukhova NV, et al. Comprehensive evaluation of tamoxifen

sequential biotransformation by the human cytochrome P450 system in vitro: Promi-

nent roles for CYP3A and CYP2D6. J Pharmacol Exp Ther 2004; 310:1062–1075.

52. Jin Y, Desta Z, Stearns V, et al. CYP2D6 genotype, antidepressant use, and tamoxifen

metabolism during adjuvant breast cancer treatment. J Natl Cancer Inst 2005; 97:30–39.

53. Goetz MP, Rae JM, Suman VJ, et al. Pharmacogenetics of tamoxifen bio-

transformation is associated with clinical outcomes of efficacy and hot flashes. J Clin

Oncol 2005; 23:9312–9318.

24 Dawood


3

Role of Genetic Variability in Breast

Cancer Treatment Outcomes

Kandace L. Amend

Department of Oncological Sciences, Mount Sinai School of Medicine,New York, New York, U.S.A.

Ji-Yeob Choi and Christine B. Ambrosone

Department of Cancer Prevention and Control, Roswell ParkCancer Institute, Buffalo, New York, U.S.A.

INTRODUCTION

Variability in toxicity and response to the same dose of medication may occur

among cancer patients and contribute to the unpredictability of clinical out-

comes. In addition to the potential importance of clinical factors in determining

drug effects (1), inherited differences in drug metabolism are likely to have a

profound effect on treatment outcomes (2). Pharmacogenetics, the study of the

contribution of genetic variation to these interindividual differences, may allow

for the individualization of therapy and improved prediction of pharmacody-

namic events, including host toxicity and treatment efficacy. To date, the role of

pharmacogenetics in breast cancer therapy has not been clearly defined. Herein

we discuss the drugs most commonly used to treat breast cancer and describe

their metabolic pathways, as well as provide epidemiologic support for the role

of pharmacogenetics in breast cancer outcomes.

25


BREAST CANCER THERAPEUTICS

The most frequently used agents to treat breast cancer include cyclophosphamide

(CP) in combination with doxorubicin [Adriamycin1 (A)] or with methotrexate

(MTX) and 5-flourouracil (5FU) (3). The taxanes paclitaxel (Taxol1) and

docetaxel (Taxotere1) have an important role in the treatment of early-stage

breast cancer as well as metastatic disease. In patients with estrogen receptor

(ER)-positive tumors, tamoxifen (TAM) reduces the risk of recurrence by 31%

and breast cancer morality by one-third (4), and randomized trials have indicated

that treatment with third-generation aromatase inhibitors (AIs) is now an

important option for the adjuvant treatment of hormone receptor–positive breast

cancer (5,6).

Multiple enzymatic pathways are involved in the metabolism and detox-

ification of chemotherapeutic agents, and common genetic polymorphisms in

drug metabolizing enzymes are likely to contribute to interindividual variability

in toxicity and cancer recurrence (7).

Cyclophosphamide

Cyclophosphamide requires multistep activation before it can function as an

antitumor alkylating agent. The drug is first oxidized in the liver to the active

4-hydroxy metabolite mainly by CYP2B6, although other enzymes, including

CYP2A6, CYP3A4, CYP2C8, CYPC29, and CYP2C19, are involved. Drug

intermediates are transported in blood and decompose spontaneously to phos-

phoramide mustard and acrolein, both of which can react with nucleophilic

groups, forming DNA cross-links and inhibition of protein synthesis (8). Inac-

tivating glutathionyl conjugation reactions of C intermediates are catalyzed by

glutathione S-transferases (GSTs) GSTA1 and GSTP1 (9). Variability in the

activity of these activating and detoxifying enzymes is likely to affect ultimate

tumor tissue dose and efficacy of treatment.

The human cytochrome P450 CYP2B6 gene is involved in the metabolic

activation of a number of clinically important chemotherapeutic drugs for breast

cancer, including CP (10,11) as well as the antioestrogen TAM (12). Several

single nucleotide polymorphisms (SNPs) in CYP2B6 have been described, some

having functional significance (13) and affecting the pharmacokinetic parame-

ters of CP. In cell culture, enzymatic activities in microsomes from COS-1 cells

expressing a K262R amino acid substitution (CYP2B6*4, CYP2B6*6, and

CYP2B6*7) showed increased values for Vmax and Vmax/Km compared with that

of the wild type (CYP2B6*1) (14), while a second study reported that the amino

acid substitution Q172H resulting from the CYP2B6*6 allele enhanced

7-ethoxycoumarin O-deethylase activity (15). A study by Lang et al. (2001)

identified a total of nine SNPs by sequencing DNA from 35 subjects and

reported extensive variability in the expression and activity of CYP2B6, where

CYP2B6 expression was significantly reduced in carriers of the R487T, Q172H,

26 Amend et al.


and K262R polymorphisms (16). Ethnic variation in allelic variants and CYP2B6

expression has been also been observed (17,18).

CYP3A4 also plays a large role in the metabolism of CP, and the CYP3A4

gene has a number of known variant alleles, including some having functional

significance (19). A single base substitution in the 50 promoter region of the gene

(20), present in 9% of European-Americans and 53% of African-Americans (21),

with the CYP3A4*1B allele associated with expression levels is 1.4-fold higher

than that in common alleles (22). Using enzyme kinetic analyses to evaluate

intersample variation in activation of CP by CYP3A and 2B in human liver, both

Chang et al. (1993) and Roy et al. (1999) found considerable interindividual

differences in CYP3A and CYP2B levels and variation over a wide range in CP

hydroxylation (10,23). Variation in CYP3A enzyme levels and expression may

be related, in part, to genetic differences or to other chemotherapeutic and

natural agents that have been shown to increase CYP3A4 expression (24).

CYP3A5 accounts for at least 50% of the total CYP3A content in most

Caucasians and African-Americans and is thus considered an important con-

tributor to interindividual differences in CYP3A-dependent drug clearance (25).

Although CYP3A5 is polymorphically expressed at high levels in a minority of

Caucasians, the expression in human liver is more frequent in African-

Americans (60%) than in Caucasians (33%) (25). Reduced expression in some

individuals may be due to a common polymorphism in intron 3 of CYP3A5

(A6986G), which generates a cryptic splice site (CYP3A*6) resulting in a

truncated amino acid (25,26).

Genetic polymorphisms have been described in several of the CYP2C

family members, including CYP2C8, CYP2C29, CYP2C18, and CYP2C19,

although the catalytic contribution of CYP2C8 and CYP2C18 to activation of CP

is likely minor compared with the contributions of CYP2C9 and CYP2C19. In

vitro, CYP2C19 was found to bioactivate CP in human liver microsomes (27),

and while both CYP2C9 and CYP2C19 were found to possess CP 4-hydroxylase

activity in vitro, CYP2C19 exhibited higher CP 4-hydroxylase activity following

only CYP2B6 (23). Although known variants in CYP2C9 and CYP2C19 exist,

little is known regarding their contribution to pharmacokinetic variability in

relation to breast cancer prognosis. In a study by Xie et al., CYP2C9*2,

CYP2C9*3, CYP2C19*2, and CYP2C19*3 were not significantly correlated with

the metabolism of CP in 19 human liver specimens (11).

Phase II drug metabolism is catalyzed by GSTs, which consist of several

isoenzymes, including mu (GSTM1), pi (GSTP1), alpha (GSTA1), theta

(GSTT1), and zeta. The GSTs detoxify many chemotherapeutic drugs, including

CP, are induced by oxidative stress, and detoxify the reactive oxygen species

(ROS) that are produced by chemotherapeutic drugs and radiation therapy.

GSTA1, the enzyme most active in glutathione conjugation reactions with CP

intermediates, has a polymorphism in the 50 promoter region of the gene (28).

Coles et al. (2001) identified substitutions designated as GSTA1*A (�567, �69,

and �52) and GSTA1*B (�631, �567, and �69) (29), with the GSTA1*B variant

Role of Genetic Variability in Breast Cancer Treatment Outcomes 27


showing decreased hepatic expression and increased GSTA2 expression com-

pared with GSTA1*A homozygotes. Given that GSTA1 increases the conjugation

of CP intermediates (9), these differences in hepatic expression may affect the

detoxification of CP metabolites and, ultimately, the outcomes of treatment and

survival.

For the GSTP1 gene, two variant alleles, GSTP1*B and GSTP1*C, have

been detected (30). A single amino acid change from isoleucine to valine occurs

at codon 105 (Ile105Val) of the GSTP1 gene, with genotype frequencies in

Caucasians 51% homozygous for Ile/Ile, 43% heterozygous for Ile/Val, and 6%

homozygous for the variant allele Val/Val (31). The allelic variant 105Val

exhibited lower glutathione conjugation of the alkylating agent thiopeta com-

pared with the isoleucine allele (32). GSTP1 variants may also differ in activity

toward CP, structurally similar to thiotepa, and patients with GSTP1 allelic

variants may have reduced activity in removal of active CP.

Anthracyclines

The commonly used anthracyclines include doxorubicin, daunorubicin, and

epirubicin, although doxorubicin and epirubicin are used more frequently for the

treatment of breast cancer. Several different mechanisms have been proposed for

anthracycline tumor activity, including free radical generation, DNA binding and

alkylation, DNA cross-linking, interference with DNA unwinding or strand

separation, direct membrane effects, inhibition of topoisomerase II, and induc-

tion of apoptosis (33). Doxorubicin, like other anthracyclines, results in the

formation of quinone-mediated free radicals, which have the capacity to cause

oxidative damage and cytotoxicity, resulting in significant cardiotoxicity as a

result of the production of ROS (34). Cellular oxidoreductases reduce Adria-

mycin to a semiquinone radical that is subsequently reoxidized by oxygen to a

superoxide anion and the parent quinine (35–37). Superoxide anions can dis-

mutate to form hydrogen peroxide and/or react with nitric oxide to form per-

oxynitrite. Lipid peroxides resulting from doxorubicin can further break down to

yield hydroxyalkenals, which are substrates for glutathione-conjugating iso-

zymes (38), indicating the potential importance of GSTs in mediating therapeutic

outcomes.

Carbonyl reductase (CBR) is a monomeric, NADPH-dependent oxidor-

eductase that catalyzes the reduction of a variety of carbonyl compounds,

including doxorubicin and daunorubicin. In humans, two CBRs have been

identified, CBR1 and CBR3. An SNP has been identified in CBR3 (CBR3

V244M), where the M244 isoform results in higher Vmax and is 100% more

efficient in catalyzing the reduction of substrate (39). To our knowledge, CBRs

have not been investigated in relation to breast cancer outcomes following

treatment with anthracyclines.

In addition to its mechanistic role in cytotoxicity of anthracyclines, it is well

documented that oxidative stress plays a role in the cell-killing abilities of many

28 Amend et al.


other chemotherapy agents, as well as in radiation therapy. Specifically for breast

cancer, CP, doxorubicin (Adriamycin), and paclitaxel and radiation therapy gen-

erate ROS and induce apoptosis, which may contribute to treatment efficacy and

increased survival. ROS can cause mitochondrial permeability, and once the

mitochondrial membrane barrier function is lost, a number of other factors con-

tribute to cell death. While ROS, among other factors, induce or facilitate mito-

chondrial permeability, glutathione and antioxidant enzymes such as manganese

superoxide dismutase (MnSOD), catalase (CAT), and glutathione peroxidase

(GPX1) inhibit it (40). In the mitochondrion, MnSOD catalyzes the dismutation of

two superoxide radicals, producing H2O2 and oxygen. If H2O2 is not neutralized, it

may contribute to further generation of ROS by a reaction catalyzed by myelo-

peroxidase (MPO). Thus, levels of MnSOD, CAT, GPX1, and MPO are important

in determining the amount of ROS or detoxification of H2O2. It is plausible that

interindividual variability resulting from polymorphisms in these genes may

influence the efficacy of cancer treatment and survival.

Two genetic polymorphisms appear to affect mitochondrial levels of

MnSOD. A single-base substitution variant in the MnSOD gene results in an

amino acid substitution from alanine to valine at position 9 (ala-9) in the

mitochondrial targeting sequence (41). The substitution alters the secondary

structure of the protein, affecting the cellular distribution of the enzyme and

transport of MnSOD in the mitochondrion, where it would be biologically

available (41). In the mitochondria, the MnSOD Ala variant has been shown to

generate more active MnSOD protein compared with the Val allele (42). A

second polymorphic variant has been identified in MnSOD, and Ile or Thr at

amino acid position 58 (43) and the Thr58MnSOD variant was shown to exhibit

only half the enzymatic activity of Ile58MnSOD (43,44). It can be hypothesized

that the variant (ala-9 and Thr58MnSOD) would confer reduced protection of

breast tumor cells from ROS produced during therapy. Thus, lesser MnSOD

activity due to polymorphisms could increase treatment efficacy but also result

in greater toxicity due to ROS.

CAT is most abundant in the liver, kidneys, and erythrocytes, where it

decomposes H2O2 to H2O and O2. A common functional polymorphism

(C-262T) alters transcription factor binding and the expression of CAT in

erythrocytes. We previously found that individuals with C alleles had sig-

nificantly higher CAT levels compared with those with T/T genotypes (45).

H2O2, if not neutralized, may contribute to further generation of ROS by a

reaction catalyzed by MPO. MPO generates ROS endogenously by functioning

as an antimicrobial enzyme, catalyzing a reaction between H2O2 and chloride to

generate hypochlorous acid (HOCl), a potent oxidizing agent. HOCl further

reacts with other biological molecules to generate secondary radicals (46). The

promoter region of the MPO gene has a G-to-A base substitution at position 463,

which creates high- (G) and low-expression (A) alleles (47). Variability in genes

in this oxidative stress-related pathway may influence ultimate levels of ROS

and, thereby, tumor cell kill.



5-Fluorouracil and Methotrexate

The agents 5FU and MTX are involved in folate metabolism, in the conversion

of homocysteine to methionine, and in purine and pyrimidine synthesis. The

activated metabolites of 5FU, similar to pyrimidine nucleotides, competitively

and reversibly bind fluoropyrimidine thymidylase synthase (TS) inhibitors,

forming covalently bound complexes with the TS enzyme, thus inhibiting the

enzyme’s activity (48). Folate antagonists, such as MTX, are structurally similar

to folate and inhibit the action of various enzymes such as dihydrofolate

reductase (DHFR) (49). Their action is accomplished through incorporation as

false precursors in DNA or RNA or through inhibition of proteins involved in

nucleotide metabolism.

Common polymorphisms have been reported in many enzymes and trans-

port proteins involved in intracellular folate metabolism. Using a candidate gene

approach, pharmacogenetic research, to date, has focused on TS because of its role

as a drug target, dihydropyrimidine dehydrogenase (DPD) because of its

involvement in pyrimidine degradation, and methylenetetrahydrofolate reductase

(MTHFR) because of its key involvement in the control of the flux of intracellular

folate metabolites (50). A genetic polymorphism in the 50 regulatory region of the

TS gene promoter, consisting of either double (2R) or triple (3R) repeats of a 28-bp

sequence (51), has been shown to influence TS expression, with higher expression

in 3R/3R tumors (52). A G>C polymorphism has been identified within the

second repeat of 3R alleles that alters the transcriptional activation of the TS gene

constructs bearing this genotype (53). A 6-bp deletion (1494del6) within the

30UTR (54) has been also associated with decreased mRNA stability (55). Several

clinical studies have shown the potential contribution of the above TS germinal

polymorphism in the prediction of tumor responsiveness and/or toxicity following

treatment by fluoropyrimidines (56). However, the role of TS polymorphism as a

predictor of treatment outcome is still unclear.

MTHFR catalyzes the conversion of 5,10-methyltetrahydrofolate to 5-methyl-

tetrahydrofolate, which is an essential cofactor in the biosynthesis of deoxy-

thymidine monophosphate. A common C677T transition in exon 4 and an A1298C

transition in exon 7 of the MTHFR gene result in lower specific activity (57,58).

Most studies, to date, have examined the association between 5FU and MTHFR

polymorphisms in the context of colorectal cancer, but Toffoli et al. (59) reported

that of six patients who developed severe acute toxicity in the first cycle of adjuvant

CMF, five had the MTHFR 677 T/T genotype and one had the C/C genotype.

Because 5FU has a relatively narrow therapeutic index, toxicity increases

as the dose is increased, and more than 80% of administered 5FU is catabolized

by DPD (1). Patients with complete or partial deficiency of DPD often experi-

ence severe or even life-threatening toxicity after the administration of 5FU (60).

At least 11 of 33 mutations identified in the dihydropyrimidine dehydrogenase

(DPYD) gene have been detected in patients suffering from severe 5FU-associated

toxicity (61–63).

30 Amend et al.


There have been few studies of MTHFR and TS polymorphisms in relation

to breast cancer treatment outcomes. In a study with 19 human cancer cell lines,

including breast cancer, Etienne et al. (64) found that there was a marked trend

for greater 5FU efficacy in mutated MTHFR 1298C cell lines compared with

those with A/A genotypes (P ¼ 0.055), but MTHFR C677T and TS 2R/3R

polymorphisms were not associated with 5FU sensitivity. Largillier et al. (65)

noted that genetic polymorphisms of MTHFR (677C>T and 1298A>C), DPD

(IVS14þ1G>A), and TS were associated with efficacy and toxicity associated

with capecitabine, an oral 5FU prodrug, in 105 breast cancer patients receiving

capecitabine as monotherapy.

Although MTX has been widely used to treat many cancers, including

leukemia, lymphomas, and breast cancer, as well as some autoimmune diseases,

there has been limited research to evaluate the effects of genetic variability on

treatment outcomes. Associations have been noted for polymorphisms in folate-

related genes in leukemia (66–69), and Sohn et al. (70) found that in vitro

transfection of mutant 677T MTHFR cDNA in breast and colon cancer cells

increased the chemosensitivity of these cells to 5FU but decreased the chemo-

sensitivity of breast cancer cells to MTX, suggesting at least a modulating role of

this genotype on cellular fluouropyrimidine sensitivity.

Taxanes

The taxanes paclitaxel and docetaxel interfere with microtubular disassembly,

ultimately resulting in DNA fragmentation and features of apoptosis (71). The

contributions of genetic variability to paclitaxel metabolism and breast cancer

outcomes is still unclear. CYP2C8 is thought to be the main CYP-metabolizing

enzyme of paclitaxel, while CYP3A4 and CYP3A5 play more minor roles in drug

metabolism (19). At least three known variants in CYP2C8 exist, and the CYP2C8*3

variant allele has been associated with decreased paclitaxel 6-ahydroxylase activityin human cell lines and human liver microsomes (72). In a recent study of

paclitaxel metabolism in breast cancer, however, CYP2C8 genotypes (CYP2C8*3

and CYP2C8*4) were not significantly associated with paclitaxel clearance (73).

Ethnic variation exists in the distribution of CYC2C8 alleles, with the CYP2C8*2

allele more common in African-Americans than Caucasians (18% vs. 0%,

respectively) and the CYP2C*3 allele more common in whites (13%) than

African-Americans (2%). The CYPC2*5 allele is found in 9% of Asians (74).

Antiestrogens

The pharmacology and metabolism of TAM is complex, and differences in

patient outcomes could be due to individual variation in the metabolism of the

drug. TAM is metabolized primarily by CYP3A4, CYP2D6, CYP2C9,

SULT1A1, and UGT2B15 (75–78). The main metabolites of TAM are

N-desmethyltamoxifen (formed by CYP3A4) and endoxifen and 4-OH-TAM



(formed by CYP2D6) (8,76). CYP2D6 is polymorphic with 46 known allelic

variants, which are categorized as resulting in abolished, decreased, normal,

increased, or qualitatively altered catalytic activity (79). The major polymorphic

variant CYP2D6 alleles include CYP2D6*2, CYP2D6*4, CYP2D6*5,

CYP2D6*10, CYP2D6*17, and CYP2D6*41 (79).

Human sulfotransferase 1A1 (SULT1A1) catalyzes the sulfation of a

variety of phenolic and estrogenic compounds, including endogenous and

environmental estrogens and the 4-hydroxy metabolite of TAM, 4-OH TAM

(80). An exon 7 polymorphism in SULT1A1 (SULT1A1*2) results in a translated

protein with approximately twofold lower catalytic activity and decreased

thermostability than the common allele (SULT1A1*1) (81). We found, as

hypothesized, that women treated with TAM who were homozygous for the

variant SULT1A1*2/*2 allele had approximately three times the risk of death

compared with those homozygous for the common allele or heterozygotes (82).

A wide range of drugs are substrates for glucuronosyltransferases (UGTs)

and this pathway has been estimated to account for approximately one-third of

all drugs metabolized by phase II drug metabolizing enzymes (2). SULT1A1 and

UGT2B15 have overlapping substrate specificity for 4-OH TAM. A functional

polymorphism in UGT2B15, a G?T substitution at codon 85 within the putative

substrate recognition site of the gene, results in higher Vmax compared with the

common allele (83). In breast cancer patients treated with TAM, the combined

effects of SULT1A1 and UGT2B15 were evaluated, and an increasing number of

variant SULT1A1 and UGT2B15 alleles was associated with poorer survival (84).

Aromatase Inhibitors

AIs, including exemestane, anastrozole, and letrozole, have come into clinical

use more recently, developed primarily for hormone-dependent breast cancer in

postmenopausal women. AIs act by inhibiting the cytochrome P450 enzyme,

aromatase (CYP19), which catalyzes the conversion of androgens to estrogens.

AIs block this enzyme, decreasing circulating levels of estrogens. It has been

demonstrated that exemestane is metabolized by CYP3A4 but is also metabo-

lized by aldoketoreductases (85). Letrazole is metabolized by CYP2A6 and

CYP3A4 to a pharmacologically inactive carbinol metabolite (86), and anas-

trazole is metabolized by N-deaklylation and hydroxylation reactions followed

by glucuronidation. However, specific isoforms involved and the effects on drug

levels have not yet been identified.

MOLECULAR EPIDEMIOLOGIC STUDIES OF PHARMACOGENETICSAND BREAST CANCER OUTCOMES

To date, most analytical studies of pharmacogenetics and breast cancer outcomes

have been conducted among patients receiving heterogeneous chemotherapy

treatments, although most chemotherapy regimens included CP. Several studies

32 Amend et al.


have evaluated polymorphisms in phase II or GST enzymes in relation to

prognosis after treatment for breast cancer. In a study of 92 women with

advanced breast cancer, the GSTM1-null genotype was not significantly asso-

ciated with tumor characteristics, disease-free survival (DFS), or overall survival

(OS) (87), but in a retrospective study with a larger sample size, we found that

women with low-activity homozygous genotypes for the GSTA1*B allele had

better survival, with the hazard of death reduced by 30% (88,89), and women

with low-activity GSTP1 Val/Val genotype had a similar risk reduction (89).

Deletion polymorphisms in GSTM1 and GSTT1, genes responsive to ROS and

resulting from radiation or chemotherapy and subsequent lipid peroxidation,

were associated with a reduced hazard of death compared with women with both

alleles present [HR ¼ 0.59; 95% confidence interval (CI), 0.36–0.97; and HR ¼0.51; 95% confidence interval (CI), 0.29–0.90] (90). In a study in China (91),

women with GSTP1 Val/Val genotypes experienced a 60% reduction in

overallmortality (HR ¼ 0.4; 95% confidence interval (CI), 0.2–0.8), although

no associations were found with GSTM1 or GSTT1 genotypes. A smaller study

(n ¼ 85) evaluated polymorphisms in both phase I and phase II metabolism of

CP among women with metastatic or inflammatory breast cancer treated with

high-dose CP, cisplatin, and carmustine. Patients with the GSTM1-null genotype

had prolonged OS, while those with variants in CYP3A4*1B and CYP3A5*1 had

shortened survival. To our knowledge, only one genotype association study has

focused on paclitaxel metabolism. In that study of 93 patients (73), there were no

significant associations between ABCB1, ABCG2, CYP1B1, CYP3A4, CYP3A5,

and CYP2C8 and paclitaxel clearance. However, patients homozygous for

CYP1B1*3 Leu/Leu genotype had significantly longer progression-free survival

(PFS) compared with patients with at least one Val allele (Table 1).

Few studies have examined the effects of MTHFR genotypes on breast

cancer survival. In the Shanghai Breast Cancer Study (92), no significant

associations were found between MTHFR genotypes and the risk of death. A

second study evaluated these same genotypes in African-American and Cauca-

sian women from the Baltimore area (93) and found that carriers of the low-

activity allele at codon 1298 (A/C or C/C) had significantly poorer survival,

while those with the variant allele at codon 677 (C/T or T/T) had improved

survival. Associations were stronger in patients with ER-negative tumors, and

African-American women had improved survival if they were carriers of either

variant allele (A1298C, Pinteraction ¼ 0.088, C677T, Pinteraction ¼ 0.026).

Molecular epidemiologic studies have indicated that genetic variation in

phases I and II conjugating enzymes can influence the efficacy of TAM therapy

for breast cancer. In a recent study in Sweden (94), patients homozygous for

CYP2D6*4 alleles experienced significantly better DFS compared with patients

with at least *1 allele (P ¼ 0.05). In a group of women randomized to 5, rather

than two years of TAM, patients homozygous for CYP3A5*3 experienced

improved relapse-free survival (RFS) (HR ¼ 0.02; CI, 0.07–0.55). However, this

finding should be interpreted with caution since the number of patients within



Tab

le1

Epidem

iologic

StudiesofGenetic

Variationin

Chem

otherapeuticMetabolizingGenes

andBreastCancerOutcomes

Genes

SNP

Participants

Treatment

Outcome

Risk(95%

CI)

Reference

Cyclophospham

ide

GSTM1

GSTM1-null

genotype

92patients

CAF

OSorDFS

GSTM1notsignificant

87

GSTA1

GSTA1*B/*B

245patients

CPbased

5-yrsurvival

GSTA1*B/*BHR

¼0.3

(0.1–0.8)

88

GSTP1

GSTPVal/Val

240patients

CPbased

OS

GSTPVal/ValHR¼

0.3

(0.1–1.0)

89

GSTP1

GSTPVal/Val

1034patients

CPbased

OS

GSTPVal/ValHR¼

0.4

(0.2–0.8)

91

6genes,CYPs,

ABCB1,

ABCG2

8SNPs

93patients

High-dose

paclitaxel

and

CPandA

PFS

CYP1B1*3Leu/Leu

Longer

PFS(P

¼0.037)

73

17genes,CYPs,

METs,GSTs

29SNPs

85patients

CPandcisplatin

andcarm

ustine

PK

ofCP

andOS

GSTM1-nullgenotype;betterOS;atleastone

variantallelein

CYP3A4*1BorCYP3A5*1

(98)

Methotrexateand5-fluorouracil

MTHFR

MTHFRC677T;

MTHFR

A1298C

1067patients

Combination

(mostreceived

5FU)

OS

MTHFRgenotypes

notsignificantwithOS

92

MTHFR

MTHFRC677T;

MTHFR

A1298C

248patients

Combination

OS

MTHFR1298A/C

orC/C

HR

¼2.05

(1.05–4.0);

MTHFR677C/T

orT/T

HR¼

0.65(0.31–1.35)

(93)

Tam

oxifen

CYP3A5;

CYP2D6;

SULT1A1;

UGT2B15

CYP3A5*3;

CYP2D6*4;

SULT1A1*2;

UGT2B15*2

677patients

(238

randomized

toTAM)

TAM

RFS

CYP3A5*3/*3in

5-year

TAM

groupHR

¼0.20(0.07–0.55);

CYP2D6*4/*4better

RFS(n

¼677)(P

¼0.05)

94

CYP3A5;

CYP2D6

CYP3A5*3;

CYP2D6*4;

223patients

TAM

RFSorDFS

orOS

CYP2D6*4/*4

HR

¼2.71(1.15–6.41)forRFS;

HR

¼2.44(1.22–4.90)forDFS(both

ratios

unadjusted)

(95)

34 Amend et al.


Tab

le1

Epidem

iologic

StudiesofGenetic

Variationin

Chem

otherapeuticMetabolizingGenes

andBreastCancerOutcomes

(Continued

)

Genes

SNP

Participants

Treatment

Outcome

Risk(95%

CI)

Reference

CYP2D6;

UGT2B15;

SUL1A1

CYP2D6*4;

UGT2B15*2;

SULT1A1*2

337patients

(162patients

treatedwith

TAM)

TAM

OSorPFS

CYP2D6NS;

UGT2B15*2andSULT1A1*2

HR

¼4.4

(1.17–16.55)in

TAM

group

84

SULT1A1

SULT1A1*2

377patients

(160treated

withTAM)

TAM

OS

SULT1A1*2/*2

HR

¼2.9

(1.1–7.6)in

TAM

group

(82)

Oxidativestress

GSTM1/GSTT1

GSTM1and

GSTT1-null

genotypes

251patients

Combination

OSandDFS

GSTM1-nullgenotype

HR

¼0.59(0.36–0.97;

GSTT1-nullgenotype

HR

¼0.51(0.29–0.90)

90

MPO,MnSOD,

CAT

MnSOD

Ala-9Val,

CAT-262C?T,

MPO-463G?A

279patients

Combination

OS

MPO

GG

genotype

HR

¼0.60(0.38–0.95),

MnSOD

CCgenotypeHR¼

0.70(0.40–1.24),

genotypes

combined

HR

¼0.33(0.13–0.80)

96

MnSOD

MnSOD-102C?T

291patients

(230treated

withchem

o-

therapy)

Combination

RFS

MnSOD

CTþ

TTHR¼

0.65(0.40–1.05)

forchem

otherapygroup

97

Abbreviations:

CAF,Cyclophospham

ide/doxorubicin/fluorouracil;A,doxorubicin;CP,cyclophospham

ide;

T,DFS,disease-freesurvival;NS,notsignificant;OS,

overallsurvival;PFS,progression-freesurvival;RFS,relapse-freesurvival;TAM,tamoxifen;5FU,5-fluorouracil;GST,glutathioneS-transferase;

SNP,single

nucleotidepolymorphism;METs,

Metallothioneins;

MTHFR,methylenetetrahydrofolate

reductase;

MPO,myeloperoxidase;

MnSOD,manganesesuperoxidedis-

mutase;CAT,catalase;PT,Pharmacokinetics.



this group was small. No statistically significant differences were found for

CYP2D6, SULT1A1, or UGT2B15 and the length of TAM treatment. In contrast,

Goetz et al. (95) found no significant associations between CYP3A5*3 and breast

cancer outcomes; however, women with the CYP2D6 *4/*4 genotype (n ¼ 13)

had statistically significant worse RFS (P ¼ 0.023) and DFS (P ¼ 0.12) but not

OS (P ¼ 0.169). However, after adjustment for nodal status and tumor size,

associations were no longer statistically significant. These results are in agree-

ment with our previous report, wherein we noted no significant associations

between the CYP2D6*4 genotype and OS or PFS among breast cancer patients

treated with TAM (84).

In addition to CYP isoforms, phase II enzymes, including sulfotransferase

1A1 (SULT1A1) and UDP-glucuronosyltransferase isoform 2B15 (UGT2B15),

have been implicated in the metabolism of TAM, and genetic variability in these

genes together may influence breast cancer survival. In a cohort of women with

breast cancer who were treated with TAM, those who were homozygous for the

low-activity SULT1A1*2 allele had a threefold higher risk of death compared

with those with at least *1 common allele (HR ¼ 2.9; 1.1–7.6), while no sig-

nificant associations were found among women who did not receive TAM (82).

When considering the independent and combined effects of SULT1A1 and

UGT2B15 genotypes on OS and recurrence among women treated with TAM,

there was a nonsignificant trend toward increased risk of disease recurrence and

poorer survival with increasing number of UGT2B15*2 alleles (P ¼ 0.11) (84).

However, there was a significant trend toward decreased OS among women

treated with TAM, with increasing number of variant alleles when UGT2B15 and

SULT1A1 genotypes were combined (P ¼ 0.03), although these risk estimates

may be unstable because only 16 cases had two variant alleles.

We previously found that genotypes for MnSOD and MPO resulting in

higher levels of ROS were associated with better survival in a study of women

from Arkansas, and the combined effects of MnSOD and MPO were greater than

those for either polymorphism alone (HR ¼ 0.33; CI, 0.13–0.80) (96). An

additional MnSOD variant, located in the promoter region of the gene, was also

associated with a significant decrease in RFS (CTþTT genotype HR ¼ 0.65, CI,

0.40–1.05), although no significant associations were found with OS (97).

In the context of a clinical trial for breast cancer with CAF versus CMF

(SWOG 8897) (unpublished data, presented at the San Antonio Breast Cancer

Symposium, 2006), as in the Arkansas study, MPO genotypes encoding for

higher activity were associated with better survival.

FUTURE DIRECTIONS

Studies, to date, indicate that interindividual variability in drug metabolism may

affect treatment-related toxicities and DFS. However, much of the data are

inconsistent, and it is likely that the evaluation of single genes in complex

pathways presents only part of a complicated picture. Thus, future studies of

36 Amend et al.


pharmacogenetics and treatment outcomes should focus on entire drug metab-

olism pathways in large sample sets receiving homogeneous treatment regimens.

Furthermore, assessment of variability across an entire gene, through construc-

tion of haplotype blocks and tag SNPs, may give more complete information

than a study of single SNPs. It is likely that, in the future, other, more systemic

markers of inherited drug metabolism capabilities may better predict treatment

outcomes, and scans of variability across the entire genome may reveal factors

that are likely to determine treatment outcomes. The consideration of individual

genotypes, as well as tumor characteristics, in therapeutic decision making holds

promise for individualized treatment for breast cancer in the future.

REFERENCES

1. Heggie GD, Sommadossi JP, Cross DS, et al. Clinical pharmacokinetics of 5-

fluorouracil and its metabolites in plasma, urine, and bile. Cancer Res 1987; 47(8):

2203–2206.

2. Evans WE, Relling MV. Pharmacogenomics: translating functional genomics into

rational therapeutics. Science 1999; 286(5439):487–491.

3. DeVita VT Jr., Hellman S, Rosenberg SA (Eds.). Cancer: Principles and Practice of

Oncology. 5th ed. New York, NY: Lippincott-Raven, 1997.

4. Early Breast Cancer Trialists’ Collaborative Group (EBCTG). Effects of chemo-

therapy and hormonal therapy for early breast cancer on recurrence and 15-year

survival: an overview of the randomised trials. Lancet 2005; 365:1687–1717.

5. Cuzick J. Aromatase inhibitors for breast cancer prevention. J Clin Oncol 2005;

23(8):1636–1643.

6. Conte P, Frassoldati A. Aromatase inhibitors in the adjuvant treatment of post-

menopausal women with early breast cancer: putting safety issues into perspective.

Breast J 2007; 13(1):28–35.

7. Choi JY, Nowell SA, Blanco JG, et al. The role of genetic variability in drug

metabolism pathways in breast cancer prognosis. Pharmacogenomics 2006; 7(4):

613–624.

8. Scripture CD, Sparreboom A, Figg WD. Modulation of cytochrome P450 activity:

implications for cancer therapy. Lancet Oncol 2005; 6(10):780–789.

9. Dirven HA, van Ommen B, van Bladeren PJ. Involvement of human glutathione

S-transferase isoenzymes in the conjugation of cyclophosphamide metabolites with

glutathione. Cancer Res 1994; 54(23):6215–6220.

10. Chang TK, Weber GF, Crespi CL, et al. Differential activation of cyclophosphamide

and ifosphamide by cytochromes P-450 2B and 3A in human liver microsomes.

Cancer Res 1993; 53(23):5629–5637.

11. Xie HJ, Yasar U, Lundgren S, et al. Role of polymorphic human CYP2B6 in

cyclophosphamide bioactivation. Pharmacogenomics J 2003; 3(1):53–61.

12. White IN, De MF, Gibbs AH, et al. Species differences in the covalent binding of

[14C]tamoxifen to liver microsomes and the forms of cytochrome P450 involved.

Biochem Pharmacol 1995; 49(8):1035–1042.

13. http://www.imm.ki.se/CYPalleles/cyp2b6.htm (accessed January 2007). 2007. 7. Ref

Type: Electronic Citation



14. Jinno H, Tanaka-Kagawa T, Ohno A, et al. Functional characterization of cyto-

chrome P450 2B6 allelic variants. Drug Metab Dispos 2003; 31(4):398–403.

15. Ariyoshi N, Miyazaki M, Toide K, et al. A single nucleotide polymorphism of

CYP2b6 found in Japanese enhances catalytic activity by autoactivation. Biochem

Biophys Res Commun 2001; 281(5):1256–1260.

16. Lang T, Klein K, Fischer J, et al. Extensive genetic polymorphism in the human

CYP2B6 gene with impact on expression and function in human liver. Pharmaco-

genetics 2001; 11(5):399–415.

17. Hiratsuka M, Takekuma Y, Endo N, et al. Allele and genotype frequencies of

CYP2B6 and CYP3A5 in the Japanese population. Eur J Clin Pharmacol 2002; 58(6):

417–421.

18. Lamba V, Lamba J, Yasuda K, et al. Hepatic CYP2B6 expression: gender and ethnic

differences and relationship to CYP2B6 genotype and CAR (constitutive androstane

receptor) expression. J Pharmacol Exp Ther 2003; 307(3):906–922.

19. Daly AK. Pharmacogenetics of the major polymorphic metabolizing enzymes.

Fundam Clin Pharmacol 2003; 17(1):27–41.

20. Rebbeck TR, Jaffe JM, Walker AH, et al. Modification of clinical presentation of

prostate tumors by a novel genetic variant in CYP3A4. J Natl Cancer Inst 1998;

90(16):1225–1229.

21. Walker AH, Jaffe JM, Gunasegaram S, et al. Characterization of an allelic variant

in the nifedipine-specific element of CYP3A4: ethnic distribution and implications

for prostate cancer risk. Mutations in brief no. 191. Hum Mutat 1998; 12(4):289

(online).

22. Amirimani B, Walker AH, Weber BL, et al. RESPONSE: re: modification of clinical

presentation of prostate tumors by a novel genetic variant in CYP3A4. J Natl Cancer

Inst 1999; 91(18):1588–1590.

23. Roy P, Yu LJ, Crespi CL, et al. Development of a substrate-activity based approach

to identify the major human liver P-450 catalysts of cyclophosphamide and ifosfa-

mide activation based on cDNA-expressed activities and liver microsomal P-450

profiles. Drug Metab Dispos 1999; 27(6):655–666.

24. Raucy JL. Regulation of CYP3A4 expression in human hepatocytes by pharma-

ceuticals and natural products. Drug Metab Dispos 2003; 31(5):533–539.

25. Kuehl P, Zhang J, Lin Y, et al. Sequence diversity in CYP3A promoters and char-

acterization of the genetic basis of polymorphic CYP3A5 expression. Nat Genet

2001; 27(4):383–391.

26. Hustert E, Haberl M, Burk O, et al. The genetic determinants of the CYP3A5

polymorphism. Pharmacogenetics 2001; 11(9):773–779.

27. Griskevicius L, Yasar U, Sandberg M, et al. Bioactivation of cyclophosphamide: the

role of polymorphic CYP2C enzymes. Eur J Clin Pharmacol 2003; 59(2):103–109.

28. Morel F, Rauch C, Coles B, et al. The human glutathione transferase alpha locus:

genomic organization of the gene cluster and functional characterization of the

genetic polymorphism in the hGSTA1 promoter. Pharmacogenetics 2002; 12(4):

277–286.

29. Coles BF, Morel F, Rauch C, et al. Effect of polymorphism in the human glutathione

S-transferase A1 promoter on hepatic GSTA1 and GSTA2 expression. Pharmaco-

genetics 2001; 11(8):663–669.

30. Ali-Osman F, Akande O, Antoun G, et al. Molecular cloning, characterization, and

expression in Escherichia coli of full-length cDNAs of three human glutathione

38 Amend et al.


S-transferase Pi gene variants. Evidence for differential catalytic activity of the

encoded proteins. J Biol Chem 1997; 272(15):10004–10012.

31. Harries LW, Stubbins MJ, Forman D, et al. Identification of genetic polymorphisms

at the glutathione S-transferase Pi locus and association with susceptibility to

bladder, testicular and prostate cancer. Carcinogenesis 1997; 18(4):641–644.

32. Srivastava SK, Singhal SS, Hu X, et al. Differential catalytic efficiency of allelic

variants of human glutathione S-transferase Pi in catalyzing the glutathione conju-

gation of thiotepa. Arch Biochem Biophys 1999; 366(1):89–94.

33. Minotti G, Menna P, Salvatorelli E, et al. Anthracyclines: molecular advances and

pharmacologic developments in antitumor activity and cardiotoxicity. Pharmacol

Rev 2004; 56(2):185–229.

34. Ravid A, Rocker D, Machlenkin A, et al. 1,25-Dihydroxyvitamin D3 enhances the

susceptibility of breast cancer cells to doxorubicin-induced oxidative damage.

Cancer Res 1999; 59(4):862–867.

35. Kong Q, Lillehei KO. Antioxidant inhibitors for cancer therapy. Med Hypotheses

1998; 51(5):405–409.

36. Kalyanaraman B, Perez-Reyes E, Mason RP. Spin-trapping and direct electron spin

resonance investigations of the redox metabolism of quinone anticancer drugs.

Biochim Biophys Acta 1980; 630(1):119–130.

37. Conklin KA. Dietary antioxidants during cancer chemotherapy: impact on chemo-

therapeutic effectiveness and development of side effects. Nutr Cancer 2000; 37(1):

1–18.

38. Sellin S, Holmquist B, Mannervik B, et al. Oxidation and reduction of 4-hydroxy-

alkenals catalyzed by isozymes of human alcohol dehydrogenase. Biochemistry

1991; 30(9):2514–2518.

39. Lakhman SS, Ghosh D, Blanco JG. Functional significance of a natural allelic

variant of human carbonyl reductase 3 (CBR3). Drug Metab Dispos 2005; 33(2):

254–257.

40. Costantini P, Jacotot E, Decaudin D, et al. Mitochondrion as a novel target of

anticancer chemotherapy. J Natl Cancer Inst 2000; 92(13):1042–1053.

41. Shimoda-Matsubayashi S, Matsumine H, Kobayashi T, et al. Structural dimorphism

in the mitochondrial targeting sequence in the human manganese superoxide dis-

mutase gene. A predictive evidence for conformational change to influence mito-

chondrial transport and a study of allelic association in Parkinson’s disease. Biochem

Biophys Res Commun 1996; 226(2):561–565.

42. Sutton A, Khoury H, Prip-Buus C, et al. The Ala16Val genetic dimorphism modu-

lates the import of human manganese superoxide dismutase into rat liver mito-

chondria. Pharmacogenetics 2003; 13(3):145–157.

43. Ho YS, Crapo JD. Isolation and characterization of complementary DNAs encoding

human manganese-containing superoxide dismutase. FEBS Lett 1988; 229(2):

256–260.

44. Borgstahl GE, Parge HE, Hickey MJ, et al. Human mitochondrial manganese

superoxide dismutase polymorphic variant Ile58Thr reduces activity by destabilizing

the tetrameric interface. Biochemistry 1996; 35(14):4287–4297.

45. Forsberg L, Lyrenas L, de FU, et al. A common functional C-T substitution poly-

morphism in the promoter region of the human catalase gene influences transcription

factor binding, reporter gene transcription and is correlated to blood catalase levels.

Free Radic Biol Med 2001; 30(5):500–505.



46. Klebanoff SJ. Myeloperoxidase: occurrence and biological function. In: Everse J,

Everse KE, Grisham MB, eds. Peroxides in Chemistry. Boca Raton: CRC

Press, 1991.

47. Piedrafita FJ, Molander RB, Vansant G, et al. An Alu element in the myeloperox-

idase promoter contains a composite SP1-thyroid hormone-retinoic acid response

element. J Biol Chem 1996; 271(24):14412–14420.

48. Chabner BA, Ryan DP, Paz-Ares LP, et al. Antineoplastic agents. In: Hardman JG,

Limbird LE, Gilman AG, eds. Goodman and Gilman’s the Pharmacological Basis of

Therapeutics. New York, NY: McGraw-Hill, 2001:1389–1460.

49. Calvert H. An overview of folate metabolism: features relevant to the action and

toxicities of antifolate anticancer agents. Semin Oncol 1999; 26(2 suppl 6):3–10.

50. Robien K, Boynton A, Ulrich CM. Pharmacogenetics of folate-related drug targets in

cancer treatment. Pharmacogenomics 2005; 6(7):673–689.

51. Horie N, Aiba H, Oguro K, et al. Functional analysis and DNA polymorphism of the

tandemly repeated sequences in the 50-terminal regulatory region of the human gene

for thymidylate synthase. Cell Struct Funct 1995; 20(3):191–197.

52. Pullarkat ST, Stoehlmacher J, Ghaderi V, et al. Thymidylate synthase gene poly-

morphism determines response and toxicity of 5-FU chemotherapy. Pharmacoge-

nomics J 2001; 1(1):65–70.

53. Mandola MV, Stoehlmacher J, Muller-Weeks S, et al. A novel single nucleotide

polymorphism within the 50 tandem repeat polymorphism of the thymidylate syn-

thase gene abolishes USF-1 binding and alters transcriptional activity. Cancer Res

2003; 63(11):2898–2904.

54. Ulrich CM, Bigler J, Velicer CM, et al. Searching expressed sequence tag databases:

discovery and confirmation of a common polymorphism in the thymidylate synthase

gene. Cancer Epidemiol Biomarkers Prev 2000; 9(12):1381–1385.

55. Mandola MV, Stoehlmacher J, Zhang W, et al. A 6 bp polymorphism in the thy-

midylate synthase gene causes message instability and is associated with decreased

intratumoral TS mRNA levels. Pharmacogenetics 2004; 14(5):319–327.

56. Danenberg PV. Pharmacogenomics of thymidylate synthase in cancer treatment.

Front Biosci 2004; 9:2484–2494.

57. Stern LL, Bagley PJ, Rosenberg IH, et al. Conversion of 5-formyltetrahydrofolic acid

to 5-methyltetrahydrofolic acid is unimpaired in folate-adequate persons homo-

zygous for the C677T mutation in the methylenetetrahydrofolate reductase gene.

J Nutr 2000; 130(9):2238–2242.

58. Weisberg I, Tran P, Christensen B, et al. A second genetic polymorphism in

methylenetetrahydrofolate reductase (MTHFR) associated with decreased enzyme

activity. Mol Genet Metab 1998; 64(3):169–172.

59. Toffoli G, Veronesi A, Boiocchi M, et al. MTHFR gene polymorphism and severe

toxicity during adjuvant treatment of early breast cancer with cyclophosphamide,

methotrexate, and fluorouracil (CMF). Ann Oncol 2000; 11(3):373–374.

60. Etienne MC, Lagrange JL, Dassonville O, et al. Population study of

dihydropyrimidine dehydrogenase in cancer patients. J Clin Oncol 1994; 12(11):

2248–2253.

61. van Kuilenburg AB, Haasjes J, Richel DJ, et al. Clinical implications of dihy-

dropyrimidine dehydrogenase (DPD) deficiency in patients with severe 5-

fluorouracil-associated toxicity: identification of new mutations in the DPD gene.

Clin Cancer Res 2000; 6(12):4705–4712.

40 Amend et al.


62. van Kuilenburg AB, De Abreu RA, van Gennip AH. Pharmacogenetic and clinical

aspects of dihydropyrimidine dehydrogenase deficiency. Ann Clin Biochem 2003;

40(pt 1):41–45.

63. Collie-Duguid ES, Etienne MC, Milano G, et al. Known variant DPYD alleles do not

explain DPD deficiency in cancer patients. Pharmacogenetics 2000; 10(3):217–223.

64. Etienne MC, Ilc K, Formento JL, et al. Thymidylate synthase and methylenete-

trahydrofolate reductase gene polymorphisms: relationships with 5-fluorouracil

sensitivity. Br J Cancer 2004; 90(2):526–534.

65. Largillier R, Etienne-Grimaldi MC, Formento JL, et al. Pharmacogenetics of

capecitabine in advanced breast cancer patients. Clin Cancer Res 2006; 12(18):

5496–5502.

66. Laverdiere C, Chiasson S, Costea I, et al. Polymorphism G80A in the reduced folate

carrier gene and its relationship to methotrexate plasma levels and outcome of

childhood acute lymphoblastic leukemia. Blood 2002; 100(10):3832–3834.

67. Costea I, Moghrabi A, Laverdiere C, et al. Folate cycle gene variants and chemo-

therapy toxicity in pediatric patients with acute lymphoblastic leukemia. Haemato-

logica 2006; 91(8):1113–1116.

68. Seidemann K, Book M, Zimmermann M, et al. MTHFR 677 (C–>T) polymorphism

is not relevant for prognosis or therapy-associated toxicity in pediatric NHL:

results from 484 patients of multicenter trial NHL-BFM 95. Ann Hematol 2006;

85(5):291–300.

69. Jazbec J, Kitanovski L, Aplenc R, et al. No evidence of association of methyl-

enetetrahydrofolate reductase polymorphism with occurrence of second neoplasms

after treatment of childhood leukemia. Leuk Lymphoma 2005; 46(6):893–897.

70. Sohn KJ, Croxford R, Yates Z, et al. Effect of the methylenetetrahydrofolate

reductase C677T polymorphism on chemosensitivity of colon and breast cancer cells

to 5-fluorouracil and methotrexate. J Natl Cancer Inst 2004; 96(2):134–144.

71. Schiff PB, Fant J, Horwitz SB. Promotion of microtubule assembly in vitro by taxol.

Nature 1979; 277(5698):665–667.

72. Soyama A, Saito Y, Hanioka N, et al. Non-synonymous single nucleotide alterations

found in the CYP2C8 gene result in reduced in vitro paclitaxel metabolism. Biol

Pharm Bull 2001; 24(12):1427–1430.

73. Marsh S, Somlo G, Li X, et al. Pharmacogenetic analysis of paclitaxel transport and

metabolism genes in breast cancer. Pharmacogenomics J 2007.

74. van Schaik RH. Implications of cytochrome P450 genetic polymorphisms on the

toxicity of antitumor agents. Ther Drug Monit 2004; 26(2):236–240.

75. Crewe HK, Ellis SW, Lennard MS, et al. Variable contribution of cytochromes P450

2D6, 2C9 and 3A4 to the 4-hydroxylation of tamoxifen by human liver microsomes.

Biochem Pharmacol 1997; 53(2):171–178.

76. Dehal SS, Kupfer D. CYP2D6 catalyzes tamoxifen 4-hydroxylation in human liver.

Cancer Res 1997; 57(16):3402–3406.

77. Stearns V, Johnson MD, Rae JM, et al. Active tamoxifen metabolite plasma con-

centrations after coadministration of tamoxifen and the selective serotonin reuptake

inhibitor paroxetine. J Natl Cancer Inst 2003; 95(23):1758–1764.

78. Nishiyama T, Ogura K, Nakano H, et al. Reverse geometrical selectivity in glu-

curonidation and sulfation of cis- and trans-4-hydroxytamoxifens by human liver

UDP-glucuronosyltransferases and sulfotransferases. Biochem Pharmacol 2002;

63(10):1817–1830.



79. Ingelman-Sundberg M. Genetic polymorphisms of cytochrome P450 2D6

(CYP2D6): clinical consequences, evolutionary aspects and functional diversity.

Pharmacogenomics J 2005; 5(1):6–13.

80. Falany CN, Wheeler J, Oh TS, et al. Steroid sulfation by expressed human cytosolic

sulfotransferases. J Steroid Biochem Mol Biol 1994; 48(4):369–375.

81. Raftogianis RB, Wood TC, Otterness DM, et al. Phenol sulfotransferase pharma-

cogenetics in humans: association of common SULT1A1 alleles with TS PST phe-

notype. Biochem Biophys Res Commun 1997; 239(1):298–304.

82. Nowell S, Sweeney C, Winters M, et al. Association between sulfotransferase 1A1

genotype and survival of breast cancer patients receiving tamoxifen therapy. J Natl

Cancer Inst 2002; 94(21):1635–1640.

83. Levesque E, Beaulieu M, Green MD, et al. Isolation and characterization of

UGT2B15(Y85): a UDP-glucuronosyltransferase encoded by a polymorphic gene.

Pharmacogenetics 1997; 7(4):317–325.

84. Nowell SA, Ahn J, Rae JM, et al. Association of genetic variation in tamoxifen-

metabolizing enzymes with overall survival and recurrence of disease in breast

cancer patients. Breast Cancer Res Treat 2005; 91(3):249–258.

85. Buzdar AU, Robertson JF, Eiermann W, et al. An overview of the pharmacology and

pharmacokinetics of the newer generation aromatase inhibitors anastrozole, letro-

zole, and exemestane. Cancer 2002; 95(9):2006–2016.

86. Sioufi A, Gauducheau N, Pineau V, et al. Absolute bioavailability of letrozole in

healthy postmenopausal women. Biopharm Drug Dispos 1997; 18(9):779–789.

87. Lizard-Nacol S, Coudert B, Colosetti P, et al. Glutathione S-transferase M1 null

genotype: lack of association with tumour characteristics and survival in advanced

breast cancer. Breast Cancer Res 1999; 1(1):81–87.

88. Sweeney C, Ambrosone CB, Joseph L, et al. Association between a glutathione

S-transferase A1 promoter polymorphism and survival after breast cancer treatment.

Int J Cancer 2003; 103(6):810–814.

89. Sweeney C, McClure GY, Fares MY, et al. Association between survival after

treatment for breast cancer and glutathione S-transferase P1 Ile105Val poly-

morphism. Cancer Res 2000; 60(20):5621–5624.

90. Ambrosone CB, Sweeney C, Coles BF, et al. Polymorphisms in glutathione

S-transferases (GSTM1 and GSTT1) and survival after treatment for breast cancer.

Cancer Res 2001; 61(19):7130–7135.

91. Yang G, Shu XO, Ruan ZX, et al. Genetic polymorphisms in glutathione-S-trans-

ferase genes (GSTM1, GSTT1, GSTP1) and survival after chemotherapy for invasive

breast carcinoma. Cancer 2005; 103(1):52–58.

92. Shrubsole MJ, Shu XO, Ruan ZX, et al. MTHFR genotypes and breast cancer sur-

vival after surgery and chemotherapy: a report from the Shanghai Breast Cancer

Study. Breast Cancer Res Treat 2005; 91(1):73–79.

93. Martin DN, Boersma BJ, Howe TM, et al. Association of MTHFR gene poly-

morphisms with breast cancer survival. BMC Cancer 2006; 6:257.

94. Wegman P, Elingarami S, Carstensen J, et al. Genetic variants of CYP3A5,

CYP2D6, SULT1A1, UGT2B15 and tamoxifen response in postmenopausal patients

with breast cancer. Breast Cancer Res 2007; 9(1):R7.



Oncol 2005; 23(36):9312–9318.

42 Amend et al.


96. Ambrosone CB, Ahn J, Singh KK, et al. Polymorphisms in genes related to oxidative

stress (MPO, MnSOD, CAT) and survival after treatment for breast cancer. Cancer

Res 2005; 65(3):1105–1111.

97. Martin RC, Ahn J, Nowell SA, et al. Association between manganese superoxide

dismutase promoter gene polymorphism and breast cancer survival. Breast Cancer

Res 2006; 8(4):R45.

98. Petros WP, Hopkins PJ, Spruill S, et al. Associations between drug metabolism

genotype, chemotherapy pharmacokinetics, and overall survival in patients with

breast cancer. J Clin Oncol 2005; 23(25):6117–6125.



4

The Role of Epigenetics in Breast Cancer:

Implications for Diagnosis, Prognosis,

and Treatment

Amy M. Dworkin, Tim H.-M. Huang, and Amanda E. Toland

Human Cancer Genetics Program, Department of Molecular Virology,Immunology, and Medical Genetics, The Ohio State University,

Columbus, Ohio, U.S.A.

INTRODUCTION

In the past 10 years, the field of epigenetics has expanded into an essential

component of clinical cancer research and therapeutics. Epigenetic modifications

in DNA or chromatin are stably transmitted over rounds of cell division but do

not result in any permanent alterations to the underlying DNA sequence. These

changes, such as histone modifications and DNA methylation of tumor sup-

pressor genes, are considered a hallmark of certain cancers, including breast

cancer. Unlike germline mutations, epigenetic modifications can potentially be

reversed, making them very appealing for preventative care and therapeutics

against cancers (1). Because of this, demethylating agents are currently being

evaluated for efficacy in the treatment of breast cancer.

There are distinct mechanisms that initiate and sustain epigenetic mod-

ifications (2–8). Of these, DNA methylation and posttranslational modifications

of histone proteins are the best understood. DNA methylation is a covalent

addition of a methyl group to DNA, usually to a cytosine located 50 to guanasine

(CpG dinucleotide). CpG dinucleotides are underrepresented in the genome,

45


except for small clusters, referred to as CpG islands, located in or near the

promoter of approximately half of all genes (1,9–11). In addition, other epi-

genetic modifications have recently been explored, including the Polycomb

group (PcG) proteins, which repress gene function and can only be overcome by

germline differentiation processes, and small noncoding RNA molecules, which

regulate gene expression by targeting RNA degradation (3). Small noncoding

RNA molecules have recently been found to also target gene promoters and

result in transcriptional gene silencing (11,12).

DNA methylation plays a variety of roles in normal cells: silencing of

transposable elements and viral sequences, maintenance of chromosomal

integrity, X chromosome inactivation, and transcriptional regulation of genes

(1). Methylation is established and maintained by at least four DNA methyl-

transferases (DNMTs), DNMT1, DNMT3A, DNMT3B, and DNMT3L, which

catalyze the covalent addition of methyl groups to cytosine in the CpG dinu-

cleotide. Methylation of CpG islands causes transcriptional inactivation (Fig. 1).

These methyltransferases and their isoforms have been demonstrated to be

essential for development; a knockout of any of these enzymes in mouse

embryos results in embryonic or perinatal lethality (1). Functionally, there are

two types of methyltransferases, maintenance and de novo. Maintenance meth-

yltransferases are essential for copying DNA methylation patterns during cell

division, and de novo methyltransferases introduce new patterns of methylation

during early development. There are duplications of functions between these

methylases so that perturbation of both types may be necessary to change overall

methylation patterns (6). DNMT1 functions as a maintenance methyltransferase

by copying DNA methylation patterns to daughter strands, following replication.

It is the most abundant and catalytically active form. DNMT3A and DNMT3B

act on both unmethylated and hemimethylated DNA. However, they are thought

to have different methylation targets depending on the cell type and stage of

development (13). In vitro knockouts of DNMT3a or DNMT3b each exhibit

distinct phenotypes. DNMT3a-null cells lack methylation at imprinting loci,

while DNMT3b-null cells result in loss of DNA methylation on endogenous

Figure 1 Methylation of CpG islands. Unmethylated CpG islands (in light gray) allow

for transcription, while methylated CpG islands (in dark gray) repress transcription.

46 Dworkin et al.


C-type retroviral DNA. Deletion of both DNMT3a and DNMT3b abolishes de

novo methylation, while DNMT1 deletion produces bulk DNA demethylation

(7). Of the methyltransferases, DNMT3L is unique. It stimulates de novo

methylation by DNMT3a, does not enhance its binding to DNA, and alone does

not bind to DNA (14).

Along with methylation, posttranslational modification of histones plays

an important role in epigenetics. Histone proteins associate with DNA to form

nucleosomes, permitting large amounts of DNA to be neatly packaged into the

nucleus. There are two configurations of chromatin, heterochromatin and

euchromatin. Heterochromatin is condensed and transcriptionally inactive, while

euchromatin has an “open” configuration and is favorable for gene transcription

(Fig. 2). The N-terminal tails of histones “stick out” of the nucleosome and are

subject to posttranslational modification such as acetylation, phosphorylation,

methylation, ubiquitination, and sumoylation (2,4,7,15). The pattern of histone

modification creates a “code,” which is read by proteins involved in chromatin

remodeling and the dynamics of gene transcription (6,11). Acetylation and

deacteylation are the most common types of histone modification. Histone

deacetylases (HDACs) are a class of enzymes that remove acetyl groups from a

e-N-acetyl lysine amino acid on a histone. Its action is opposite to that of histone

acetyltransferases (HATs). Deacetylation removes acetyl groups from histone

tails, causing the histones to wrap more tightly around the DNA and interfering

with transcription by blocking access to transcription factors. The overall result

of histone deacetylation is a global (nonspecific) reduction in gene expression

(Table 1).

METHYLATION IN CANCER

Cancer development is complex due to progressive changes of the genome that

result in altering normal cellular processes such as cell growth, apoptosis, and

angiogenesis (12). Both genetic and epigenetic modifications play a strong role

in tumorigenesis (7). In classic tumor biology, cells lose both copies of tumor

Figure 2 Histone modification and chromatin remodeling. Histone modifications, like

acetylation, change the affinity of the histones to DNA. The acetylated histones on the left

are in an open euchromatin structure and can be transcribed. The unacetylated histones on

the right are in a heterochromatin structure and repress transcription. Source: Cartoon

compliments of Kristi Bennett.

The Role of Epigenetics in Breast Cancer 47


Table 1 Histone Modifiers

Histone modifier Mechanism Catalysis Role in transcription

Acetylation Introduction of the

acetyl group to

the lysine residue

at the N terminus

of the histone

protein

HATs Enhances transcription

(removes positive

charge and reduces

affinity between

histones and DNA)

Deacetylation Removal of the

acetyl group

from the lysine

residue at the N

terminus of the

histone protein

HDACs Represses transcription

Phosphorylation Addition of

phosphate group

Histone kinases Enhances transcription

(neutralizes histones’

basic charge and

reduces affinity to

DNA)

Dephosphorylation Removal of

phosphate group

Histone

phosphatases

Represses transcription

Methylation Addition of 1, 2, or

3 methyl groups

on either lysine

or arginine

residues

HMTs Silences genes

Demethylation Removal of methyl

groups

Cytochrome P450

family of liver

enzymes

Activates genes

Ubiquitination Covalent

attachment

of ubiquitin

Enzymatic cascade,

including

ubiquitin ligase,

E1 ubiquitin

activating

enzyme, and

ATP

Labels protein for

degradation

Deubiquitination Removal of

ubiquitin

Isopeptidases Reverses

ubiquitination

Sumoylation Covalent

attachment

of SUMO to

lysine residues

Enzymatic cascade,

including SUMO-

activating enzyme

and ATP

Can mediate gene

silencing through

recruitment of HDAC

and heterochromatin

protein 1

Abbreviations: HATs, histone acetyltransferases; HDACs, histone deacetylases; HMT, histone

methyltranferases; SUMO, small ubiquitin-like modifier; ATP, adenosine triphosphate.

48 Dworkin et al.


suppressor genes through mutation or deletion. Epigenetic modifications, such

as DNA methylation, serve a similar role to somatic mutations, leading to

gene silencing. Overall, tumors tend to exhibit hypomethylation of repeat

elements and pericentromeric regions and hypermethylation of normally

unmethylated CpG islands. Gene-specific hypermethylation of CpG islands is

associated with chromosomal instability, resulting in increased mutation rates

and abnormal gene expression. For example, hypermethylation of critical

genes, such as tumor suppressors, adhesion molecules, DNA repair genes,

inhibitors of angiogenesis, and inhibitors of metastasis, is observed in

tumorigenesis (7). Hypermethylation of DNA repair genes (e.g., hMLH1) is

associated with genomic instability through disruption of chromosome repli-

cation control (7). Unlike gene mutations, DNA methylation can be reversed

by DNMT inhibitors.

An important component of the cancer epigenome is an altered DNA

methylation pattern of global demethylation and localized promoter hyper-

methylation. Aberrant methylation patterns are seen in almost every cancer type

(7,16). In addition to aberrant methylation, tumors also show altered expression

of HATs, making them another potential target for therapy (11). It is believed

that histone modifications precede promoter hypermethylation to initiate tran-

scriptional silencing of tumor suppressor genes. Repressive histone mod-

ifications are first established. Over time the density of DNA methylation

increases in the gene promoter regions, which creates a “permanent” mode of

silencing (17).

THE ROLE OF EPIGENETICS IN BREAST CANCER

In the United States, breast cancer is one of the most common cancers, with

approximately 200,000 women diagnosed with breast cancer each year (http://

www.cancer.org). In addition to well-understood DNA mutations in breast

tumors, epigenetic alterations of gene expression are key contributors (16,18–24).

A large number of genes are found to be hypermethylated in breast tumors but

not in normal breast tissue. Many of these genes have well-characterized func-

tions in tumor development. Genes commonly methylated in breast cancer and

their potential roles in tumor development are listed in Table 2. Four key genes,

RASSF1A, BRCA1, ESR1, and ESR2, are further discussed in detail.

RASSF1A

The CpG island of the Ras-associated domain family 1A gene (RASSF1) is

hypermethylated in 60% to 77% of breast cancers, and its transcripts are fre-

quently inactivated in cancer cell lines and primary tissues (13,22). When cells

lacking RASSF1A expression are treated with a DNMT inhibitor, such as 5-aza-

20-deoxycytidine, expression can be reactivated (25). RASSF1A is involved in



growth regulation and apoptotic pathways. Specifically, RASSF1A is a Ras

effector and induces apoptosis through its interactions with pro-apoptotic kinase

MST1. Mouse knockout studies show that RASSF1a–/– mice are prone to

spontaneous development of lung adenomas, lymphomas, and breast adeno-

carcinomas. Knockout RASSF1A mice are prone to early spontaneous tumori-

genesis and show a severe tumor susceptibility phenotype compared to wild-type

mice (25). In addition to breast tumors, hypermethylation of RASSF1A can be

detected in nonmalignant breast cells and patient sera. In one study, RASSF1A

hypermethylation in sera of breast cancer patients was detected in six out of 26

cases. These data suggest that RASSF1A hypermethylation may be a promising

biomarker to screen putative cancer patients (25).

BRCA1

Germline mutations of BRCA1 and BRCA2 are responsible for familial breast

cancers. Somatic mutations in sporadic cases are rare, but chromosomal losses

occur in 30% to 50% of sporadic tumors, respectively (26). BRCA1 acts as a

tumor suppressor gene for both breast and ovarian cancer (16). It encodes a

multifunctional protein involved in DNA repair, cell cycle checkpoint control,

protein ubiquitinylation, and chromatin remodeling (19). In vitro studies showed

that decreased BRCA1 expression in cells led to increased levels of tumor

growth, while increased expression of BRCA1 led to growth arrest and apoptosis.

Inactivation of BRCA1 by promoter methylation is seen in 7% to 31% of

Table 2 Genes Methylated in Breast Cancer

Gene Cancer hallmark contribution Reference

RASSF1A Apoptosis 15

CYP26A1 Microenvironment, encodes for cytochrome p450 enzymes 17

KCNAB1 Microenvironment 17

SNCA Microenvironment 17

RARb Growth regulation 15

TWIST Cell lineage and differentiation 15

Cyclin D2 Cell cycle regulation 15

APC Cell cycle regulation and apoptosis 8

BRCA1 Cell cycle regulation 8,15

BRCA2 Tumor suppressor gene 15

p16 Cell cycle regulation 8,15

DAPK Apoptosis 8

GSTP1 Cellular metabolism and carcinogen detoxification 8

TMS1 Apoptosis 8

PR Growth regulation 8

CDH1 Angiogenesis 8

WWOX Apoptosis 38

50 Dworkin et al.


sporadic breast cancer cases (27). It is believed that aberrant methylation pairs

with loss of heterozygosity to reduce BRCA1 expression in invasive sporadic

breast tumors (16,19,22,26,28). The magnitude of the decrease of functional

BRCA1 protein correlates with disease prognosis (19,22). Phenotypically,

BRCA1-methylated tumors are similar to tumors with germline mutations.

Higher-grade tumors tend to show complete loss of BRCA1 protein. In contrast to

BRCA1, BRCA2 does not show as high a degree of promoter methylation (16).

ESR1 and ESR2

Along with other genes, the estrogen receptors (ERs) ERa and ERb have been

implicated in breast cancer development. ERa is encoded by ESR1, and when

estrogen is activated, it stimulates cell proliferation. ERb is encoded by ESR2

and is known to inhibit the proliferation and invasion of breast cancer cells. ERaand ERb both have promoter-associated CpG islands that can be abnormally

methylated in breast cancer, but ESR2 methylation has been less well studied

(16,18). Almost all breast cancers show some degree of DNA methylation of the

ESR1 gene promoter, but this methylation is only associated with gene repres-

sion in ~30% of tumors (16,18). Approximately 66% of breast cancers express

ERa. A fraction of breast cancers that are initially ERa positive lose ER

expression during tumor progression, but it is unclear if this is due to methylation

or other causes (18).

The presence or absence of ERs plays an important role in both therapy

and survival (16,18). ER-positive breast cancer can be treated with antiestrogenic

drugs, such as tamoxifen, which binds to the ER, preventing estrogen from

binding. This results in a decrease in growth of ER-positive breast cancer cells.

ER-negative cells are no longer responsive to estrogen; therefore, antiestrogenic

drugs have no effect (18). Reversal of ER-negative status is an attractive target

for breast cancer therapy.

BIOMARKERS AND DETECTION

Gene-specific epigenetic changes for breast cancer are likely to occur early in

tumorigenesis and have potential to be used as early detection and prevention

methods (2). As epigenetic modifications become clear biomarkers for breast

cancer treatment and therapeutic intervention, it is important to understand the

many different techniques available for studying and detecting the presence of

methylation, histone modifications, and microRNAs (11).

New, promising high-throughput methylation detection methods are

available, which enable researchers and clinicians to identify an “epigenetic

signature” specific to breast cancer. Biomarkers based on unique DNA meth-

ylation and histone modification patterns of breast tumors may be used in the

future for diagnosis and early detection. Currently, breast cancer detection relies

on various screening methods. Cytology is the main standard for identification



of abnormalities typical of cellular transformation. Quantitative multiplex

methylation-specific polymerase chain reaction (PCR) (QM-MSP) is a highly

sensitive method to quantitate cumulative gene promoter hypermethylation in

samples where DNA is limiting. First, gene-specific primer pairs are used to

coamplify genes, independent of their methylation status. Next, gene-specific

primers (methylated and unmethylated) conjugated with fluorescent labels are

combined with the first PCR product. The signal generated is proportional to the

extent of DNA amplification (15). Since it is known that multigene methylation

of CpG islands is common in early breast cancer, a panel of nine genes were

tested for their sensitivity and specificity of detecting premalignant changes

using QM-MSP. These genes included RASSF1A, RARb, TWIST, HIN1, Cyclin

D2, APC, BRCA1, BRCA2, and p16. Tests were done using ductal lavage, nipple

aspiration fluids, and fine needle aspirates.

Using this detection panel, the rate of detection of breast cancer cells rises

from 43% sensitivity with cytologic examination alone to 71.4% sensitivity (15).

Hypermethylation of genes that are commonly methylated in breast cancer in

sera in breast cancer patients has also been used to detect early malignant

changes. Sera methylation can be detected using methylation-specific PCR (MS-

PCR), a noninvasive PCR-based technique. Early studies indicate that MS-PCR

is a promising approach to screen putative cancer patients (25). This technique

utilizes primers designed for methylated or unmethylated bisulfite-modified

DNA (29). RASSF1A gene hypermethylation has been detected using MS-PCR

in sera of ovarian cancer patients, with 100% specificity (25).

Recurrence rates for breast cancer in the same breast have been recorded as

high as 40% despite negative pathological margins. Researchers believe that a

primary tumor may serve as a locus from which methylation density progres-

sively diffuses outward to surrounding tissues. CYP26A1, KCNAB1, RASSF1A,

and SNCA have found to be frequently methylated in adjacent normal tissue. A

methylation profile containing genes such as these, or those identified in other

studies (Table 2), could be used by clinicians to determine whether patients

should undergo radiological therapy or other treatments to prevent future

recurrence (17).

MassARRAY and pyrosequencing, two additional methods of detecting

methylation, are currently being developed for use in clinical settings.

MassARRAY is a highly sensitive technique that uses base-specific cleavage and

matrix-assisted laser desorption/ionization time-to-flight mass spectrometry

(MALDI-TOF MS). It is capable of detecting DNA methylation levels as low as

5%. It is suitable for testing methylation patterns on various sources, including

archival tissues and laser capture microdissected specimens (1). Pyrosequencing

is a locus-specific quantitative method that utilizes the detection of pyrophos-

phate, which is liberated from incorporated nucleotides by DNA polymerase

during strand elongation. Free pyrophosphates are converted to adenosine tri-

phosphate (ATP), which provides energy for the oxidation of luciferin to then

generate light. Nucleotides are added sequentially to enable base calling.

52 Dworkin et al.


Pyrosequencing has two major advantages over MS-PCR. First, the data are

actual nucleotide sequences rather than fluorescence data. Second, pyrose-

quencing can detect partially methylated sequences that are outside of the priming

sites (30).

Commonly used methods for detecting histone modifications usually rely

on chromatin immunoprecipitation (ChIP). In ChiP, DNA is cross-linked to

chromatin, followed by co-immunoprecipitation using antibodies against specific

histone residues or chromatin proteins. “ChIP” DNA can be used in microarrays,

cloning, and sequencing (11). Techniques for detecting histone modifications

have not yet been adapted for clinical application.

CLINICAL APPLICATION

Breast cancer prevention, treatments, and diagnostics are being developed to

target epigenetic changes leading to breast cancer. Treatments for breast cancer

currently being studied extensively focus on reversing aberrant DNA methyl-

ation and histone acetylation of tumor suppressor genes and genes involved in

therapeutic response. Combinations of epigenetic-targeted therapies with con-

ventional chemotherapeutic agents may provide a way to resensitize drug-

resistant tumors (Table 3).

Methylation Therapy

DNA methylation in proliferating cells is dependent on continued expression of

DNMTs. Treatments that focus on inhibiting the expression of these proteins

would therefore result in progressive reduction in DNA methylation in newly

divided cells. There are three main classes of DNA methylation inhibitors:

nucleoside inhibitors, nonnucleoside weak inhibitors, and rationally designed

inhibitors. Nucleoside inhibitors trap DNMTs for degradation. However, they

require DNA incorporation and active DNA synthesis, which limits the drug in

hypoproliferating cells. 5-Azacytidine (5-aza-dC) and its deoxy analog are two

of the first hypomethylating agents approved by the U.S. Food and Drug

Administration for cancer treatment (31). They work by two mechanisms: (i) as

demethylators by incorporating into CpG sites opposite a methylated CpG site on

the template strand and (ii) by binding to the cytosine in the catalytic domain of

the DNMT1 enzyme. Within a few hours of treatment with 5-aza-dC, there is

rapid passive loss of methylation (32). Nonnucleoside inhibitors are not as well

understood and are limited by their low level of hypomethylation induction.

Nonnucleoside inhibitors were first discovered for their use in reducing autor-

eactivity for autoimmune diseases. During this study researchers discovered their

potential as demethylating agents. Researchers are currently using clinically

approved nonnucleoside inhibitors as a starting point to study their use as

demethylating agents. Rationally designed inhibitors are also in development.

These drugs are designed on the basis of knowledge of specific chemical



Table 3 Drugs in Use for Breast Cancer Treatment

Nucleoside inhibitors

Drug Use Research level Reference

5-Azacytidine Trap DNMTs for degradation Clinical use 31,40

5-aza-

20deoxycytidineTrap DNMTs for degradation Clinical use 31

Zebularine Unknown mechanism Shown promise in vitro 31

50-fluoro-20deoxycytidine

Unknown mechanism In clinical trials 31

Nonnucleoside inhibitors


Procainamide Mechanism not well

understood but proven to

decrease expression of

DNMT1 and DNMT3a

Clinically used as

an antihypertensive

drug

31,41,42

Hydralazine Mechanism not well

understood but proven to

decrease expression of

DNMT1 and DNMT3a

Clinically used as a

vasodilator

31,41,42

Rationally designed inhibitors


NSC303530 Blocks the active site of DNMTs Basic research 31,43

NSC401077 Blocks the active site of DNMTs Basic research 31,43

HDAC inhibitors

Drug Use Advantages Disadvantages

Research

level Reference

LAQ824 HDAC

inhibitor

Unknown Important to

pay attention

to schedule of

administration

Preclinical

level; studies

on hemato-

logical

malignancies

40

Magnesium

Valporate

and

Hydralazine

HDAC and

methyl-

transferase

inhibitor

Increases

efficacy of

conventional

cytotoxic

agents

Unknown Phase III

clinical

trials

44

N-hydroxy-2-

propenamide

HDAC

inhibitor

Suppresses

mammary

tumor growth

in cells

Unknown Research

level

l45

Abbreviations: DNMTs, DNA methyltransferases; HDAC, histone deacetylase.

54 Dworkin et al.


responses in the body, tailoring combinations of these to fit a treatment profile.

An example of a rational design involves the use of three-dimensional infor-

mation to target the drug. Limitations of DNA methylation inhibitors include the

cooperation between different DNMTs such that the drugs must inhibit several of

them simultaneously. All demethylating agents suffer from nonspecificity and

toxicity. The next phase of drug development is how to target gene-specific

hypomethylation (e.g., unmethylated oligonucleotides) (31).

Histone Modification Therapy

Histone deacetylases (HDACs) are also targets for therapy (Fig. 3). HDACs are

important because of their role in cell cycle arrest and inhibition of apoptosis.

These drugs have been studied in isolation and synergistically with DNMTs.

Therapies that target both DNMTs and HDACs act globally, which can cause

reactivation of other genes that are typically silenced, such as oncogenes (18,33–37).

ER Therapy

Drugs are in development for different types of breast cancer. A therapy that could

restore gene expression of ERS1 (for ER-negative cancer patients) could rees-

tablish cancer cell growth regulation through estrogen. After reexpression of

ERS1, antiestrogenic drugs could subsequently be used. A number of drugs,

including DNMT inhibitors and HDAC inhibitors, are being used to reactivate ER

expression (Table 4). DNMTs and HDACs work synergistically to silence gene

expression of ER. One problem found in treatments that target DNMTs and HDACs

is that alone they may not be enough to reverse ER-promoter hypermethylation.

Figure 3 Drug impact on histone modifications. HDAC modifies histones and leads to

chromatin remodeling so that genes are transcriptionally inactive. HAT can reverse the

effects of HDAC to acetylate histones and open the confirmation into an active state.

Drugs such as Scriptaid, TSA, LAQ824, and N-hydroxy-2-propenamide can act on this

pathway to block HDACs and keep DNA transcriptionally active. Abbreviations: HDAC,

histone deactylase; HAT, histone acetyltransferase; TSA, trichostatin A. Source: Cartoon

compliments of Kristi Bennett.



Tamoxifen is a commonly used drug for ERa-positive breast cancers.

Unfortunately, acquired resistance of tamoxifen occurs in nearly 40% of patients.

Therefore, it is important to identify patients most likely to respond to

Tamoxifen and to identify individuals likely to acquire resistance. One prom-

ising biomarker for Tamoxifen response is WWOX. Preliminary studies show

that WWOX expression levels predict tamoxifen resistance better than the two

previously known biomarkers, PR and HER2 (38). WWOX appears to mediate

tamoxifen sensitivity, and its expression is reduced in a large fraction (63.2%) of

breast cancers (39). The primary mechanism of downregulation of WWOX is

through DNA hypermethylation of its regulatory region.

FUTURE DIRECTIONS

The next 10 years are likely to bring significant advances for breast cancer

treatment. Therapeutics that target methylation and histone modifications may

play a leading role in treating breast cancer. Since epigenetic modifications can

also be used as biomarkers, targeted therapies may some day be used as pre-

ventative measures.

REFERENCES

1. Brena RM, Huang TH, Plass C. Quantitative assessment of DNA methylation:

potential applications for disease diagnosis, classification, and prognosis in clinical

settings. J Mol Med 2006; 84(5):365–377.

2. Balch C, Montgomery JS, Paik HI, et al. New anti-cancer strategies: epigenetic

therapies and biomarkers. Front Biosci 2005; 10:1897–1931.

3. Luczak MW, Jagodzinski PP. The role of DNA methylation in cancer development.

Folia Histochem Cytobiol 2006; 44(3):143–154.

Table 4 Drugs Involved in Reactivation of ER Expression

Treatment Result Reference

5-aza-Dc

(methyltransferase

inhibitor)

Partial demethylation of ER CpG islands,

reexpression of ER mRNA, and synthesis

of functional ER protein

18,36,37,46

Antisense oligonucleotides

(for DNMT1)

ER gene reexpression and estrogen

responsiveness reestablished

18,35

TSA (HDAC inhibitor) Reexpression of ERa protein 18,33

DNMT and HDAC

inhibitor

Reactivation ERa protein and expression 18,37

Scriptaid (HDAC inhibitor) Reexpression of functional ER in vivo 18,34

Scriptaid and 5-aza-Dc More effective in inducing ER expression

than just Scriptaid or 5-aza-Dc alone

18,34

Abbreviations: ER, estrogen receptor; DNMT, DNA methyltransferase; HDAC, histone deacetylase;

TSA, trichostatin A.

56 Dworkin et al.


4. Miyamoto K, Ushijima T. Diagnostic and therapeutic applications of epigenetics. Jpn

J Clin Oncol 2005; 35(6):293–301.

5. Reamon-Buettner SM, Borlak J. A new paradigm in toxicology and teratology:

altering gene activity in the absence of DNA sequence variation. Reprod Toxicol

2007; 24(1):20–30.

6. Szyf M. Therapeutic implications of DNA methylation. Future Oncol 2005; 1(1):

125–135.

7. Ting AH, McGarvey KM, Baylin SB. The cancer epigenome—components and

functional correlates. Genes Dev 2006; 20(23):3215–3231.

8. Widschwendter M, Jones PA. DNA methylation and breast carcinogenesis. Onco-

gene 2002; 21(35):5462–5482.

9. Hellebrekers DM, Griffioen AW, van Engeland M. Dual targeting of epigenetic

therapy in cancer. Biochim Biophys Acta 2007; 1775(1):76–91.

10. Hellebrekers DM, Jair KW, Vire E, et al. Angiostatic activity of DNA methyl-

transferase inhibitors. Mol Cancer Ther 2006; 5(2):467–475.

11. Curt Balch TH-MH, Nephew KP. High-throughput assessments of epigenetics in

human disease. In: Kim S, Tang H, Mardis ER, eds. Advances in Genome Sequencing

Technology and Algorithms, Norwood, MA: Artech House Publishers, 2007.

12. Han J, Kim D, Morris KV. Promoter-associated RNA is required for RNA-directed

transcriptional gene silencing in human cells. Proc Natl Acad Sci U.S.A. 2007; 104(30):

12422–12427.

13. Oka M, Rodi�c N, Graddy J, et al. CpG sites preferentially methylated by Dnmt3a in

vivo. J Biol Chem 2006; 281(15):9901–9908.

14. Ooi SK, Qiu C, Bernstein E, et al. DNMT3L connects unmethylated lysine 4 of

histone H3 to de novo methylation of DNA. Nature 2007; 448(7154):714–717.

15. Fackler MJ, Malone K, Zhang Z, et al. Quantitative multiplex methylation-specific

PCR analysis doubles detection of tumor cells in breast ductal fluid. Clin Cancer Res

2006; 12(11 Pt 1):3306–3310.

16. Campan M, Weisenberger DJ, Laird PW. DNA methylation profiles of female ste-

roid hormone-driven human malignancies. Curr Top Microbiol Immunol 2006;

310:141–178.

17. Yan PS, Venkataramu C, Ibrahim A, et al. Mapping geographic zones of cancer risk

with epigenetic biomarkers in normal breast tissue. Clin Cancer Res 2006; 12(22):

6626–6636.

18. Giacinti L, Claudio PP, Lopez M, et al. Epigenetic information and estrogen receptor

alpha expression in breast cancer. Oncologist 2006; 11(1):1–8.

19. Mirza S, Sharma G, Prasad CP, et al. Promoter hypermethylation of TMS1, BRCA1,

ERalpha and PRB in serum and tumor DNA of invasive ductal breast carcinoma

patients. Life Sci 2007; 81(4):280–287.

20. Sharma D, Blum J, Yang X, et al. Release of methyl CpG binding proteins and histone

deacetylase 1 from the Estrogen receptor alpha (ER) promoter upon reactivation in

ER-negative human breast cancer cells. Mol Endocrinol 2005; 19(7):1740–1751.

21. Sui M, Huang Y, Park BH, et al. Estrogen receptor alpha mediates breast cancer cell

resistance to paclitaxel through inhibition of apoptotic cell death. Cancer Res 2007;

67(11):5337–5344.

22. Vincent-Salomon A, Ganem-Elbaz C, Manie E, et al. X inactive-specific transcript

RNA coating and genetic instability of the X chromosome in BRCA1 breast tumors.

Cancer Res 2007; 67(11):5134–5140.



23. Visvanathan K, Sukumar S, Davidson NE. Epigenetic biomarkers and breast cancer:

cause for optimism. Clin Cancer Res 2006; 12(22):6591–6593.

24. Zhou Q, Davidson NE. Silencing estrogen receptor alpha in breast cancer cells.

Cancer Biol Ther 2006; 5(7):848–849.

25. Pfeifer GP, Dammann R. Methylation of the tumor suppressor gene RASSF1A in

human tumors. Biochemistry (Mosc) 2005; 70(5):576–583.

26. Esteller M, Silva JM, Dominguez G, et al. Promoter hypermethylation and BRCA1

inactivation in sporadic breast and ovarian tumors. J Natl Cancer Inst 2000; 92(7):

564–569.

27. Wei M, Grushko TA, Dignam J, et al. BRCA1 promoter methylation in sporadic

breast cancer is associated with reduced BRCA1 copy number and chromosome 17

aneusomy. Cancer Res 2005; 65(23):10692–10699.

28. Catteau A, Morris JR. BRCA1 methylation: a significant role in tumour develop-

ment? Semin Cancer Biol 2002; 12(5):359–371.

29. Herman JG, Graff JR, Myohanen S, et al. Methylation-specific PCR: a novel PCR

assay for methylation status of CpG islands. Proc Natl Acad Sci U.S.A. 1996; 93(18):

9821–9826.

30. Shames DS, Minna JD, Gazdar AF. Methods for detecting DNA methylation in

tumors: from bench to bedside. Cancer Lett 2007; 251(2):187–198.

31. Issa JP.DNAmethylation as a therapeutic target in cancer.ClinCancerRes2007; 13(6):

1634–1637.

32. Christman JK. 5-Azacytidine and 5-aza-20-deoxycytidine as inhibitors of DNA

methylation: mechanistic studies and their implications for cancer therapy. Oncogene

2002; 21(35):5483–5495.

33. Kawai H, Li H, Avraham S, et al. Overexpression of histone deacetylase HDAC1

modulates breast cancer progression by negative regulation of estrogen receptor

alpha. Int J Cancer 2003; 107(3):353–358.

34. Keen JC, Yan L, Mack KM, et al. A novel histone deacetylase inhibitor, scriptaid,

enhances expression of functional estrogen receptor alpha (ER) in ER negative

human breast cancer cells in combination with 5-aza 20-deoxycytidine. Breast CancerRes Treat 2003; 81(3):177–186.

35. Yan L, Nass SJ, Smith D, et al. Specific inhibition of DNMT1 by antisense oligo-

nucleotides induces re-expression of estrogen receptor-alpha (ER) in ER-negative

human breast cancer cell lines. Cancer Biol Ther 2003; 2(5):552–556.

36. Yang X, Ferguson AT, Nass SJ, et al. Transcriptional activation of estrogen receptor

alpha in human breast cancer cells by histone deacetylase inhibition. Cancer Res

2000; 60(24):6890–6894.

37. Yang X, Phillips DL, Ferguson AT, et al. Synergistic activation of functional

estrogen receptor (ER)-alpha by DNA methyltransferase and histone deacetylase

inhibition in human ER-alpha-negative breast cancer cells. Cancer Res 2001; 61(19):

7025–7029.

38. Guler G, Iliopoulos D, Guler N, et al. Wwox and Ap2g expression levels predict

tamoxifen response. Clin Cancer Res 2007; 13(20).

39. Guler G, Uner A, Guler N, et al. The fragile genes FHIT and WWOX are inactivated

coordinately in invasive breast carcinoma. Cancer 2004; 100(8):1605–1614.

40. Hurtubise A, Momparler RL. Effect of histone deacetylase inhibitor LAQ824 on

antineoplastic action of 5-Aza-20-deoxycytidine (decitabine) on human breast car-

cinoma cells. Cancer Chemother Pharmacol 2006; 58(5):618–625.

58 Dworkin et al.


41. Chuang JC, Yoo CB, Kwan JM, et al. Comparison of biological effects of non-

nucleoside DNA methylation inhibitors versus 5-aza-20-deoxycytidine. Mol Cancer

Ther 2005; 4(10):1515–1520.

42. Cornacchia E, Golbus J, Maybaum J, et al. Hydralazine and procainamide inhibit

T cell DNA methylation and induce autoreactivity. J Immunol 1988; 140(7):

2197–2200.

43. Siedlecki P, Boy RG, Musch T, et al. Discovery of two novel, small-molecule

inhibitors of DNA methylation. J Med Chem 2006; 49(2):678–683.

44. Arce C, Perez-Plasencia C, Gonzalez-Fierro A, et al. A proof-of-principle study of

epigenetic therapy added to neoadjuvant doxorubicin cyclophosphamide for locally

advanced breast cancer. PLoS ONE 2006; 1:e98.

45. Joung KE, Min KN, An JY, et al. Potent in vivo anti-breast cancer activity of IN-

2001, a novel inhibitor of histone deacetylase, in MMTV/c-Neu mice. Cancer Res

2006; 66(10):5394–5402.

46. Ferguson AT, Vertino PM, Spitzner JR, et al. Role of estrogen receptor gene

demethylation and DNA methyltransferase. DNA adduct formation in 5-aza-

20deoxycytidine-induced cytotoxicity in human breast cancer cells. J Biol Chem

1997; 272(51):32260–32206.



5

Comparative Genomic Hybridization

and Copy Number Abnormalities in

Breast Cancer

Nicholas Wang and Joe Gray

Life Sciences Division, Lawrence Berkeley National Laboratory,Berkeley, California, U.S.A.

INTRODUCTION

The progression of normal breast epithelial cells from a normal state toward one

characterized by uncontrolled growth and metastatic behavior is caused by the

deregulation of key cellular processes and signaling pathways. These alterations

in normal cellular behavior are rooted in the accumulation of genomic and

epigenomic lesions that impact hallmarks of cancer, such as the ability of the cell

to control proliferation, undergo apoptosis, increase motility leading to invasion,

and alter angiogenesis. A suite of technologies has now been developed to assess

genomic and epigenomic aberrations that contribute to cancer progression. The

application of these has shown that genome copy number abnormalities (CNAs)

are among the most frequent genomic aberrations. Remarkably, these studies

have revealed that 10% to 15% of the genes in a typical carcinoma tumor may be

deregulated by recurrent genome CNAs and regions. Some of these genes

influence disease progress and so may be assessed to facilitate prognosis. Others

influence response to therapy and so may be assessed as predictive markers.

Some of these enable oncogenic processes, on which tumors depend for survival,

and so are candidate therapeutic targets. In most cases, array comparative

61


genomic hybridization (CGH) is the method of choice for their discovery and

may be used in some setting for clinical assessments of these abnormalities.

Accordingly, we review here several of the current array CGH technologies

available and the considerations needed when determining the technology most

applicable to a given study.

CGH analyses of thousands of primary breast tumors and cell lines have

now defined recurrent CNAs throughout the genome and identified genes

encoded in these aberrant regions that are deregulated by the abnormalities (1,2–4).

In breast cancer, for example, 10% to 15% of known genes are deregulated by

these CNAs (2–4). The aberrations tend to fall into two general classes that may

contribute in different ways to cancer pathophysiology: low-level CNAs that

occur through the gain or loss of extended regions of single chromosomes or

extended regions thereof and high-level amplification or homozygous deletions

that tend to be narrower in genomic extent.

Most genes are deregulated by low-level aberrations. The contributions of

these genes to cancer pathophysiology are not well understood; however, evi-

dence is now emerging that these may provide, en ensemble, a general metabolic

advantage (4–6). These aberrations typically occur relatively early in the carci-

nogenesis process. It has been proposed that these aberrations are selected during

successful transition through a telomere crisis (7).

High-level amplification and narrow homozygous deletions provide strong

evidence of active selection during cancer progression and seem to occur later

during cancer progression. In breast cancer, regions of high-level amplification

encode well-studied oncogenes, such as ERBB2 (17q12) (8), FGFR1 (8p11) (9),

PVT1 and MYC (8q24) (10,11), CCND1 and EMS1 (11q13) (12), and ABI1 and

ZNF217 (20q13) (13,14), while deletions contribute to the inactivation of tumor

suppressor genes such as TP53 (17p13) (15), CDKN2A (9p21) (16), and BRCA1

(13q14) (17). While it is often assumed that these oncogenes and tumor sup-

pressor genes are the sole targets for these aberrations, a growing body of

functional data suggests that several genes in each region of recurrent abnor-

mality may be coselected by the event. In breast cancer, for example, regions of

amplification in which more than one gene appears to be functionally important

include 8p11, 8q24, 11q13, 17q21, and 20q13 (4). Table 1 summarizes published

studies for genes in these regions of amplification. The concept of multigene

selection is also supported by studies in mouse models, where the extents of

amplicons in mice and humans are similar. Similarities in the extent of ampli-

fication around ERBB2 in humans and in polyomavirus middle T-induced

mammary tumors in mice are particularly striking (18). Similar studies are

emerging at other organ sites (19). As a result, it is critically important that the

extent of each region be defined as precisely as possible. High-resolution CGH

analyses are now contributing to this process.

Several recurrent high-level CNAs are associated with clinical endpoints,

including reduced survival duration and response to aberration-targeted therapies.

Assays for recurrent aberrations or genes deregulated by CNA abnormalities have

62 Wang and Gray


already been incorporated into molecular indicators of outcome. In addition,

several therapeutic agents are now being developed to attack genes deregulated

fully or in part by recurrent genome abnormalities so that assays for these targets

may be useful as predictors of response. In breast cancer, the most prominent

examples are trastuzumab (20,21) and lapatinib (22) (NEJM PMID 17192538),

developed to attack tumors driven by amplification of ERBB2. In addition,

therapies have been developed to treat tumors arising from deletion or inactivation

of TP53 (23) and activation of elements of the PI3-kinase pathway, including

PIK3CA, PTEN, AKT, and ZNF217 (24–27).

The ability to detect and define the extents of CNAs has steadily improved

with the evolution of CGH analysis technologies (Table 2). The first CGH

studies mapped CNAs onto metaphase chromosomes, but the resolution of such

analyses was limited to about 10 Mbp by the complex packaging of the DNA

into chromosomes (28). The introduction of array CGH in which CNAs were

mapped onto arrays of cloned probes substantially increased resolution. Early

arrays comprised hundreds to thousands of probes and improved resolution to

about 1 Mbp (29,30). Current cloned probe arrays interrogate the genome at

approximately single-gene resolution (5). Today, several commercial platforms

map abnormalities onto arrays of oligonucleotide probes with subgene resolution

(either on flat substrates or attached to microbeads) (31), and some allow

analyses in an allele-specific manner (32). Applications of these platforms are

further refining the extents of recurrent genomic aberrations but also reveal the

somewhat surprising existence of germline copy number polymorphisms that

must be taken into account when interpreting CGH measurements. Several

aspects of these important array CGH analysis technologies are reviewed in the

following sections.

Table 1 Array CGH Analysis Platform Summary

Platform

composition Reference Density

Allele

specific Source

cDNA 5 40K No Academic

BAC 68 25K No Academic/Spectral

Genomics

Oligonucleotide http://affymetrix.com 950K Yes Affymetrix

Oligonucleotide http://www.

nimblegen.com

400K No Nimblegen

Oligonucleotide http://www.chem.

agilent.com

240K No Agilent

Oligonucleotide http://www.illumina.

com

1M Yes Illumina

MIP 41 20K Yes Academic

Abbreviations: CGH, comparative genomic hybridization; BAC, bacterial artificial chromosome;

MIP, molecular inversion probe.

Comparative Genomic Hybridization in Breast Cancer 63


COMPARATIVE GENOMIC HYBRIDIZATION

CGH is the primary technology used for assessment of genome copy number.

Typically, analysis proceeds by hybridizing differentially labeled tumor and nor-

mal DNA to a representation of the normal genome, although some platforms

perform tumor and normal reference hybridization in separate representations. The

ratio of tumor label to normal label, hybridized to each locus on the representation

is then measured as an estimate of relative copy number at that locus.

The first CGH experiments employed metaphase chromosomes as the

representation onto which information was mapped (28,33). While useful, the

genomic resolution was limited by the nonlinear organization of DNA along

metaphase chromosomes. Thus, chromosome CGH has now been largely sup-

planted by array CGH, in which information is mapped onto arrays of DNA

probes (Fig. 1). The initial array CGH platforms comprised large genomic cloned

probes, such as yeast artificial chromosomes (YACs) (34), bacterial artificial

chromosomes (BACs), and P1-derived artificial chromosomes (PACs) (29), or

shorter cDNA sequences spotted on glass slides (35). More recently, synthetic

oligonucleotide probes have been used successfully (36–39). Several platforms

are summarized in Table 1 and in the following sections.

Cloned probe arrays are typically available from large academic cores with

the experience and resources to maintain and print the arrays. These arrays are

developed using the large collections of cDNAs (35) or BACs (29,30,40)

developed during the course of the human genome program. The advantages of

Table 2 Data Analysis Algorithms

Platform Program Reference Type

Required

software

Two-color arrays aCGH 54 HMM R

GLAD 56 Adaptive weight

smoothing

R

CLAC 57 Clustering along

chromosomes

R

DNAcopy 55 Segmentation R

Single-color arrays

(Affymetrix)

dChipSNP 37 HMM none

CARAT 69 none

CNAG 70 Regression none

GEMCA 71 Regression none

Abbreviations: aCGH, array comparative genomic hybridization; HMM, hidden Markov model;

GLAD, Gaussian model–based approach; CART, copy number analysis with regression and tree; CLAC,

clustering along each chromosome; SNP, single nucleotide polymorphism; CNAG, Copy Number

Analyzer for GeneChip; GEMCA, Genotyping microarray based copy number variation analysis.

64 Wang and Gray


this approach are that the clones are readily available in the public domain and

the signals produced during hybridization are large so that measurement preci-

sion is relatively high. The disadvantages are that the clones are not perfectly

curated and mapped, so some may be mismapped. In addition, managing the

large number is daunting, and the ultimate resolution of the analysis platforms is

limited by the relatively genomic extents of the cloned probes.

Commercial vendors, such as Affymetrix, Agilent, Illumina, and Nim-

bleGen, have begun offering oligonucleotide-based arrays with which to study

genomewide CNAs (41–43). The Illumina arrays are unique in that the oligo-

nucleotide probes are attached to microbeads (39), while the NimbleGen arrays

are synthesized on demand and so can be easily customized (44). The smaller

target sequences, 25 to 85 bp, also offer the possibility of higher resolution than

is possible using cloned probe arrays. Some protocols generate relative copy

number information by comparing tumor and normal hybridization signals using

two-color cohybridization approaches, while others rely on hybridizations of

tumor and normal reference DNA to separate arrays. The latter approach is

technically simpler but requires additional normalization steps to account for

array-to-array differences, introducing additional experimental variables when

analyzing data sets.

Figure 1 The use of a dual-color hybridization protocol for the detection of genome copy

number aberrations. Both a normal sample (representing two copies of the genome) and a

tumor sample are randomly labeled with differential fluorescent dyes. They are then

cohybridized onto an array platform (BACs or oligonucleotides) with the subsequent log2dye ratios of the tumor sample compared with the normal sample used to detect copy

number changes. Abbreviation: BACs, bacterial artificial chromosomes.



Since each oligonucleotide probe is synthesized to be complementary to a

known sequence, oligonucleotide array platforms do not suffer as much from

incorrect mapping, and probes can be selected to be uniquely represented in the

genome, thereby minimizing cross-hybridization artifacts. The smaller sequen-

ces also are useful in detecting single nucleotide polymorphisms (SNPs), since

hybridization rates to 25 mer are significantly reduced by a single nucleotide

mismatch (37,45). The additional information provides researchers the ability to

detect loss of heterozygosity (LOH) in regions of neutral copy change and also

provide data for use in association studies (42). Regions of LOH that do not carry

imbalances in copy number may provide evidence of selective pressures dupli-

cating one allele coupled with the loss of the other. This is an important

mechanism enabling removal of wild-type alleles while selecting mutations.

Allele-specific selection can be either positive or negative. For example, loss of

the wild-type allele and duplication of an inactivating mutation might lead to

inactivation of a tumor suppressor gene. The high frequency of LOH on 17p

where p53 is located is an example of this type of selection (Fig. 2). On the other

hand, amplification of a polymorphism that increases gene activity may con-

tribute an advantage to the tumor. Amplification of a specific polymorphism in

the aurora kinase STK15 is one example (46). The utility of allele-specific copy

number analysis may also contribute to the identification of polymorphisms that

confer increased susceptibility to cancer, since evidence is now emerging that

tumors may increase the number of copies of polymorphisms that confer

increased risk and may decrease the number of copies of polymorphisms that

confer decreased risk of developing cancer (47). The frequency of allele-specific

copy increases and decreases in individual tumors is substantial, as illustrated in

Figure 3.

While commercial oligonucleotide arrays offer advantages, the smaller

target sequences also present some disadvantages. Foremost among these is the

relatively low signal-to-noise ratio generated during hybridization. Because the

hybridization sequences are smaller, there is less nucleic acid sequence available

to label, leading to less intense labeling. This typically leads to higher variability

between probe sets in oligonucleotide arrays, and the need to combine multiple

data points to estimate copy number, which in turn reduces the actual resolution

of these arrays (38,48). The increased variability compared to BAC arrays also

makes the detection of low-level changes, such as single-copy gains and losses,

more difficult to detect, although performance with oligonucleotide arrays is

improving steadily. Typically, microgram quantities of starting DNA are needed

for an analysis.

The most recent array CGH analysis platform does not operate through

hybridization of tumor and normal DNA to a normal representation. Instead, it

utilizes a molecular inversion probe (MIP) interrogation strategy to perform

allele-specific copy number analysis (45). In this approach, each locus is inter-

rogated by a linear MIP probe that has sequence homology to approximately

20-bp regions flanking a single base to be tested. Each MIP also contains a unique

66 Wang and Gray


tag sequence and universal polymerase chain reaction (PCR) primer-binding

sites facing away from each other so that PCR with primers to the universal

sequences will not amplify. The first step in MIP CGH is to hybridize the linear

MIPs to the DNA sample to be interrogated. After hybridization, the reaction

mix is separated into multiple tubes, each containing polymerase, ligase, and one

of the four nucleotides. The polymerase and ligase then fill the gaps in each

reaction mix that contains a nucleotide complementary to the base at the test

locus. The linear DNA and probes are then eliminated, reactions are mixed, and

the ligated probes are amplified and labeled using PCR with the universal pri-

mers. The resulting PCR products are hybridized to an array comprising oligo-

nucleotides homologous to the tag sequences in the MIPs. To date, MIP analyses

interrogating over 20,000 loci have been reported (45). This approach reports

Figure 2 LOH detection using Affymetrix GeneChip1 Human Mapping 50K Array Xba

240. A panel of breast cancer cell lines is individually hybridized to oligonucleotide

arrays, and copy number changes and LOH are determined using dChipSNP. (A) The

expected result of a genomic deletion is the presence of LOH, as seen on chromosome 8.

(B) However, chromosome 17 shows much higher levels of LOH compared with deletions,

suggesting that these regions have undergone genomic changes such as a deletion or

duplication event or genomic conversion of one homolog so that you maintain normal

copy number but not allelic information. This may result in the loss of a functional allele

of a tumor suppressor gene or an active gain-of-function allele in the case of an oncogene.

These genomic changes are only evident using a platform capable of detecting allelic

changes. Abbreviation: LOH, loss of heterozygositiy.



Figure 3 Allelic copy number changes in breast cancer cell lines from a 50K MIP

array platform. Allele copy numbers are represented as the higher copy allele and

lower copy allele. (A) Allele copy number data from HCC38 from chromosome 5.

The lack of copies of the lower allele is indicative of LOH. The array detects a region

of LOH without copy number change. (B) Allele copy number data from ZR75.30

from chromosome 8. 8q displays a region of monoallelic amplification in the region

that contains MYC. Abbreviations: MIP, molecular inversion probe; LOH, loss of

heterozygositiy.

68 Wang and Gray


copy number in an allele-specific manner and works well with FFPE material

since only ~40 bp of intact DNA is needed at each locus for proper analysis.

Typically, only about 75 ng of DNA is needed for an analysis since the reaction

itself contains a PCR amplification step. The approach has the advantage that the

tag sequences and arrays can be designed to have uniform hybridization con-

ditions to copy number and the hybridization does not depend on the quality of

the hybridized DNA.

DNA REQUIREMENTS

One of the most important factors when considering a clinical CGH tool is the

quality and amount of DNA available for analysis. Depending on which array

platform is chosen, the amounts needed for successful hybridization can range

from 50 ng to 3 mg. In addition, some protocols have been developed that allow

analysis of material from formalin-fixed, paraffin-embedded (FFPE) samples

(39,49). The largest amounts of starting material generally are used in protocols

in which entire genomes are labeled and hybridized to the array, as is usual for

array CGH using BAC or Agilent oligonucleotide arrays. In these experiments,

the starting DNA is typically labeled with fluorescent nucleotides using random

priming. For BAC arrays, unlabeled repeat-rich DNA is included to suppress

hybridization of repeated sequences to repetitive sequences that are present in

the BAC probes. Larger amounts of DNA are needed to “drive” these whole-

genome hybridization reactions. The amount of starting material needed for

analysis is reduced if the DNA is amplified or reduced in complexity—for

example, through a PCR amplification step preceded by restriction digestion that

reduces the complexity of the hybridized DNA (50). This complexity reduction

results in an increase of template sequences, which can utilized in downstream

hybridization steps. Both academically and commercially developed platforms

result in data sets that are able to detect gains and losses with the caveat that

variations and noise may differ between platforms (Fig. 4).

While there is a range of starting DNA amounts between the various

platforms, recent protocols have been developed that can amplify a large

quantity of DNA from small starting amounts. These applications can be critical

when dealing with samples that are limiting. These whole-genome amplifica-

tions are useful for both intact, high-quality DNA and degraded samples and

involve the use of j29 enzyme or PCR (51,52). Alternately, the MIP platform has

a targeted PCR amplification inherent to the protocol, so it is particularly tolerant

of small amounts of starting DNA (45). To date, successful data sets have been

generated using as little material as 1 ng or a single cell (53). However, as with

any random amplification procedure, the protocols may artificially create mis-

representations of the genome, resulting in false positives for amplifications and

deletions. Determining the amount and quality of starting material along with the



Figu

re4

Chromosomal

copynumber

changes

inthebreastcancercelllineMDAMB453.DNA

from

MDAMB453isapplied

tothreeseparate

platform

s:(A

)a3KBACarray,(B)a20KMIP

array,and(C

)a50KXbaIAffymetrixSNParray.D

atafromeach

platform

arescaled

usingadjacentdata

pointsso

thattheresultingresolutionsarecomparable.Thedataarethen

smoothed

usingtheGLADsoftwarepackage.Abbreviations:BAC,bacterial

artificialchromosome;MIP,molecularinversionprobe;SNP,singlenucleotidepolymorphism;GLAD,Gaussianmodel–based

approach.

70 Wang and Gray


type of information required will help determine the appropriate array CGH

platform and DNA preparation technique to utilize.

ABNORMALITY DETECTION AND INTERPRETATION

Computational strategies to identify copy number gains and losses are a critical

part of array CGH. The first step in the analysis of data is the identification of

any features with anomalous intensity readings. These could result from a high

background, errors in printing, contamination, or air bubbles prohibiting

hybridization of the sample to the array. After removal of these spurious data

points, a normalization step is applied to calculate the tumor hybridization

intensity along the genome, relative to the normal hybridization intensity. A

statistical analysis is then applied to identify genome segments with different

copy number levels. Several strategies have been developed for this purpose and

are summarized in Table 2. These include the hidden Markov model (HMM)

(54), circular binary segmentation (55), a Gaussian model–based approach

(GLAD) (56), hierarchical clustering along each chromosome (CLAC) (57), an

expectation maximization–based method (58), a Bayesian model (59), and a

wavelet approach (60). Several of these algorithms have been made available as

R packages through Bioconductor (www.bioconductor.org). Comparsions of

several of these methods have now been published (61,62).

Detection of the differences in copy number between tumor and normal

samples is increasingly straightforward and precise as CGH procedures improve.

However, the interpretation of CGH analyses requires additional interpretation.

First, high-resolution array CGH analyses are important now that focal, germline

genome copy number polymorphisms are relatively common (36). These may be

scored as CNAs if the samples for tumor versus normal CGH analyses are not

from the same individual. In addition, it is tempting to think that the regions of

the genome not detected as aberrant during CGH analyses are present in cancer

genomes in two copies. This is often not the case since tumor genomes may be

highly aneuploid. Thus, a CGH comparison of a perfectly tetraploid tumor with a

diploid normal sample will not show any evidence of abnormality. Thus, if

absolute copy number assessments are important, it will be necessary to normalize

the CGH analyses using information about absolute copy number at one or a few

loci measured using fluorescent in situ hybridization (FISH) or quantitative PCR.

CONCLUSIONS AND FUTURE DIRECTIONS

Array CGH technologies are becoming increasingly powerful tools for the

identification of recurrent genome CNAs that are associated with outcome or

response to therapy in breast and other cancers (63). The ability to detect genetic

copy number changes of both large and small scales will be critical in evaluating

the most effective therapies on an individual patient basis. In the short term,

array CGH will likely be accomplished using commercial oligonucleotide

analysis platforms that allow assessment of allele-specific copy number using



material from FFPE samples. In the long term, however, array CGH may be

supplanted by high-throughput sequencing technologies that allow assessment of

both copy number and DNA sequence through deep sequencing. It is already clear

that shotgun sequencing, if done at sufficient redundancy, offers the same or better

resolution than CGH platforms (64,65). Next-generation sequencing platforms

based on massively parallel single-molecule sequencing promise sufficiently low-

cost sequencing, so that these platforms may become a viable alternative to array

CGH (66,67). Today, this is not cost effective, but costs of high-throughput

sequencing are falling rapidly, so these approaches should be followed closely.

REFERENCES

1. Knuutila S, Autio K, Aalto Y. Online access to CGH data of DNA sequence copy

number changes. Am J Pathol 2000; 157:689.

2. Neve RM, Chin K, Fridlyand J, et al. A collection of breast cancer cell lines for the

study of functionally distinct cancer subtypes. Cancer Cell 2006; 10:515–527.

3. Pollack JR, Sorlie T, Perou CM, et al. Microarray analysis reveals a major direct role

of DNA copy number alteration in the transcriptional program of human breast

tumors. Proc Natl Acad Sci U.S.A. 2002; 99:12963–12968.

4. Chin K, DeVries S, Fridlyand J, et al. Genomic and transcriptional aberrations linked

to breast cancer pathophysiologies. Cancer Cell 2006; 10:529–541.

5. Bergamaschi A, Kim YH, Wang P, et al. Distinct patterns of DNA copy number

alteration are associated with different clinicopathological features and gene-expression

subtypes of breast cancer. Genes Chromosomes Cancer 2006; 45:1033–1040.

6. Callagy G, Pharoah P, Chin SF, et al. Identification and validation of prognostic

markers in breast cancer with the complementary use of array-CGH and tissue

microarrays. J Pathol 2005; 205:388–396.

7. Chin K, de Solorzano CO, Knowles D, et al. In situ analyses of genome instability in

breast cancer. Nat Genet 2004; 36:984–988.

8. Slamon DJ, Godolphin W, Jones LA, et al. Studies of the HER-2/neu proto-oncogene

in human breast and ovarian cancer. Science 1989; 244:707–712.

9. Theillet C, Adelaide J, Louason G, et al. FGFRI and PLAT genes and DNA

amplification at 8p12 in breast and ovarian cancers. Genes Chromosomes Cancer

1993; 7:219–226.

10. Escot C, Theillet C, Lidereau R, et al. Genetic alteration of the c-myc protooncogene

(MYC) in human primary breast carcinomas. Proc Natl Acad Sci U.S.A. 1986;

83:4834–4838.

11. Guan Y, Kuo WL, Stilwell JL, et al. Amplification of PVT1 contributes to the

pathophysiology of ovarian and breast cancer. Clin Cancer Res 2007; 13:5745–5755.

12. Schuuring E, Verhoeven E, Mooi WJ, et al. Identification and cloning of two

overexpressed genes, U21B31/PRAD1 and EMS1, within the amplified chromosome

11q13 region in human carcinomas. Oncogene 1992; 7:355–361.

13. Anzick SL, Kononen J, Walker RL, et al. AIB1, a steroid receptor coactivator

amplified in breast and ovarian cancer. Science 1997; 277:965–968.

14. Tanner MM, Tirkkonen M, Kallioniemi A, et al. Amplification of chromosomal

region 20q13 in invasive breast cancer: prognostic implications. Clin Cancer Res

1995; 1:1455–1461.

72 Wang and Gray


15. Varley JM, Brammar WJ, Lane DP, et al. Loss of chromosome 17p13 sequences and

mutation of p53 in human breast carcinomas. Oncogene 1991; 6:413–421.

16. Kamb A, Gruis NA, Weaver-Feldhaus J, et al. A cell cycle regulator potentially

involved in genesis of many tumor types. Science 1994; 264:436–440.

17. Friedman LS, Ostermeyer EA, Lynch ED, et al. The search for BRCA1. Cancer Res

1994; 54:6374–6382.

18. Hodgson JG, Malek T, Bornstein S, et al. Copy number aberrations in mouse breast

tumors reveal loci and genes important in tumorigenic receptor tyrosine kinase

signaling. Cancer Res 2005; 65:9695–9704.

19. Maser RS, Choudhury B, Campbell PJ, et al. Chromosomally unstable mouse tumors

have genomic alterations similar to diverse human cancers. Nature 2007; 447:966–971.

20. Yeon CH, Pegram MD. Anti-erbB-2 antibody trastuzumab in the treatment of HER2-

amplified breast cancer. Invest New Drugs 2005; 23:391–409.

21. Vogel CL, Cobleigh MA, Tripathy D, et al. First-line Herceptin monotherapy in

metastatic breast cancer. Oncology 2001; 61(suppl 2):37–42.

22. Geyer CE, Forster J, Lindquist D, et al. Lapatinib plus capecitabine for HER2-

positive advanced breast cancer. N Engl J Med 2006; 355(26):2733–2743.

23. Kumar S, Walia V, Ray M, et al. p53 in breast cancer: mutation and counter-

measures. Front Biosci 2007; 12:4168–4178.

24. Jonsson G, Staaf J, Olsson E, et al. High-resolution genomic profiles of breast cancer

cell lines assessed by tiling BAC array comparative genomic hybridization. Genes

Chromosomes Cancer 2007; 46:543–558.

25. Wu G, Xing M, Mambo E, et al. Somatic mutation and gain of copy number of

PIK3CA in human breast cancer. Breast Cancer Res 2005; 7:R609–R616.

26. Huang G, Krig S, Kowbel D, et al. ZNF217 suppresses cell death associated with

chemotherapy and telomere dysfunction. Hum Mol Genet 2005; 14:3219–3125.

27. Bellacosa A, de Feo D, Godwin AK, et al. Molecular alterations of the AKT2

oncogene in ovarian and breast carcinomas. Int J Cancer 1995; 64:280–285.

28. Kallioniemi A, Kallioniemi OP, Sudar D, et al. Comparative genomic hybridization

for molecular cytogenetic analysis of solid tumors. Science 1992; 258:818–821.

29. Pinkel D, Segraves R, Sudar D, et al. High-resolution analysis of DNA copy number

variation using comparative genomic hybridization to microarrays. Nat Genet 1998;

20:207–211.

30. Snijders AM, Nowak N, Segraves R, et al. Assembly of microarrays for genome-

wide measurement of DNA copy number. Nat Genet 2001; 29:263–264.

31. Toruner GA, Streck DL, Schwalb MN, et al. An oligonucleotide based array-CGH

system for detection of genome wide copy number changes including subtelomeric

regions for genetic evaluation of mental retardation. Am J Med Genet A 2007;

143:824–829.

32. George RE, Attiyeh EF, Li S, et al. Genome-wide analysis of neuroblastomas using

high-density single nucleotide polymorphism arrays. PLoS ONE 2007; 2:e255.

33. Kallioniemi A, Kallioniemi OP, Piper J, et al. Detection and mapping of amplified

DNA sequences in breast cancer by comparative genomic hybridization. Proc Natl

Acad Sci U.S.A. 1994; 91:2156–2160.

34. Solinas-Toldo S, Lampel S, Stilgenbauer S, et al. Matrix-based comparative genomic

hybridization: biochips to screen for genomic imbalances. Genes Chromosomes

Cancer 1997; 20:399–407.



35. Pollack JR, Perou CM, Alizadeh AA, et al. Genome-wide analysis of DNA copy-

number changes using cDNA microarrays. Nat Genet 1999; 23:41–46.

36. Lucito R, Healy J, Alexander J, et al. Representational oligonucleotide microarray

analysis: a high-resolution method to detect genome copy number variation. Genome

Res 2003; 13:2291–2305.

37. Zhao X, Li C, Paez JG, et al. An integrated view of copy number and allelic

alterations in the cancer genome using single nucleotide polymorphism arrays.

Cancer Res 2004; 64:3060–3071.

38. Barrett MT, Scheffer A, Ben-Dor A, et al. Comparative genomic hybridization using

oligonucleotide microarrays and total genomic DNA. Proc Natl Acad Sci U.S.A.

2004; 101:17765–17770.

39. Oosting J, Lips EH, van Eijk R, et al. High-resolution copy number analysis of

paraffin-embedded archival tissue using SNP BeadArrays. Genome Res 2007; 17:

368–376.

40. Hodgson G, Hager JH, Volik S, et al. Genome scanning with array CGH delineates

regional alterations in mouse islet carcinomas. Nat Genet 2001; 29:459–464.

41. Iljin K, Wolf M, Edgren H, et al. TMPRSS2 fusions with oncogenic ETS factors in

prostate cancer involve unbalanced genomic rearrangements and are associated with

HDAC1 and epigenetic reprogramming. Cancer Res 2006; 66:10242–10246.

42. Zhou X, Mok SC, Chen Z, et al. Concurrent analysis of loss of heterozygosity (LOH)

and copy number abnormality (CNA) for oral premalignancy progression using the

Affymetrix 10K SNP mapping array. Hum Genet 2004; 115:327–330.

43. Sharp AJ, Hansen S, Selzer RR, et al. Discovery of previously unidentified genomic

disorders from the duplication architecture of the human genome. Nat Genet 2006;

38:1038–1042.

44. Selzer RR, Richmond TA, Pofahl NJ, et al. Analysis of chromosome breakpoints in

neuroblastoma at sub-kilobase resolution using fine-tiling oligonucleotide array

CGH. Genes Chromosomes Cancer 2005; 44:305–319.

45. Wang Y, Moorhead M, Karlin-Neumann G, et al. Allele quantification using

molecular inversion probes (MIP). Nucleic Acids Res 2005; 33:e183.

46. Ewart-Toland A, Briassouli P, de Koning JP, et al. Identification of Stk6/STK15 as a

candidate low-penetrance tumor-susceptibility gene in mouse and human. Nat Genet

2003; 34:403–412.

47. Balmain A, Gray J, Ponder B. The genetics and genomics of cancer. Nat Genet 2003;

33:238–244.

48. Carvalho B, Ouwerkerk E, Meijer GA, et al. High-resolution microarray comparative

genomic hybridisation analysis using spotted oligonucleotides. J Clin Pathol 2004;

57:644–646.

49. Nassiri M, Gugic D, Olczyk J, et al. Preservation of skin DNA for oligonucleotide

array CGH studies: a feasibility study. Arch Dermatol Res 2007; 299:353–357.

50. Hicks J, Muthuswamy L, Krasnitz A, et al. High-resolution ROMA CGH and FISH

analysis of aneuploid and diploid breast tumors. Cold Spring Harb Symp Quant Biol

2005; 70:51–63.

51. Guillaud-Bataille M, Valent A, Soularue P, et al. Detecting single DNA copy number

variations in complex genomes using one nanogram of starting DNA and BAC-array

CGH. Nucleic Acids Res 2004; 32:e112.

74 Wang and Gray


52. Lage JM, Leamon JH, Pejovic T, et al. Whole genome analysis of genetic alterations

in small DNA samples using hyperbranched strand displacement amplification and

array-CGH. Genome Res 2003; 13:294–307.

53. Le Caignec C, Spits C, Sermon K, et al. Single-cell chromosomal imbalances

detection by array CGH. Nucleic Acids Res 2006; 34:e68.

54. Fridlyand J, Snijders A, Pinkel D, et al. Hidden Markov models approach to the

analysis of array CGH data. J Multivar Anal 2004; 90:132–153.

55. Olshen AB, Venkatraman ES, Lucito R, et al. Circular binary segmentation for the

analysis of array-based DNA copy number data. Biostatistics 2004; 5:557–572.

56. Hupe P, Stransky N, Thiery JP, et al. Analysis of array CGH data: from signal ratio

to gain and loss of DNA regions. Bioinformatics 2004; 20:3413–3122.

57. Wang P, Kim Y, Pollack J, et al. A method for calling gains and losses in array CGH

data. Biostatistics 2005; 6:45–58.

58. Myers CL, Dunham MJ, Kung SY, et al. Accurate detection of aneuploidies in array

CGH and gene expression microarray data. Bioinformatics 2004; 20:3533–3543.

59. Daruwala RS, Rudra A, Ostrer H, et al. A versatile statistical analysis algorithm to detect

genome copy number variation. Proc Natl Acad Sci U.S.A. 2004; 101:16292–16297.

60. Hsu L, Self SG, Grove D, et al. Denoising array-based comparative genomic

hybridization data using wavelets. Biostatistics 2005; 6:211–226.

61. Lai WR, Johnson MD, Kucherlapati R, et al. Comparative analysis of algorithms for

identifying amplifications and deletions in array CGH data. Bioinformatics 2005;

21:3763–3770.

62. Willenbrock H, Fridlyand J. A comparison study: applying segmentation to array

CGH data for downstream analyses. Bioinformatics 2005; 21:4084–4091.

63. Gunn SR, Robetorye RS, Mohammed MS. Comparative genomic hybridization arrays

in clinical pathology: progress and challenges. Mol Diagn Ther 2007; 11:73–77.

64. Volik S, Raphael BJ, Huang G, et al. Decoding the fine-scale structure of a breast

cancer genome and transcriptome. Genome Res 2006; 16:394–404.

65. Volik S, Zhao S, Chin K, et al. End-sequence profiling: sequence-based analysis of

aberrant genomes. Proc Natl Acad Sci U.S.A. 2003; 100:7696–7701.

66. Albert TJ, Molla MN, Muzny DM, et al. Direct selection of human genomic loci by

microarray hybridization. Nat Methods 2007; 4:903–905.

67. Bennett ST, Barnes C, Cox A, et al. Toward the 1,000 dollars human genome.

Pharmacogenomics 2005; 6:373–382.

68. Ishkanian AS, Malloff CA, Watson SK, et al. A tiling resolution DNA microarray

with complete coverage of the human genome. Nat Genet 2004; 36:299–303.

69. Huang J, Wei W, Chen J, et al. CARAT: a novel method for allelic detection of DNA

copy number changes using high density oligonucleotide arrays. BMC Bioinformatics

2006; 7(83):83.

70. Nannya Y, Sanada M, Nakazaki K, et al. A robust algorithm for copy number

detection using high-density oligonucleotide single nucleotide polymorphism geno-

typing arrays. Cancer Res 2005; 65(14):6071–6079.

71. Komura D, Shen F, Ishikawa S, et al. Genome-wide detection of human copy number

variations using high-density DNA oligonucleotide arrays. Genome Res 2006; 16(12):

1575–1584.



6

Gene Expression Profiling as an Emerging

Diagnostic Tool to Personalize

Chemotherapy Selection for

Early Stage Breast Cancer

Cornelia Liedtke and Lajos Pusztai

Department of Breast Medical Oncology, The University of Texas,Anderson Cancer Center, Houston, Texas, U.S.A.

INTRODUCTION

The selection of adjuvant chemotherapy for breast cancer is currently more of an

art than a science. Several combination chemotherapy regimens can improve

disease-free and overall survival of stage I to stage III breast cancer when used as

adjuvant (postoperative) or neoadjuvant (preoperative) therapy (1,2). However,

no chemotherapy regimen is universally effective for all patients, and it is cur-

rently impossible to predict which of these regimens will actually work for a

particular individual. Choosing the best possible adjuvant or neoadjuvant che-

motherapy regimen for an individual is important because the right choice may

make the difference between cure and relapse. Results from traditionally

designed randomized clinical trials offer little guidance for individualized

treatment selection. These studies compare the activity of different treatments to

establish which is the most effective for the entire study population. This

approach ignores the fact that many patients are cured with earlier generation

adjuvant regimens and therefore are not well served by the administration of

77


longer, potentially more toxic, and more expensive second- or third-generation

combination therapies. These studies also cannot reveal whether some cancers

are particularly sensitive to a specific drug (or a combination of drugs) and

therefore can only be cured by that treatment. The most effective treatment for

an individual may or may not be the same regimen that is the most effective for

the whole population. These theoretical concerns and the increasing number of

adjuvant chemotherapy options, as well as the advent of new molecular ana-

lytical techniques, fuel a renewed interest in developing chemotherapy response

predictors that may lead to more personalized treatment recommendations.

THE ROLE OF NEOADJUVANT CHEMOTHERAPYIN THE TREATMENT OF BREAST CANCER

Administration of chemotherapy before surgical resection of the cancer (pre-

operative or neoadjuvant chemotherapy) provides a unique opportunity to

identify molecular markers of response. Preoperative chemotherapy is the cur-

rent standard of care for locally advanced breast cancer because this treatment

often renders previously inoperable cancers amenable to surgery due to frequent

and substantial tumor responses (3). Preoperative chemotherapy is also

increasingly used in the management of operable breast cancer because it can

reduce tumor size and allow for more limited surgery and possibly improved

cosmetic outcome (4). Importantly, there is no statistically significant difference

in survival between those who receive chemotherapy before surgery (i.e., neo-

adjuvant) and those who are treated with the same regimen postoperatively (i.e.,

adjuvant) (5–7). Therefore, preoperative chemotherapy is an appropriate option

for most patients who present with a breast cancer that requires systemic therapy.

Furthermore, eradication of all invasive cancer from the breast and regional

lymph nodes by the preoperative therapy [i.e., pathologic complete response,

(pCR)], is associated with excellent long-term, disease-free, and overall survival

(7,8). Therefore, one could propose that the regimen most likely to induce pCR is

the most effective treatment for a particular individual, and that such treatment

could be ideally suited to that patient either as adjuvant or neoadjuvant therapy.

This is the rationale why pCR is a commonly used clinical endpoint for breast

cancer response predictors. However, it is important to realize that most patients

with less than pCR had at least some clinical response to chemotherapy, and

therefore cannot be considered truly resistant to treatment, but rather, cases with

pCR represent extreme chemotherapy sensitivity.

CURRENT CHEMOTHERAPY RESPONSE PREDICTORS

There are several clinical and pathologic variables of breast cancer that are con-

sistently associated with a greater likelihood of response to neoadjuvant che-

motherapy. These include estrogen receptor (ER)-negative status, smaller tumor

size, increased proliferative activity, and high nuclear or histological grade (9–12).

78 Liedtke and Pusztai


Tumor size, grade, and ER-status are routinely available parameters that can be

combined into a multivariate response prediction model (13). One such model

is freely available at our institutional Web site, http://www.mdanderson.org/

Care_Centers/BreastCenter/dIndex.cfm?pn¼448442B2-3EA5-4BAC-

98310076A9553E63. However, clinical variable–based predictions are limited to

predict response to chemotherapy in general and lack regimen specificity.

A large number of molecular markers were proposed as potential predictors

of response to various chemotherapy agents (11,12,14–16). However, currently

there are no molecular markers of response that could be used in the clinic to select

patients for a particular therapy. Biomarker studies were traditionally conducted as

correlative science projects attached to therapeutic trials as optional components.

This has imposed serious limitations on these studies. Sample sizes for marker

discovery were limited by the availability of specimens that has lead to under-

powered observations and introduced potential bias due to subset analysis. Most

commonly, only associations between marker and response were reported, but no

formal prediction models were developed. The methodologies to detect a marker

were rarely standardized, and definitions for test-positive and test-negative cases

often differed from study to study, even for the same marker.

In addition to these technical limitations, there are important theoretical

considerations, which also suggest that single gene molecular markers expressed

by the cancer may have limited predictive value. Factors that reside outside the

neoplastic cells could influence drug response. The importance of interactions

between stroma and neoplastic cells, including neovascularization, changes in

interstitial fluid pressure, and immune/inflammatory response are increasingly

recognized as potential determinants of prognosis and possibly response to

therapy (17,18). Rates of drug metabolism are also variable and can have an

effect on drug activity. Cellular response to toxic insults is a complex molecular

process that involves simultaneous induction of several proapoptotic and sur-

vival pathways, including activation of repair and drug elimination mechanisms

(19,20). Tumor response may depend on the balance of these pathways, and

therefore any single gene may contain only limited information about the

complexity of the mechanisms that determine response to chemotherapy.

GENE EXPRESSION PROFILING AS A NOVEL TOOLFOR RESPONSE PREDICTION

Gene expression profiling with DNA microarrays represents a relatively recent

tissue analysis tool that was developed in the early 1990s (21). This technology

enables investigators to measure, in a semiquantitative manner in a single

experiment, the expression of almost all mRNAs expressed in a small piece of

cancer tissue. The rationale behind pharmacogenomica research is that by using

aPharmacogenomics is defined in this review as the use of mRNA expression profiles of the tumor to

predict drug efficacy. Pharmacogenetics is defined as the study of DNA variations in the germline to

predict toxicity or treatment response.

Gene Expression Profiling to Personalize Chemotherapy Selection 79


this technology, one can identify individually predictive genes and combine

these into a multivariate prediction model that will be more accurate than any

single gene alone. Multivariate outcome prediction models are not new to

medicine. Independent and individual weakly predictive clinical variables have

been combined into useful and validated prognostic or predictive models. There

are several established statistical methods for combining variables into predic-

tion models, including the commonly used logistic regression analysis and

the Cox proportional hazards regression model (22). Predictors could be

binary—that is, predicting response versus no response—or could predict the

probability of response on a continuous scale. The Adjuvant Online (http://www

.adjuvantonline.com) software represents a freely available validated multivariate

prognostic prediction tool for breast cancer.

The general strategy for multigene marker development in neoadjuvant

studies is to compare the transcriptional profiles of cancers that responded to

therapy with those that did not and to identify differentially expressed genes that

can be combined into a prediction model. There are several caveats that make

pharmacogenomic predictor discovery challenging.

1. The multiple comparison problem inherent to microarray analysis is well

known. It refers to the large number of variables (i.e., genes) that are

compared between two small data sets. As a consequence, observation of

many small p values may be due to chance. This could result in spuriously

identifying genes as associated with response when in fact they are not.

Several methods used to adjust for this phenomenon, at least to some extent,

by calculating false discovery rates for particular p valueswere described (23).

2. Another confounder relates to the coordinated expression of thousands of genes.

Individual transcripts do not represent independent variables, but rather their

expression is highly correlated with one another. Many phenotypic character-

istics of breast cancer are also associated with, and likely caused by, the coor-

dinated expression of thousands of genes. For example, ER-positive cancers

differ fromER-negative tumors in theexpressionof thousandsofgenes (24).The

gene expression pattern of high-grade tumors is also different from low-grade

tumors (25). These large-scale gene expression patterns associated with clinical

phenotypes have a profound influence on the pharmacogenomic discovery

process, because ER-negative, high-grade breast cancers are more sensitive

to many different types of chemotherapies compared with ER-positive and

low-grade tumors. Unadjusted pharmacogenomic comparison of responders

with nonresponders will yield gene lists that are dominated by ER- and grade-

associated genes, and the resulting pharmacogenomic predictormay represent

a predictor of clinical phenotype. When the discovery sample size is large, it

is possible to adjust for these phenotypic confounders; however, in small

studies, these adjustments may be difficult to do.

3. The probability that a supervised pharmacogenomic discovery approach, when

responders are compared with nonresponders, will lead to regimen-specific



predictors depends on to what extent the response groups are balanced for

strong phenotypic markers and on the extent of molecular differences that

determine drug-specific response. If drug sensitivity is influenced by the

two- to threefold higher or lower expression of a few dozen genes, these

differences may not be readily discovered through supervised pharmaco-

genomic analysis of small data sets. The modest gene expression differences

between responders and nonresponders can be masked by the larger scale

molecular differences due to any phenotypic imbalance between the

response groups and the technical noise of microarray experiments. For

example, when 24,000 measurements are performed (e.g., Affymetrix

U133A gene chip) and the overall concordance is 97.98% in a technical

replicate (i.e., same RNA profiled twice), 1.31% of all measurements can

have twofold or greater variation. This means that 314 genes decreased or

increased twofold from one experiment to another due to technical noise

alone (26). If the sample size is large, this technical noise can be separated

from the true signal, but when sample sizes are small it is more difficult to

separate noise from the true signal.

PHARMACOGENOMIC MARKER DEVELOPMENT PROCESS

Ideally, the development process of molecular markers should resemble the

development stages of therapeutic agents (Fig. 1) (27). In this context, a phase I

marker discovery study would demonstrate the technical feasibility of developing a

predictor and assess the reproducibility of marker results. The second phase

Figure 1 Schema of the developmental stages of molecular markers.



includes phase II validation studies of statistically adequate power to estimate

more precisely the predictive performance of the test in independent validation

cases. The next level of marker evaluation would include the demonstration of

clinical value by showing that a particular clinical endpoint improves when the

new test is used compared with the current standard, which could be no testing

or another type of testing. This may require a prospective phase III marker

validation study. Currently, no pharmacogenomic response predictors have

completed all three levels of clinical evaluation. Most published studies represent

phase I marker discovery and phase II marker validation studies.

FEASIBILITY OF MICROARRAY-BASED DIAGNOSTIC TESTS

Technical Reproducibility

The reproducibility of microarray measurements has been examined systemati-

cally by the Microarray Quality Control Project (MAQC), which included the US

Food and Drug Administration and 51 academic and industry collaborators. The

goal of this project was to assess the within- and between-laboratory technical

reproducibility of microarray measurements and compare the results across

different microarray platforms. Four different RNA samples were profiled in five

replicates on seven different array platforms, three different laboratories repre-

senting each platform. The same RNA samples were also profiled by three

different reverse-transcription polymerase chain reaction (RT-PCR) methods.

The most important finding of this collaborative effort was that the microarray

measurements were highly reproducible within and across the different com-

mercially available microarray platforms (28,29). The median coefficient of

variation (CV) for within-laboratory replicates ranged from 5% to 15% for the

various platforms and it was 10% to 20% for between-laboratory replicates. Most

importantly, the concordance between the microarray based mRNA measure-

ments and RT-PCR results were also high. The correlation coefficients for gene

expression ratios measured by arrays and by RT-PCR in two different samples

ranged from 0.79 to 0.92 for several hundreds of genes.

Between-platform variation in gene expression measurements is due to

sequence variations in the probe sets that target the same gene at different

locations. Factors that determine signal intensities generated by distinct probes

for the same gene include GC content of the sequence, sequence length, cross-

match opportunities, and the location of the probe sequence in relation to the

30end of the target gene. The concordance between signal intensity for probes

that target the same gene on different platforms is directly related to sequence

homology (30,31). Probes with complete sequence match yield highly concor-

dant results across platforms.

Few studies examined the within- and cross-platform reproducibility of

multigene prediction scores, as opposed to the reproducibility of single gene

measurements. Two reports indicated high reproducibility of pharmacogenomic



prediction results in replicate experiments from the same RNA, when the same

experimental procedures were used (26,32). Furthermore, when random noise

was added to corrupt the gene expression data, a remarkable degree of robustness

was observed in the face of noise.

However, caution must be exercised when trying to reproduce prediction

results across different platforms. There are multiple sources of variation that

will decrease reproducibility including differences in probe set and different

normalization methods. Essentially, all published results that attempted cross-

platform testing of gene signatures reported diminished, but not completely

lost, classification accuracy on data generated by other than the original

platform (33).

Preanalytical Sources of Variation

Another potentially important source of variation in microarray data is pre-

analytical variations, including differences in tissue sampling and RNA degra-

dation. There are several different techniques to obtain tissue from a patient for

pharmacogenomic studies, including excisional biopsy (with or without micro-

dissection), fine needle aspiration (FNA), and core needle biopsy (CNB). Both

needle biopsy techniques yield sufficient RNA for gene expression profiling with

DNA microarrays (1–3 mg) (34–36). However, different sampling methods may

result in different cellular composition and distinct gene expression profiles (37).

FNA specimens contain mostly neoplastic cells with few infiltrating leukocytes,

whereas the tumor cell content of CNB can vary from 10% to 90%. It is

important to consider the different amount of stromal cell that might be present

in a specimen, because the presence or absence of stromal genes might influence

prognostic or predictive gene signatures. Also, in studies comparing pre- and

postchemotherapy samples, an important source of variation may be the different

cellular composition of the specimens.

RNA is an easily degradable molecule and therefore was considered

suboptimal target for marker measurements compared with DNA or proteins.

The development of single step universal RNA stabilizing and preserving

reagents (RNA later, Ambion) has completely changed the field of RNA-based

diagnostics (38). Today, RNA can be stabilized within seconds after a fresh

tissue biopsy has been obtained and stored at room temperature for several days

(if needed) due to readily available commercial RNA stabilizing solutions.

Under these circumstances, RNA degradation from properly collected fresh

tissues is now rare. Interestingly, it is now more cumbersome to preserve pro-

teins for future analysis than it is to preserve RNA. A single step, universal

protein preserving solution is yet to be discovered. However, it is important to

realize that surgical resection methods (i.e., length of warm ischemia) and the

length of time between removing a piece of tissue from the body and placing it

into RNA preservative have a major effect on RNA quality (39).



STATISTICAL DESIGN OF PHARMACOGENOMICDISCOVERY AND VALIDATION STUDIES

Clinical biomarker discovery and validation is governed by similar statistical

rules as therapeutic clinical trials (40). Estimation of discovery sample size is

essential to develop a reliable pharmacogenomic predictor, and appropriately

powered independent validation studies are needed to convincingly demonstrate

its predictive values. There are several statistical methods that could be used to

estimate the number of patients needed to develop a clinical outcome predictor

(41). One simple approach is to power the study to discover individual genes that

show a predefined level of differential expression (often expressed in statistical

terms as standard effect size) between outcome groups. However, the number

and extent of differentially expressed genes are often unknown in advance, and

the multiple comparisons (i.e., thousands of different measurements on each

case) represent a major bias in this approach. Another approach takes into

account accumulating preliminary results to estimate the sample size needed for

discovery (42). With multivariate prediction algorithms, it is expected that the

predictive accuracy increases as the predictor is developed from and trained on

larger and larger data sets until it reaches a plateau (i.e., best possible predictor

within the constrains of the technology). The general principle is to fit a learning

curve into existing data by incrementally increasing the training set size and

extrapolate results to larger sample sizes. This is an intuitively appealing

approach, but it requires a substantial amount of preliminary results from many

dozens of cases. It also implies that discovery sample size will vary depending

on the difficulty of separating the two response groups on the basis of gene

expression results. When we applied this method to human breast cancer gene

expression data to estimate the number of cases needed for discovery of a

chemotherapy response predictor, the results indicated that 80 to 100 cases

would yield a predictor as good as extending the sample size to 200 (32).

To assess the accuracy of the predictor, it is necessary to test it on inde-

pendent patient cohorts. Often, there are not enough cases available for valida-

tion, and investigators report cross-validation results. In cross-validation, some

samples are left out from the marker development process and later on used to

test the accuracy of the predictor, and this process is repeated many times. This

leave-one-out cross-validation method gives a reliable estimate of the overall

accuracy of the marker development process, but it is not the validation of a

particular predictor. The genes that contribute to the predictors differ from

iteration to iteration. The final single best predictor that is developed from the

entire training data must be tested on completely separate cases to assess its

predictive value. The goal of an independent validation study is to (i) define the

true sensitivity, specificity, and the positive and negative predictive values (PPV

and NPV) with narrow confidence intervals (CI) and (ii) to prove clinical utility

of the test. Since the estimated performance characteristics of the test are known

from the cross-validation results, the design of validation studies is more



straightforward. Sample size for a single arm validation studies can be calculated

on the basis of the known historical response rate in unselected patients and the

estimated PPV of the genomic predictor. For example, if we assume that the

response rate in unselected patients is 30% and the true PPV of the test is 60%

(two-fold greater chance of response in “test-positive cases”), the sample

size could be calculated as follows: The lower bound of a two-sided 95% CI

should not be less than 0.3 (since the expected maximum pCR rate is 30% in

unselected cases). This implies that the standard error for the PPV is less than 0.1

and therefore the study must include at least 24 “test-positive” patients. The

proportion of patients with response is expected to be around 30%, and an

accurate prediction algorithm should generate a proportion that test positive also

about 0.20 to 0.40, this equates to a total sample size of 80 to 120 patients for

validation. If the predictor performs better and the true PPV is higher than 60%,

fewer patients may be sufficient. A randomized trial design could simultaneously

address response rates in selected and unselected patients and could also assess

treatment specificity of the predictor.

BREAST CANCER PHARMACOGENOMIC RESPONSE PREDICTORS

Supervised Pharmacogenomic Predictors Developed from Human Data

Results from several phase I genomic response marker discovery studies were

reported that compared gene expression profiles of cases with response and no-

response, and used the genes that were differentially expressed to construct

multigene predictors. Some of these markers were also assessed in small inde-

pendent validation cohorts. Table 1 summarizes the currently available literature

in this field. Chang et al. presented the first study that suggested the feasibility of

developing multigene predictors of response to neoadjuvant chemotherapy. They

developed a 92-gene predictor for clinical response to docetaxel using gene

expression data from 20 patients with locally advanced breast cancer (36).

Leave-one-out partial cross-validation indicated a predictive accuracy of 88%

(95% CI, 68–97%), sensitivity of 85% (95% CI, 55–98%), and specificity of 91%

(95% CI, 59–100%). Larger scale independent validation of this marker set has

not yet been reported. Another small study examined 24 breast cancer patients

treated with neoadjuvant paclitaxel, 5-fluoruracil, doxorubicin, and cyclo-

phosphamide (T/FAC) chemotherapy and developed a 74-gene predictor for

pCR (43). This multigene predictor model, tested on an independent cohort of

12 patients, showed a predictive accuracy of 78% (95%CI, 52–94%), PPV of 100%

(95% CI, 29–100%), NPV of 73% (95% CI, 45–92%), sensitivity of 43% (95%

CI, 10–82%), and specificity of 100% (95% CI, 72–100%). A larger follow-up

study included 82 cases for marker discovery and examined 780 distinct phar-

macogenomic predictors in cross-validation (32). Many of the distinct predictors

demonstrated statistically equal performance in complete cross-validation. This

(text continues on page 91)



Tab

le1

OverviewofPublished

StudiesonPharmacogenomicPredictorsofResponse

toNeoadjuvantChem

otherapyin

BreastCancer

Patientcharacteristicsand

treatm

ent

Method/hypothesis

Geneexpression

platform

Results

Comment

Changet

al.,

Lancet2003

(36)

24(¼

dev.)and6(¼

val.)pts.

withlocallyadvancedbreast

cancer

?4cycles

3-w

eekly

docetaxel

Development,

LOOCV

and

independent

validationofa

geneexpression

profile

predictivefor

clinical

response

(�25%

RD)

HgU95-A

v2

(12,625genes)

92differentially

expressed

genes

(p<

0.001)from

different

functional

classes;

Combinationinto

predictionmodel

LOOCV:88%

accu-

racy,92%

PPV,83%

NPV,85%

sensitiv-

ity,91%

specificity

Independentvalidation:

correctclassification

ofsixpts.withsen-

sitivetumors

Firststudydeveloping

amultigene

predictorfor

response

to

chem

otherapyin

breastcancer

patients

Ayerset

al.,JC

O

2004(43)

24(¼

dev.)and12(¼

val.)pts.

withprimarybreastcancer

?3cycles

3-w

eekly

or12

cycles

weekly

paclitaxel

followed

by4cycles

ofFAC

(T/FAC)

Developmentand

independent

validationof

geneexpression

predictorfor

pCR

cDNA

arrays

(30,721genes)

Developmentof

74-gene-multigene

predictor


78%

accuracy,100%

PPV,73%

NPV,

43%

sensitivity,

100%

specificity

Nosinglegenesuffi-

cientlyassociated

withpCRto

serveas

singlevalid

marker

Proofofprinciple

forbuildingmulti-

genepredictorsfor

chem

otherapy

response



Tab

le1

OverviewofPublished


toNeoadjuvantChem

otherapyinBreastCancer(Continued

)


treatm

ent

Method/hypothesis

Geneexpression

platform

Results

Comment

Iwao-K

oizumi,

JCO

2005(44)

44(¼

dev.)and26(¼

val.)pts.

withstageII/IV

breastcancer

?4cycles

ofdocetaxel

Developmentand

independent

validationof

geneexpression

predictorfor

clinical

response

(WHO

criteria)

ATAC-PCR(2,453

selected

genes)

Developmentof

85-genesignature


81%

accuracy,73%

PPV,91%

NPV,

92%

sensitivity,71%

specificity

Genes

controllingthe

cellularredox

environmentasso-

ciated

withdoce-

taxel

resistance

?supported

by

invitro

results

Hannem

annet

al.,

JCO

2005(45)

48pts.withLABC

?cycles

ofdoxorubicin/

docetaxel

(AD,n¼

24)

ordoxorubicin/cyclo-

phospham

ide(A

C,n¼

24)

Identificationof

geneexpression

predictorfor

pCRornPCR

forresponse

to

AD

and/orAC

Proprietaryarray

(18,432genes)

Failure

toidentify

dif-

ferentially

expressed

genes

ordefinemul-

tigenepredictor

Hierarchical

clustering:

pCR/npCR

samples

did

notcluster

dis-

tinctly

compared

toRD

Comparisonofgene

expressionbetween

pre-and

postchem

otherapy

samples:

more

transcriptional

changes

among

sensitivecompared

toresistanttumors

Gianniet

al.,JC

O

2005(46)

89(¼

dev.)and82(¼

val.)pts.

withLABC

?3cycles

of3-w

eekly

doxorubicin/paclitaxel

followed

by12cycles

weekly

paclitaxel

?3cycles

3-w

eekly

or

12cycles

weekly

paclitaxel

followed

by4cycles

of

FAC

(T/FAC)

Identificationof

geneexpression

markersthat

predictresponse

Correlation

betweenRSand

response

RT-PCR

(384

genes)

AffymetrixHG-

U133Plus2.0

86genes

correlating

withpCR(p

<0.05)

form

ingthreeclus-

ters

(ER,prolifera-

tionandim

mune-

relatedgenecluster)

RScorrelated

withpCR

86-genemodel

devel-

oped

andsuccess-

fullyvalidated

(Continued

)



Tab

le1

OverviewofPublished


toNeoadjuvantChem


)


treatm

ent

Method/hypothesis

Geneexpression

platform

Results

Comment

Folgueira

etal.,

ClinCancerRes

2005(47)

38(¼

dev.)and13(¼

val.)

31(¼

dev.)and13(¼

val.)

?4cycles

3-w

eekly

doxorubicin/cyclo-

phospham

ide(A

C)

Developmentand

independent

validationof

geneexpression

predictorfor

pCR

692cD

NA

microarray

4,608gene

platform

25differentially

expressed

transcripts

betweenresponders

andnonresponders,

3geneclassifier

could

notbevalidated;

successfuldevelopment

andvalidation

(LOOCV

andinde-

pendentdataset)of

distinct

three-gene

classifier

Dressman

etal.,

ClinCancerRes

2006(48)

37pts.withstageIIB/III

breastcancer

?4cycles

ofliposomal

doxorubicin/paclitaxel

þlocalwhole

breast

hyperthermia

Definitionofgene

signature

for

(a)inflam

matory

breastcancer

(IBC),(b)tumor

hypoxia,and(c)

clinical

response

andvalidationin

twoindependent

datasets

Affymetrix

HG-U

133

Plus2.0

Identificationandvali-

dation(LOOCV)of

(a)22-genesignature

forIBC

(b)18-genesignature

fortumorhypoxi

Failure

toidentify

gene

signature

predicting

clinical

response

Developmentand

successfulvalidation

(2independent

datasets)ofa

signature

predictive

forlymphnode

persistence



Tab

le1

OverviewofPublished


toNeoadjuvantChem


)


treatm

ent

Method/hypothesis

Geneexpression

platform

Results

Comment

Thuerigen

etal.,

JCO

2006(49)

52(¼

dev/GEsD

oc)

and48

(¼val./GEDoc)

pts.with

breastcancer

?5cycles

biweekly

gem

ci-

tabine/epirubicin

followed

by

4cycles

biweekly

docetaxel

(GEsD

oc)

?6cycles

of3-w

eekly

gem

citabine/epirubicin/

docetaxel

(GEDoc)

Developmentand

independent

validationof

geneexpression

predictorfor

pCR

Oligonucleotide

array(21329

genes)

512-gene-signature

predictiveofpCR

with:independent

validation:88%

accuracy,64%

PPV,

95%

NPV,78%

sensitivity,90%

specificity

HER2/neu

expression

independently

predictiveof

response

Parket

al.,Breast

CancerRes

Treat

2006(50)

21pts.withstageII/IIIbreast

cancer

?4cycles

of3-w

eekly

FEC

followed

by12cycles

ofweekly

paclitaxel

Developmentand

LOOCV

of

ABC

transporter

geneexpression

profiles

predictivefor

pCR

AffymetrixHG-

U133Plus2.0

(49ABC

trans-

porters)

11differentially

expressed

ABCtransportersmulti-

genepredictormodelof

differentially

expressed

ABCtransporterspre-

dictiveofpCR

LOOCV:93%

accu-

racy,93%

PPV,94%

NPV,88%

sensitiv-

ity,96%

specificity

Cleatoret

al.,

BreastCancer

Res

Treat

2006

(51)

40pts.withprimarybreast

cancer

?6cycles

doxorubicin/

cyclophospham

ide(A

C)

Developmentand

LOOCV

of

predictorfor

clinical

response

Affymetrix

HG-U

133A

253differentially

expressed

genes;

75and178genes

over-

expressed

inresistant

andsensitivetumors,

respectively

Developmentand

LOOCVofgenepre-

dictorforresponse

Previouslyidentified

92-genedocetaxel

classifier

(36)pre-

dictedonly

39–61%

tumors

correctly

?geneexpression

pattern

fordocetaxel

isnotpredictivefor

ACresponse

(Continued

)



Tab

le1

OverviewofPublished


toNeoadjuvantChem


)


treatm

ent

Method/hypothesis

Geneexpression

platform

Results

Comment

Hesset

al.,JC

O

2006(32)

82(¼

dev.)and51(¼

val.)pts.

withprimarybreastcancer

?3cycles

3-w

eekly

or

12cycles

weekly

paclitaxel

followed

by4cycles

ofFAC

(T/FAC)

Developmentand

independent

validationof

geneexpression

predictorfor

pCR

Affymetrix

HG-U

133A

Identificationof30-

genepredictor

Validationfivefold

cross-validationand

permutationtesting


76%

accuracy,52%

PPV,96%

NPV,

92%

sensitivity,71%

specificity

Follow-upstudy

of(43)

Predictorpreserved

its

predictiveaccuracy

when

applied

to

replicate

samples

Largeststudyto

date

Minaet

al.,Breast

CancerRes

Treat

2006(52)

45pts.withstageII/III

breast

cancer(including6pts.with

inflam

matory

BC)

?3cycles

biweekly

doxor-

ubicin

and6cycles

weekly

docetaxel

Identificationof

genes

correlat-

ingwithpCR

CorrelationofpCR

withRSorER-

relatedgenes

Identificationof

genes

correlating

withinflam

ma-

tory

phenotype

RT-PCR

assay

(274genes)

22of274candidategenes

correlated

withpCR(p

<0.05)form

ingthreelarge

clusters(angiogenesis,

proliferation,andinva-

sion-related

genes)

Lackofcorrelation

betweenpCR

and

RS/ER-related

genes

24/274genes

correlated

withinflam

matory

BC

Abbreviations:ABCtransporter,ATP-bindingcassette

transporter;dev.,development/trainingset;IBC,inflam

matory

breastcancer;LABC,locallyadvancedbreast

cancer;LOOCV,leave-one-outcross-validation;npCR,nearpCR;NPV,negativepredictivevalue;

pCR,pathological

complete

response;PPV,positivepredictive

value;

pts.,patients;RD,residual

disease;RS,recurrence

score

(OncotypeDX#)(53);RT-PCR,reverse

transcriptionpolymerasechainreaction;val.,validationset.



observation is confirmed by similar reports in the genomic literature, which

indicates that several different but equally good predictors can be developed

from the same data. This is due to the coordinated expression of many thousands

of genes. Any combination from a particular group of coexpressed genes can

carry similar information. However, a nominally best 30-probe set genomic

predictor of pCR (DLDA-30) was selected for independent validation on 51

cases, and its performance was compared with a multivariate clinical variable–

based predictor including age, estrogen receptor status, and grade. The genomic

predictor showed a sensitivity of 92% that was significantly higher than that of

the clinical predictor (61%). The NPV (96% vs. 86%) and area under the

receiving operator curve (0.877 vs. 0.811) were also slightly better. Several other

smaller studies demonstrated similar proof of concept that this supervised

approach to marker development is feasible (Table 1).

Genomic Predictors from Cell Line Models

An interesting alternative approach is to use cancer models to develop chemo-

therapy response predictors and test the performance of these in human data. The

gene expression profiles of breast cancer cell lines are different from those of

human breast cancers, but nevertheless, they recapitulate many biological and

genomic properties of primary breast cancers (54). Several groups reported drug-

specific transcriptional changes in response to chemotherapy exposure in vitro

that might be exploited as predictors of therapy response in patients (55).

However, the predictive value of these in vitro generated signatures in patients

has not yet been studied extensively. Nevertheless, this is an important research

direction that if working can have important implications for individualizing

treatment regimens for patients. In fact one could envisage custom-assembled

chemotherapy regimens based on sensitivity profiles for individual drugs. It is

rapidly becoming feasible to test such concepts in silico. Several groups have

made their complete gene expression data from neoadjuvant biopsies publicly

available. Cell line-based predictive signatures can be tested on these already

existing and clinically annotated human data sets (56).

Gene Expression Changes in Serial Biopsies

Similar to cell lines, transcriptional response to chemotherapy was also observed

in serial biopsies before and during chemotherapy. The rationale is that these

changes may be more predictive of response than baseline gene expression

profiles. However, the few small studies that examined this approach reported

very heterogeneous transcriptional response to chemotherapy (57). In essence,

different genes change in different individuals. Sotiriou et al. examined gene

expression changes after one complete cycle of neoadjuvant chemotherapy and

demonstrated that tissue samples, obtained from patients with good response,

exhibited much greater changes in gene expression than those from poor



responders (34). These observations were corroborated by results obtained by

Hannemann et al., who compared gene expression profiling on pretreatment

biopsies (n ¼ 46) and posttreatment tumors (n ¼ 17) obtained at time of surgery.

They also demonstrated that neoadjuvant chemotherapy causes major changes in

gene expression in those cases who responded to treatment compared with those

who did not (58,59). Several technical and theoretical concerns limit our

enthusiasm for this approach to predictor development. It is difficult to obtain

serial biopsies at regular intervals from patients who undergo chemotherapy.

There is substantial variation in the actual timing of the biopsy compared with

the planned time of obtaining the follow-up biopsy. For example, biopsies

planned at 24 hours after chemotherapy are in fact usually collected during an

18- to 36-hour window. Since the kinetics of important gene expression changes

are unknown, this introduces unpredictable variation into already highly variable

data. Also, there are altogether missing time points, which further reduce the

already small sample size for discovery. These factors increase the risk for false

discovery from serially collected small pharmacogenomic data sets. More

importantly, the practical value of a response predictor that is applied to a biopsy

after chemotherapy was given is not particularly appealing.

FUTURE POTENTIAL OF PHARMACOGENOMICS

Pharmacogenomic predictive marker discovery is a new science. It is relatively

less mature than the field of prognostic gene signatures because of the lack of

preexisting tissue banks from patients who received preoperative chemotherapy.

All human biopsies had to be collected in prospective clinical trials in the past

few years, unlike the readily available frozen and archived tumor banks that were

suitable for prognostic signature discovery and validation. However, several

such data sets now exist and the field has likely entered an accelerated phase of

development.

We have learned several important lessons. (i) It is clear that the gene

expression profiles of breast cancers with extreme chemotherapy sensitivity are

different from those with lesser sensitivity. Unfortunately, these gene expression

differences correspond to major clinical pathological phenotypes such as ER-

negative status and high grade that somewhat limits the clinical value of the

current predictors. However, as larger data sets become available, it will be

possible to attempt to develop predictors after stratification by known clinical

pathological variables. (ii) It is also clear from many different reports in the

genomic literature that there is no unique, single best signature for any particular

classification problem. There are several distinct predictors that can perform

equally well (60). (iii) The availability of large, public, clinically annotated,

comprehensive gene expression databases started to revolutionize the biomarker

discovery process. It is now possible to test in silico any mRNA-based marker

for prognostic or chemotherapy or endocrine therapy predictive values, respec-

tively, on data sets that were prospectively assembled for this purpose (61). As



more neoadjuvant data sets become available from patient cohorts treated with

different treatment regimens, it will be possible to examine the regimen spe-

cificity of the various predictors. (iv) The most important long-term benefit from

pharmacogenomic research may not be the current predictive signatures or the

ones that will be developed in the near future, which still represent relatively

crude attempts to match patients with existing drugs, but rather the large

molecular data sets that this research generates. These databases could allow the

identification of novel drug targets in human data and the development of drugs

that target subsets of patients with particular genomic abnormalities.

REFERENCES

1. National Comprehensive Cancer Network. Available at: http://www.nccn.org. Accessed

June 2007.

2. Goldhirsch A, Wood WC, Gelber RD, et al. Meeting highlights: updated interna-

tional expert consensus on the primary therapy of early breast cancer. J Clin Oncol

2003; 21(17):3357–3365.

3. Scholl SM, Fourquet A, Asselain B, et al. Neoadjuvant versus adjuvant chemo-

therapy in premenopausal patients with tumours considered too large for breast

conserving surgery: preliminary results of a randomised trial: S6. Eur J Cancer 1994;

30A(5):645–652.

4. Boughey JC, Peintinger F, Meric-Bernstam F, et al. Impact of preoperative versus

postoperative chemotherapy on the extent and number of surgical procedures in

patients treated in randomized clinical trials for breast cancer. Ann Surg 2006; 244(3):

464–470.

5. Makris A, Powles TJ, Ashley SE, et al. A reduction in the requirements for mas-

tectomy in a randomized trial of neoadjuvant chemoendocrine therapy in primary

breast cancer. Ann Oncol 1998; 9(11):1179–1184.

6. Mauriac L, Durand M, Avril A, et al. Effects of primary chemotherapy in conser-

vative treatment of breast cancer patients with operable tumors larger than 3 cm.

Results of a randomized trial in a single centre. Ann Oncol 1991; 2(5):347–354.


outcome of women with operable breast cancer. J Clin Oncol 1998; 16(8):2672–2685.

8. Kuerer HM, Newman LA, Smith TL, et al. Clinical course of breast cancer patients

with complete pathologic primary tumor and axillary lymph node response to

doxorubicin-based neoadjuvant chemotherapy. J Clin Oncol 1999; 17(2):460–469.

9. Abu-Farsakh H, Sneige N, Atkinson EN, et al. Pathologic predictors of tumor

response to preoperative chemotherapy in locally advanced breast carcinoma. Breast

J 1995; 1(2):96–101.

10. Colleoni M, Minchella I, Mazzarol G, et al. Response to primary chemotherapy in

breast cancer patients with tumors not expressing estrogen and progesterone recep-

tors. Ann Oncol 2000; 11(8):1057–1059.

11. Chang J, Powles TJ, Allred DC, et al. Biologic markers as predictors of clinical

outcome from systemic therapy for primary operable breast cancer. J Clin Oncol

1999; 17(10):3058–3063.

12. Ellis P, Smith I, Ashley S, et al. Clinical prognostic and predictive factors for

primary chemotherapy in operable breast cancer. J Clin Oncol 1998; 16(1):107–114.



13. Rouzier R, Pusztai L, Delaloge S, et al. Nomograms to predict pathologic complete

response and metastasis-free survival after preoperative chemotherapy for breast

cancer. J Clin Oncol 2005; 23(33):8331–8339.

14. Rozan S, Vincent-Salomon A, Zafrani B, et al. No significant predictive value of

c-erbB-2 or p53 expression regarding sensitivity to primary chemotherapy or

radiotherapy in breast cancer. Int J Cancer 1998; 79(1):27–33.

15. Petit T, Borel C, Ghnassia JP, et al. Chemotherapy response of breast cancer depends

on HER-2 status and anthracycline dose intensity in the neoadjuvant setting. Clin

Cancer Res 2001; 7(6):1577–1581.

16. Zhang F, Yang Y, Smith T, et al. Correlation between HER2 expression and response

to neoadjuvant FAC chemotherapy in breast cancer. Cancer 2003; 97(7):1758–1765.

17. Silberstein GB. Tumour-stromal interactions. Role of the stroma in mammary

development. Breast Cancer Res 2001; 3(4):218–223.

18. Coussens LM, Werb Z. Inflammation and cancer. Nature 2002; 420(6917):860–867.

19. Burns TF, El-Deiry WS. Microarray analysis of p53 target gene expression patterns

in the spleen and thymus in response to ionizing radiation. Cancer Biol Ther 2003; 2(4):

431–443.

20. Reed JC. Dysregulation of apoptosis in cancer. J Clin Oncol 1999; 17(9):2941–2953.

21. Fodor SP, Rava RP, Huang XC, et al. Multiplexed biochemical assays with bio-

logical chips. Nature 1993; 364(6437):555–556.

22. Kattan MW, Scardino PT. Prediction of progression: nomograms of clinical utility.

Clin Prostate Cancer 2002; 1(2):90–96.

23. Pounds S, Morris SW. Estimating the occurrence of false positive and false negatives

in microarray studies by approximating and partitioning the empirical distribution of

p-values. Bioinformatics 2003; 19(10):1236–1242.

24. Pusztai L, Ayers M, Stec J, et al. Gene expression profiles obtained from fine-needle

aspirations of breast cancer reliably identify routine prognostic markers and reveal

large-scale molecular differences between estrogen-negative and estrogen-positive

tumors. Clin Cancer Res 2003; 9(7):2406–2415.

25. Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast cancer:

understanding the molecular basis of histologic grade to improve prognosis. J Natl

Cancer Inst 2006; 98(4):262–272.

26. Anderson K, Hess KR, Kapoor M, et al. Reproducibility of gene expression signature

based predictions in replicate experiments. Clin Cancer Res 2006; 12(6):1721–1727.

27. Pusztai L. Perspectives and challenges of clinical pharmacogenomics in cancer.

Pharmacogenomics 2004; 5(5):451–454.

28. MAQC Consortium, . The MicroArray Quality Control (MAQC) project shows inter-

and intraplatform reproducibility of gene expression measurements. Nat Biotechnol

2006; 24(9):1151–1161.

29. Klebanov L, Yakovlev A. How high is the level of technical noise in microarray

data? Biol Direct 2007; 2(1):9.

30. Mecham BH, Klus GT, Strovel J, et al. Sequence-matched probes produce increased

cross-platform consistency and more reproducible biological results in microarray-

based gene expression measurements. Nucleic Acids Res 2004; 32(9):e74.

31. Ji Y, Coombes K, Zhang J, et al. RefSeq refinements of UniGene-based gene

matching improve the correlation of expression measurements between two micro-

array platforms. Appl Bioinformatics 2006; 5(2):89–98.



32. Hess KR, Anderson K, Symmans WF, et al. Pharmacogenomic predictor of sensi-

tivity to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin,

and cyclophosphamide in breast cancer. J Clin Oncol 2006; 24(26):4236–4244.

33. Stec J, Wang J, Coombes K, et al. Comparison of the predictive accuracy of DNA

array-based multigene classifiers across cDNA arrays and Affymetrix GeneChips.

J Mol Diagn 2005; 7(3):357–367.

34. Sotiriou C, Powles TJ, Dowsett M, et al. Gene expression profiles derived from fine

needle aspiration correlate with response to systemic chemotherapy in breast cancer.

Breast Cancer Res 2002; 4(3):R3.

35. Assersohn L, Gangi L, Zhao Y, et al. The feasibility of using fine needle aspiration

from primary breast cancers for cDNA microarray analyses. Clin Cancer Res 2002;

8(3):794–801.

36. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the pre-

diction of therapeutic response to docetaxel in patients with breast cancer. Lancet

2003; 362(9381):362–369.

37. Symmans WF, Ayers M, Clark EA, et al. Total RNA yield and microarray gene

expression profiles from fine-needle aspiration biopsy and core-needle biopsy

samples of breast carcinoma. Cancer 2003; 97(12):2960–2671.

38. Mutter GL, Zahrieh D, Liu C, et al. Comparison of frozen and RNALater solid tissue

storage methods for use in RNA expressionmicroarrays. BMCGenomics 2004; 5(1):88.

39. Dash A, Maine IP, Varambally S, et al. Changes in differential gene expression

because of warm ischemia time of radical prostatectomy specimens. Am J Pathol

2002; 161(5):1743–1748.

40. Simon R. Roadmap for developing and validating therapeutically relevant genomic

classifiers. J Clin Oncol 2005; 23(29):7332–7341.

41. Pusztai L, Hess KR. Clinical trial design for microarray predictive marker discovery

and assessment. Ann Oncol 2004; 15(12):1731–1737.

42. Mukherjee S, Tamayo P, Rogers S, et al. Estimating dataset size requirements for

classifying DNA microarray data. J Comput Biol 2003; 10(2):119–142.



cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22(12):

2284–2293.

44. Iwao-Koizumi K, Matoba R, Ueno N, et al. Prediction of docetaxel response in human

breast cancer by gene expression profiling. J Clin Oncol 2005; 23(3):422–431.

45. Hannemann J, Oosterkamp HM, Bosch CA, et al. Changes in gene expression

associated with response to neoadjuvant chemotherapy in breast cancer. J Clin Oncol

2005; 23(15):3331–3342.

46. Gianni L, Zambetti M, Clark K, et al. Gene expression profiles in paraffin-embedded

core biopsy tissue predict response to chemotherapy in women with locally advanced


47. Folgueira MA, Carraro DM, Brentani H, et al. Gene expression profile associated

with response to doxorubicin-based therapy in breast cancer. Clin Cancer Res 2005;

11(20):7434–7443.

48. Dressman HK, Hans C, Bild A, et al. Gene expression profiles of multiple breast

cancer phenotypes and response to neoadjuvant chemotherapy. Clin Cancer Res

2006; 12(3 pt 1):819–826.



49. Thuerigen O, Schneeweiss A, Toedt G, et al. Gene expression signature predicting

pathologic complete response with gemcitabine, epirubicin, and docetaxel in primary


50. Park S, Shimizu C, Shimoyama T, et al. Gene expression profiling of ATP-binding

cassette (ABC) transporters as a predictor of the pathologic response to neoadjuvant

chemotherapy in breast cancer patients. Breast Cancer Res Treat 2006; 99(1):9–17.

51. Cleator S, Tsimelzon A, Ashworth A, et al. Gene expression patterns for doxorubicin

(Adriamycin) and cyclophosphamide (cytoxan) (AC) response and resistance. Breast

Cancer Res Treat 2006; 95(3):229–233.

52. Mina L, Soule SE, Badve S, et al. Predicting response to primary chemotherapy:

gene expression profiling of paraffin-embedded core biopsy tissue. Breast Cancer

Res Treat 2007; 103(2):197–208.

53. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifen-

treated, node-negative breast cancer. N Engl J Med 2004; 351(27):2817–2826.

54. Neve RM, Chin K, Fridlyand J, et al. A collection of breast cancer cell lines for the

study of functionally distinct cancer subtypes. Cancer Cell 2006; 10(6):515–527.

55. Troester MA, Hoadley KA, Sorlie T, et al. Cell-type-specific responses to chemo-

therapeutics in breast cancer. Cancer Res 2004; 64(12):4218–4226.


motherapeutics. Nat Med 2006; 12(11):1294–1300.

57. Buchholz TA, Stivers DN, Stec J, et al. Global gene expression changes during

neoadjuvant chemotherapy for human breast cancer. Cancer J 2002; 8(6):461–468.

58. Hannemann J, Oosterkamp HM, Bosch CAJ, et al. Changes in gene expression

profiling due to primary chemotherapy in patients with locally advanced breast

cancer. Proc Am Soc Clin Oncol 2004; 22(14S):502 (abstract).

59. Modlich O, Prisack HB, Munnes M, et al. Immediate gene expression changes after

the first course of neoadjuvant chemotherapy in patients with primary breast cancer

disease. Clin Cancer Res 2004; 10(19):6418–6431.

60. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based pre-

dictors for breast cancer. N Engl J Med 2006; 355(6):560–569.

61. Andre F, Hatzis C, Anderson K, et al. Microtubule-associated protein-tau is a

bifunctional predictor of endocrine sensitivity and chemotherapy resistance in

estrogen receptor-positive breast cancer. Clin Cancer Res 2007; 13(7):2061–2067.


[gajendra][6�9-Standard][D:/informa_Publishing/H5293_Leyland_Jones_112035/z_production/z_3B2_3D_files/978-1-4200-5293-0_CH0007_O.3d] [27/3/08/13:22:5] [97–106]

7

Gene Expression–Based Predictors

of Prognosis and Response to

Chemotherapy in Breast Cancer

Soonmyung Paik

Division of Pathology, NSABP Foundation, Pittsburgh, Pennsylvania, U.S.A.

INTRODUCTION

With the explosion of genomic technology, numerous studies have demonstrated

the potential clinical utility of gene expression profiles as a predictor of prog-

nosis and response to therapy. Of particular interest is the demonstration that

there is a general link between poor prognostic features and response to che-

motherapy. In addition, linearity of RNA-based assays shows a strong promise

that they can be very useful as predictive assays for targeted therapies.

Since many excellent reviews have been written on this subject, this

chapter will concentrate mainly on conceptual background to aid the clinical

readers understand what is going on in the field rather than going into details.

GENE EXPRESSION PROFILING PROVIDESPROGNOSTIC INFORMATION

Gene expression profiling using microarray platforms allows interrogation of

expression levels of essentially all known genes in the tumor cells (1). In most

studies, gross tumor rich area is homogenized to extract RNA. Therefore, the

profile that is achieved is mean expression levels of each gene among different

97


cell population in the tumor, including host stromal cells or inflammatory infil-

trates. It is also important to realize that current methods do not have sensitivity to

allow absolute measurement of gene expression levels of all genes printed on the

microarray. Instead only about 50% of genes will have enough signal above

background level (so-called present call). Different microarray platforms use DNA

probes of various lengths (ranging from 22-mer oligonucleotide in Affymetrix

GeneChip, 60-mer oligonucleotides for Agilent arrays, and average 1 kilobase for

cDNA arrrays) representing different positions in each gene between the arrays.

Sensitivity and dynamic range for measuring a specific gene could be widely

different among different array platforms. Thus different studies using different

platforms may result in identifying different set of genes representing same clin-

ical phenotype (2). Obviously most studies also differ in the study population.

Until recently microarray gene expression profiling required fresh or snap frozen

tissue, seriously limiting the scale of studies conducted so far.

It is not entirely surprising to find gene expression profiling to provide

prognostic information, since after all morphology-based prognostic factors such

as tumor histological grade is essentially end result of gene expression changes in

the tumor cells. Two seminal papers using microarray gene expression profiling,

one by Sorlie et al. (3) and the other by van ’t Veer et al. (4), clearly demonstrated

the association of gene expression profiles and prognosis of patients diagnosed

with early breast cancer. However, these two studies used very different

approaches. In the seminal paper published in 2000, Perou et al. used unsupervised

analysis of gene expression data and identified distinct biological subtypes

(normal-like, luminal A and B, basal-like, and HER2) (5). Subsequently Sorlie

et al. demonstrated that these subtypes confer differing prognosis with basal-like

and HER2 associated with poor clinical outcome (3). On the other hand, van ’t

Veer et al. conducted a supervised analysis to develop prognostic index based on

expression levels of 70 genes identified from microarray analysis (4). This was

then validated in a larger cohort (6). Recently Dr. Perou’s group has demonstrated

that there is a significant overlap of information between the two approaches

using same dataset—meaning those classified as good prognosis using 70-gene

van ’t Veer set was found to be largely populated by luminal A subtype (2).

While studies described above had enormous impact on how we view

breast cancer, despite recent clearance of 70-gene Mammaprint1 assay by the

U.S. Food and Drug Administration (FDA), actual clinical implementation has

been impeded by the need for fresh or snap frozen tumor tissue because of the

requirement for highest quality RNA as a starting material for the microarray

gene expression profiling.

DEVELOPMENT AND VALIDATION OF OncotypeDxTM

RECURRENCE SCORE ASSAY

RNA isolated from formalin-fixed, paraffin-embedded tissue (FFPET) samples

is a poor material for gene expression profiling. Masuda et al. analyzed the RNA

extracted from FFPET for chemical modifications (2,7). In freshly made FFPET

98 Paik


that were fixed and processed in ideal conditions (fixed in 10% buffered formalin

at 48C), RNAs are fairly well preserved. Although the extracted RNAs did not

show any sign of degradation compared with fresh samples, they were poor

substrates for cDNA synthesis and subsequent polymerase chain reaction (PCR)

amplification, so that only PCR amplification of short targets was possible. The

investigators found addition of monomethylol (CH2OH) groups to all four bases

to varying degrees and some adenines dimerized through metheylene bridging.

In addition to the chemical modification, however, RNAs in FFPET continue to

be degraded or fragmented over time during storage for reasons unclear. Cronin

et al. systematically examined the quality of RNA extracted from breast cancer

FFPET specimens taken at different times. RNA from FFPET archived for

approximately 1 year were less fragmented than RNA archived for nearly 6 or

17 years (8).

There have been reports on the use of reverse transcription polymerase

chain reaction (RT-PCR) to measure expression levels of few genes using RNA

isolated from FFPET, but large-scale RT-PCR was not reported until recently.

Cronin et al. from Genomic Health Inc. has invested on developing high quality

controlled robotic assays for RT-PCR and published proof of concept study

(8). In the latter study, while RNAs isolated from old formalin fixed tissue were

heavily fragmented with average size under 200 base pairs and absolute signal

was much less for older samples, normalization of signal using a set of reference

genes could largely correct these problems.

Encouraged by these results, the National Surgical Adjuvant Breast and

Bowel Project (NSABP) has engaged in a collaborative study with Genomic

Health Inc. to develop prognostic gene expression assay for node negative (N�)

estrogen receptor–positive (ERþ) breast cancer treated with tamoxifen (9).

While there are many other outstanding clinical questions that need to be

addressed, we were compelled to first address the question of who among N�ERþ tamoxifen treated patients need chemotherapy. In addition, there were two

independent tamoxifen treated cohorts (tamoxifen arms of B-14 and B-20 trials)

in the NSABP tissue bank so that discovery and validation could be performed

without difficulty. We wanted to make sure that by the time of reporting, the

assay was validated in a completely independent cohort. At that time, technology

and cost limited Genomic Health investigators to examine only 250 genes

maximum in RNA extracted from 60 mm thick section from a paraffin block.

Therefore, existing cancer literature and microarray data were surveyed and

250 candidate genes were selected. Before tumor blocks from B-20 tamoxifen

treated cohort were examined, two other cohorts, one from mixed patient

population and the other from those with more than 10 positive nodes were

examined with the same assay (10). We did not believe that there will be dif-

ferent prognostic genes based on nodal status and therefore selected the prog-

nostic genes that were equally prognostic across all three cohorts as a basis for

prognostic model building. From the beginning, because of the continuous nature

of RNA expression data, we believed strongly that a linear prediction model

needed to be developed.

Predictors of Prognosis and Response to Chemotherapy 99


Correlating clinical outcome with expression levels yielded many genes

from these three studies, and 16 top-performing genes were identified for final

model building and validation. Relative expression levels of the 16 genes are

measured in relationship to average expression levels of 5 reference genes.

While a majority of genes comprising 16 genes are ER (ER, PGR, BCL2,

SCUBE2) and proliferation (Ki67, STK15, Survivin, CCNB1, MYBL2) related,

there are other genes (HER2, GRB7, MMP11, CTSL2, GSTM1, CD68, BACG1)

(9). The unscaled recurrence score (RSu) was calculated with the use of coef-

ficient that are defined on the basis of regression analysis of gene expression.

Recurrence score (RS) was rescaled from the unscaled recurrence score (Rsu) as

follows: RS ¼ 0 if Rsu < 0; RS ¼ 20 � (Rsu–6.7) if 0 � Rsu � 100; and RS ¼100 if Rsu > 100 (9). This resulted in RS ranging from 0 to 100.

Final validation of the RS was achieved by examination of its performance

in a completely independent cohort from NSABP trial B-14, which was not used

in the model building process (9). The validation study was conducted with

rigorous predefined statistical analysis plan with prespecified outcome end

points and cutoffs for RS. The predefined low-risk group (RS below 18) had

significant better prognosis compared with higher risk groups. More importantly,

as shown in the published report, there is a linear relationship between

RS and risk of distant failure at 10 years among N� ERþ tamoxifen-treated

patients (9).

To test the applicability of RS in community oncology setting, Habel et al.

conducted a nested case-control study in breast cancer patients diagnosed from

1985 to 1994 at 14 Kaiser hospitals (11). Eligibility included negative nodes, age

less than 75 years, and no chemotherapy. Cases died of breast cancer prior to

2002. Up to three controls were matched to each case on age, race, tamoxifen

treatment, facility, diagnosis year, and follow-up time. Among 4964 potentially

eligible patients, 220 cases and 570 matched controls were identified. Median

age was 59 years (range, 28–74 years); 30.9% were tamoxifen treated (mostly

after 1988). Median follow-up time was 4.9 years for cases (time to death) and

12.9 years for controls. RS was significantly associated with breast death in ERþpatients treated with or without tamoxifen (p ¼ 0.002). This study is important

since it provides validation of RS assay used in a community setting.

PREDICTION OF RESPONSE TO CHEMOTHERAPY

Among gene expression–based markers, only RS has been tested whether it

predicts response to chemotherapy in adjuvant setting (12). Since many of the

genes comprising 21 genes in the RS assay are ER or proliferation related, we

hypothesized early on that RS would be predictive of response to chemotherapy.

Existence of chemotherapy arm in trial B-20 allowed us to ask this

question (in fact this exactly was the reason why we first decided to work with

N� ERþ tamoxifen-treated cohort since we knew that we could not only dis-

cover and validate the prognostic index, but also test whether it predicts response

100 Paik


to chemotherapy using banked materials readily available to us) (12). There were

651 evaluable patients (227 randomized to tamoxifen and 434 randomized to

tamoxifen plus chemotherapy). Patients with tumors who had high RS greater

than 30 had a large absolute benefit of chemotherapy [with an absolute increase

in distant recurrence free survival (DRFS) at 10 years of 27.6 � 8.0%, mean �SE]. Patients with tumors who had low RS less than 18 derived minimal, if any,

benefit from chemotherapy (with an absolute increase in DRFS at 10 years of

–1.1 � 2.2%, mean � SE). The test for interaction between chemotherapy

treatment and RS was statistically significant (p ¼ 0.038). Of particular

importance is the finding that there is a linear relationship between RS and

degree of benefit from chemotherapy when RS was examined as a continuous

variable in Cox Model.

While the relationship between RS and chemotherapy response has not

been validated in an independent cohort as was done for prognostic aspect of the

assay, supportive evidences exist to allow its use in clinical decision making.

The most important data are from a cohort of patients treated with neoadjuvant

chemotherapy at Milan reported by Gianni et al. (13). In this cohort, higher RS

correlated with higher incidence of complete pathological response (pCR). In

addition, due to the fact that many genes in RS are proliferation and ER related, we

can make reasonable assumption that B-20 data are true. Since the decision to get

chemotherapy has to be based on evaluation of baseline risk and assessment of

the degree of potential benefit from adding chemotherapy, it is reasonable to

conclude that those with lower RS may not need chemotherapy since their

baseline risk is low on the basis of B-14 data, and expected benefit is low on the

basis of B-20 data. One headache is the fact that because of the continuous nature

of the RS, it is difficult to pinpoint from which RS score patients start to gain

benefit from chemotherapy. Clinical oncologists are not used to this kind of data.

Uncertainties regarding the degree of benefit from chemotherapy in patients with

intermediate range of RS resulted in launching a large randomized clinical trial

named Trial Assigning IndividuaLized Options for Treatment (TAILORx) by

North American Breast Cancer Intergroup (14). Many questioned about the

reasons for the definition of the intermediate risk group in TAILORx trial being

different from the definition used in the original publication in B-14 cohort. This

was due to the fact that primary clinical end point in TAILORx is disease-free

survival instead of distant disease-free survival and consideration of 95% con-

fidence intervals in the baseline risk assessment. Again one has to realize that

these cutoffs are totally arbitrary.

PREDICTION OF RESPONSE TO NEOADJUVANT CHEMOTHERAPY

Preoperative systemic therapy trials provide unique opportunities to assess

sensitivity of tumor cells to systemic therapy in vivo while treatment is ongoing.

Numerous studies have demonstrated the correlation of pCR and long-term

outcome in breast cancer patients, potentially eliminating the need for long-term



followup. Together with ease of obtaining fresh pretreatment core biopsies

allowing high throughput assessment of molecular markers, preoperative trials

may allow speedy discovery and validation of predictive markers for response to

systemic therapy. So it is not surprising that many investigators attempted to

utilize neoadjuvant setting to develop predictive markers of response to che-

motherapy using microarray-based gene expression profiling. Regrettably the

cumulative experience has so far been somewhat disappointing. No profile has

been found to have enough accuracy to be directly clinically applicable, although

Hess et al. demonstrated that gene expression–based predictor still provided

better prediction than clinical predictors (15). But more importantly, there has

been no study with large enough sample size to provide conclusive answers as to

whether one could reliably predict pCR using gene expression levels.

It probably is somewhat naıve to expect gene expression profile alone to

predict pCR. For example, one would expect a smaller tumor would have higher

chance of getting into pCR than a bigger tumor with same gene expression

phenotype. Host factors such as drug metabolism and distribution may influence

the chance for pCR. Prediction of combination regimens also may be more

difficult than predicting response to single regimen. However, in the latter sit-

uation, because of low pCR rate, it is very difficult to develop predictor unless

sample size is very high. In this regard, perhaps what investigators have achieved

so far is as far as they can get, unless these other factors are integrated into

algorithm development. In addition, pCR itself is not a perfect measure of

response either. As many as 80% of tumors show clinical response on taxane-

based neoadjuvant chemotherapy, and therefore among those not achieving pCR,

there is a continuous spectrum of residual burden in the breast. So it may make

more sense to correlate gene expression with residual tumor burden than pCR,

which has a somewhat arbitrary cutoff.

Despite these limitations, certain consensus has emerged that are in

agreement with what we found in B-20 study—that the tumors with poor

prognostic signatures have higher chance of responding to chemotherapy. In

addition to above described study by Gianni et al. (13), Rouzier et al. (using gene

expression array) (16) and Carey et al. (using IHC surrogates of intrinsic sub-

types) (17) has examined whether Perou’s intrinsic subtypes correlate with

chance of pCR. As expected from B-20 data, basal and HER2 subtypes have

higher chance of pCR compared to luminal subtypes. So there is a general

agreement in the field that genetic program that determines the prognosis is

tightly linked with general sensitivity to chemotherapy.

ASSESSMENT OF RESIDUAL RISK AFTER CHEMOTHERAPY

Despite the fact that gene expression data alone have not been shown to be able

to predict pCR with high accuracy, gene expression profiling can be still

important in patients treated with neoadjuvant chemotherapy. In an unpublished

study, using pretreatment core biopsy samples from NSABP B-27, we found that

102 Paik


the combination of prognostic profile with pCR may be used to define residual

clinical risk after chemotherapy, since those with pCR had a very good prognosis

despite having a poor prognostic gene signature. This means that patients with

poor prognostic gene signature who did not achieve pCR are at high risk of

recurrence after chemotherapy. Such data will allow identification of patients

who will require more than chemotherapy without the need for waiting until they

develop recurrences and opens door for a concept for post-neoadjuvant targeted

therapy trials.

CLINICAL UTILITY OF RS ASSAY

How Is RS Compared to NCCN or St. Gallen Clinical Practice Guidelines?

In B-14 study, less than 10% of the tamoxifen treated patients are classified as

low risk by either NCCN or St. Gallen clincial guidelines. These patients did

have excellent prognosis with less than 10% distant failure within 10 years. In

contrast, RS classified 51% of the same cohort into low-risk category with very

similar 10-year failure rate. Therefore 40% more ERþ tamoxifen-treated

patients would be spared from chemotherapy if RS is used in comparison to

NCCN or St. Gallen clinical guidelines. In addition, unlike NCCN and St. Gallen

guidelines, which assumed that treatment benefit is equal in all subgroups,

NSABP B-20 data showed that the benefit from chemotherapy is nearly zero in

low-risk group defined by RS.

How Is RS Compared to Adjuvant on Line Algorithm?

The high cost of the RS assay (estimated at $3400) can only be justified by the

additional information that it provides compared with other assays. Bryant and

colleagues, in collaboration with Peter Ravdin, compared RS with Adjuvant on

Line (AOL) in the NSABP B-14 cohort in a presentation made at the 2005

St. Gallen Breast Cancer Symposium (personal communications with the

late Dr. John Bryant). The AOL program provides individualized estimates of

recurrence risk on the basis of information such as patient age, tumor size, nodal

involvement, and histological grade. It is actually not that straightforward to

compare RS and AOL, since AOL does not provide risk scores. Therefore, it is

difficult to utilize the numerical output from AOL when comparisons are made

to RS. To compare the two, Dr. Bryant rank ordered output from AOL so that

similar proportion of patients belong to low- (50%), intermediate- (25%), and

high-risk (25%) categories as defined by RS. Rank-ordered AOL (R-AOL)

performed remarkably well as a prognostic factor (in B-14) as well as predictive

factor for benefit from chemotherapy (in B-20). RS and R-AOL were both

independently prognostic in B-14 in multivariate analysis, suggesting that both

gene expression signature and clinical features influence patient prognosis.

However, agreement between the two assays was only 48%. RS was able to tease



out patients with poor prognosis from the R-AOL low-risk category, so that the

10-year distant relapse rate was 5.6% for patients categorized as low risk by both

R-AOL and RS. R-AOL low-risk but RS intermediate- and high-risk patients had

a 12.9% distant relapse rate at 10 years in the tamoxifen-treated cohort (p ¼0.04). For R-AOL intermediate- or high-risk patients, the distant relapse rate was

8.9% when the RS risk was low, whereas the distant relapse rate was 30.7%

when the RS risk was intermediate or high (p < 0.0001). These data suggest that

the RS adds to existing prognostic markers and it would be ideal to incorporate

RS and AOL into a single algorithm. However, at this point we don’t have a

clear way to actually implement this finding into AOL in clinical use.

Applicability in Node-Positive Patients

One question about RS is whether it will work for Nþ ERþ patients. There is no

reason why it should not work when it actually works even in 10þ node patients.

However, risk estimates for RS will be different between N� and Nþ patients

although overall direction of the data would be the same, so one cannot simply

apply the current RS algorithm to Nþ patients without a study.

Categorical Vs. Continuous Variable

One of the biggest problems in the field currently in my view is the use of RS as

a categorical variable (low-, intermediate-, and high-risk categories), which was

arbitrarily defined for validation study to ease the understanding of the data,

because clinicians are not used to seeing a continuous data (9). It is very

important to recognize the continuous nature of the RS. For example, a patient

with score 17 (categorized as low risk) has nearly identical risk as a patient with

score 19 (categorized as intermediate risk). On the other hand, a patient with RS

of 3 would have much different prognosis than the one with RS of 17, although

they are both categorized as low risk in the published report.

FUTURE PERSPECTIVE

The problem of current gene expression profiling methods is the need for fresh or

snap frozen tissue and/or high cost. There are alternative methods that may be

able to overcome these barriers. For example, Quantiplex assay (Panomics)

based on combination of branched DNA amplification and fluorescence-labeled

beads may allow multiplexing as many as 30 genes at a time in a 96 well plate

with reasonable amount of input RNA extracted from FFPET. cDNA-mediated

annealing, selection, extension, and ligation (DASL) assays from Illumina

is another promising example (18). These assays cost fraction of the cost for

RT-PCR or microarray-based methods and can utilize FFPET as starting mate-

rials. Replacement of standard markers such as ER and HER2 by these assays or

QRT-PCR is desirable since they will provide more linear prediction of response.

104 Paik


Availability of gene expression profiling methods for FFPET opens the

possibility that archived materials from adjuvant chemotherapy trials can be

analyzed, and therefore current approach of using gene expression analysis in

neoadjuvant setting as a mean to identify predictive markers will become less

important. The latter approach always will require validation in adjuvant setting

since the FDA will not approve marker based on pCR as a surrogate—therefore

it makes more sense to perform discovery and validation of predictive markers

purely in the adjuvant setting.

REFERENCES

1. Schena M, Shalon D, Davis RW, et al. Quantitative monitoring of gene expression

patterns with a complementary DNA microarray. Science 1995; 270:467–470.

2. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based predictors

for breast cancer. N Engl J Med 2006; 355:560–569.

3. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas

distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A

2001; 98:10869–10874.

4. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts



Nature 2000; 406:747–752.



7. Masuda N, Ohnishi T, Kawamoto S, et al. Analysis of chemical modification of RNA

from formalin-fixed samples and optimization of molecular biology applications for

such samples. Nucleic Acids Res 1999; 27:4436–4443.

8. Cronin M, Pho M, Dutta D, et al. Measurement of gene expression in archival

paraffin-embedded tissues: development and performance of a 92-gene reverse

transcriptase-polymerase chain reaction assay. Am J Pathol 2004; 164:35–42.


treated, node-negative breast cancer. N Engl J Med 2004; 351:2817–2826.

10. Cobleigh MA, Tabesh B, Bitterman P, et al. Tumor gene expression and prognosis in

breast cancer patients with 10 or more positive lymph nodes. Clin Cancer Res 2005;

11:8623–8631.

11. Habel LA, Shak S, Jacobs MK, et al. A population-based study of tumor gene

expression and risk of breast cancer death among lymph node-negative patients.

Breast Cancer Res 2006; 8:R25.

12. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in women

with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol 2006;

24:3726–3734.



breast cancer. J Clin Oncol 2005; 23:7265–7277.

14. Sparano JA. TAILORx: trial assigning individualized options for treatment (Rx).

Clin Breast Cancer 2006; 7:347–350.



15. Hess KR, Anderson K, Symmans WF, et al. Pharmacogenomic predictor of sensitivity

to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin, and cyclo-

phosphamide in breast cancer. J Clin Oncol 2006; 24:4236–4244.

16. Rouzier R, Perou CM, Symmans WF, et al. Breast cancer molecular subtypes respond

differently to preoperative chemotherapy. Clin Cancer Res 2005; 11:5678–5685.

17. Carey LA, Dees EC, Sawyer L, et al. The triple negative paradox: primary tumor

chemosensitivity of breast cancer subtypes. Clin Cancer Res 2007; 13:2329–2334.

18. Bibikova M, Talantov D, Chudin E, et al. Quantitative gene expression profiling in

formalin-fixed, paraffin-embedded tissues using universal bead arrays. Am J Pathol

2004; 165:1799–1807.

106 Paik


8

Utilization of Genomic Signatures for

Personalized Treatment of Breast Cancer

P. Kelly Marcom, Carey K. Anders,Geoffrey S. Ginsburg, and Anil Potti

Duke Institute for Genome Sciences & Policy, Department of Medicine,Duke University Medical Center, Durham, North Carolina, U.S.A.

Joseph R. Nevins

Duke Institute for Genome Sciences & Policy, Department of MolecularGenetics and Microbiology, Duke University Medical Center, Durham,

North Carolina, U.S.A.

INTRODUCTION

The challenge of personalized cancer treatment lies in matching the most effective

therapy to the characteristics of an individual patient’s tumor. The selection

of therapy for both early- and late-stage breast cancer, as is the case for most

cancer therapy, is still largely empiric and guided by large, population-based,

randomized clinical trials. Althoughwidely adopted, this approach is inadequate for

the selection of individualized chemotherapy regimens. Estimates of benefit are

extrapolations from the effects observed in these large trials and are not necessarily

applicable to individual patients.

The use of genomic data, in particular gene expression profiles derived from

DNA microarray analysis, has the promise of providing the basis for truly indi-

vidualized therapy. This approach has the potential to transform the treatment of

107


cancer from a population-based approach to an individualized one directed toward

unique patient characteristics. This includes the use of gene expression data to

identify subtypes of cancer not previously recognized by traditional methods of

analysis. Profiling genomic patterns to identify new subclasses of tumors has

previously been described in distinguishing differences between acute myeloid

leukemia and acute lymphoblastic leukemia (1,2). Similar studies have employed

gene expression data to identify distinct subtypes of breast cancer with prognostic

significance (3).

A landmark study in breast cancer published in 2002 described the use of a

DNA microarray-based, 70-gene expression profile as a prognostic factor in

breast cancer (4). This genomic marker has demonstrated predictive value

stratifying stage I and II breast cancer patients into two broad subgroups, rela-

tively high-risk versus low-risk groups as defined by a 10-year risk of breast

cancer recurrence. Various other studies have described the development of

expression profiles that have prognostic value in breast cancer (5–9).

The Netherlands 70-gene predictor has provided the basis for an expanded

European study that aims to measure its performance in a more diverse set of

patients (10). The parallel, prospective clinical trial MINDACT (Microarray In

Node-negative Disease may Avoid ChemoTherapy) aims to measure the effec-

tiveness of this genomic predictor in guiding adjuvant chemotherapy when

compared to predictions based solely on the traditional clinicopathologic fea-

tures. Likewise, a large prospective trial in the United States, TAILORx (Trial

Assigning IndividuaLized Options for Treatment), is currently evaluating the

Oncotype DX test (Genomic Health, Inc., Redwood City, California, U.S.),

a polymerase chain reaction (PCR)-based assay of 21 genes correlated with

10-year breast cancer recurrence risk (11), to guide the administration of hor-

monal therapy with or without chemotherapy among patients with estrogen

receptor (ER)-positive, early-stage breast cancer.

Of equal, if not greater, importance in achieving the goal of personalized

treatment of breast cancer is the ability to predict response to specific chemo-

therapeutics, particularly the standard-of-care cytotoxic regimens incorporated

into routine clinical practice. Various studies have described approaches to

developing genomic predictors of response to chemotherapeutic agents prom-

ising a more effective strategy of assigning chemotherapeutic regimens to

patients, as guided by unique characteristics of their breast tumors (12–14).

The overall aim of this chapter is to highlight both opportunities presented

by recent advances in the development of genomic-based approaches within the

field of breast cancer and corresponding strategies for implementation of these

techniques into clinical practice. Rather than relying on historic paradigms of

chemotherapeutic selection as guided by broad, population-based studies, the

goal of this unique methodology is to improve patient outcome, and ultimately

patient survival, through the optimal selection of patient-specific chemo-

therapeutic agents as guided by genomic-based strategies.

108 Marcom et al.


CYTOTOXIC THERAPIES FOR EARLY-STAGE BREAST CANCER

Currently, the standard regimens for the adjuvant treatment of early-stage breast

cancer combine anthracyclines, taxanes, and cyclophosphamide in limited per-

mutations of combination, dose, and administration schedule. The addition of a

taxane, either sequentially or in combination, to a core anthracycline/cyclo-

phosphamide regimen in the setting of node-positive breast cancer has been

shown to improve disease-free survival (DFS) by 4% to 7% in four separate

randomized trials (15–18). However, it is unclear whether all breast cancer

subtypes are equally affected by this therapeutic approach. Recent work exam-

ining a series of breast cancer adjuvant trials conducted over the past two dec-

ades further confirms the importance of defining breast cancer subtypes and their

relationship to treatment benefit. This work demonstrated a lack of benefit for

this broad treatment approach across a subset of ER-positive patients (19).

Optimizing chemotherapy outcomes through the manipulation of dosing

schedules remains controversial. In CALGB 9741 (Cancer and Leukemia Group

B Intergroup 9741), decreasing the dosing interval from every three weeks to

every two weeks [i.e., dose-dense (DD) scheduling] was shown to improve DFS

and overall survival (OS) at three-year median follow-up [risk ratio (RR) ¼ 0.74,

p ¼ 0.010] (20). However, an update at five-year median follow-up showed a

less impressive overall improvement in DFS (RR ¼ 0.80, p ¼ 0.012) and

perhaps no improvement in ER-positive disease (21). Moreover, a randomized

Phase III trial from the Gruppo Oncologico Nord Ovest–Mammella Inter

Gruppo (GONO–MIG) examining DD scheduling of a non-taxane-containing

regimen did not show a benefit in DFS or OS (22). In an accompanying edi-

torial to this publication, the ambiguity regarding the benefit of DD therapy

was acknowledged and emphasis was placed on the need to define biologically

relevant breast cancer subtypes so that individualized treatment strategies could

be developed (23).

A ROLE FOR PREOPERATIVE SYSTEMIC CHEMOTHERAPYTO EVALUATE GENOMIC STRATEGIES

Preoperative, systemic chemotherapy (PST), also known as neoadjuvant therapy,

is an effective means for assessing the sensitivity of an individual breast tumor

to chemotherapy. Multiple trials of preoperative systemic chemotherapy have

been performed. Initially, this approach was primarily used for locally advanced,

inoperable, or inflammatory breast cancer. However, subsequent trials have

evaluated PST in the setting of smaller, operable breast tumors both to increase

breast conservation rates and to assess sensitivity to chemotherapy.

Over the past two decades, PST for early-stage breast cancer has been

shown to be a safe and effective method for assessing the chemosensitivity of

primary breast cancers. Sequencing systemic therapy first, with the consequent

Genomic Signatures for Personalized Treatment of Breast Cancer 109


three- to six-month delay in surgery, has not been shown to adversely affect

outcomes (24). Assessment of pathologic response to PST, both in the breast and

regional lymph nodes, provides the most powerful prognostic information

available. Patients who obtain a complete pathologic response in the breast and

lymph nodes have a greater than 95% survival at seven years (24,25). Although

PST provides a powerful tool for in vivo assessment of chemosensitivity,

exploiting the information gained in the aforementioned trials to improve

management of individual patients has been more difficult.

Studies employing a PST approach have yielded conflicting results illus-

trating the heterogeneity of primary breast tumors and responsiveness to dif-

ferent chemotherapeutic agents. This observation begs the question on whether

or not breast tumors commonly possess intrinsic ‘pan-chemotherapy’ respon-

siveness or resistance. For example, patients responsive to preoperative treat-

ment with CVAP (Cytoxan, vincristine, Adriamycin, and prednisone) were

either continued on the same regimen or crossed over to docetaxel (26).

Responding patients who received docetaxel had a significantly higher patho-

logic complete response (pCR) rate than those who continued on the CVAP

regimen, supporting differential breast tumor sensitivity to anthracyclines and

taxanes. Additionally, the pCR rate in the NSABP (National Surgical Adjuvant

Breast and Bowel Project) B-27 clinical trial was doubled by the addition of

preoperative docetaxel to preoperative doxorubicin/cyclophosphamide (AC)

chemotherapy (26.1% vs. 12.9%, p < 0.0001) supporting this conclusion.

However, the addition of preoperative docetaxel did not yield a statistically

significant difference in overall DFS when added to preoperative AC alone

[hazard ratio (HR) ¼ 0.9, p ¼ 0.22]. The question remains open on whether

patients only partially responsive to AC, who achieved a pCR following

administration of docetaxel, might have achieved a pCR with docetaxel alone,

thus avoiding the risk of cardiotoxicity inherent to anthracyclines. Unfortunately,

this study is somewhat flawed by the concurrent administration of tamoxifen

with chemotherapy in patients with known ER-positive breast tumors (25). An

additional study by the GEPARDUO (German Preoperative Adriamycin Doce-

taxel) study compared DD doxorubicin/docetaxel (AT) with AC followed by

docetaxel, illustrating improved response among patients receiving cyclo-

phosphamide. This study, however, was more widely interpreted as a compari-

son of sequential versus DD therapy (27).

The largest trials investigating response to PST with AC come from the

NSABP B-18 and the aforementioned B-27 clinical trials, where the pCR rates,

allowing for residual in situ disease, were consistently approximately 13%

(25,28). Data for single-agent paclitaxel are more limited; however, a study

conducted at the M.D. Anderson Cancer Center reported pCR rates of 6% when

excluding in situ diseases (29). A comprehensive review of these and other PST

trials supports that the pCR rate in the breast with most doublet chemotherapies

is approximately 10% to 15% (24). The National Comprehensive Cancer

110 Marcom et al.


Network (NCCN) guidelines support that standard adjuvant regimens can also be

used as preoperative therapy (http://www.nccn.org).

Although response to PST is a powerful tool for assessing tumor chemo-

sensitivity, it is a post hoc evaluation and requires the exposure of individual

patients to toxic and potentially ineffective therapies. Consequently, it has been

difficult to exploit the knowledge gained about both tumor sensitivity and

resistance to the patient’s benefit. In addition, to our knowledge, no studies have

examined the influence of PST on the postoperative management of patients who

have minimal or no response to PST.

The relative importance of including either an anthracycline or taxane or

both as part of the adjuvant treatment regimen of early stage breast cancer is an

area of active research. These two classes of agents, in combination with

cyclophosphamide, provide the foundation for the treatment of early-stage breast

cancer, and several clinical trials provide insight into the current use of these

agents. A recently published randomized phase III trial compared AC with

docetaxel/cyclophosphamide (TC) (30). While this trial demonstrated improve-

ment in DFS favoring TC (86% vs. 80%, p ¼ 0.015), the confidence interval (CI)

approached 1.0 (CI ¼ 0.50–0.94) and there was no difference in OS despite

median follow-up of 5.5 years. Interestingly, a previous trial comparing AT with

AC showed no difference in DFS at four years (31). Although both of these trials

enrolled a lower-risk population of patients, reconciling differences in outcome

across these two studies is difficult. An ongoing trial in patients with fewer than

four positive axillary nodes is comparing AC with single-agent paclitaxel as

adjuvant therapy (CALGB 40101), in part on the basis of responses observed in

the aforementioned PST trials. In summary, although many active chemo-

therapeutic agents suitable for the treatment of early-stage breast cancer exist,

data available in 2007 do not provide the necessary guidance to select the most

effective chemotherapeutic agent for each unique patient. This observation

speaks to the importance of developing novel methods employing genomic

analyses of tumor biology to individualize and optimize patient care.

GENE EXPRESSION PROFILES TO PREDICT THERAPEUTIC RESPONSE

Given the current status of neoadjuvant and adjuvant chemotherapy for breast

cancer, an opportunity exists to use genomic information to improve the selection

of patients who would derive the greatest benefit from treatment with cytotoxic

chemotherapy. That is, prognostic tools that can identify patients clearly in need of

further treatment can help to refine the decisions based now on clinical infor-

mation. Additionally, the development of predictive signatures specific for

response to individual cytotoxic agents would provide an invaluable clinical tool

to enable the better use of available standard-of-care agents. Historically, pre-

dictive biomarkers in the setting of breast cancer have played an important role in

selecting patients for targeted therapies (i.e., trastuzumab and endocrine therapies



including tamoxifen and aromatase inhibitors). However, validated biomarkers for

commonly used cytotoxic agents are not yet clinically available. The importance

of cytotoxic agents, even with the advent of the targeted molecular therapy era,

cannot be discounted. To date, many targeted therapies demonstrate greatest

efficacy when combined with chemotherapeutic agents. To address this relevant

clinical need, a variety of studies have been performed with the aim of developing

predictors of sensitivity for a spectrum of cytotoxic chemotherapies.

Gene expression profiling of breast cancers in the context of preoperative

chemotherapy has shown promise in providing predictive signatures of response

to specific agents (12–14). The availability of clinical datasets provides an

opportunity to move ahead with the prospective assessment of gene expression

profiling as a means for guiding PST. This methodology can also simultaneously

provide insight into breast cancer biology, as the investigation of profiles will

shed light on the deregulation of molecular pathways.

Several studies have correlated breast cancer gene expression profiles with

response to preoperative chemotherapy. Chang et al. performed gene expression

profiling on locally advanced breast cancers treated with docetaxel to identify

genes predictive of response to this therapeutic modality (32). Similar work has

been done with preoperative doxorubicin therapy and sequential taxane/anthra-

cycline therapy (33,34). While some studies have suggested that single markers

such as NF-kB may predict resistance to anthracycline therapy, most have found

that a multigene classifier is needed. In comparing a signature for doxorubicin/

cyclophosphamide sensitivity with a signature for docetaxel sensitivity, Cleator

et al. concluded that unique signatures for sensitivity to different chemotherapy

regimens are likely to exist (33).

More recently, studies of PST have examined the pCR rates for clinically

and molecularly defined breast cancer subtypes. These retrospective analyses of

PST trials demonstrate marked differences in pCR rates depending on breast

cancer subtype (35,36). A recent publication by the M.D. Anderson Breast Group

has retrospectively defined gene expression profiles that predict response to a

sequential regimen of paclitaxel (Taxol) followed by 5-fluorouracil, doxorubicin

(Adriamycin), and cyclophosphamide (T-FAC) (13).

As an alternate strategy, drug sensitivity data from the NCI-60 panel of

cancer cell lines, coupled with baseline Affymetrix gene expression data unique

to these cells, has been utilized to generate a series of gene expression signatures

with the potential to predict sensitivity to various chemotherapeutic agents (37)

(Fig. 1). To test the capacity of the in vitro docetaxel sensitivity predictor to

accurately identify the patients who responded to docetaxel, we evaluated data

from a previously published neoadjuvant study that linked gene expression data

with clinical response to docetaxel (32). The in vitro–generated profile correctly

predicted docetaxel response in 22 of 24 patient samples, achieving an overall

accuracy of 91.6%. Findings were further validated in an ovarian cancer dataset,

where accuracy for predicting the response to salvage docetaxel therapy

exceeded 85% (37).

112 Marcom et al.


Given that the in vitro developed gene expression profiles predicted

clinical response to docetaxel, the approach has been extended to test the ability

of additional signatures to predict response to commonly used salvage therapies

in an independent dataset of ovarian cancer patients previously treated with

Adriamycin. As shown in Figure 1B, each of these predictors was capable of

accurately predicting the response to the drugs in patient samples, achieving an

accuracy in excess of 81% overall. Importantly, in these examples, the overall

clinical response rate varied from 25% to 45%. Using the expression signatures

to predict patients likely to respond would essentially increase the “effective”

response rate (the positive predictive value for chemosensitivity for any given

drug) to greater than 85%, by selecting those patients likely to respond, based on

the drug sensitivity predictor. These data suggest that the genomic-based pre-

dictors of chemotherapeutic response provide a unique opportunity to guide the

selection of a cytotoxic agent most effective for an individual patient—the

essence of personalized medicine.

Importantly, it is also evident from further analyses that while there are

overlaps in the predicted sensitivities to the chemotherapeutic agents among

ovarian patients, there are also distinct groups of patients that are predicted to be

sensitive to various single-agent salvage therapies (Fig. 2). Many therapeutic

regimens make use of combinations of chemotherapeutic cytotoxic agents,

raising a question on the extent that signatures of individual therapeutic response

will also predict response to a combination of agents. To address this question,

Figure 1 Signatures that predict response to cytotoxic chemotherapies. Heatmaps of

expression profiles selected to predict sensitivity to the indicated chemotherapies. Each

column in the figure represents individual samples. Each row represents an individual

gene, ordered from top to bottom according to regression coefficients. Shown below is

prediction of single agent therapy response in patient samples using cell line based

expression signatures of chemosensitivity. Source: Adapted from Potti et al., Nature

Medicine 12:1294 –1300, 2006.



data from a breast neoadjuvant trial were utilized to evaluate the ability of single-

agent signatures for T-FAC to predict patient response measured by pCR (34).

Utilization of individual chemotherapy sensitivity signatures to assess patient

response illustrated a significant distinction between the responders (n ¼ 13) and

nonresponders (n ¼ 38), with the exception of 5-flourouracil. Importantly, the

combined probability of sensitivity to all four agents in the T-FAC neoadjuvant

regimen was calculated using the probability theorem. It was clear from this

analysis that the prediction of response based on a combined probability of

sensitivity, built from individual chemosensitivity predictions, yielded a statis-

tically significant (p < 0.0001, Mann–Whitney U test) distinction between the

responders and nonresponders.

A STRATEGY FOR USING CHEMOTHERAPYRESPONSE PREDICTORS IN PST

Although promising, studies to date have included post hoc gene expression

profiling of tumor samples. The ability to effectively select the most appropriate

preoperative systemic cytotoxic agents for the treatment of early-stage breast

cancer must ultimately be tested and validated in prospective clinical trials. This

rationale approach will not only provide an opportunity to confirm biologically

relevant subtypes, but also establish their predictive capacities regarding

response to specific, conventional cytotoxic agents.

Figure 2 Patterns of predicted sensitivity to commonly used chemotherapies. A

collection of breast cancer samples represented by microarray data was used to predict

the likely sensitivity to the indicated agents based on the signatures at the top of the

figure. Predictions are plotted as a heatmap representing probability of sensitivity.

Adapted from Potti et al., Nature Medicine 12:1294 –1300, 2006.

114 Marcom et al.


The development of genomic signatures predictive of response to com-

monly employed chemotherapeutics offers an opportunity to significantly

expand the identification of combinations of cytotoxic agents best suited for an

individual patient. With this in mind, several prospective clinical studies have

been proposed to test the capability of genomic predictors to most effectively

match cytotoxic agents with individual patients. This concept is best evaluated in

the context of neoadjuvant trials where hypotheses regarding the effectiveness of

genomic predictions can be accurately assessed through pathologic measure-

ments of tumor response.

We note that the availability of chemotherapy response predictors provides

an opportunity to employ genomic predictors in present-day practice, where

patients are commonly treated with one or more therapeutic regimens of equal

efficacy (Fig. 3). The ability to select the most effective therapy, among a panel

of standard-of-care agents, is a rational strategy because standard treatment

procedures would not be greatly altered. The goal of these trials would be to

more effectively utilize currently available standard agents. For example, a

breast neoadjuvant study has been designed and is currently being implemented

by employing this approach. Patients with early-stage breast cancer will be

treated with one of two current standard-of-care therapies for which genomic

predictors are available. This study design would compare the response rates

for the randomly or doctor-selected A or B regimens versus assignment to

therapy A or B via genomic predictor of response. The success of this study

could then set the stage for larger studies utilizing genomic predictors of

chemotherapy sensitivity to more accurately predict response to cytotoxic

therapies. The next phase of clinical studies would then focus on novel ther-

apeutics (i.e., targeted therapies) for the individual patients unlikely to respond

to either regimen.

Figure 3 Schematic diagram of prospective clinical trial that would evaluate the

capacity of predictors of chemotherapy response to improve response rates.



OTHER GENOMIC OPPORTUNITIES TO GUIDEBREAST CANCER THERAPEUTICS

The successful development and validation of predictors of response to standard

chemotherapeutics will foster the more efficient use of these agents through the

identification of patients with the greatest likelihood of response. Concurrently,

these strategies will also identify a subgroup of patients resistant to all available

standard-of-care chemotherapeutics, necessitating the implementation of tar-

geted strategies. This is a critical point as it is of little benefit to patients to

identify resistance if a plan for alternative therapeutic strategies does not exist.

Currently, most patients resistant to standard chemotherapeutic agents are tra-

ditionally treated with second- or third-line chemotherapeutic agents or are

enrolled on a clinical trial. The more effective alternative is to identify patients

likely to be resistant prior to initiation of therapy so that opportunities for

alternative treatments might be employed in the first-line setting.

A viable option is to exploit the wealth of pathway-specific therapeutics

that have been developed over the past two decades. To this goal, gene

expression patterns reflecting deregulation of oncogenic pathways have been

developed, as depicted in Figure 4 (38). These oncogenic pathway signatures

have been shown to accurately predict the status of pathway deregulation in a

series of both tumors derived from murine models and human tumors.

These results demonstrate an ability to predict pathway profiles unique to

individual human tumors, thus characterizing the genetic events associated with

tumor development. Importantly, the assay of gene expression profiles provides

a measure of the consequence of the oncogenic process, irrespective of how the

pathway might have been altered. Thus, even if the known oncogene is not

mutated, but rather another component of the pathway is altered, the gene

expression profile will detect this alteration.

The analysis of oncogenic pathways through gene expression profiling

offers an opportunity to identify new therapeutic options for patients by pro-

viding a potential basis for guiding the use of pathway-specific drugs. To

demonstrate the value of pathway prediction to guide drug selection, pathway

deregulation was mapped among a series of breast cancer cell lines to screen

potential therapeutic drugs. In parallel with pathway status mapping, the cell

lines were treated with a variety of drugs known to target specific activities

within given oncogenic pathways. A variety of available reagents known to

target oncogenic pathways for which signatures have been developed [i.e., a far-

nesyl transferase inhibitor (FTS) and an Src inhibitor (SU6656)] were employed.

In each case, a clear relationship was demonstrated between prediction of pathway

deregulation and sensitivity to the respective targeted therapeutic. These prelim-

inary, yet promising, results have been further validated in a larger set of 50 cancer

cell lines, including lung, ovarian, and melanoma tumor types.

Taken together, these data demonstrate a potential approach to identify

therapeutic options for chemotherapy-resistant patients, as well as identify novel

116 Marcom et al.


combinations for chemotherapy-sensitive patients. This unique approach repre-

sents a potential strategy to optimize effective treatments for cancer patients and

warrants future, prospective, validations trials in both the adjuvant and metastatic

settings.

Pathway signatures have the potential to identify patients most sensitive

to standard-of-care cytotoxic drugs, as well as those resistant to standard

regimens. This approach creates an opportunity to combine chemotherapy

sensitivity profiles with concurrent oncogenic pathway deregulation within an

individual patient’s tumor. As depicted in Figure 5, predictions of sensitivity to

Figure 4 Patterns of predicted cell signaling pathway activation. A collection of

breast cancer samples represented by microarray data was used to predict the likely state

of activation of the indicated cell signaling pathways based on the signatures at the top of

the figure developed to represent pathway activation. Predictions are plotted as a heatmap

representing probability of pathway activation. A Kaplan-Meier survival analysis is

shown at the bottom reflecting the overall survival for patients in each cluster defined by

pathway patterns. Source: Adapted from Bild et al., Nature 439:353–357, 2006.



various cytotoxic agents can be combined with predictions of oncogenic

pathway deregulation with the goal of defining the most optimal regimen. For

example, a patient whose tumor is resistant to docetaxel may concurrently

illustrate a high probability of Src pathway deregulation. Src inhibitors are

currently being tested in clinical trials in a variety of solid tumors, including

breast tumors. The selection of a therapy targeted against the Src pathway may

prove more efficacious for this individual patient compared with the empiric

selection of a cytotoxic agent. In addition, this strategy could be employed to

identify novel combinations of therapies—either multiple cytotoxic agents or

cytotoxic agents combined with targeted agents—as guided by a more rational

and informed approach.

IMPACT AND FUTURE

The practice of oncology continually faces the challenge of matching the optimal

therapeutic regimen with the most appropriate patient, balancing relative benefit

with risk, to achieve the most favorable outcome. This challenge is often

daunting with marginal success rates in many advanced disease contexts, likely

reflecting the enormous complexity of the disease process coupled with an

inability to properly guide the use of available therapeutics.

The importance of selecting patients who will respond to a given thera-

peutic agent is perhaps best illustrated by the example of trastuzumab. In the

absence of selection as guided by Her2 protein overexpression or gene ampli-

fication, the overall response rate among breast cancer patients is on the order of

Figure 5 Combining information regarding status of individual patient’s tumors.

The figure depicts an analysis of a set of breast tumors for predicted sensitivity to various

chemotherapies (top figure) and then the pattern of pathway activation in relation to one

of these (docetaxel sensitivity).

118 Marcom et al.


10%. In contrast, for patients selected on the basis of Her2 overexpression or

amplification, the overall response rate when combined with chemotherapy

approached 50% to 60% (39–41). The use of trastuzumab in the treatment of

both early-stage and advanced breast cancer is guided by Her2 protein over-

expression of gene amplification. Gene expression signatures predicting

response to various cytotoxic chemotherapeutic agents provide an equally

important opportunity to optimize response to various cytotoxic agents.

Importantly, clinically-available, validated predictors of chemotherapy

response could provide an invaluable tool in present day practice. Commonly,

patients are treated with one or more therapeutic regimens that, on a population

basis, have individually demonstrated equal efficacy. Validated chemotherapy

predictive of chemotherapy sensitivity provides an opportunity to select the most

effective therapy among a panel of standard-of-care agents individualized to a

particular patient’s tumor biology. It is essential that this approach be incorpo-

rated into modern day clinical trials. Ultimately, this novel concept has the

potential to augment future treatment strategies incorporating likelihood of

sensitivity to common cytotoxic agents, as well as various targeted therapeutic

agents. Incorporating this approach into the routine management of patients with

various malignancies has the potential to more rationally assign combination

therapies, both cytotoxic and targeted agents, as rationally guided by each

individual’s unique tumor biology.

SUMMARY

Treatment within the field of oncology, in particular that of breast cancer, is

shifting away from a broad, population-based approach toward individualized

care as driven by unique tumor biology. Validated signatures of chemotherapy

sensitivity, coupled with an understanding of oncogenic pathway signaling

within a given breast tumor, provides an opportunity to improve the utilization of

standard-of-care therapies. This approach also provides an opportunity to take

the next step toward personalized cancer care—incorporation of targeted

agents—as guided by advances in genomic technology.

REFERENCES

1. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell

lymphoma identified by gene expression profiling. Nature 2000; 403:503–511.

2. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer: class dis-

covery and class prediction by gene expression monitoring. Science 1999; 286:531–537.

3. Perou CM, Jeffrey SS, van de Rijn M, et al. Distinctive gene expression patterns in

human mammary epithelial cells and breast cancers 1999; 96:9212–9217.

4. van ’t Veer L, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts





outcomes. Lancet 2003; 361:1590–1596.

6. Pittman J, Huang E, Dressman H, et al. Integrated modeling of clinical and gene

expression information for personalized prediction of disease outcomes. Proc Natl

Acad Sci U S A 2004; 101:8431–8436.

7. Ma XJ, Wang Z, Ryan PD, et al. A two-gene expression ratio predicts clinical outcome

in breast cancer patients treated with tamoxifen. Cancer Cell 2004; 5:607–616.

8. Chang HY, Sneddon JB, Alizadeh AA, et al. Gene expression signature of fibroblast

serum response predicts human cancer progression: similarities between tumors and

wounds. PLOS Biol 2004; 2:206–214.

9. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumors.

Nature 2000; 406:747–752.

10. van de Vijver MJ, He YD, van’T Veer LJ, et al. A gene-expression signature as a






2003; 362:362–369.



and cyclophosphamide in breast cancer. J Clin Oncol 2006; 24:4236–4244.

14. Modlich O, Prisack HB, Munnes M, et al. Predictors of primary breast cancers

responsiveness to preoperative epirubicin/cyclophosphamide-based chemotherapy:

translation of microarray data into clinically useful predictive signatures. J Transl

Med 2005; 3:1–18.

15. Mamounas EP, Bryant J, Lembersky B, et al. Paclitaxel after doxorubicin plus

cyclophosphamide as adjuvant chemotherapy for node-positive breast cancer: results

from NSABP B-28. J Clin Oncol 2005; 23:3686–3696.

16. Henderson IC, Berry DA, Demetri GD, et al. Improved outcomes from adding

sequential Paclitaxel but not from escalating Doxorubicin dose in an adjuvant che-

motherapy regimen for patients with node-positive primary breast cancer. J Clin

Oncol 2003; 21:976–983.

17. Martin M, Pienkowski T, Mackey J, et al. Adjuvant Docetaxel for node-positive

breast cancer 2005; 52:2302–2313.

18. Roche H, Fumoleau P, Spielmann M, et al. Five years analysis of the PACS 01 trial:

6 cycles of FEC100 vs 3 cycles of FEC100 followed by 3 cycles of docetaxel (D) for

the adjuvant treatment of node positive breast cancer. Breast Cancer Res Treatment

2004; 88:27 (abstr).

19. Berry DA, Cirrincione C, Henderson IC, et al. Estrogen-receptor status and outcomes

of modern chemotherapy for patients with node-positive breast cancer. JAMA 2006;

295:1658–1667.

20. Citron ML, Berry DA, Cirrincione C, et al. Randomized trial of dose-dense versus

conventionally scheduled and sequential versus concurrent combination chemo-

therapy as postoperative adjuvant treatment of node-positive primary breast cancer:

first report of Intergroup Trial C9741/Cancer and Leukemia Group B Trial 9741.

J Clin Oncol 2003; 21:1431–1439.

120 Marcom et al.


21. Hudis C, Citron M, Berry D, et al. Five year follow-up of INT C9741: dose-dense

chemotherapy is safe and effective. Presented at: The 28th Annual San Antonio

Breast Cancer Symposium, December 8–11, 2005.

22. Venturini M, Del Mastro L, Aitini E, et al. Dose-dense adjuvant chemotherapy in

early breast cancer patients: results from a randomized trial. J Natl Cancer Inst 2005;

97:1724–1733.

23. Lin NU, Gelman R, Winer EP. Dose density in breast cancer: a simple message?

J Natl Cancer Inst 2005; 97:1712–1714.

24. Kaufmann M, Hortobagyi GN, Goldhirsch A, et al. Recommendations from an

international expert panel on the use of neoadjuvant (primary) systemic treatment of

operable breast cancer: an update. J Clin Oncol 2006; 24:1940–1949.

25. Bear HD, Anderson S, Smith RE, et al. Sequential preoperative or postoperative

docetaxel added to preoperative doxorubicin plus cyclophosphamide for operable

breast cancer: national Surgical Adjuvant Breast and Bowel Project Protocol B-27.

J Clin Oncol 2006; JCO.2005.04:1665.

26. Smith IC, Heys SD, Hutcheon AW, et al. Neoadjuvant chemotherapy in breast cancer:

significantly enhanced response with docetaxel. J Clin Oncol 2002; 20:1456–1466.

27. von Minckwitz G, Raab G, Caputo A, et al. Doxorubicin with cyclophosphamide

followed by docetaxel every 21 days compared with doxorubicin and docetaxel every

14 days as preoperative treatment in operable breast cancer: the GEPARDUO study

of the German Breast Group. J Clin Oncol 2005; 23:2676–2685.


outcome of women with operable breast cancer. J Clin Oncol 1998; 16:2672–2685.

29. Buzdar AU, Singletary SE, Theriault RL, et al. Prospective evaluation of paclitaxel

versus combination chemotherapy with fluorouracil, doxorubicin, and cyclo-

phosphamide as neoadjuvant therapy in patients with operable breast cancer. J Clin

Oncol 1999; 17:3412–3417.

30. Jones SE, Savin MA, Holmes FA, et al. Phase III trial comparing doxorubicin plus

cyclophosphamide with docetaxel plus cyclophosphamide as adjuvant therapy for

operable breast cancer. J Clin Oncol 2006; 24:5381–5387.

31. Goldstein L, O’Neill A, Sparano J, et al. E2197: Phase III AT (doxorubicin/docetaxel)

vs. AC (doxorubicin/cyclophosphamide) in the adjuvant treatment of node positive and

high risk node negative breast cancer. American Society of Clinical Oncology Annual

Meeting, May 13–17, 2005.

32. Chang JC, Wooten EC, Tsimelzon A, et al. Patterns of resistance and incomplete

response to docetaxel by gene expression profiling in breast cancer patients. J Clin

Oncol 2005; 23:1169–1177.

33. Cleator S, Tsimelzon A, Ashworth A, et al. Gene expression patterns for doxorubicin

(Adriamycin) and cyclophosphamide (Cytoxan) (AC) response and resistance. Breast

Cancer Res Treat 2006 95:229–233.




35. Guarneri V, Frassoldati A, Giovannelli S, et al. Primary systemic therapy for

operable breast cancer: a review of clinical trials and perspectives. Cancer Lett 2007;

248:175–185.

36. Rouzier R, Perou CM, Symmans WF, et al. Breast cancer molecular subtypes respond

differently to preoperative chemotherapy. Clin Cancer Res 2005; 11(16):5678–5685.



37. Potti A, Lancaster JM, Dressman HK, et al. A genomic strategy to guide the use of

chemotherapeutic drugs in solid tumors. Nat Med 2006; 12:1294–1300.

38. Bild A, Yao G, Chang JT, et al. Oncogenic pathway signatures in human cancers as a

guide to targeted therapies. Nature 2006; 439:353–357.

39. Chan A, Martin M, Untch M, et al. Vinorelbine plus trastuzumab combination as

first-line therapy for HER 2-positive metastatic breast cancer patients: an interna-

tional phase II trial. Br J Cancer 2006; 95:788–793.

40. Marty M, Cognetti F, Maraninchi D, et al. Randomized phase II trial of the efficacy

and safety of trastuzumab combined with docetaxel in patients with human epider-

mal growth factor receptor 2-positive metastatic breast cancer administered as first-

line treatment: the M77001 study group. J Clin Oncol 2005; 23:4265–4274.

41. Robert N, Leyland-Jones B, Asmar L, et al. Randomized phase III study of trastuzumab,

paclitaxel, and carboplatin compared with trastuzumab and paclitaxel in women with

HER-2-overexpressing metastatic breast cancer. J Clin Oncol 2006; 24:2786–2792.

122 Marcom et al.


9

Personalized Medicine by the Use of

Microarray Gene Expression Profiling

Stella Mook and Laura J. van ’t Veer

Department of Pathology, The Netherlands Cancer Institute,Amsterdam, The Netherlands

Fatima Cardoso

Department of Medical Oncology, Institute Jules Bordet, Brussels, Belgium

INTRODUCTION

For accurate management of early breast cancer, several important decisions

need to be made by patients and clinicians regarding the type of surgery, the

need for adjuvant radiotherapy, and particularly the type of adjuvant systemic

therapy (chemotherapy and/or endocrine therapy). The decision about adjuvant

systemic therapy is of paramount importance since breast cancer is mainly a

systemic disease, with a majority of deaths occurring as a consequence of distant

metastases. This decision depends on the likelihood of recurrence; patient-

related factors such as biological age, menopausal status, performance status,

and co-morbidities; and tumor-related factors such as size, grade, lymph node

status, and hormonal and human epidermal growth factor receptor 2 (HER2)

status. Most of these variables are combined into prognostic models or guidelines

such as the St. Gallen consensus or the National Comprehensive Cancer Network

(NCCN) guidelines, in order to assist patients and clinicians in the decision-

making process (1–3). Although these guidelines provide valuable information

about prognosis in certain breast cancer populations, the prediction of disease

123


outcome for the individual patient is far from accurate. In reality, patients with

similar clinicopathological characteristics can have strikingly different outcomes.

Because of these limitations, and because of the recognition of the incurable

nature of metastatic breast cancer, clinicians consider prescribing adjuvant

chemotherapy in the majority of early-stage breast cancer patients to reduce the

risk of relapse. However, a large proportion of these patients is probably cured

with locoregional treatment alone or in combination with adjuvant endocrine

therapy, thus unnecessarily exposing many patients to chemotherapy side effects.

The identification of robust and reliable prognostic markers that accurately select

patients not requiring adjuvant chemotherapy is essential to decrease the problem

of overtreatment. Additionally, a more accurate patient selection will also con-

tribute to the decrease in undertreatment, a less common but dangerous problem.

The introduction of high-throughput techniques, such as microarray-based

gene expression profiling and sequencing of the human genome, has led to a novel

approach in breast cancer research. Measuring the expression level of thousands of

genes in one experiment using microarray technology enables us to better

understand the biology and heterogeneity of breast cancer. The first study to

examine gene expression patterns in breast cancer showed the existence of at least

four molecular subtypes of breast cancer—luminal A, luminal B, basal, and HER2

positive—distinguished by extensive differences in gene expression (4). Several

subsequent studies confirmed these findings (5–8). In addition, microarray tech-

niques have been used to identify profiles that correlate with disease outcome

(prognostic profile) or with response to treatment (predictive profile) (9–16).

One of the first prognostic profiles was the 70-gene profile (Mamma-

PrintTM) identified by van ’t Veer et al. (13). This dichotomous microarray

classifier can accurately distinguish breast cancer patients who have a high

likelihood of remaining free of distant metastases (good profile) from those who

are at high risk of developing distant metastases (poor profile) (Fig. 1).

DEVELOPMENT AND FIRST RETROSPECTIVE VALIDATIONOF THE 70-GENE PROFILE (MAMMAPRINTTM)

The 70-gene prognostic profile was developed in a training set of 78 lymph

node–negative, invasive breast tumors less than 5 cm in diameter. Patients were

younger than 55 years at diagnosis, had no prior malignancies, and were treated

in the Netherlands Cancer Institute (NKI) with breast conserving therapy or

mastectomy. Five out of 78 patients received adjuvant systemic treatment

comprising chemotherapy (n ¼ 3) or endocrine therapy (n ¼ 2). Thirty-four

patients developed distant metastases within five years of diagnosis, while the

remaining 44 patients remained free of distant metastases. The mean follow-up

of patients who remained free of distant metastases was 8.7 years, and the mean

time to metastases was 2.5 years.

For all 78 samples, the percentage of tumor cells in a hematoxylin- and

eosin-stained section, before and after cutting sections for RNA isolation, was at

124 Mook et al.


least 50%. RNA was isolated and labeled with a fluorescent dye. Reference RNA

consisted of an equal amount of RNA from all tumors. Both sample RNA and

reference RNA were hybridized on an oligonucleotide microarray platform

containing approximately 25,000 genes (produced by Agilent Technologies, Palo

Alto, California, U.S.) (17). After hybridization, microarray slides were washed

and scanned using a confocal laser scanner (Agilent). Fluorescence intensities

were quantified, corrected for background noise, and normalized (13).

Using a statistical method called “supervised classification,” 231 genes

that were differentially expressed in tumors that metastasized compared with

tumors that did not were selected. These 231 genes were then ranked on the basis

of their correlation coefficient with disease outcome. Using a leave-one-out

cross-validation procedure, the top 70 of these genes appeared to be the optimal

set of genes to accurately discriminate between tumors with a good disease

outcome (free from distant metastases for at least five years) and tumors with a

poor disease outcome (distant metastases within five years of diagnosis).

All 78 tumors were ranked according to their correlation with the average

expression of the 70 genes of the patients who did not develop distant metas-

tases. The sensitivity was optimized by setting a threshold, which resulted in the

Figure 1 Steps in 70-gene microarray analysis. Source: From Ref. 32. copyright# 2002

Massachusetts Medical Society. All rights reserved. Adopted with permission.

Microarray Gene Expression Profiling 125


misclassification of 3 out of 34 patients with a poor disease outcome being

classified as good prognosis and therefore would erroneously have had chemo-

therapy withheld (9% misclassification). Subsequently, the 70-gene profile was

validated in 19 breast cancer tumors—7 tumors from patients with a good

clinical outcome (no distant metastases within five years of diagnosis) and

12 tumors from patients who did develop distant metastases within five years of

diagnosis. Using the 70-gene profile, only 2 out of 19 tumors were classified

incorrectly, thereby confirming the initial results (13,18).

The same group of investigators performed a first retrospective validation

study using a consecutive series of 295 breast tumors from NKI (144 lymph

node–positive and 151 lymph node–negative tumors). All patients were younger

than 53 years at diagnosis and were treated with locoregional therapy alone

(56%) or in combination with adjuvant systemic therapy comprising chemo-

therapy alone (31%), hormonal therapy alone (7%), or a combination (7%). In

this consecutive series of 295 patients, 61 were also part of the previous series

used to develop the 70-gene profile (14).

All 295 tumors were classified as good or poor profile by calculating the

correlation coefficient with the established average expression level of the

70 genes in the previously studied good-outcome group. For the 234 samples that

were not included in the previous study, tumors with a correlation coefficient

above the threshold of 0.4 were classified as good profile; for the 61 tumors

that were also part of the previous study, the threshold was adapted to 0.55 to

correct for overfitting (18).

The 70-gene profile accurately classified 115 tumors (39%) as a good-

prognosis group (good profile) with a 10-year overall survival of 95% (�2.6%)

and 180 tumors (61%) as a poor-prognosis group (poor profile) with a 10-year

overall survival of 55% (�4.4%). Interestingly, the 70-gene profile was not only

associated with disease outcome in the lymph node–negative patients [hazard

ratio (HR) for distant metastases 5.5; 95% confidence interval (CI) 2.5–12.2; p <0.001], but also strongly associated with disease outcome in the group of 144

lymph node–positive patients (HR for distant metastases 4.5; 95% CI 2.0–10.2;

p < 0.001). Furthermore, the only significant independent factors for prediction

of distant metastases were a poor-prognosis profile, a larger tumor, the presence

of vascular invasion, and no chemotherapy treatment. The 70-gene profile was

by far the most powerful predictor of distant metastases, with an adjusted HR of

4.6 (95% CI 2.3–9.2; p < 0.001) (14,18).

THE PROGNOSTIC VALUE OF THE 70-GENEPROFILE IN A CLINICAL CONTEXT

To assess the potential added value of this prognostic tool in a clinical context,

the 70-gene profile was compared with the St. Gallen consensus guidelines or

the NIH criteria used at that time (3,19). In comparison with these guidelines, the

70-gene profile assigned more patients to the good-prognosis group than the

St. Gallen consensus guidelines or the NIH criteria (40% vs. 15% and 7%,

126 Mook et al.


respectively). Moreover, patients identified by the 70-gene profile as having

good prognosis were more likely to remain free of distant metastases than

patients identified as having good prognosis according to the St. Gallen con-

sensus guidelines or the NIH criteria. On the other hand, patients who were

assigned to the poor-prognosis group by the 70-gene profile had a higher risk of

developing distant metastases as compared with the poor-prognosis group

classified by the St. Gallen consensus guidelines or the NIH criteria.

The 70-gene profile was built to have a minimum of misclassified patients

with a poor disease outcome; in other words, the low-risk group should be

maximized without causing considerable undertreatment. Consequently, in this

retrospective validation series, the 70-gene profile had a high negative predictive

value of 96% for the 234 new patients and a positive predictive value of only

38%. That is to say, of all patients with a good prognosis profile, 4% will still

develop distant metastases and would erroneously have had chemotherapy

withheld, whereas of all patients classified as having a poor prognosis, 38% will

indeed develop distant metastases. Consequently, treatment decisions based on

the 70-gene profile would still lead to overtreatment; but since the total pro-

portion of patients identified as having a poor profile is much smaller than the

proportion of high-risk patients according to the St. Gallen consensus guidelines

or the NIH criteria, overtreatment would be reduced by 25% to 30%. Further-

more, the overall prediction of disease outcome and consequent treatment

decision according to the 70-gene profile seems to be more accurate.

INDEPENDENT RETROSPECTIVE VALIDATIONBY THE TRANSBIG CONSORTIUM

To further substantiate the prognostic value of the 70-gene profile, the European

Union’s sixth Framework Project TRANSBIG (http://www.breastinternational-

group.org/TransBIG.aspx) conducted an independent retrospective validation

study (20). Frozen tumor material and clinical data were collected from

302 patients from five different cancer centers in the United Kingdom, Sweden,

and France. Patients were younger than 61 years at diagnosis, had a lymph node–

negative T1 or T2 tumor, and had not received adjuvant systemic therapy. The

patients were treated before 1999, and the median follow-up in this series was

13.6 years. The frozen tumor samples were sent to Agendia, the microarray facility

in Amsterdam, where RNA was isolated and hybridized on a custom-designed

microarray, measuring the gene expression level of the 70 genes in triplicate.

Tumors were classified as poor profile if the correlation coefficient of the average

expression was under 0.4. The microarray facility had no access to the clinical

data, and the researchers collecting the clinical data were blinded for the 70-gene

profile results. Only an independent statistical office had access to both the clinical

and the microarray data simultaneously. Eighty percent of the samples were

centrally reviewed for estrogen receptor (ER) status and tumor grade, and the

centers were independently audited for quality of data control.



This independent study confirmed that the 70-gene profile can accurately

select patients who are at high risk of developing distant metastases (poor prog-

nosis) from patients who are at low risk of developing distant metastases and

hence can be safely spared chemotherapy (good profile), with HRs of 2.79 (95%

CI 1.60–4.87) and 2.32 (95% CI 1.35–4.0) for distant metastases–free survival and

overall survival, respectively. Moreover, patients with a good profile had a 10-year

overall survival of 89% compared with 69% in patients with a poor profile. The

70-gene profile outperformed several prognostic methods such as the Nottingham

Prognostic Index and the St. Gallen consensus guidelines. The performance of the

70-gene profile was also compared with Adjuvant! Online. The Adjuvant! Soft-

ware—available at www.adjuvantonline.com—calculated a 10-year overall sur-

vival probability based on clinicopathological criteria (the patient’s age, tumor

size, tumor grade, ER status, and comorbidity) (21,22). Patients were considered

as having low clinical risk when the 10-year breast cancer–specific survival

probability calculated by Adjuvant! was more than 88% if their tumors were ER

positive and more than 92% if they were ER negative. The rationale for these two

cutoffs was the assumption that patients with ER-positive tumors would now

receive hormonal therapy, with an average absolute benefit of 4%. In the dis-

cordant cases, the 70-gene profile provided much stronger prognostic information,

with almost identical survival rates for the low and high clinical risk groups within

each 70-gene profile risk, suggesting that the 70-gene profile predicts disease

outcome independently of clinical and pathological characteristics (Fig. 2) (20).

Figure 2 Kaplan–Meier curve by clinical and 70-gene expression signature risk groups.

Overall survival. Source: From Ref. 20. By permission of Oxford University Press.

128 Mook et al.


Another important finding of this validation study is the time dependency

of the 70-gene profile. The median follow-up in this series was twice as long as

the follow-up of the initial series. To determine the influence of the time of

distant metastases on the prognostic value of the profile, HRs were calculated

with arbitrary censoring of all observations at different time points. The adjusted

HR for distant metastases at 5 years was 4.7 in comparison with an adjusted HR

of 3.5 at 10 years, suggesting a better prediction of early (within five years of

diagnosis) metastases by the 70-gene profile (18,23,24).

OTHER PROGNOSTIC PROFILES IN BREAST CANCER

Several other prognostic gene expression signatures were developed, such as the

76-gene profile, the wound signature, and the genomic grade index (15,25,26).

The 76-gene prognostic profile was developed on a different microarray platform

(Affymetrix, Santa Clara, California, U.S.) and using a slightly different meth-

odology; the genes associated with relapse of the disease were selected sepa-

rately for ER-positive and ER-negative patients. The genes selected in each

group were then combined in a 76-gene profile (15). A retrospective validation

study confirmed the prognostic value of this profile in early-stage breast cancer

patients (27). Recently, the 76-gene profile has also been validated independ-

ently by the TRANSBIG consortium in the same patient series, using the same

methodology as the validation of the 70-gene profile (24). Although only

three genes were common to both profiles, the biological pathways represented

are the same, and the prognostic value of both the 70-gene profile and the

76-gene profile was validated in this patient series (20,24). Since the profiles

perform equally, and the 70-gene profile has proven to be reproducible and

robust and is available as a Food and Drug Administration (FDA)-cleared

diagnostic test (28), the TRANSBIG consortium decided to use the 70-gene

profile in a large prospective clinical trial, the Microarray In Node-negative

Disease may Avoid ChemoTherapy (MINDACT) trial.

IMPLEMENTATION OF THE 70-GENE PROFILE:THE RASTER TRIAL AND THE LOGISTICS PILOT STUDY

In addition to scientific evidence provided by validation studies, implementation

of these microarray profiles in both prospective clinical trials and in daily practice

requires logistical feasibility. The requirement of fresh tissue as the source of high-

quality RNA has several consequences for tumor sample handling. Traditional

fixation of fresh tissue in formaldehyde, for example, causes degradation of RNA,

thereby making tumor samples unsuitable for microarray experiments. In addition,

to avoid interlaboratory variability, fresh tumor samples have to be shipped to one

microarray facility. Consequently, for the implementation of microarray profiles,

local procedures have to be adapted, and close collaboration between pathologists,

surgeons, and medical oncologists is of outmost importance.



Recently, two multicenter feasibility studies were conducted, the RASTER

study and a logistics pilot study (18,29). The first study was coordinated by NKI

with financial support from the Dutch Health Care Insurance Board. The aim of

the study was to assess the feasibility of collecting good-quality tissue from

several community hospitals in the Netherlands to perform the 70-gene profile

test. For all patients included in this study, tumor samples from excised speci-

mens were obtained and placed in a commercially available preservation fluid

(RNAlater1, Qiagen GmBH, Hilden, Germany) at room temperature. Sub-

sequently, the samples were sent by conventional mail to NKI, where they

were snap-frozen and stored at –808C. The 70-gene profile was performed at the

microarray facility in Amsterdam. Preliminary results showed that it is feasible

to adapt local procedures and to collect good-quality tissue for microarray testing

in a multicentric setting (18,29,30).

In the RASTER study, tissue was preserved temporarily in RNAlater and

sent by conventional mail at room temperature. Preservation of tumor tissue in

RNAlater may influence several processes in the tissue, such as protein levels.

Since one of the additional aims of the MINDACT trial is the establishment of a

biological materials bank for future research, including proteomics, storage of

tissue in RNAlater is not suitable for the MINDACT trial and fresh, frozen tissue

will be mandatory (31). Therefore, a second feasibility study was conducted to

test the collection, storage, and shipment of fresh, frozen tissue in six European

hospitals. Women younger than 71 years, with early-stage breast cancer, were

included. After surgery, the pathologist obtained representative tumor samples

using a 6-mm biopsy puncher. Tumor samples were snap-frozen in liquid

nitrogen and stored at –808C within one hour of surgery. The samples were

shipped to Amsterdam on dry ice by a contracted courier specializing in the

transportation of frozen tissue. At Agendia, RNA was isolated when the samples

were representative of the tumor (i.e. tumor cells � 50%). After measurement of

the quality and quantity of the RNA, the samples were hybridized on the 70-gene

profile platform. Results showed that in general it is feasible to collect good-

quality fresh, frozen tissue for microarray tests in a multicentric and multina-

tional setting. Additionally, in the course of this feasibility study, procedures

were adapted and optimized for use in the MINDACT trial (18).

PROSPECTIVE VALIDATION OF THE 70-GENEPROFILE IN THE MINDACT TRIAL

Despite the completion of all necessary technical validation procedures and the

FDA clearance as the in vitro diagnostic multivariate index assay (IVDMIA),

including approval of clinical utility, the clinical validation of the 70-gene

profile can still be further optimized in currently diagnosed, adjuvantly treated

populations to achieve level-1 evidence. The TRANSBIG consortium has

decided to further validate the 70-gene prognostic alongside the classical clin-

icopathological factors (31). This trial, the MINDACT trial [EORTC (European

130 Mook et al.


Organization for Research and Treatment of Cancer) 10041 BIG (Breast Inter-

national Group) 3-04], will enroll 6000 women between 18 and 71 years of age,

with primary invasive breast cancer at diagnosis. Tumors should be unilateral,

stage T1, T2, or operable T3; carcinomas in situ are eligible provided invasive

cancer is present. Patients should be treated with breast conserving treatment or

mastectomy and should have a negative sentinel node or axillary clearance.

Patients with primary malignancies or previous chemotherapy or radiotherapy

will not be included. The main goal of the trial is to prove that the 70-gene

profile will safely assign chemotherapy to a smaller proportion of node-negative

breast cancer patients. Therefore, all patients included in MINDACT will have a

risk assessment based on the 70-gene profile and according to the currently used

clinicopathological criteria. This last risk assessment is provided by an updated

version of Adjuvant! Online. If, according to both risk assessments, the patients’

risk of developing distant metastases is low (an estimated 13% of all patients), no

chemotherapy will be given. If both methods classify the patients’ risk of devel-

oping distant metastases as high (an estimated 55% of the patients), chemotherapy

will be proposed. If the risk assessments are discordant (an estimated 32% of

all patients), patients will be randomized to the clinicopathological criteria or

the 70-gene profile to be used for adjuvant chemotherapy decision making

(Fig. 3). The group of most interest to answer the primary research question

comprises patients with a low-risk 70-gene profile (good profile) and high-risk

Figure 3 MINDACT study design. Source: From Ref. 18. With permission.

Abbreviations: R-T, treatment decision randomization: CT, chemotherapy.



clinicopathological criteria who will be randomized to use the 70-gene profile

for the treatment decision and thus receive no chemotherapy. In this group a

null hypothesis of a five-year distant metastases–free survival of 92% will be

tested (31).

Additional study objectives are related to the type of chemotherapy and

endocrine therapy. Patients who are to receive chemotherapy may be randomized

between an anthracycline-based regimen and a regimen of docetaxel and cape-

citabine in order to compare the efficacy and safety of both regimens. The latter

regimen may have increased efficacy with less long-term side effects such as

cardiotoxicity and leukemia. Post-menopausal (spontaneous or induced) patients

with hormone receptor–positive tumors will be randomized for endocrine ther-

apy with two years of tamoxifen, followed by five years of letrozole or seven

years of letrozole upfront.

In addition to these three research questions, whole genome arrays will be

performed for all 6000 patients, and frozen and paraffin-embedded tumor sam-

ples together with blood samples will be collected from all patients. These

microarray data and the biological materials will be stored in an independent

biobank and will provide a valuable source of information for future research

that may result in the identification of new prognostic or predictive biomarkers

and new drug targets.

FUTURE PROSPECTS

The MINDACT trial is the first randomized clinical trial that will prospectively

evaluate the prognostic value of a microarray-based prognostic tool, the 70-gene

profile or Mammaprint, in an adjuvantly treated population. The implementation

of molecular diagnostics based on the true biology of a patient’s tumor will

greatly expedite the introduction of personalized medicine.

The details of the rationale, design, and logistics of the MINDACT trial are

described in references 18,23, and 31, and the authors strongly recommend these

articles to the reader.

ACKNOWLEDGMENTS

The studies mentioned in this chapter were supported by the European Com-

mission Framework Programme VI, the Center of Biomedical Genetics, the

Dutch Health Care Insurance Board, the Dutch National Genomic Initiative:

Cancer Genomics Program, the Breast Cancer Research Foundation, the EBCC-

Breast Cancer Working Group–association sans but lucratif (asbl), the Fondation

Belge Contre Le Cancer, the S. G. Komen for the Cure Foundation, Jacqueline

Seroussi Memorial Foundation for Cancer Research, and the European Organ-

isation for Research and Treatment of Cancer (EORTC)–Breast Cancer Group.

S. Mook was partially supported by the traineeship program of the TRANSBIG

network. The authors thank the numerous individuals who have contributed

132 Mook et al.


to the studies mentioned in this review, especially those from the TRANSBIG

consortium and the EORTC, and all the patients who have and still are participating

in these studies.

CONFLICTS OF INTEREST

Dr. L. J. van ’t Veer is a named inventor on a patent application for Mammaprint

and reports holding equity in Agendia BV.

REFERENCES

1. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in

Oncology. Breast Cancer V.I.2008. Available at: http://www.nccn.org/professionals/

physician_gls/PDF/breast.pdf.

2. Carlson RW, McCormick B. Update: NCCN breast cancer clinical practice guide-

lines. J Natl Compr Canc Netw 2005; 3(suppl 1):S7–S11.

3. Goldhirsch A, Glick JH, Gelber RD, et al. Meeting highlights: international expert

consensus on the primary therapy of early breast cancer 2005. Ann Oncol 2005;

16(10):1569–1583.


Nature 2000; 406(6797):747–752.

5. Hu Z, Fan C, Oh DS, et al. The molecular portraits of breast tumors are conserved

across microarray platforms. BMC Genomics 2006; 796.

6. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carci-

nomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci

U S A. 2001; 98(19):10869–10874.

7. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes

in independent gene expression data sets. Proc Natl Acad Sci U S A. 2003; 100(14):

8418–8423.

8. Sotiriou C, Neo SY, McShane LM, et al. Breast cancer classification and prognosis

based on gene expression profiles from a population-based study. Proc Natl Acad Sci

U.S.A. 2003; 100(18):10393–10398.



2003; 362(9381):362–369.


outcomes. Lancet 2003; 361(9369):1590–1596.

11. Jansen MP, Foekens JA, van SI, et al. Molecular classification of tamoxifen-resistant

breast carcinomas by gene expression profiling. J Clin Oncol 2005; 23(4):732–740.


treated, node-negative breast cancer. New England J Med 2004; 351(27):2817–2826.


clinical outcome of breast cancer. Nature 2002; 415(6871):530–536.

14. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A Gene-expression signature as a

predictor of survival in breast cancer. New England J Med 2002; 347(25):1999–2009.

15. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant metastasis

of lymph-node-negative primary breast cancer. Lancet 2005; 365(9460):671–679.



16. West M, Blanchette C, Dressman H, et al. Predicting the clinical status of human

breast cancer by using gene expression profiles. Proc Natl Acad Sci U.S.A. 2001;

98(20):11462–11467.

17. Hughes TR, Mao M, Jones AR, et al. Expression profiling using microarrays fab-

ricated by an ink-jet oligonucleotide synthesizer. Nat Biotech 2001; 19(4):342–347.

18. Mook S, ’van ’t Veer LJ, Rutgers EJ, et al. Individualization of therapy using

Mammaprint: from development to the MINDACT trial. Cancer Genomics Proteo-

mics 2007; 4(3):147–155.

19. Eifel P, Axelson JA, Costa J, et al. National Institutes of Health Consensus Devel-

opment Conference Statement: adjuvant therapy for breast cancer, November 1–3,

2000. JNCI Cancer Spectrum 2001; 93(13):979–989.

20. Buyse M, Loi S, van ’t Veer L, et al. Validation and clinical utility of a 70-gene

prognostic signature for women with node-negative breast cancer. JNCI Cancer

Spectrum 2006; 98(17):1183–1192.

21. Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the

prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23(12):

2716–2725.

22. Ravdin PM, Siminoff LA, Davis GJ, et al. Computer program to assist in making

decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol

2001; 19(4):980–991.

23. Cardoso F, van ’t Veer L, Rutgers E, et al. Clinical application of the 70-gene profile

(MammaprintTM): the MINDACT trial. J Clin Oncol 2008; 26(5):729–735.

24. Desmedt C, Piette F, Loi S, et al. Strong time dependence of the 76-gene prognostic

signature for node-negative breast cancer patients in the TRANSBIG multicenter

independent validation series. Clin Cancer Res 2007; 13(11):3207–3214.

25. Chang HY, Nuyten DSA, Sneddon JB, et al. Robustness, scalability, and integration

of a wound-response gene expression signature in predicting breast cancer survival.

Proc Natl Acad Sci U S A. 2005; 102(10):3738–3743.


understanding the molecular basis of histologic grade to improve prognosis. JNCI

Cancer Spectrum 2006; 98(4):262–272.

27. Foekens JA, Atkins D, Zhang Y, et al. Multicenter validation of a gene expression-

based prognostic signature in lymph node-negative primary breast cancer. J Clin

Oncol 2006; 24(11):1665–1671.

28. Couzin J. DIAGNOSTICS: amid debate, gene-based cancer test approved. Science

2007; 315(5814):924.

29. Bueno-de-Mesquita JM, van Hakten WH, Retel VP, et al. use of 70-gene signature to

predict prognosis of patients with node-negative breast cancer: a prospective com-

munity-based feasibility study (RASTER). Lancet Oncol 2008; 8: 1079–1087.

30. van de Vijver M. Gene-expression profiling and the future of adjuvant therapy.

Oncologist 2005; 10(suppl 2):30–34.

31. Bogaerts J, Cardoso F, Buyse M, et al. Gene signature evaluation as a prognostic

tool: challenges in the design of the MINDACT trial. Nat Clin Pract Oncol 2006; 3(10):

540–551.

32. Sauter G, Simon R. Predictive molecular pathology. N Engl J Med 347 2002; (25):

1995–1996.

134 Mook et al.


10

Predictive Markers for Targeted Breast

Cancer Treatment

Hans Christian B. Pedersena and John M. S. Bartlett

Endocrine Cancer Group, Cancer Research Centre,Western General Hospital, Edinburgh, United Kingdom

INTRODUCTION

Currently selection of breast cancer treatment is guided predominantly by patient

prognosis using classical pathological assessment of tumors, with higher risk

patient groups being offered more aggressive therapy. The choice of treatment

regimen is guided by a very small number of predictive biomarkers (Table 1) (1).

However, it is now clearly recognized that not only may the risks of treatment

outweigh the benefits in some patient groups but that not all patients are equal

with respect to their response to and benefit gained from exposure to novel and

established therapies. There is increasing interest in the identification of bio-

logical markers to improve the targeting of established therapies to those cancers

likely to respond. In addition, development of new molecular targeted drugs has

led to an urgent need for research into predictive markers that can distinguish

patients who benefit from treatment. New proteomic technologies are leading to

markers being discovered at an increasing rate, with protein expression, cellular

localization, and posttranslational modification being recognized as those of

potential importance. Many new immunohistochemical (IHC) markers are

aHans Christian B. Pedersen is employed on a Marie Curie Transfer of Knowledge project between

Dako A/S, Glostrup, Denmark, and University of Edinburgh.

135


currently being tested for predictive potential for various therapies in breast

cancer. Through large-scale gene expression studies using microarrays, it has

been realized that tumors have molecular signatures that may determine response

to treatment and patient prognosis. Many candidate-predictive and prognostic

signatures in breast cancer are being explored, and some are already in use to aid

patient management (e.g., OncoType Dx1, Genomic Health, Redwood City,

California, U.S.). The challenge for the future is to identify both the biological

and diagnostic determinates of treatment response in order to allow matching of

appropriate treatments to specific patients.

Future breast cancer treatment is likely to be guided by predictive markers

along with genomic signatures, treating patients with mixtures of molecular

targeted therapeutics. This chapter will review current IHC and fluorescent in

situ hybridization (FISH)-based predictive markers in breast and the challenges

associated with the development and implementation of predictive markers in

future.

CURRENT PREDICTIVE MARKERS IN BREAST CANCER: LESSONS LEARNT

A predictive marker is defined as a factor that indicates sensitivity or resistance

to a specific treatment (2). In contrast, a prognostic marker predicts a tumor’s

future behavior independent of systemic adjuvant treatment. Some markers can

be both predictive and prognostic, as exemplified by the estrogen receptor a(ERa) expression, which both predicts response to endocrine treatment and

correlates with prognosis.

Hormonal Therapy and Hormone Receptors

ERa and progesterone receptor (PR) status were the first predictive markers used

in the clinic to select patients for breast cancer treatment. Today, IHC ERa or PR

staining is done routinely on patient tumor samples to predict response to

endocrine therapy. More correctly, the assay is used to exclude ERa- or PR-

negative patients from ERa- or PR-targeted therapy, as they will have a low

probability of responding to treatment (3). Historically however, ERa was first

discovered in the late 1960s on its ability to bind radiolabeled ligand (4,5).

Table 1 Current FDA-Approved Predictive Markers in Breast

Predictive Marker Method Drug

ERa IHC Tamoxifen, AIs

PR IHC Tamoxifen, AIs

HER2/neu IHC Trastuzumab (Herceptin)

HER2/neu FISH Trastuzumab (Herceptin)

Abbreviations: ERa, estrogen receptor a; PR, progesterone receptor; AIs,aromatase inhibitors; FISH, fluorescence in situ hybridization.

136 Pedersen and Bartlett


Within 10 years of its initial discovery, assays had been developed that allowed

ER to be used a marker for hormonal responsiveness and clinical aggressiveness

of the tumor (6). Many years passed before it was realized that only ERa-positive patients could benefit from treatment and for these finding to be applied

as clinical practice (3). It is now almost universally accepted that ERa testing has

a central role in breast cancer management, but it is worth noting that no pro-

spective trial has yet tested the predictive value of ERa or PR as predictive

biomarkers. Later evidence has emerged supporting testing for PR as well as

ERa (7). PR is a transcriptional target of ligand-bound ERa, and it is has been

proved that including the PR score in the ERa score provides stronger evidence

for true ERa positivity and therefore makes patients more likely to respond to

therapy (8–10). However, controversy remains over the value of supplementary

markers to ERa, and in several countries, PR testing in not routinely performed.

The drugs administered to ERa- or PR-positive patients have been through

many generations through the last three decades, from pure ERa-antagonist(fulvestrant) to ERa–partial antagonist tamoxifen and drugs inhibiting the enzyme

aromatase that synthesise oestrogen (11). Current clinical practice in breast cancer

management utilizes a combination adjuvant tamoxifen followed up by aromatase

inhibitors (AIs) or AIs alone. However, ER or PR do not select between these

therapies, and despite evidence that PR is predictive of poor response on tamox-

ifen, there is no evidence, to date, that AIs circumvent this resistance.

Despite the long history of the ERa or PR status as a predictive and

prognostic marker, no standardized protocols for fixation and IHC staining are

yet widely applied. Conflicting evidence exists for when patients are considered

ER negative and therefore will not respond to treatment, but above 10% positive-

staining tumor cells are generally considered ER positive (12). Perhaps because

ERa has been available for testing as a predictive marker for such a long

time, before emphasis was put on standardizing IHC testing, this has lead to the

development of a multitude of different protocols using a variety of antibodies.

ER IHC results have been shown to be highly influenced by the efficiency of the

antigen-retrieval step (13). External quality assurance initiatives such as the

United Kingdom National External Quality Assessment Scheme for Immuno-

cytochemistry (UK NEQAS-ICC) exist now to improve quality across clinical

laboratories through external assessment of analytical and interpretative per-

formance; such schemes are however not uniformly successful.

In addition to standardizing ERa IHC for scoring tumors as positive or

negative, it has been suggested that quantification of ERa levels may be nec-

essary in some clinical settings because tumors expressing low or intermediate

ERa levels have been shown to benefit from chemotherapy in addition to

endocrine therapy (14). Current ERa IHC protocols are semiquantitative, so

alternatives such as radiolabeled ligand-binding assays or quantitative poly-

merase chain reaction (qPCR) have been suggested. Gene expression of ER and

PR correlate well with protein expression, and it has been suggested to be an

alternative to IHC as it is sensitive and quantitative over a large dynamic

Predictive Markers for Targeted Breast Cancer Treatment 137


range (15). It has been suggested that more quantitative methods of measuring

ERa and PR may provide additional information to guide choice of therapy. We

will discuss this in detail later (12).

HER2/neu and Trastuzumab (Herceptin1)

HER2/neu or human epidermal growth factor receptor 2, a product of the proto-

oncogene HER2 first discovered in 1985, belongs to the HER family of tyrosine

kinase receptors (16–18). Numerous studies have documented HER2/neu-

positive breast cancers display more aggressive disease and shortened disease-

free survival (19). Using recombinant technologies, trastuzumab (Herceptin1,

Genentech, South San Francisco, California, U.S.), a monoclonal immunoglo-

bulin G1 class, humanized murine antibody, was developed to specifically target

patients with breast cancer that overexpressed the HER2/neu protein (20,21).

Clinical trials demonstrated the efficiency of trastuzumab for treating metastatic

breast cancers and subsequent trials in combination therapy (21–25) for HER2

positive disease. Again, HER2/neu, like ERa, is more predictive of which

patients will not respond to treatment.

As in the case for endocrine therapy, most predictive marker development

followed drug development, leading to some confusion. In initial clinical trials,

patient selection was based on an IHC assay using the mouse monoclonal

antibody that had been humanized to create trastuzumab (26,27). Today most

patients are selected for trastuzumab therapy on the basis of HER2/neu

expression levels in tumor cells (Fig. 1). In 2001, HER2/neu testing achieved

standard care of practice in the American Society of Clinical Oncology (ASCO)

breast cancer clinical practice guidelines. In most laboratories, tumors are

screened with FDA-approved IHC tests, HercepTest1 (Dako, Glostrup, Denmark)

or HER2/neu (CB11) PATHWAY1 (Ventana Medical Systems, Inc., Tucson,

Arizona, U.S.). Ambiguous cases are then analyzed with FDA-approved HER2/

neu FISH from Vysis (Downers Grove, Illinois, U.S.), Dako, or Ventana. Alter-

natively, in some laboratories, FISH is the only method used for HER2/neu testing.

HER2/neu testing for trastuzumab therapy has undergone considerable scrutiny

in numerous studies, because evidence emerged indicated that the IHC tests

generated considerable amount of false positives and negatives (28–31). This

discrepancy is due in part to lack of adherence to protocols, variable fixation

methods, and subjectivity related to tumor scoring.

Only 20% to 35% of HER2/neu positive tumors respond to treatment,

depending on the context the drug is given in, and studies have demonstrated that

this may be attributed to a combination of factors, including lack of pathway

dependence, receptor being inactive despite overexpressed, or the induction of

resistance through various mechanisms (32). Controversy persists with regard to

the optimal method for HER2/neu testing and interpretation of results (33). It has

been reported that it may be more cost effective to test all tumors with HER2/neu

FISH, avoiding false positives and false negatives (28). This is a possibility, as



FISH and, more critically, chromogenic in situ hybridization (CISH) technology

become more widely distributed, and an agreement can be made on the best

choice of HER2/neu testing.

DEVELOPMENT OF PREDICTIVE MARKERS

The successful targeting of HER2/neu with trastuzumab inspired the develop-

ment of a wave of new therapeutic antibodies and kinase inhibitors that were

aimed at inhibiting growth factor–regulated signaling. However, repeating the

success of HER2/neu testing for trastuzumab therapy has proven difficult.

Inhibitors of mammalian Target Of Rapamycin (mTOR, e.g., temsirolimus,

Toricel1, Wyeth Pharmaceuticals, Inc., Madison, New Jersey, U.S.), epidermal

growth factor receptor (EGFR, e.g., cetuximab, Erbitux1, ImClone Systems, Inc.,

Branchburg, New Jersey, U.S.) and vascular endothelial growth factor receptor

(VEGFR, e.g., bevacizumab, Avastin1, Genentech) have shown modest or no

results in clinical trials as single agent therapies in breast cancer (34–36). This may

be partly attributed to lack of predictive markers of response in combination with

drugs simply not being effective as single agents. However, this failure may also

Figure 1 Current HER2/neu testing in breast cancer. IHC staining performed on patient

tumor samples and scored using the HercepTest. Patient testing 0 to 1þ go on to receive

other treatments. Patients testing 2þ go on to HER2/neu FISH testing and if amplified for

HER2/neu receive trastuzumab. Patients scoring 3þ receive trastuzumab. Alternatively all

patients undergo HER2/neu FISH and patients scored as amplified receive trastuzumab.

Abbreviations: IHC, immunohistochemical; FISH, fluorescence in situ hybridization.



reflect poorly designed approaches toward the discovery of predictive biomarkers.

Predictive markers for EGFR-targeted therapy have been in focus for the past five

years, and we will therefore briefly go through the history of EGFR-targeted

therapy and what has been learnt for future predictive marker development. In

addition, a model for codevelopment of predictive markers in relation to tradi-

tional drug development is presented.

EGFR—A Cautionary Tale

EGFR or HER1 belongs to the same family of tyrosine kinase receptors as

HER2/neu and EGFR is considered an attractive target for therapeutic inter-

vention in many cancer forms due to the receptor being frequently mutated and

overexpressed (34). EGFR overexpression is associated with reduced disease-

free survival and resistance to chemotherapy (37). EGFR inhibitors (e.g., erlo-

tinib and gefitinib) and therapeutic antibodies against the extracellular domain of

the receptor (e.g., cetuximab, matuzumab, and panitumumab) are currently in

clinical trials or late-stage development (38). Most clinical trials have inves-

tigated EGFR inhibitors in cancers of the lung and colon, and cetuximab has

been approved for treatment of colorectal cancer (34). Clinical trials have also

been conducted in breast cancer, with limited success (39–47). In several of these

trials, patients have been selected for trial on the basis of EGFR overexpression.

However, this was later shown to be a mistake because EGFR expression, as

currently evaluated, did not correlate with clinical response (48,49) or with

breast cancer cell line inhibition (50). Clinical trials in breast cancer have clearly

demonstrated that EGFR inhibitors and therapeutic antibodies are targeting

EGFR receptors in both tumor and skin from patients, so the lack of effect is not

due to the drug’s failure to reach the target or the target’s lack of inhibition (41).

So EGFR inhibitors hit their target, and certain cancer patients clearly benefit

from these drugs. Therefore, it is surprising that although EGFR-targeted therapy

has been in focus for the past decade and drugs are now being administered to

patients, no FDA-approved test or profile exists at present to guide clinicians on

the use of EGFR inhibitors. Studies have shown EGFR gene copy number,

mutations, and downstream signalling to be associated with clinical response in a

series of lung cancer trials (51–55). However, these were retrospective, poorly

powered studies, which are of insufficient impact to influence molecular diag-

nostics. Taken together, what has emerged from clinical trials in breast and other

cancers along with findings from in vitro studies is that EGFR inhibition is not

similar to HER2/neu-targeted therapy in breast cancer, where a single, well-

characterized molecular event can be used to target treatment. Therefore, what

constitutes a good predictive marker for EGFR-targeted therapy in breast cancer

is currently not clear. For future clinical trials with EGFR inhibitors to be suc-

cessful and current, approved EGFR inhibitors to be administered to the right

patients, it will be necessary to understand better the biology of EGFR-directed

signaling and the relative importance of mutations, gene copy number, and



downstream signaling pathways have on therapeutic response. These findings

can then be used to setup adequately designed molecular pathology and trans-

lational studies within clinical trials that will address predictive markers for

EGFR-targeted therapy.

Predictive Markers in Drug Development

Future clinical trials should utilize findings from HER2/neu and EGFR-targeted

therapy. We have learnt that overexpression or amplification alone does not

necessarily mean that protein can be targeted with drugs or that it is indicative of

aberrant activation. Posttranslational modifications such as phosphorylation,

ubiquitination and acetylation, and localization are critical for protein function.

Furthermore, what has become clear is that preclinical models do not reflect the

complexity of human tumors well, and, therefore, predictive marker selection

must be validated in clinical settings. Despite all these challenges, drug devel-

opment must try and entail predictive marker development so that stratification

of patients can be made possible, thus avoiding unnecessary side effects and

improving the chance of a trial being successful. Without an increased focus on

predictive marker–directed patient selection, many of the drugs currently in

clinical trials run the risk of failing to reach the clinic despite being efficacious

for a subpopulation of patients. Therefore, it should be cost effective for phar-

maceutical companies to build predictive marker strategies in drug development,

as this may improve the success rate of drug development in clinicals. In

addition, it would be useful to develop predictive markers of response to many of

the drugs already in the market, thus improving patient care and reducing

treatment costs by only treating patients who benefit from the treatment.

The major risk we face at present is that we will continue to use poorly

designed retrospective analyses of existing trial populations to seek to identify

markers on the basis of incomplete understanding of the biology of drug

response. This is, in part, due to the fact that we have no reference criteria

against which to judge biomarker development. For ERa, despite no level-1

evidence being available, it is unthinkable that this biomarker be regarded as

anything less than essential to the management of breast cancer in the twenty-

first century. For HER2, only after the central importance of this biomarker was

established were steps put in place to produce robust, accurate, and clinically

viable test methods. However, clearly for future biomarkers, a structured

approach will improve the likelihood of selecting appropriate markers. For this

to succeed, we need to progress from the current model to one where, as with

drug development, critical attention is paid to each step of the process by which

novel biomarkers are developed (56).

To develop and validate predictive markers, a number of hurdles have to be

passed before they can be considered for clinical practice. These processes should

be addressed in a structured manner to maximise the possibility of success. For

example, the development of robust and accurate diagnostic assays should



precede, not follow, the identification of a potential clinical role for biomarkers

(56). To illustrate this, we have adapted a biomarker development scheme from

Pepe et al. and Hayes et al. (57,58) into four phases (Figure 2).

l Phase 1: Discovery stage. In order to discover predictive markers of

response, high-throughput gene expression microarrays or proteomic

technologies are often used. The primary output of this phase is to develop

a set of potential predictive markers that allow selecting patients who

respond to treatment. Ideally, samples from the neoadjuvant setting should

be used so that information about the influence that treatment has on the

markers selected can be obtained.l Phase 2: Development stage. In this stage, the predictive marker assay now

needs make the transition from the research setting to the clinical diag-

nostic laboratory. At this stage, it is required to develop the assay for

correct determination of predictive marker of response to a level where it

Figure 2 Development of IHC- and FISH-based predictive markers. Abbreviations: IHC,

immunohistochemical; FISH, fluorescence in situ hybridization.



can be used by clinical institutions, according to guidelines put forward by

ASCO and EORTC (The European Organization for Research and

Treatment of Cancer) (58,59). This means that proper instructions need to

be in place for all aspects of predictive marker testing, including sample

handling and pretreatment, protocols and standards for ensuring uniform

testing across institutions, and external quality assurance. It is desirable

that a set of positive controls are developed for interlaboratory variation to

be accounted for in the same fashion as the HercepTest comes with pos-

itive control cell lines.l Phase 3: Validation phase. In order to validate markers found in phase 1,

retrospective validation studies need to be set up. When setting up clinical

trial at this stage, it is important that one of the endpoints of the trial is to

assess predictive markers. One of the primary outcomes of this phase is to

evaluate how well the predictive markers perform in independent

validation studies and what predictive markers perform compared better

with traditional diagnostic markers such as nodal status, grade, etc.l Phase 4: Clinical utility phase. If success is obtained in phase 3, the marker

can be move to phase 4, with clinical studies utilizing randomized

prospective screening with predictive markers. The primary outcome of

phase 4 is to evaluate if patients selected with predictive markers were the

ones that benefited from treatment and if it is ethically defendable to select

patients on the basis of the results of the clinical studies. The predictive

marker then needs to go through regulatory approval for the defined

indication before being applied in the clinic.

Predictive marker development requires the joint effort of multiple aca-

demic institutions and industry to cover all the aspects of development. It takes

many years and large investments, and the risk for failure is high because of the

nature of drug development. That may be the reason why so few new markers

have been introduced in recent years despite major research and development

efforts and new techniques.

Future Perspectives

Cancer is a disease that involves the complex interplay of multiple genes in a

framework of mutations in the cancer cell. In addition, tumor cells have inter-

actions with other cells, thereby further promoting growth and invasiveness.

Patients respond differently to treatment based on their tumor characteristics.

Despite recent challenges, we foresee the strong likelihood that current research

and drug development will introduce more predictive markers into IHC labo-

ratories in the same fashion as trastuzumab has resulted in HER2/neu being

routinely tested for (see above). This foresight is based on the premise that more

care and more robust approaches are applied to the development of predictive

biomarkers. It is likely that within two to five years, a panel of markers will



be tested on every tumor sample in order to decide optimal therapy regimen.

Currently only three predictive IHC markers (ERa, PR, and HER2/neu) are tested

for routinely in breast, but already some pathologists use additional markers to

inform decisions, e.g., p53, Ki67, and TOP2A. As more and more predictive IHC

markers are introduced in clinical laboratories, this will lead to an increased

pressure on pathology laboratories to deliver fast and accurate decisions. How can

this be achieved and what technological advances can improve IHC testing?

Current IHC testing in pathology laboratories is subjective and semi-

quantitative, as slides are scored manually and using chromogenic dyes (56–61).

The use of computer-assisted analysis of predictive markers has been proven to

reduce slide-scoring variability among pathologists (31). While FISH/CISH

analysis is more robust, there remains room for improvement in this area also.

We foresee the introduction of image analysis as a standard tool for IHC scoring,

which will begin the process of moving the analytical phase from the microscope

to the computer screen.

As analytical challenges increase, accurate quantitation will become

increasingly important. Already evidence suggests that additional information

could be derived if it were possible to robustly quantify ERa and Ki67. There is,

therefore, a strong opportunity for systems such as the Automated QUantitative

Analysis (AQUA1) from HistoRX, Inc., New Haven, Connecticut, U.S., to pave

the way for a switch from traditional chromogenic staining techniques to fluo-

rescent staining techniques which are potentially quantitative (wide dynamic

range) (62). A further advantage is that fluorescent labels can be multiplexed

together, yielding spatial information and quantization of proteins in different

compartments of the cell. New techniques being developed, such as the prox-

imity ligation in situ assay, show promising results for quantitative analysis of

specific protein interactions in FFPE, i.e., formalin-fixed, paraffin-embedded,

tissue (63). In some cases, this may be superior to the measurement of protein

concentrations or posttranslational modifications, as most proteins take part in

multiprotein complexes and their inclusion in the complex indicates relevant

activity. In addition, several protein-interaction inhibitors are currently in clinical

trials, and, therefore, measurement of the amount of target (interacting proteins) in

the tissue may be a predictive marker for responsiveness to the drug.

There remains a key question: Will we be testing protein and DNA

markers on slides at all in 5 to 10 years? There are a number of developments in

process now miniaturizing assays onto small chips that will measure hundreds of

protein or DNA markers quantitatively and in a reproducible fashion. Such

technological advances will change the pathology workflow in the future so that

every cancer sample coming into the clinic will be analyzed by the pathologist,

tumor cells microdissected into solution, and DNA, protein, and RNA extracted

and applied onto analytical chips. The readout from these analyses could then be

used to determine optimal therapy regimen for the patient and hopefully cure the

disease. However, until these techniques have been properly validated and

developed for clinical use and trials have proven their value, we are likely to rely



on traditional in situ techniques for protein analysis (IHC) and molecular

determination of gene amplifications (FISH/CISH).

In summary, introduction of quality control, automated platforms, and

computer-assisted scoring will lead to higher throughput and more accurate

decisions with regard to treatment and facilitate the development of future

predictive markers.

CONCLUSION

As our understanding of cancer as a disease progresses and the tools to inves-

tigate cancer biology develops, the complexity of the disease is unraveling.

Breast cancers are now being classified not only on the basis of traditional

pathology traits but also according to molecular phenotype and, more recently,

gene expression profiles. New molecular targeted therapeutics will lead to more

predictive markers being introduced, and these markers must be prospectively

validated, internally and externally quality assured, and rigorously quantitative to

allow correct decisions for treatment. Therefore, an evolution of the pathology

laboratory is likely to take place in the near future, leading to a switch to

fluorescent-based techniques, digital imaging of slides, and computerized scor-

ing of predictive markers. On a longer time horizon, new technologies being

developed and applied are likely to change cancer diagnostics so that molecular

phenotyping of patients will take center stage in patient management. In the not

too distant future, there is the possibility that pathologists will type and stage

tumors and simultaneously identify key areas of tissue for automated micro-

dissection and processing. The subsequent process of molecular phenotyping,

which may include kinomic, proteomic, and transcriptomic assays performed

simultaneously could be carried out in specialised laboratories using robotics and

microfluidics on high-density protein and gene arrays to determine optimal

treatment regimens for tumors with specific molecular profiles linked to drug

response. Such a future vision may represent the more radical view of what could

be accomplished by linking of current concepts and technologies to maximize

patient benefit from therapeutic options. However, we must dream of such

futures if we are to maximize progress in the area of molecular pathology for the

benefit of the most important people, our patients.

REFERENCES

1. Edwards BK, Brown ML, Wingo PA, et al. Annual report to the nation on the status

of cancer, 1975-2002, featuring population-based trends in cancer treatment. J Natl

Cancer Inst 2005; 97(19):1407–1427.

2. Duffy MJ. Predictive markers in breast and other cancers: A review. Clin Chem

2005; 51(3):494–503.

3. Osborne CK. Steroid hormone receptors in breast cancer management. Breast Cancer

Res Treat 1998; 51(3):227–238.



4. Jensen EV, DeSombre ER. Mechanism of action of the female sex hormones. Annu

Rev Biochem 1972; 41:203–230.

5. Gorski J, Toft D, Shyamala G, et al. Hormone receptors: studies on the interaction of

estrogen with the uterus. Recent Prog Horm Res 1968; 24:45–80.

6. Osborne CK, Yochmowitz MG, Knight WA III, et al. The value of estrogen and

progesterone receptors in the treatment of breast cancer. Cancer 1980; 46(12 suppl):

2884–2888.

7. Horwitz KB, McGuire WL. Predicting response to endocrine therapy in human

breast cancer: a hypothesis. Science 1975; 189(4204):726–727.

8. Elledge RM, Green S, Pugh R, et al. Estrogen receptor (ER) and progesterone

receptor (PgR), by ligand-binding assay compared with ER, PgR and pS2, by

immuno-histochemistry in predicting response to tamoxifen in metastatic breast

cancer: a Southwest Oncology Group Study. Int J Cancer 2000; 89(2):111–117.

9. Ravdin PM, Green S, Dorr TM, et al. Prognostic significance of progesterone

receptor levels in estrogen receptor-positive patients with metastatic breast cancer

treated with tamoxifen: results of a prospective Southwest Oncology Group study.

J Clin Oncol 1992; 10(8):1284–1291.

10. Bardou VJ, Arpino G, Elledge RM, et al. Progesterone receptor status significantly

improves outcome prediction over estrogen receptor status alone for adjuvant endocrine

therapy in two large breast cancer databases. J Clin Oncol 2003; 21(10): 1973–1979.

11. Jordan VC, Brodie AM. Development and evolution of therapies targeted to the

estrogen receptor for the treatment and prevention of breast cancer. Steroids 2007;

72(1):7–25.

12. Allred DC, Harvey JM, Berardo M, et al. Prognostic and predictive factors in breast

cancer by immunohistochemical analysis. Mod Pathol 1998; 11(2):155–168.

13. Rhodes A, Jasani B, Barnes DM, et al. Reliability of immunohistochemical dem-

onstration of oestrogen receptors in routine practice: interlaboratory variance in the

sensitivity of detection and evaluation of scoring systems. J Clin Pathol 2000; 53(2):

125–130.

14. Schnitt SJ. Estrogen receptor testing of breast cancer in current clinical practice:

what’s the question? J Clin Oncol 2006; 24(12):1797–1799.



16. Coussens L, Yang-Feng TL, Liao YC, et al. Tyrosine kinase receptor with extensive

homology to EGF receptor shares chromosomal location with neu oncogene. Science

1985; 230(4730):1132–1139.

17. King CR, Kraus MH, Aaronson SA. Amplification of a novel v-erbB-related gene in

a human mammary carcinoma. Science 1985; 229(4717):974–976.

18. Yamamoto T, Ikawa S, Akiyama T, et al. Similarity of protein encoded by the human

c-erb-B-2 gene to epidermal growth factor receptor. Nature 1986; 319(6050):230–234.

19. Ross JS, Fletcher JA, Bloom KJ, et al. Targeted therapy in breast cancer: the HER-2/

neu gene and protein. Mol Cell Proteomics 2004; 3(4):379–398.

20. Carter P, Presta L, Gorman CM, et al. Humanization of an anti-p185HER2 antibody

for human cancer therapy. Proc Natl Acad Sci U S A 1992; 89(10):4285–4289.

21. Hortobagyi GN. Overview of treatment results with trastuzumab (Herceptin) in

metastatic breast cancer. Semin Oncol 2001; 28(6 suppl 18):43–47.

22. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al. Trastuzumab after adjuvant

chemotherapy inHER2-positive breast cancer. N Engl JMed 2005; 353(16):1659–1672.



23. Vogel CL, Cobleigh MA, Tripathy D, et al. Efficacy and safety of trastuzumab as a

single agent in first-line treatment of HER2-overexpressing metastatic breast cancer.

J Clin Oncol 2002; 20(3):719–726.

24. Cobleigh MA, Vogel CL, Tripathy D, et al. Multinational study of the efficacy and

safety of humanized anti-HER2 monoclonal antibody in women who have HER2-

overexpressing metastatic breast cancer that has progressed after chemotherapy for

metastatic disease. J Clin Oncol 1999; 17(9):2639–2648.

25. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal

antibody against HER2 for metastatic breast cancer that overexpresses HER2.

N Engl J Med 2001; 344(11):783–792.

26. Pegram MD, Lipton A, Hayes DF, et al. Phase II study of receptor-enhanced

chemosensitivity using recombinant humanized anti-p185HER2/neu monoclonal

antibody plus cisplatin in patients with HER2/neu-overexpressing metastatic breast

cancer refractory to chemotherapy treatment. J Clin Oncol 1998; 16(8):2659–2671.

27. Baselga J, Tripathy D, Mendelsohn J, et al. Phase II study of weekly intravenous

recombinant humanized anti-p185HER2 monoclonal antibody in patients with HER2/

neu-overexpressing metastatic breast cancer. J Clin Oncol 1996; 14(3):737–744.

28. Fornier M, Risio M, Van Poznak C, et al. HER2 testing and correlation with efficacy

of trastuzumab therapy. Oncology (Williston Park) 2002; 16(10):1340–1342.

29. Press MF, Slamon DJ, Flom KJ, et al. Evaluation of HER-2/neu gene amplification and

overexpression: comparison of frequently used assay methods in a molecularly char-

acterized cohort of breast cancer specimens. J Clin Oncol 2002; 20(14): 3095–3105.

30. Lebeau A, Deimling D, Kaltz C, et al. Her-2/neu analysis in archival tissue samples

of human breast cancer: comparison of immunohistochemistry and fluorescence in

situ hybridization. J Clin Oncol 2001; 19(2):354–363.

31. Wang S, Saboorian MH, Frenkel E, et al. Laboratory assessment of the status of

Her-2/neu protein and oncogene in breast cancer specimens: comparison of immu-

nohistochemistry assay with fluorescence in situ hybridisation assays. J Clin Pathol

2000; 53(5):374–381.

32. Moasser MM. Targeting the function of the HER2 oncogene in human cancer

therapeutics. Oncogene 2007; 26(46):6577–6592.

33. Bartlett JM. Pharmacodiagnostic testing in breast cancer: focus on HER2 and tras-

tuzumab therapy. Am J Pharmacogenomics 2005; 5(5):303–315.

34. Scaltriti M, Baselga J. The epidermal growth factor receptor pathway: a model for

targeted therapy. Clin Cancer Res 2006; 12(18):5268–5272.

35. Chan S, Scheulen ME, Johnston S, et al. Phase II study of temsirolimus (CCI-779), a

novel inhibitor of mTOR, in heavily pretreated patients with locally advanced or

metastatic breast cancer. J Clin Oncol 2005; 23(23):5314–5322.

36. Schneider BP, Sledge GW, Jr. Drug insight: VEGF as a therapeutic target for breast

cancer. Nat Clin Pract Oncol 2007; 4(3):181–189.

37. Tsutsui S, Kataoka A, Ohno S, et al. Prognostic and predictive value of epidermal

growth factor receptor in recurrent breast cancer. Clin Cancer Res 2002; 8(11):

3454–3460.

38. Johnston JB, Navaratnam S, Pitz MW, et al. Targeting the EGFR pathway for cancer

therapy. Curr Med Chem 2006; 13(29):3483–3492.

39. Moulder SL, Arteaga CL. A Phase I/II Trial of trastuzumab and gefitinib in patients

with Metastatic Breast Cancer that overexpresses HER2/neu (ErbB-2). Clin Breast

Cancer 2003; 4(2):142–145.



40. Tan AR, Yang X, Hewitt SM, et al. Evaluation of biologic end points and pharma-

cokinetics in patients with metastatic breast cancer after treatment with erlotinib, an

epidermal growth factor receptor tyrosine kinase inhibitor. J Clin Oncol 2004; 22(15):

3080–3090.

41. Baselga J, Albanell J, Ruiz A, et al. Phase II and tumor pharmacodynamic study of

gefitinib in patients with advanced breast cancer. J Clin Oncol 2005; 23(23):5323–5333.

42. Fountzilas G, Pectasides D, Kalogera-Fountzila A, et al. Paclitaxel and carboplatin as

first-line chemotherapy combined with gefitinib (IRESSA) in patients with advanced

breast cancer: a phase I/II study conducted by the Hellenic Cooperative Oncology

Group. Breast Cancer Res Treat 2005; 92(1):1–9.

43. Gasparini G, Sarmiento R, Amici S, et al. Gefitinib (ZD1839) combined with weekly

epirubicin in patients with metastatic breast cancer: a phase I study with biological

correlate. Ann Oncol 2005; 16(12):1867–1873.

44. Polychronis A, Sinnett HD, Hadjiminas D, et al. Preoperative gefitinib versus

gefitinib and anastrozole in postmenopausal patients with oestrogen-receptor positive

and epidermal-growth-factor-receptor-positive primary breast cancer: a double-blind

placebo-controlled phase II randomised trial. Lancet Oncol 2005; 6(6):383–391.

45. von Minckwitz G, Jonat W, Fasching P, et al. A multicentre phase II study on

gefitinib in taxane- and anthracycline-pretreated metastatic breast cancer. Breast

Cancer Res Treat 2005; 89(2):165–172.

46. Ciardiello F, Troiani T, Caputo F, et al. Phase II study of gefitinib in combination

with docetaxel as first-line therapy in metastatic breast cancer. Br J Cancer 2006;

94(11):1604–1609.

47. Modi S, D’Andrea G, Norton L, et al. A phase I study of cetuximab/paclitaxel in

patients with advanced-stage breast cancer. Clin Breast Cancer 2006; 7(3):270–277.

48. Chung KY, Shia J, Kemeny NE, et al. Cetuximab shows activity in colorectal cancer

patients with tumors that do not express the epidermal growth factor receptor by

immunohistochemistry. J Clin Oncol 2005; 23(9):1803–1810.

49. Vincenzi B, Santini D, Rabitti C, et al. Cetuximab and irinotecan as third-line

therapy in advanced colorectal cancer patients: a single centre phase II trial. Br

J Cancer 2006; 94(6):792–797.

50. Campiglio M, Locatelli A, Olgiati C, et al. Inhibition of proliferation and induction

of apoptosis in breast cancer cells by the epidermal growth factor receptor (EGFR)

tyrosine kinase inhibitor ZD1839 (‘Iressa’) is independent of EGFR expression level.

J Cell Physiol 2004; 198(2):259–268.

51. Uramoto H, Mitsudomi T. Which biomarker predicts benefit from EGFR-TKI

treatment for patients with lung cancer? Br J Cancer 2007; 96(6):857–863.

52. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth

factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib.

N Engl J Med 2004; 350(21):2129–2139.

53. Yun CH, Boggon TJ, Li Y, et al. Structures of lung cancer-derived EGFR mutants

and inhibitor complexes: mechanism of activation and insights into differential

inhibitor sensitivity. Cancer Cell 2007; 11(3):217–227.

54. Benvenuti S, Sartore-Bianchi A, Di Nicolantonio F, et al. Oncogenic activation of

the RAS/RAF signaling pathway impairs the response of metastatic colorectal can-

cers to anti-epidermal growth factor receptor antibody therapies. Cancer Res 2007;

67(6):2643–2648.



55. Pao W, Wang TY, Riely GJ, et al. KRAS mutations and primary resistance of lung

adenocarcinomas to gefitinib or erlotinib. PLoS Med 2005; 2(1):E17.

56. Bartlett JM, Forsyth A. Detection of HER2 gene amplification by fluorescence in

situ hybridization in breast cancer. Methods Mol Med 2006; 120:309–322.

57. Pepe MS, Etzioni R, Feng Z, et al. Phases of biomarker development for early

detection of cancer. J Natl Cancer Inst 2001; 93(14):1054–1061.

58. Hayes DF, Bast RC, Desch CE, et al. Tumor marker utility grading system:

a framework to evaluate clinical utility of tumor markers. J Natl Cancer Inst 1996;

88(20):1456–1466.

59. Sweep FC, Fritsche HA, Gion M, et al. Considerations on development, validation,

application, and quality control of immuno(metric) biomarker assays in clinical

cancer research: an EORTC-NCI working group report. Int J Oncol 2003; 23(6):

1715–1726.

60. Bartlett J, Mallon E, Cooke T. The clinical evaluation of HER-2 status: which test to

use? J Pathol 2003; 199(4):411–417.

61. Bartlett JM, Going JJ, Mallon EA, et al. Evaluating HER2 amplification and over-

expression in breast cancer. J Pathol 2001; 195(4):422–428.

62. Rimm DL. What brown cannot do for you. Nat Biotechnol 2006; 24(8):914–916.

63. Soderberg O, GullbergM, Jarvius M, et al. Direct observation of individual endogenous

protein complexes in situ by proximity ligation. Nat Methods 2006; 3(12):995–1000.



11

Circulating Tumor Cells in Individualizing

Breast Cancer Therapy

James M. Reuben

Department of Hematopathology, The University of TexasM. D. Anderson Cancer Center, Houston, Texas, U.S.A.

Massimo Cristofanilli

Department of Breast Medical Oncology, The University of TexasM. D. Anderson Cancer Center, Houston, Texas, U.S.A.

INTRODUCTION

Breast cancer is the most frequently diagnosed nonskin cancer among women

in the United States and is second only to lung cancer in causing cancer-related

deaths among them (1). The vast majority of deaths are due to recurrent

metastatic disease. Occult dissemination of tumor cells is the main cause of

recurrent metastatic breast cancer (MBC) in patients who have undergone

resection of their primary tumor (2). Approximately 5% of patients with breast

cancer have clinically detectable metastases at the time of initial diagnosis, and

a further 30% to 40% of patients who appear clinically free of metastases

harbor occult metastases (3,4). The formation of metastatic colonies is a

continuous process, commencing early during the growth of the primary tumor.

Metastasis occurs through a cascade of linked sequential steps involving

multiple host-tumor interactions. This complex process requires the cells to

enter the circulation, arrest at the distant vascular bed, extravasate into the

organ interstitium and parenchyma, and proliferate as a secondary colony.

151


Several experimental studies suggest that during each stage of the process, only

the fittest tumor cells survive (2).

The first report on tumor cells in the peripheral circulation was attributed

to Ashworth in 1869 (5). Since then, the existence, origin, and clinical signifi-

cance of circulating tumor cells (CTCs) have been debated. The introduction of

sensitive and specific immunohistochemical techniques in the late 1970s led to

renewed interest in the detection of CTCs and their possible association with

minimal residual disease in solid malignancies (6,7). However, there is little

understanding of the genetic and phenotypic characteristics of CTCs in periph-

eral blood. It has been speculated that mammary stem cells represent the cellular

origins of cancer because they exist quiescently over long periods and can

accumulate multiple mutations over the lifetime, ultimately giving rise to tumors

when stimulated to proliferate (8–10). It was recently reported that highly

tumorigenic cells possessing properties consistent with those of stem or pro-

genitor cells could be isolated from human breast cancers (9). Furthermore,

transplantation of as many as several hundred of these cells into the cleared

mammary fat pads of etoposide-treated mice resulted in the growth of breast

tumors (10). Cancer stem cells not only exist in human breast cancer but also

have different biologic features than the more differentiated cells that constitute

the majority of cells in human breast cancers. These cancer stem cells may be

detectable in the peripheral blood and bone marrow of patients with MBC and

may account for their worse prognosis. Hence the development of therapies

specifically targeting cancer stem cells may have great benefits for patients with

CTCs who possess the properties of cancer stem cells.

When CTCs migrate to distal sites, they can potentially give rise to

microscopic lesions that result in metastasis.Microscopic disease may be present in

lymph nodes, bone marrow (primary breast cancer), and peripheral blood (meta-

static disease) at the time of breast cancer diagnosis (3,4,11,12). Most studies have

demonstrated that the detection of microscopic disease in patients with breast

cancer contributes prognostic information and, in selected cases, can predict the

efficacy of treatments (12). In patients with primary breast cancer, the detection of

microscopic disease in lymph nodes and bone marrow has led to a better under-

standing of the role of minimal residual disease in establishing metastasis.

Over the past few years, immunomagnetic separation technology, with its

higher level of sensitivity and specificity, has been used to improve the detection

of CTCs compared with the detection of occult CTCs by reverse transcriptase-

polymerase chain reaction (RT-PCR) (13–21). In this chapter, we review the

methods of detecting CTCs, the prognostic implications of CTCs, and the

potential means of exploiting CTCs in developing individualized therapy.

DETECTION OF CIRCULATING TUMOR CELLS

Advances in technology have facilitated the detection of even very small num-

bers of CTCs in the peripheral blood of cancer patients. The principle of these

methods is based on the fact that breast cancer cells express epithelial cell

152 Reuben and Cristofanilli


adhesion molecule-1 (EpCAM). EpCAM, a 40-kDa glycoprotein, is present on

most epithelial carcinomas, and recent studies indicate that this molecule has a

major morphoregulatory function, relevant not only to epithelial tissue devel-

opment but also to carcinogenesis and tumor progression (22,23).

Detection of CTCs by the CellSearchTM System

The only assay approved by the U.S. Food and Drug Administration is the

CellSearchTM system (Veridex, LLC, Warren, New Jersey, U.S.). With this assay,

a small sample of peripheral blood is allowed to react with ferrofluids coated with

antibodies to EpCAM. Next, the sample is processed through a magnetic field to

retain the EpCAM-positive cells, while the EpCAM-negative cells, primarily of

hematopoietic origin, are eluted from the column and discarded. Thereafter, the

enriched EpCAM-positive cells are allowed to react with the nucleic acid dye

40,6-doamidino-2-phenylindole (DAPI), monoclonal antibodies specific for leu-

kocytes (anti-CD45 antibodies conjugated with allophycocyanin), and epithelial

cells (phycoerythrin-conjugated antibodies to cytokeratin 8, 18, and 19) and

analyzed by the CellSpotterTM Analyzer (Veridex). The CellSpotter Analyzer is a

semiautomated fluorescence-based microscopy system that enables computer-

generated reconstruction of cellular images. The analyzer interrogates each cell

image to determine if it meets the very stringent requirements of an algorithm to

be classified as a CTC. The algorithm insures that a CTC expresses EpCAM and

not leukocyte lineage-specific antigens (represented by CD45), exhibits cyto-

plasmic expression of cytokeratin, and contains a nucleus that binds DAPI.

Absence of any of these characteristics disqualifies a cell image as a CTC. In

identifying CTCs, the procedure identifies cell objects that possess some, but not

all, of the required characteristics to be a CTC and are labeled as “unclassified”

objects (24). The stringent criteria of CellSearch for identifying a CTC are

laudable; however, the merits of the assay continue to be debated as it minimizes

the scientific or clinical significance of unclassified objects.

Although CellSearch is a very reproducible and reliable method for the

identification and enumeration of CTCs in the peripheral blood of patients with

breast cancer, its clinical utility has been limited by its high cost. The semi-

automated assay is also somewhat labor intensive. In addition, the need to per-

meabilize the cell membrane of a viable cell to introduce DAPI and anticytokeratin

antibodies to label intracellular structures further deters the possibility of inter-

rogating viable cells for their growth potential, colony formation, and genomics,

among others. To address these issues, the CellProfileTM kit was introduced that

permits the retrieval of EpCAM-positive cells prior to cell permeabilization and

interrogation for genomic or proteomic profiles of the CTCs.

Isolation of CTCs by the AutoMACSTM System

EpCAM, also known as human epithelial cell antigen (HEA) or CD326, is

expressed on the majority of tumor cells of epithelial cell origin but not on

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 153


circulating B cells, T cells, or monocytes. In another assay that exploits this

characteristic of CTCs, peripheral blood mononuclear cells are incubated with

magnetic beads coated with anti-CD326 (Miltenyi Biotec, Inc., Auburn, Califor-

nia, U.S.) prior to passage through a magnetic column to enrich for CTCs. Typ-

ically, mononuclear cells are isolated from peripheral blood, bone marrow, pleural

fluid, paracentesis fluid, or other tissues, mixed with anti-CD326 microbeads, and

incubated at 48C to 88C for 15 minutes. After washing, the cell pellet is resus-

pended and loaded onto the magnetic column of an automatic magnetic cell

sorting (AutoMACSTM) system (Miltenyi Biotec) (25). CD326-positive CTCs are

isolated using the positive selection protocol and can then be centrifuged onto

glass slides to determine the morphology, viability, and purity of the preparation.

In one study, Papanicolaou (Pap)-stained cytospin smears were analyzed in

detail to identify tumor cells on the basis of conventional cytologic criteria. In

cytospin slides of enriched CTCs, large cells with cytomorphological features

suspicious for tumor cells were detected, and subsequently the cytospin slides

were immunocytochemically stained with the avidin-biotin complex peroxidase

technique using pancytokeratin antibody for the identification of tumor cells

(Fig. 1A). (The immunostains of CTCs were performed and interpreted by

Dr. Savitri Krishnamurthy, Department of Pathology, The University of Texas

M. D. Anderson Cancer Center).

A second cytospin slide of CD326-positive bone marrow tumor cells from

the same patient was stained with a cocktail of anticytokeratin antibodies using

the avidin-biotin complex peroxidase technique for the identification of tumor

cells (Fig. 1B). In the majority of cases, the suspicious cells were positive for

cytokeratin, thereby confirming the presence of micrometastatic carcinoma.

In addition to bone marrow, it is possible to isolate CTCs from the

paracentesis fluid of a patient with MBC. Figure 2 shows a cluster of CTCs

Figure 1A A cytospin slide of CTCs from the bone marrow of a patient with primary

breast cancer using anti-CD326 antibody–coated microbeads and reacted with Pap stain.

The figure shows a cluster of CTCs with cellular morphology consistent with that of

epithelial cells. Abbreviations: CTCs, circulating tumor cells; Pap, Papanicolaou.



isolated from the paracentesis fluid of a patient with MBC that were reacted with

Pap stain.

Isolation of CTCs by the EasySepTM System

Another immunomagnetic cell selection system is EasySepTM (StemCell Tech-

nologies, Inc., Vancouver, Canada), which combines the specificity of mono-

clonal antibodies with the simplicity of a column-free magnetic system. In this

procedure, highly enriched cells are obtained by positively selecting the cells of

interest with antibodies to a specific cell-surface antigen. The cells targeted for

Figure 1B CD326-positive CTCs are cytokeratin-positive epithelial cells. Abbreviation:

CTCs, circulating tumor cells.

Figure 2 CD326-positive CTCs from the paracentesis fluid of a patient with MBC.

Abbreviations: CTCs, circulating tumor cells; MBC, metastatic breast cancer.



selection are cross-linked to EasySep nanoparticles via the formation of tetra-

meric antibody complexes (StemCell Technologies) in a standard centrifuge

tube, which is then placed in the EasySep magnetic chamber. The handheld

magnetic chamber is gently inverted to discard the cells that are not bound to

the specific nanoparticle complexes that adhere to the walls of the tube in the

magnetic chamber. To obtain a highly pure population, the residual cells in the

tube are rinsed twice and then harvested by removing the tube from the chamber.

Enriched CTCs can then be used for further interrogation for the differential

expression of EpCAM or other markers of interest.

Positive selection is most effective when a specimen has a generous

complement of the desired cell population. When a specimen, such as peripheral

blood, contains few or inadequate numbers of the cells of interest, a negative-

selection approach can be taken to enrich for these cells. One negative-selection

technique is the RosetteSepTM (StemCell Technologies) method that uses a

highly purified combination of mouse and rat monoclonal antibodies. When

reacted with peripheral whole blood, these antibodies bind in bispecific tetra-

meric antibody complexes directed against cell surface antigens on human

hematopoietic cells (CD2, CD16, CD19, CD36, CD38, CD45, and CD66b) and

glycophorin A on red blood cells, or erythrocytes. The unwanted cells of hem-

atopoietic origin form rosettes by cross-linking to multiple erythrocytes that can

be easily removed by performing a typical Ficoll-PaqueTM (GE Amersham-

Pharmacia Diagnostics, Uppsala, Sweden) density gradient procedure for the

separation of mononuclear cells from peripheral blood. The rosettes, free

erythrocytes, and granulocytes form a pellet, while the unlabeled, desired CTCs

are collected from the interface between the plasma and the buoyant density

medium. Figures 3A and 3B represent cytospin slides of CTCs isolated from the

Figure 3A Pap stain of a cluster of CTCs isolated from the peripheral blood of a patient

with MBC using RosetteSepTM. Abbreviations: Pap, Papanicolaou; CTCs, circulating

tumor cells; MBC, metastatic breast cancer.



peripheral blood of a patient with MBC that were reacted with Pap stain

(Fig. 3A) and anticytokeratin antibodies (Fig. 3A).

Detection of CTCs by AdnaGen’s Technology

In addition to magnetic separation methodologies, PCR greatly facilitates the

detection of occult tumor cells through the use of nucleic acid analysis. The

PCR-based assay, Adna Test Breast Cancer Select, developed by AdnaGen AG,

Langenhagen, Germany, makes use of RT-PCR to identify putative transcripts of

genes in EpCAM-positive cells that are isolated by a magnetic separation

method. This approach takes advantage of the prevailing state of the art of CTC

isolation and enhances the chance of detecting very low numbers of CTCs

or occult CTCs on the basis of their expression of tumor-associated genes.

AdnaGen’s two-step “combination-of-combinations principle” involves initially

the enrichment of CTCs using an antibody mix linked to magnetic particles (26).

Thereafter, mRNA is extracted from the CTCs and transcribed into cDNA before

being subjected to multiplex RT-PCR for the detection of tumor-associated

genes. Owing to the combination of different selection and tumor markers, both

the heterogeneity of the tumor cells and the possible therapy-induced deviations

in the expression patterns are taken into account. The AdnaGen system has high

analytical sensitivity (two cells in 5 mL of blood) and high specificity (>90%)

achieved through the combination of multimarker tumor cell enrichment and

multiplex gene expression profiling.

Figure 3B Cytokeratin-positive CTCs from the peripheral blood of a patient with MBC

using RosetteSepTM. Abbreviations: CTCs, circulating tumor cells; MBC, metastatic

breast cancer.



CLINICAL SIGNIFICANCE AND PROGNOSTIC VALUE OF CTCs in MBC

Tumor cells can be detected in the locoregional lymph nodes of women with

early breast cancer, and the presence of these cells has been shown to have a

negative effect on long-term prognosis (3,12). Despite evidence of the prognostic

value of CTCs in some studies, the detection of micrometastases was never

incorporated into cancer-staging protocols or considered a valuable clinical tool.

This may be the result of a combination of factors, such as variable antigen

expression in poorly differentiated tumors and reports of cytokeratin and epi-

thelial membrane antigen positivity in cells that are not of epithelial origin,

which demonstrated a need for more sensitive and specific methods of detection

than were available at the time. This need was filled to some degree by sensitive

PCR techniques (14,16,17). In the past decade, a few studies have shown that the

detection of occult disease by PCR has prognostic significance in some solid

tumors (13). However, PCR-based assays for the detection of occult tumor cells

have limitations, particularly contamination of samples, specificity of the assays,

and inability to quantify tumor cells. Moreover, PCR-based methods cannot be

used to perform functional assays. These factors have precluded the widespread

use of PCR in in vitro diagnostic applications.

Using an immunomagnetic detection approach, Austrup et al. reported the

prognostic significance of genomic alterations (e.g., c-erbB-2 overexpression)

present in CTCs purified from the blood of patients with breast cancer (27). The

authors investigated genomic imbalances, such as mutation, amplification, and

loss of heterozygosity, of 13 tumor suppressor genes and 2 proto-oncogenes

using DNA isolated from minimal residual cancer cells. The presence and the

number of genomic imbalances measured in disseminated tumor cells were

significantly associated with a worse prognosis (27).

Subsequently, Meng et al. demonstrated that CTCs recapitulate the human

epidermal growth factor receptor 2 (HER2) status of the primary tumor (28).

Furthermore, the authors demonstrated that a fraction of patients with HER2-

negative primary tumors had detectable HER2 gene amplification in their CTCs,

suggesting acquisition or selection of this phenotype during cancer progression

(37.5%; 95% confidence interval, 18.8–59.4%). Intriguingly, the initiation of

trastuzumab-based therapy in a few cases with altered HER2 status in CTCs was

associated with a clinical response (29). These important observations support

the argument that detection of CTCs and determination of their gene amplifi-

cation or expression can be used for better tailoring of therapies.

The predictive and prognostic roles of CTCs were investigated in a pro-

spective multicenter clinical trial led by researchers at the University of Texas

M. D. Anderson Cancer Center (11). In this study, the CellSearch system was

used to prospectively determine the prognostic and predictive value of CTCs in

patients with MBC who were about to start a new systemic treatment. The 177

enrolled patients underwent peripheral blood collection at monthly intervals for

up to six months after enrollment. A cutoff of 5 CTCs/7.5 mL was used to



stratify patients into positive and negative groups (positive, �5 CTCs/7.5 mL;

negative, <5 CTCs/7.5 mL). The investigators reported that patients classified as

positive had shorter progression-free survival times (2.7 months vs. 7.0 months;

P ¼ 0.0001) and shorter overall survival times (10.9 months vs. 21.9 months;

P < 0.0001) than did those classified as negative. Furthermore, the CTC status at

first follow-up after initiation of therapy (three weeks) had an even greater

association with progression-free survival time (2.1 months versus 7.0 months;

P < 0.0001) and overall survival time (8.2 months versus >18 months; P <0.0001). On multivariate Cox hazards regression analysis, CTC levels, both at

baseline and at first follow-up, were the most significant predictors of progres-

sion-free and overall survival (11).

An analysis restricted to patients who were about to start first-line systemic

therapy after the diagnosis of MBC had similar findings (30). A subsequent

investigation including patients with either measurable or evaluable MBC con-

firmed that CTCs are an independent predictor of survival that is more powerful

than standard prognostic measures, such as hormone receptor status, and mea-

sures of tumor burden (e.g., the Swenerton score or level of CA27-29) (31).

The detection of changes in CTC status may also have predictive utility.

The CTC detection rate at first follow-up was lower than at baseline, partic-

ularly in patients undergoing first-line treatment (25% versus 52%) and those

with visceral disease (28% versus 50%) (11). In patients with newly diagnosed

disease treated with chemotherapy alone (n ¼ 37), the percentage of patients

who were CTC positive decreased significantly from baseline to first follow-up

(57% to 37%, P ¼ 0.001) and first imaging-visit (at eight to nine weeks) blood

draw (57% to 31%, P ¼ 0.004) (31). More importantly, in patients receiving

trastuzumab-containing regimens, CTC-positive patients decreased from

baseline to first follow-up (59% to 7%, P ¼ 0.600) and first imaging-visit

blood draw (59% to 0%, P ¼ not applicable). This data indicates that patients

with newly diagnosed disease who are about to start first-line therapy can have

detectable CTCs and that the changes in CTC status at three to four weeks may

indicate a benefit from systemic treatment (particularly trastuzumab-based

regimens). Moreover, the determination of CTCs at baseline and follow-up

appears to be a superior prognostic marker in patients with measurable MBC

compared with standard imaging assessments (32). CTCs could be used for

assessment of the clinical benefit of systemic treatments, for prognostic

stratification, and for evaluation of patients with nonmeasurable disease in

future prospective validation trials.

STAGING MBC USING CTCs

The American Joint Committee on Cancer (AJCC) classification system, based

on a schema developed by the International Union Against Cancer (UICC)

(33–35), includes much of the traditional prognostic information used by clini-

cians when developing a comprehensive treatment plan. In brief, the AJCC



system attempts to define the disease by incorporating all aspects of cancer

distribution in terms of the primary tumor (T), lymph nodes (N), and distant

metastasis (M). A “stage” group (0, I, II, III, or IV) is then assigned on the basis

of the possible TNM permutations, with 0 reflecting minimal involvement and

IV either the maximum tumor involvement or distant metastasis. This basic

TNM staging is then further subdivided according to the time the evaluation is

performed: c ¼ clinical (before surgical treatment) and p ¼ pathologic (after

surgical specimen analysis) (10,35).

Over the past 45 years, the AJCC has regularly updated its staging stan-

dards to incorporate advances in prognostic technology. The committee’s current

work concerns the development of prognostic indices based on molecular

markers. However, until these changes are incorporated, TNM staging continues

to quantify only the physical extent of the disease. Although it covers the

approximately 2% of women who present with metastatic disease (stage IV), the

prognostic information is applicable only to in situ, local, and regional primary

breast cancers. Given the heterogeneity of the disease in primary and metastatic

settings, the potential for continued mutation, and the variety of treatment

options available, the information acquired at the time of the initial diagnosis of

breast cancer may not be as relevant to planning the treatment for recurrence in

the approximately 30% of women who have recurrent MBC years after their

initial diagnosis.

The recent demonstration that the presence of CTCs predicts the prognosis

of two subgroups of patients with MBC raises the possibility that this method

will allow for a true “biologic staging” of breast cancer (e.g., stages IVA and

IVB) (11). The main limitations of the previous study were the sample size (only

177 patients); the inclusion only of patients with measurable disease, which does

not reflect the heterogeneity of MBC; and the inclusion of patients at different

points in treatment—those with newly diagnosed disease and those undergoing

second- or third-line treatment. To overcome these limitations, a larger inter-

national validation trial, the International Stage IV Stratification Study, was

begun. The objective of this study was to confirm the prognostic value of CTC

detection in a larger and more homogeneous cohort of patients with newly

diagnosed MBC. The study aimed at enrolling 660 patients and providing a more

detailed analysis of the association between CTC detection and other factors

such as ethnicity (black compared with white and Hispanic groups) and specific

disease sites (visceral versus nonvisceral disease).

CLINICAL APPLICATION OF CTCs

Despite years of clinical research, the odds of patients with MBC achieving a

complete response remain extremely low. Thus, the main goal in the manage-

ment of MBC is palliation (36). Only a few patients who experience a complete

response after chemotherapy remain in this state for prolonged periods, although



some have remained in remission for more than 20 years (37). These long-term

survivors are usually young, have an excellent performance status, and, more

importantly, have limited metastatic disease (37,38). Most patients with MBC

respond only transiently to conventional therapies and develop evidence of

progressive disease within 12 to 24 months of starting treatment. For these

patients, systemic treatment does not result in a significant improvement in

survival time but may improve their quality of life.

At present, clinicians use three different systemic treatment modalities for

advanced breast cancer: endocrine therapy, chemotherapy, and biologic targeted

therapy (39–42). Appropriate selection of patients for these modalities is based

mostly on tissue assessment of hormone receptor status (estrogen and proges-

terone receptors) and c-erbB-2 status. In patients who lack expression of hor-

mone receptors or who demonstrate no amplification of c-erbB-2, cytotoxic

chemotherapy is used. However, no standard therapeutic regimen has been

defined. For example, the optimal schedule of chemotherapy administration in

MBC (i.e., concurrent vs. sequential) remains controversial, and the decision

must be individualized. Sledge et al. addressed this issue in a prospective study

that included 739 chemotherapy-naıve patients who were randomly assigned to

receive, at progression, doxorubicin, paclitaxel, or both (39). Although the

response rates and times to treatment failure were improved with the combina-

tion regimen, the overall survival rate was comparable in all groups. Other trials

have demonstrated a survival advantage with combination regimens over single-

agent chemotherapy, but all of these studies have shown differences in the actual

median overall survival time between 10 and 13 months across studies, sug-

gesting that even with comparable inclusion criteria, the heterogeneity that is

typical of MBC cannot be eliminated (41,42). This heterogeneity is indeed one

of the major limitations to the development of more personalized treatments for

patients. Therefore, although the data demonstrate that patients can benefit from

combination therapy, they do not clearly identify the subsets of patients who will

most benefit or who will experience only additional toxicity.

In this context, the use of CTCs to stratify patients at the time of disease

recurrence may be an appropriate way to design personalized therapeutic

approaches. CTC detection may enable a more rational selection of treatments

for patients with newly recurrent disease, and this approach could maximize

the chance of a particular combination or single new drug showing clinical

benefit and, eventually, prolonging survival. It is possible that CTC detection

could be used in the design of efficacy trials of different therapeutic

approaches. The efficacy of these treatments could be more easily assessed if

patients were stratified by their prognosis, leading to more tailored treatment

strategies. We believe that the challenge for the next generation of clinical

trials, and the responsibility for both clinical investigators and the pharma-

ceutical industry will be to incorporate these concepts into the process of drug

development.



FUTURE USES OF CTCs

The process of sorting cancer cells from other cellular components (e.g., blood

and stromal cells) in clinical samples is fundamentally important for the future of

genomic and proteomic analysis (43–45). Collecting representative tissue from

solid tumor metastases usually requires more invasive procedures that increase

the risk of complications and discomfort. Furthermore, these procedures may not

provide an adequate specimen for detailed analysis and typically cannot be

repeated for dynamic evaluation of the biologic changes during treatments.

Theoretically, CTC detection would allow specific genes [e.g., c-erbB-2, epider-

mal growth factor receptor (EGFR), and mammaglobin B (MGB)] or more global

gene expression to be analyzed while using specific targeted treatments for MBC

based on the expression of the CTCs (29,46–48). This information could then be

used to design specific treatments that more appropriately reflect the dynamics and

heterogeneity of MBC.

CONCLUSIONS

The detection of microscopic disease in the peripheral blood of patients with

MBC provides prognostic information. This information will allow appropriate

risk stratification and modification of the current staging system for advanced

disease. Furthermore, the intriguing hypothesis that CTCs are, at least in part,

representative of tumorigenic cancer stem cells may present an additional

challenge for translational research. In fact, the phenotypic evaluation of CTCs is

indicating the possibility of more critical design of targeted therapies in the

setting of MBC. Moreover, with the introduction of this novel concept, it is

expected that investigators and the pharmaceutical industry will derive benefits

from these sorting technologies that will contribute to more tailored treatments

and sophisticated trial designs.

REFERENCES

1. Surveillance Epidemiology and End Results (SEER). National Cancer Institute

Website. Available at: http://seer.cancer.gov/statfacts/html/breast.htm.

2. Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med 1971; 285

(21):1182–1186. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=

Retrieve&db=PubMed&dopt=Citation&list_uids=4938153.

3. Clare SE, Sener SF, Wilkens W, et al. Prognostic significance of occult lymph node

metastases in node-negative breast cancer. Ann Surg Oncol 1997; 4(6):447–451.

Available at http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=

PubMed&dopt=Citation&list_uids=9309332.

4. Braun S, Pantel K, Muller P, et al. Cytokeratin-positive cells in the bone marrow

and survival of patients with stage I, II, or III breast cancer. N Engl J Med 2000; 342(8):

525–533. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve

&db=PubMed&dopt=Citation&list_uids=10684910.



5. Ashworth T. A case of cancer in which cells similar to those in the tumours were

seen in the blood after death. Aust Med J 1869; 14:146.

6. Roberts S, Watne A, McGrath R. Technique and results of isolation of cancer cells

from the circulating blood. Am Med Assoc Arch Surg 1958; 76:334–346.

7. Long L, Jonasson O, Roberts S, et al. Cancer cells in blood: results of a simplified

isolation technique. Arch Surg 1960; 80:910–919. Available at http://www.ncbi.nlm.

nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=

14418112.

8. Dontu G, Al-Hajj M, Abdallah WM, et al. Stem cells in normal breast development

and breast cancer. Cell Prolif 2003; 36(suppl 1):59–72. Available at: http://www

.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_

uids=14521516.

9. Al-Hajj M, Wicha MS, Benito-Hernandez A, et al. Prospective identification of

tumorigenic breast cancer cells. Proc Natl Acad Sci U S A 2003; 100(7):3983–3988.

Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=Pub

Med&dopt=Citation&list_uids=12629218.

10. Al-Hajj M, Becker MW, Wicha M, et al. Therapeutic implications of cancer stem

cells. Curr Opin Genet Dev 2004; 14(1):43–47. Available at: http://www.ncbi.nlm.


15108804.

11. Cristofanilli M, Budd GT, Ellis MJ, et al. Circulating tumor cells, disease progres-

sion, and survival in metastatic breast cancer. N Engl J Med 2004; 351(8):781–791.



12. Reed W, Bohler PJ, Sandstad B, et al. Occult metastases in axillary lymph nodes as a

predictor of survival in node-negative breast carcinoma with long-term follow-up.

Breast J 2004; 10(3):174–180. Available at: http://www.ncbi.nlm.nih.gov/entrez/

query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15125740.

13. An XY, Egami H, Hayashi N, et al. Clinical significance of circulating cancer cells in

peripheral blood detected by reverse transcriptase-polymerase chain reaction in

patients with breast cancer. Tohoku J Exp Med 2001; 193(2):153–162. Available at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=

Citation&list_uids=11318030.

14. Bossolasco P, Ricci C, Farina G, et al. Detection of micrometastatic cells in breast cancer

by RT-pCR for the mammaglobin gene. Cancer Detect Prev 2002; 26(1):60–63.



15. Gross HJ, Verwer B, Houck D, et al. Model study detecting breast cancer cells in

peripheral blood mononuclear cells at frequencies as low as 10(-7). Proc Natl Acad

Sci U S A 1995; 92(2):537–541. Available at: http://www.ncbi.nlm.nih.gov/entrez/


16. Hawes D, Neville AM, Cote RJ. Methods used for the detection of occult metastasis.

Biomed Pharmacother 2001; 55(4):229–242. Available at: http://www.ncbi.nlm.nih.

gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=

11393810.

17. Hu XC, Chow LW. Detection of circulating breast cancer cells with multiple-marker

RT-PCR assay. Anticancer Res 2001; 21(1A):421–424. Available at: http://www



.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation

&list_uids=11299772.

18. Martin KJ, Graner E, Li Y, et al. High-sensitivity array analysis of gene expression

for the early detection of disseminated breast tumor cells in peripheral blood. Proc

Natl Acad Sci U S A 2001; 98(5):2646–2651. Available at http://www.ncbi.nlm.nih.


11226293.

19. Pachmann K, Heiss P, Demel U, et al. Detection and quantification of small numbers

of circulating tumour cells in peripheral blood using laser scanning cytometer (LSC).

Clin Chem Lab Med 2001; 39(9):811–817. Available at http://www.ncbi.nlm.nih.


11601678.

20. Racila E, Euhus D, Weiss AJ, et al. Detection and characterization of carcinoma

cells in the blood. Proc Natl Acad Sci U S A 1998; 95(8):4589–4594. Available at



21. Witzig TE, Bossy B, Kimlinger T, et al. Detection of circulating cytokeratin-positive

cells in the blood of breast cancer patients using immunomagnetic enrichment and digital

microscopy. Clin Cancer Res 2002; 8(5):1085–1091. Available at http://www.ncbi.nlm.


12006523.

22. Kagan M, Howard D, Bendele T, et al. A sample preparation and analysis system for

identification of circulating tumor cells J Clin Ligand Assay 2002; 25(1):104–110.

23. Terstappen LW, Rao C, Gross S, et al. Peripheral blood tumor cell load reflects the

clinical activity of the disease in patients with carcinoma of the breast. Int J Oncol

2000; 17(3):573–578. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?

cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=10938400.

24. Riethdorf S, Fritsche H, Muller V, et al. Detection of circulating tumor cells in

peripheral blood of patients with metastatic breast cancer: a validation study of the

CellSearch system. Clin Cancer Res 2007; 13(3):920–928. Available at: http://

clincancerres.aacrjournals.org/cgi/content/abstract/13/3/920.

25. Taubert H, Blumke K, Bilkenroth U, et al. Detection of disseminated tumor cells in

peripheral blood of patients with breast cancer: correlation to nodal status and

occurrence of metastases. Gynecol Oncol 2004; 92:256–261.

26. Lankiewicz S, Gutierrez B, Bocher O. Quantitative real-time RT-PCR of dis-

seminated tumor cells in combination with immunomagnetic cell enrichment. Mol

Biotechnol 2006; 34:15–27.

27. Austrup F, Uciechowski P, Eder C. Prognostic value of genomic alterations in

minimal residual cancer cells purified from the blood of breast cancer patients. Br J

Cancer 2001; 83:1664–1673.

28. Meng S, Tripathy D, Shete S, et al. HER-2 gene amplification can be acquired as

breast cancer progresses. Proc Natl Acad Sci U S A 2004; 101(25):9393–9398.



29. Meng S, Tripathy D, Shete S, et al. uPAR and HER-2 gene status in individual

breast cancer cells from blood and tissues. Proc Natl Acad Sci U S A 2006; 103

(46):17361–17365. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=

Retrieve&db= PubMed&dopt=Citation&list_uids=17079488.



30. Cristofanilli M, Hayes DF, Budd GT, et al. Circulating tumor cells: a novel prog-

nostic factor for newly diagnosed metastatic breast cancer. J Clin Oncol 2005; 23(7):

1420–1430. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=

Retrieve&db=PubMed&dopt=Citation&list_uids=15735118.

31. Cristofanilli M, Broglio KR, Guarneri V, et al. Circulating tumor cells in metastatic

breast cancer: biologic staging beyond tumor burden. Clin Breast Cancer 2007; 7(6):



32. Budd GT, Cristofanilli M, Ellis MJ. Circulating tumor cells versus imaging:

Predicting overall survival in metastatic breast cancer. Clin Cancer Res 2006; 12(21):

6403–6409.

33. Internation Union Against Cancer (UICC). TNM Classification of Malignant Tumours.

1st ed. Geneva, Switzerland: UICC, 1968.

34. Denoix P. Tumor, node, and metastasis (TNM). Bull Inst Nat Hyg (Paris) 1944; 1:1–69.

35. Berndt H, Titze U. TNM clinical stage classification of breast cancer. Int J Cancer

1969; 4(6):837–844. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?

cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=5378609.

36. Greene FL, Page DI, Fleming ID, eds. AJCC Cancer Staging Manual. 6th ed. New

York, NY: Springer-Verlag, 2002.

37. Ellis MJ, Hayes DF, Lippman ME. Treatment of Metastatic Disease. Philadelphia,

PA: Lippincott-Raven, 2000.

38. Swenerton KD, Legha SS, Smith T, et al. Prognostic factors in metastatic breast

cancer treated with combination chemotherapy. Cancer Res 1979; 39(5):1552–1562.



39. Bernard-Marty C, Cardoso F, Piccart MJ. Facts and controversies in systemic

treatment of metastatic breast cancer. Oncologist 2004; 9(6):617–632. Available at:



40. Sledge GW, Neuberg D, Bernardo P, et al. Phase III trial of doxorubicin, paclitaxel,

and the combination of doxorubicin and paclitaxel as front-line chemotherapy for

metastatic breast cancer: an intergroup trial (E1193). J Clin Oncol 2003; 21(4):



41. O’Shaughnessy J, Miles D, Vukelja S, et al. Superior survival with capecitabine plus

docetaxel combination therapy in anthracycline-pretreated patients with advanced

breast cancer: phase III trial results. J Clin Oncol 2002; 20(12):2812–2823. Available

at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=


42. O’Shaughnessy J, Nag S, Calderillo-Ruiz G. Gemcitabine plus paclitaxel versus

paclitaxel as first-line treatment for anthracycline pre-treated metastatic breast cancer:

interim results of a global phase III study. Proc Am Soc Clin Oncol 2003; 22:7.

43. Hancock W, Apffel A, Chakel J, et al. Integrated genomic/proteomic analysis. Anal

Chem 1999; 71(21):742A–748A. Available at: http://www.ncbi.nlm.nih.gov/entrez/


44. Marshall A, Hodgson J. DNA chips: an array of possibilities. Nat Biotechnol 1998; 16(1):

27–31. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&

db=PubMed&dopt=Citation&list_uids=9447589.



45. Page MJ, Amess B, Townsend RR, et al. Proteomic definition of normal human

luminal and myoepithelial breast cells purified from reduction mammoplasties. Proc

Natl Acad Sci U S A 1999; 96(22):12589–12594. Available at: http://www.ncbi.nlm.


10535966.

46. Smirnov DA, Zweitzig DR, Foulk BW, et al. Global gene expression profiling of

circulating tumor cells. Cancer Res 2005; 65(12):4993–4997. Available at: http://www

.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&

list_ uids=15958538.

47. Speicher MR, Carter NP. The new cytogenetics: blurring the boundaries with

molecular biology. Nat Rev Genet 2005; 6(10):782–792. Available at: http://www

.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&

list_uids=16145555.

48. Reuben JM, Lee BN, Li C. Genomic of circulating tumor cells in metastatic breast

cancer. Proc ASOC 25:32s (abstr 1002).



12

Individualization of Endocrine Therapy

in Breast Cancer

Amelia B. Zelnak and Ruth M. O’Regan

Emory Winship Cancer Institute, Atlanta, Georgia, U.S.A.

Clodia Osipo

Loyola University, Maywood, Illinois, U.S.A.

INTRODUCTION

Discovery of estrogen receptors (ER) was critical for the development of

endocrine therapy in breast cancer. Expression of ERa, the predominant iso-

form, in breast tumors of both premenopausal and postmenopausal women is a

highly predictive marker for response to antiestrogen treatment in women with

ERa-positive breast cancer. Today, endocrine therapies include the use of

antiestrogens such as tamoxifen and aromatase inhibitors such as anastrozole,

exemestane, and letrozole. Tamoxifen, the first antiestrogen to be used for the

treatment of ERa-positive breast cancer, competitively blocks the actions of

17b-estradiol (E2), the female hormone that binds and activates ERa in tumors.

In postmenopausal women, peripheral aromatization of androgens to estrogens

is the major source of plasma estrogen. Aromatase inhibitors inhibit this

reaction and consequently suppress the production of circulating estrogen in

postmenopausal women. Endocrine therapies are effective at reducing recur-

rence, increasing overall survival, and reducing contralateral breast cancer up

to 50%. However, about 50% of patients with ERa-positive breast cancer have

167


intrinsic resistance to antiestrogen therapy and therefore do not benefit. In

contrast to patients with intrinsically resistant tumors, there are patients who do

initially respond to antiestrogen therapy; however, most of these patients

develop acquired resistance during the treatment regimen. Therefore, the

current goal in breast cancer research is to elucidate the mechanisms of both

intrinsic and acquired resistance to tamoxifen and the aromatase inhibitors in

order to develop new therapeutic strategies to prevent and/or treat resistant

breast cancer.

ANTIESTROGENS

Nonsteroidal antiestrogens were initially developed as contraceptives in the

1960s. Walpole and colleagues synthesized tamoxifen (termed ICI 46, 474), a

potent antiestrogen with antifertility properties in rats. However, in humans,

tamoxifen induced ovulation in subfertile women. Therefore, the development

of tamoxifen as a contraceptive was terminated. However, Walpole also

patented the application of tamoxifen as a drug treatment for hormone-

dependent cancers. Thus, clinical trials were started to evaluate tamoxifen

against the standard endocrine treatment at the time, diethylstilbestrol (DES),

for the treatment of advanced breast cancer in postmenopausal women (1).

Tamoxifen not only was as effective as DES for the treatment of advanced

breast cancer but also had fewer side effects. Therefore, the advantage of

tamoxifen over DES was crucial for its subsequent evaluation as a treatment for

all stages of breast cancer.

In 1962, Jensen and colleagues discovered ERa. Jensen demonstrated

that E2, the circulating female hormone that promoted breast cancer growth,

binds to diverse tissue sites around a woman’s body but is retained in estrogen-

target tissues, for example, the uterus and vagina (2). The identification of

ERa as the target of E2 action in the breast and antiestrogens blocking the

binding of E2 to ERa provided a therapeutic target and an approach for the

treatment of breast cancer. Although many antiestrogens were discovered and

tested during the 1960s and 1970s, only tamoxifen was considered safe enough

for extensive clinical evaluation (3). Clinical trials ultimately demonstrated

that patients with ERa-positive breast cancer benefited the most from

tamoxifen therapy, whereas women with ERa-negative breast cancer were

found to be unaffected. During subsequent clinical trials, five years of adju-

vant tamoxifen treatment was found to be more effective than less than five

years of treatment in improving time to tumor recurrence and overall survival

(4). In contrast to the beneficial effects of tamoxifen as a treatment for breast

cancer, both laboratory and clinical results showed that tamoxifen increased

the risk of endometrial cancer in the uterus fourfold in postmenopausal women

compared with untreated women (5,6). These results strongly indicated that

tamoxifen was not a pure antiestrogen but had selective functions depending

on the target tissue.

168 Zelnak et al.


Resistance to Tamoxifen

Overview

Five years of adjuvant tamoxifen therapy is a standard of care for early-stage ERa-positive breast cancer. Although there is increasing use of aromatase inhibitors in

postmenopausal women, tamoxifen is the only approved effective therapy in pre-

menopausal women. Approximately 50% of patients with ERa-positive breast

cancer benefit from tamoxifen treatment in the advanced- and early-stage settings

(4,7). Unfortunately, the other 50% of patients do not respond, and the ones that do

respond initially eventually progress. Thus, the problem with tamoxifen therapy is

that breast tumors can either be resistant (intrinsic or de novo) prior to endocrine

treatment or become resistant (acquired) during therapy. Numerous studies

have looked at the mechanisms responsible for this resistant phenotype. They

include the loss of the ERa gene and/or protein in tumors during short-term and

long-term treatment, identification of mutations within the ERa gene, changes in

the pharmacologic response of resistant breast cancer cells to tamoxifen, activation

of ERa-mediated gene transcription in the absence of ligand, modifications in gene

transcription by changes in interactions between ERa and coregulators, and the

development of enhanced surface-to-intracellular signaling cross-talk between

growth factor receptors and ERa (Fig. 1).

Figure 1 Mechanisms of hormone resistance.

Individualization of Endocrine Therapy in Breast Cancer 169


ER-Negative Breast Cancer and Resistance

Breast cancers initially lacking expression of ERa have intrinsic resistance to

endocrine therapy. In addition, loss of ERa expression upon recurrence of breast

cancer during tamoxifen therapy has been reported in as many as 25% of tumors

(8–10). However, most of the acquired resistance during adjuvant tamoxifen

treatment in ERa-positive tumors is not due to changes in ERa expression.

Interestingly, most tamoxifen-resistant breast tumors retain ERa and respond to

secondary endocrine treatments. Overall, a loss of ERa status does not seem to

be the major mechanism for the development of acquired antiestrogen resistance.

ER Mutations

Several mutant forms of both the ERa and ERb, another ER isoform, have been

identified and previously reviewed (11–13). However, the functional signifi-

cance of these mutants as a mechanism of antiestrogen resistance is yet to be

elucidated. Most of the data demonstrate the existence of splice variants for ERaand ERb mRNAs without evidence of translation into functional proteins. Most

of the tumors express both mutant and wild-type ER with the wild type being the

predominant species. A single-point mutation in the ligand-binding domain

(LBD) of ERa has been reported (D351Y) that converts tamoxifen and ralox

ifene from antiestrogen to estrogens in in vivo and in vitro models of antiestrogen-

stimulated MCF-7 tumor cells (14,15). However, the D351Y ERa mutant has yet

to be detected in either intrinsic or acquired resistant breast tumors from patients.

In addition, mutations in the F-region of the ER have been shown to affect

the activities of both E2 and 4-hydroxytamoxifen (4OHT), the active metabolite

of tamoxifen (16). Therefore, the clinical relevance of ER mutants is unclear to

date as a mechanism of resistance to antiestrogens. However, our understanding

of ER mutations in response to antiestrogens might lead to the development of

better antiestrogens or other drugs for breast cancer treatment.

Role of Pharmacogenetics

Tamoxifen undergoes extensive primary and secondary metabolism. The sec-

ondary metabolite 4-hydroxy-N-desmethyl tamoxifen (endoxifen) is formed by

the CYP2D6-mediated oxidation of N-desmethyl tamoxifen and has been shown

to play an important role in the anticancer effect of tamoxifen. CYP2D6 geno-

typing has revealed that women can be classified as either poor, intermediate, or

extensive metabolizers of tamoxifen. Women who are poor metabolizers of

tamoxifen, and, therefore, have lower levels of endoxifen, have been shown to

have worse disease-free survival. No difference in overall survival has been

shown (17). Women who are extensive metabolizers of tamoxifen have also been

shown to experience more hot flashes. A recent study showed that women who

experience hot flashes while on tamoxifen had a hazard ratio of recurrence of

170 Zelnak et al.


0.51 compared with women without hot flashes (18). This study suggests that

vasomotor symptoms and efficacy of tamoxifen may be related to a women’s

pharmacogenetics.

Role of Coregulators in Resistance

The recruitment of coactivators or corepressors to the ER determines the switch

between ER activation and repression. In addition, the coordinated action of

ligand, ER, and coregulators determines which genes are transcribed or repressed

depending on the cellular context and thus which cells will or will not proliferate.

The overall data indicate that intricate modulation of the ER-to-coregulator ratio

in breast cancer cells could determine resistance to antiestrogens. For example,

the coactivator SRC-1 activates ER in a ligand-independent manner while

increasing 4OHT’s agonist activity (19). On the other hand, the corepressor

SMRT blocks the agonist activity of 4OHT-induced by SRC-1 (20). The other

member of the p160 coactivator family is amplified in breast cancer (AIB1,

SRC-3, and RAC3). AIB1 mRNA was found to be amplified in 60% of breast

cancers, and the protein product was found to be overexpressed in about 10% of

tumors. Recently, an association was discovered between high AIB1 and human

epidermal growth factor receptor 2 (HER2/neu), a member of the epidermal

growth factor receptor (EGFR) family of receptor tyrosine kinases and tamoxifen-

resistant breast cancers (20). The potential mechanism for this resistance requires

activation of the mitogen-activated protein kinase (MAPK) signaling cascade by

HER2/neu, which in turn potentially phosphorylates and activates AIB1. Acti-

vated AIB1 in turn can convert 4OHT from an antiestrogen to an estrogen in

regards to ERa activity. In contrast, low levels of N-CoR mRNA, a corepressor,

have been implicated with resistance to tamoxifen (21). Therefore, the identi-

fication of abnormally expressed coregulators will probably assist in the design

of future therapies that improve the efficacy of endocrine treatment without the

development of resistance.

EGFR and HER2/Neu and Antiestrogen Resistance

Numerous laboratory and clinical studies indicate that overexpression and/or

aberrant activity of the HER2/neu (erbB2) signaling pathway is associated with

antiestrogen resistance in breast cancer (22). The HER2/neu receptor is a

member of the EGFR family of receptor tyrosine kinases, which include HER3

(erbB3) and HER4 (erbB4) (23–26). Upon dimerization, the tyrosine kinase

domains located within the COOH-terminal regions of receptors are activated by

an autophosphorylation cascade on specific tyrosine residues, which activate

downstream effectors, such as MAPK and Akt, which promote cellular prolif-

eration, survival, anti-apoptosis, and transformation. HER2/neu is overexpressed

and/or amplified in 25% to 30% of breast tumors and is associated with a more

aggressive phenotype and poor prognosis (27). Patients with breast tumors

overexpressing HER2/neu exhibit much lower response rates to antiestrogen



therapy (27). Thus, it is suggested that one possible mechanism of resistance to

tamoxifen is overexpression of HER2/neu in ERa-positive breast cancers. In

addition, overexpression of EGFR and its ligands are observed in several human

cancers including breast cancer (28–30). The increased expression of EGFR is

frequently associated with tumor progression and resistance to antiestrogens.

These data suggest that EGFR and/or HER2/neu are possible targets for pre-

venting or treating antiestrogen resistance in breast cancer. Strategies to target

EGFR and HER2/neu include the use of humanized monoclonal antibodies to

the receptors (31), tyrosine kinase inhibitors that block reduction of adenosine

triphosphate (ATP) to adenosine diphospohate (ADP) þ Pi, and receptor antisense

molecules (32). Gefitinib, (ZD 1839, Iressa1), an EGFR-specific tyrosine kinase

inhibitor, has been shown to inhibit growth of breast cancer cell lines in vitro that

are resistant to tamoxifen (33). More importantly, the combination of gefitinib

and tamoxifen was shown to prevent resistance (34), demonstrating that (1)

EGFR might be a key player in the development of antiestrogen resistance and

(2) inhibiting the activity of this receptor might be therapeutically beneficial in

preventing resistance to antiestrogens. In addition to gefitinib, trastuzumab

(Herceptin1

) is a humanized monoclonal antibody directed against the ectodo-

main of the HER2/neu receptor. It has been shown to restore breast cancer cell

sensitivity to tamoxifen in HER2/neu-overexpressing cells (35).

Role of MAPK

MAPK is a serine/threonine kinase that is activated by phosphorylation cascades

originating from GTP-bound Ras, a downstream effector of EGFR and HER2/

neu. MAPK has been shown to be hyperactivated in MCF-7 breast cancer cells

overexpressing either EGFR or HER2/neu (36–38). This increased activity of

MAPK promoted an enhanced association of ERa with coactivators and

decreased interaction with corepressors, thereby enhancing hormone-dependent

gene transcription and possibly leading to tamoxifen-resistant breast cancer. The

role of the MAPK signaling pathway in tamoxifen resistance has been demon-

strated primarily in vitro using specific inhibitors such as U0126 (MAPKK,

MEK1/2 inhibitor), AG1478 (EGFR and HER2/neu inhibitors), and PD98059

(MAPK, ERK1/2 inhibitor) (36–38). The exact mechanism by which hyper-

activated MAPK converts the antiestrogenic activity of the tamoxifen-ERacomplex to more of an estrogenic one in resistant breast cancer cell is yet

unclear. However, it has been demonstrated that activation of the Ras/MAPK

pathway by EGFR/HER2/neu receptor activation could phosphorylate Ser-118 in

the AF-1 domain of ERa, thus promoting ligand-independent transcription of

ERa and possibly loss of tamoxifen-induced inhibition of ERa-mediated gene

transcription. On the basis of these preclinical models, several clinical trials are

under way to determine whether treating patients with EGFR inhibitors (i.e.,

Iressa) and/or MAPK-specific inhibitors (U0126) can effectively treat and/or

prevent antiestrogen resistance in breast tumors.

172 Zelnak et al.


Role of PI3-K/AKT Signaling

Phosphatidylinositol 3-kinase (PI3-K) is a heterodimer complex consisting of

a regulatory subunit, p85, and a catalytic subunit, p110. The regulatory subunit

p85 is phosphorylated by either receptor tyrosine kinases such as EGFR/

HER2/neu or intracellular adapter kinases such as Shc or Src and thereafter

interacts and activates the catalytic subunit p110 (39). The activated PI3-K

subsequently activates downstream effector kinases such as Akt (PKB), a

serine/threonine kinase. Akt has been shown to activate either directly or

indirectly proteins responsible for prosurvival and anti-apoptosis in cancer

cells. Several reports indicate that the PI3-K/Akt pathway interacts with the

ERa pathway, resulting in bidirectional cross talk. It has been shown that PI3-

K/Akt can mediate E2-induced transcription of cyclin D1 and entry of MCF-7

breast cancer cells into S-phase (40). In addition, the PI3-K/Akt pathway can

induce ERa phosphorylation on Ser-167 to promote ERa-mediated tran-

scription and thus protect cells against tamoxifen-induced apoptosis (41). The

accumulating data suggest that ERa-positive breast tumors with alterations of

PI3-K and Akt signaling, whether dependent or independent of EGFR/HER2/

neu, might be insensitive to antiestrogens and thus lead to resistance. Future

studies are needed to confirm whether blocking PI3-K and/or Akt signaling in

ERa-positive breast cancer will be beneficial in subverting resistance to

antiestrogens.

New SERMS

New drug discovery for selective ER modulators (SERMs) is currently driven

by the known side effects of tamoxifen, which include an increase in the

incidence of endometrial cancer, development of resistance, and the recent

report of the negative effects of hormone replacement therapy (HRT) on cor-

onary heart disease (42). Several approaches are being pursued by altering the

antiestrogenic side chain or improving the pharmacokinetics of existing mol-

ecules. Several novel antiestrogen compounds have been developed to replace

tamoxifen for ERa-positive breast cancer without the agonist actions. These

compounds have the potential to be more effective than tamoxifen, having

the advantage of reducing the incidence of endometrial cancer and possibly

acquired resistance during therapy. Two groups of antiestrogenic drugs exist

today with the potential to replace tamoxifen: (1) The SERMs that include

tamoxifen-like compounds (triphenylethylenes) such as toremifene, idoxifene,

and GW 5638 and fixed-ring compounds (benzothiophenes) such as raloxifene,

arzoxifene, EM 652, and CP 336,156, and (2) selective ER downregulators

(SERDs and pure antiestrogen) such as ICI 182,780 (Fulvestrant1). In phase II

trials in patients with tamoxifen-resistant metastatic breast cancer, the new

SERMs showed low response rates (0–15%) (43,44), suggesting cross-

resistance to tamoxifen. No difference in outcome was noted between

tamoxifen and toremifene in patients with metastatic or early-stage breast



cancer (45,46). In contrast, few clinical trials exist today evaluating the

effectiveness of the fixed-ring compounds to tamoxifen. The Study of

Tamoxifen and Raloxifene (STAR) trial compared the efficacy of raloxifene

over tamoxifen as a chemopreventive for women at high risk for breast cancer.

Both drugs reduced the risk of invasive breast cancer by approximately 50%.

Tamoxifen also lowered the incidence of lobular carcinoma in situ (LCIS) and

ductal carcinoma in situ (DCIS) by about 50 percent; however, raloxifene did

not have an effect on the incidence of LCIS and DCIS (47). Of note, the

incidence of uterine cancer was 36% lower, and the incidence of throm-

boembolic events was 29% lower in the raloxifene arm (47). Fulvestrant and

other SERDs have been shown to have high affinity for ERs compared with

tamoxifen with none of the agonist activities. In clinical trials, fulvestrant

initially showed significant promise for the treatment of advanced breast

cancer (48,49). However, in a phase III clinical trial versus tamoxifen for first-

line therapy of advanced breast cancer, fulvestrant was not superior to

tamoxifen in terms of time-to-treatment failure (50). Therefore, more work is

needed to evaluate the efficacy of SERDs in comparison with SERMs for ERa-positive breast cancer.

AROMATASE INHIBITORS

An alternate strategy to endocrine therapy, which specifically inhibits binding of

E2 to the ER, is to inhibit the production of E2 by blocking the cytochrome p450

aromatase enzyme, the rate-limiting enzyme that converts androgens (i.e., tes-

tosterone and androstenedione) to estrogens (i.e., E2 and estrone) in the adrenal

gland, surrounding stroma, and adipose tissue of the breast tumor. The main

drugs of this type are aromatase inhibitors, which include Type I (steroidal) or

Type II (nonsteroidal) (Fig. 2). Steroidal inhibitors are competitive-substrate

mimics of androstenedione. These include formestane and exemestane, which

are irreversible inhibitors that bind with high affinity to the binding site of

aromatase and are converted to a covalently bound intermediate. Nonsteroidal

inhibitors include the first-generation aromatase inhibitor aminoglutethimide and

the second-generation compounds anastrozole and letrozole. All nonsteroidal

aromatase inhibitors act by binding reversibly to the enzyme and competitively

inhibiting binding of the substrate androstenedione. The benefits of using aro-

matase inhibitors over tamoxifen are believed to be the complete deprivation of

E2 and thus better efficacy for ERa-positive breast cancer (51). Recent clinical

data have clearly demonstrated that anastrozole (52), letrozole (53), and

exemestane (54) are more effective than tamoxifen as first-line treatments in

patients with metastatic breast cancer. On the basis of clinical results (52,53),

currently both anastrozole and letrozole are approved by the Food and Drug

Administration for first-line treatment of postmenopausal, ERa-positiveadvanced breast cancer. The data from advanced breast cancer trials provided the

rationale to perform large-scale clinical trials to determine whether there is an

174 Zelnak et al.


advantage of using aromatase inhibitors over tamoxifen in the adjuvant setting.

The anastrozole, tamoxifen, and the combination of anastrozole and tamoxifen

(ATAC) trial has sufficient follow-up data to confirm that anastrozole is superior

to tamoxifen as a first-line adjuvant therapy in ERa-positive breast cancer with

regard to disease-free survival and incidence of contralateral breast cancer (55).

The Breast International Group (BIG) 1-98 study demonstrated a similar

improvement in event-free survival to confirm that letrozole is superior to

tamoxifen as first-line adjuvant hormonal therapy (56). Other trials have dem-

onstrated an improved outcome for postmenopausal patients with early-stage

breast cancer treated with two to three years of tamoxifen, followed by an

aromatase inhibitor, compared with five years of tamoxifen (57,58). The BIG 1-98

trial will help to determine whether these sequencing/switching approaches are

equivalent to upfront aromatase inhibitors.

Mechanisms of Resistance

Clinically, patients that relapse after a previous response to tamoxifen usually

have a clinical response to aromatase inhibitors (59,60). These results strongly

indicate that the ERa continues to be expressed and is functional in breast tumors

that are resistant to antiestrogens. However, although estrogen deprivation

treatment might be more effective than tamoxifen in delaying resistance, even-

tually resistance to aromatase inhibitors will also develop. To date, it is unclear

Figure 2 Structures of selective aromatase inhibitors.



whether similar mechanisms of actions that have been identified for tamoxifen

resistance are also involved in resistance to aromatase inhibitors. The exact

mechanisms contributing to aromatase inhibitor resistance has yet to be fully

elucidiated. However, in vitro studies have identified mutations within the aro-

matase gene that confers resistance to aromatase inhibitors (61). These mutations

have not yet been identified in human breast carcinomas (62). Other studies have

demonstrated that estrogen deprivation supersensitizes the breast cancer cell to

low levels of estrogen, thus creating a hypersensitive environment to overcome

estrogen deprivation resulting in resistance (63–65). In addition, results suggest

that there is increased cross talk between growth factor receptor signaling

pathways and ERa. ERa has been shown to become activated and supersensitized

by several different intracellular kinases, including MAPKs, insulin-like growth

factors, and the PI3-K/Akt pathway (41,66–69). Therefore, the data suggest that

ERa continues to be an integral part of the breast cancer cell signaling pathway

even after resistance to aromatase inhibitors has developed.

Role of Progesterone Receptor and HER2/Neu

There is emerging evidence to suggest that ER-positive cancers that do not

express the progesterone receptor (PR) and/or HER2/neu are somewhat intrin-

sically resistant to tamoxifen and perhaps hormonal therapy in general. Arpino et

al. have demonstrated an increased relapse rate in patients with ER-positive, PR-

negative cancers compared with ER-positive, PR-positive cancers, treated with

tamoxifen (70). Patients treated with tamoxifen with ER-positive, HER2-positive

metastatic breast cancers have a shorter time to treatment failure compared with

ER-positive, HER2-negative cancers (71). In fact, Arpino et al. have demon-

strated an increase in both HER1 and HER2 in ER-positive, PR-negative cancers

compared to ER-positive, PR-positive cancers, suggesting an interplay between

the ER and epidermal growth factor pathways (70).

Two very small trials demonstrated a significantly increased clinical

response rate in patients with ER-positive, HER2-positive cancers treated with

preoperative aromatase inhibitors compared to preoperative tamoxifen (72,73).

This led to a widely accepted hypothesis that aromatase inhibitors were a better

choice than tamoxifen in patients with ER-positive, HER2-positive cancers.

However, an analysis of the BIG-1-98 trial demonstrates that letrozole improves

outcome compared to tamoxifen in both ER-positive, HER2-positive cancers

(HR 0.68) and in ER-positive, HER2-negative cancers (HR 0.72) (74). A recent

subanalysis of the ATAC trial demonstrated a significantly improved outcome in

ER-positive, HER2-negative cancers (HR 0.66) but not in ER-positive, HER2-

positive cancers (HR 0.92), but this may have been due to the small number of

patients in the HER-positive group (75). Are HER2-positive cancers somewhat

resistant to not just tamoxifen but also to aromatase inhibitors? As outlined

above, a recent trial randomized patients with HR-positive, HER2-positive

metastatic breast cancers to anastrozole alone or to anastrozole plus trastuzumab

176 Zelnak et al.


(76). Although there was no significant difference in overall survival, possibly

because patients randomized to anastrozole alone could receive trastuzumab at

disease progression, the time to progression was doubled from 2.4 months in the

anastrozole-alone arm to 4.8 months in the combined arm (p ¼ 0.0016) (76). The

clinical benefit rate in the combination arm was 42%—significantly higher than

in the anastrozole alone arm. A trial that evaluated single-agent trastuzumab as

first-line therapy for patients with HER2-positive cancers demonstrated a clinical

benefit rate of 48% (77). This suggests the intriguing possibility that HR-

positive, HER2-positive cancers are driven by the HER2 pathway, which renders

the cancers partly resistant to hormonal therapies.

An initial evaluation of the ATAC trial using case report forms revealed

that TTR was longer for anastrozole in both ER-positive/PR-positive and ER-

positive/PR-negative subgroups, but the benefit was more pronounced in the ER-

positive/PR-negative subgroup [HR 0.84, 95% confidence interval (CI) 0.69–

1.02 vs. 0.43, 95% CI 0.31–0.61] (78). Importantly, the ER and PR analyses were

not performed centrally. More recently, a central analysis of about 2000 patients

on the ATAC trial demonstrated similar improvements with the use of anas-

trozole compared to tamoxifen, regardless of PR status (HR anastrozole vs.

tamoxifen 0.72 for ER-positive, PR-positive subgroups and 0.66 for ER-positive,

PR-negative subgroups) (75). In the BIG-1-98 trial, similar benefits for letrozole

compared to tamoxifen were seen in the ER-positive/PR-positive and ER-

positive/PR-negative subgroups (74).

On the basis of this data, decisions regarding whether to start a patient on

tamoxifen or an aromatase inhibtor should not be based on PR or HER2 status.

Further molecular profiling may help in the future in making decisions regarding

optimal hormonal therapies.

Role of Gene Expression Analysis

Recent studies have shown that that gene expression profiling of breast tumors

can be used to predict outcome. The initial gene expression signatures were

developed using snap-frozen tissue, limiting their clinical application (79). An

assay (Oncotype DX, Genomic Health, Inc., Redwood City, California, U.S.) has

been developed and validated in women with early-stage, node-negative breast

cancer using fixed, paraffin-embedded tissue. The 21-gene panel includes genes

involved with proliferation, invasion, and hormone response and classifies a

tumor as being low, intermediate, or high risk of recurrence based on the level of

expression (80). Additional studies have shown that a woman’s recurrence score

also predicts response to chemotherapy: women with low recurrence scores

derive little benefit from adjuvant chemotherapy, whereas women with high

recurrence scores benefit significantly from chemotherapy in addition to hor-

monal therapy. Most strikingly are the findings that breast cancers with a high

recurrence score obtain minimal, if any, benefit from five years of tamoxifen

(81). Given the fact that these cancers likely have higher expression of HER2



and lower quantitative hormone receptors, this is in keeping with the data out-

lined above, suggesting that HER2-positive cancers have intrinsic resistance to

tamoxifen. It is not known if women with intermediate recurrence scores benefit

from chemotherapy in addition to hormonal therapy (82). The TAILORx (Trial

Assigning Individualized Options for Treatment) trial is under way, randomizing

early-stage breast cancer patients with intermediate recurrence scores to che-

motherapy plus hormonal therapy versus hormonal therapy alone.

FUTURE RESEARCH AND THERAPEUTICS

It is clear that the central regulator of growth for ERa-positive breast tumors is

ERa. ER signaling is evidently not isolated from other cellular signaling path-

ways. Research studies demonstrate that there is bidirectional cross talk between

EGFR/HER2/neu, MAPK, PI3-K/Akt, and ERa, which ultimately affects the

cellular response of a cell to estrogens and mitogens. ERa integrates numerous

signals from hormones, growth factors, and intracellular kinases along with the

array of coregulators to modulate cellular physiology and tumor pathology. The

first generation antiestrogen/SERM, tamoxifen, has been effective at increasing

overall survival of patients with ERa-positive breast cancer. However, the

eventual development of resistance to tamoxifen is common because of the

multiple signaling networks affecting the function of ERa. As a result of this

limitation, intense research over the past 25 years has revealed the need for

alternate treatment strategies to tamoxifen such as newer and better SERMs with

less agonist activity in the uterus while being full antagonists in the breast.

In addition to SERMs, other therapeutic approaches have emerged such as

the use of aromatase inhibitors to prevent the synthesis of E2 and thus deprive

breast cancer cells and other cells of E2. The strategy of estrogen deprivation

may also have consequences such as resistance, bone loss, and/or dementia. In

the ATAC and BIG 1-98 adjuvant trials, changes in bone mineral density (BMD)

and increased incidence of bone fractures were reported. Longer follow-up in

these studies will help determine if the decrease in BMD stabilizes over time or

if it continues throughout the course of treatment.

Therapeutic opportunities also exist from research studies investigating the

role of growth factor receptors EGFR and HER2/neu and their intracellular

signaling communicators MAPK and PI3-K/Akt in the development of tamox-

ifen resistance. Trastuzumab, the HER2/neu-specific antibody, has been shown

to be highly effective in treating breast tumors overexpressing HER2/neu.

Moreover, research data indicate a role for HER2/neu in antiestrogen resistance,

demonstrating a rationale for using trastuzumab to treat or prevent antiestrogen-

resistant breast cancer. EGFR inhibitors and other tyrosine kinase inhibitors may

also prove to be beneficial in preventing resistance when given in combination

with tamoxifen or other SERMs or SERDs such as fulvestrant.

In addition to EGFR and HER2/neu as important modulators of tamoxifen

action, dissecting the specific interrelationship between ERa and its coregulators

178 Zelnak et al.


may open the door for novel therapeutics. An understanding of how ERa-mediated gene transcription is dysregulated in breast cancer and how this

changes in resistance during endocrine therapy will hopefully lead to improved

strategies for the use of current and future antiestrogen therapies.

REFERENCES

1. Cole MP, Jones CT, Todd ID. A new anti-oestrogenic agent in late breast cancer. An

early clinical appraisal of ICI46474. Br J Cancer 1971; 25:270–275.

2. Jensen EV, Jacobson HI. Basic guides to the mechanism of estrogen action. Recent

Prog Horm Res 1962; 18:387–414.

3. Jordan VC, Dowse LJ. Tamoxifen as an anti-tumour agent: effect on oestrogen

binding. J Endocrinol 1976; 68:297–303.

4. Fisher B, Redmond C, Brown A, et al. Adjuvant chemotherapy with and without

tamoxifen in the treatment of primary breast cancer: 5-year results from the National

Surgical Adjuvant Breast and Bowel Project Trial. J Clin Oncol 1986; 4:459–471.

5. Fisher B, Costantino JP, Redmond CK, et al. Endometrial cancer in tamoxifen-

treated breast cancer patients: findings from the National Surgical Adjuvant Breast

and Bowel Project (NSABP) B-14. J Natl Cancer Inst 1994; 86:527–537.

6. Gottardis MM, Robinson SP, Satyaswaroop PG, et al. Contrasting actions of

tamoxifen on endometrial and breast tumor growth in the athymic mouse. Cancer

Res 1988; 48:812–815.

7. Muss HB, Case LD, Atkins JN, et al. Tamoxifen versus high-dose oral medroxy-

progesterone acetate as initial endocrine therapy for patients with metastatic breast

cancer: a Piedmont Oncology Association study. J Clin Oncol 1994; 12:1630–1638.

8. Kuukasjarvi T, Kononen J, Helin H, et al. Loss of estrogen receptor in recurrent

breast cancer is associated with poor response to endocrine therapy. J Clin Oncol

1996; 14:2584–2589.

9. Bachleitner-Hofmann T, Pichler-Gebhard B, Rudas M, et al. Pattern of hormone

receptor status of secondary contralateral breast cancers in patients receiving adju-

vant tamoxifen. Clin Cancer Res 2002; 8:3427–3432.

10. Johnston SR, Saccani-Jotti G, Smith IE, et al. Changes in estrogen receptor, pro-

gesterone receptor, and pS2 expression in tamoxifen-resistant human breast cancer.

Cancer Res 1995; 55:3331–3338.

11. Hopp TA, Fuqua SA. Estrogen receptor variants. J Mammary Gland Biol Neoplasia

1998; 3:73–83.

12. Murphy LC, Dotzlaw H, Leygue E, et al. The pathophysiological role of estrogen

receptor variants in human breast cancer. J Steroid BiochemMol Biol 1998; 65:175–180.

13. Fuqua SA. The role of estrogen receptors in breast cancer metastasis. J Mammary

Gland Biol Neoplasia 2001; 6:407–417.

14. Jiang SY, Langan-Fahey SM, Stella AL, et al. Point mutation of estrogen receptor

(ER) in the ligand-binding domain changes the pharmacology of antiestrogens in ER-

negative breast cancer cells stably expressing complementary DNAs for ER. Mol

Endocrinol 1992; 6:2167–2174.

15. Liu H, Park WC, Bentrem DJ, et al. Structure-function relationships of the raloxifene-

estrogen receptor-alpha complex for regulating transforming growth factor-alpha

expression in breast cancer cells. J Biol Chem 2002; 277:9189–9198.



16. Schwartz JA, Zhong L, Deighton-Collins S, et al. Mutations targeted to a predicted helix

in the extreme carboxyl-terminal region of the human estrogen receptor-alpha alter its

response to estradiol and 4-hydroxytamoxifen. J Biol Chem 2002; 277:13202–13209.


transformation is associated with clinical outcomes of efficacy and hot flashes.

J. Clin Onc 2005; 23:9312–8.

18. Mortimer J, Flatt S, Parker B, et al. Tamoxifen, hot flashes, and breast cancer

recurrence: support for pharmacogenetics. Proc ASCO 2007 (abstr 500).

19. Smith CL, Nawaz Z, and O’Malley BW. Coactivator and corepressor regulation of

the agonist/antagonist activity of the mixed antiestrogen, 4-hydroxytamoxifen. Mol

Endocrinol 1997; 11:657–666.

20. Osborne CK, Bardou V, Hopp TA, et al. Role of the estrogen receptor coactivator

AIB1 (SRC-3) and HER-2/neu in tamoxifen resistance in breast cancer. J Natl

Cancer Inst 2003; 95:353–361.

21. Girault I, Tozlu S, Lidereau R, et al. Expression analysis of DNA methyltransferases

1, 3A, and 3B in sporadic breast carcinomas. Clin Cancer Res 2003; 9:4415–4422.

22. Kurokawa H, Arteaga CL. ErbB (HER) receptors can abrogate antiestrogen action in

human breast cancer by multiple signaling mechanisms. Clin Cancer Res 2003;

9:511S–515S.

23. Yarden Y, Sliwkowski MX. Untangling the ErbB signalling network. Nat Rev Mol

Cell Biol 2001; 2:127–137.

24. Yarden Y. The EGFR family and its ligands in human cancer. signalling mechanisms

and therapeutic opportunities. Eur J Cancer 2001; 37(suppl 4):S3–S8.

25. Yarden Y. Biology of HER2 and its importance in breast cancer. Oncology 2001;

61(suppl 2):1–13.

26. Rubin I, Yarden Y. The basic biology of HER2. Ann Oncol 2001; 12(suppl 1):S3–S8.

27. Spigel DR, Burstein HJ. HER2 overexpressing metastatic breast cancer. Curr Treat

Options Oncol 2002; 3:163–174.

28. Klijn JG, Berns PM, Schmitz PI, et al. The clinical significance of epidermal growth

factor receptor (EGF-R) in human breast cancer: a review on 5232 patients. Endocr

Rev 1992; 13:3–17.

29. Klijn JG, Look MP, Portengen H, et al. The prognostic value of epidermal growth

factor receptor (EGF-R) in primary breast cancer: results of a 10 year follow-up

study. Breast Cancer Res Treat 1994; 29:73–83.

30. Tang CK, Gong XQ, Moscatello DK, et al. Epidermal growth factor receptor vIII

enhances tumorigenicity in human breast cancer. Cancer Res 2000; 60:3081–3087.

31. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal

antibody against HER2 for metastatic breast cancer that overexpresses HER2.

N Engl J Med 2001; 344:783–792.

32. Nahta R, Hortobagyi GN, Esteva FJ. Growth factor receptors in breast cancer:

potential for therapeutic intervention. Oncologist 2003; 8:5–17.

33. Morris C. The role of EGFR-directed therapy in the treatment of breast cancer.

Breast Cancer Res Treat 2002; 75(suppl 1):S51–S55, S57–S59 (discussion).

34. Johnston SR, Head J, Pancholi S, et al. Integration of signal transduction inhibitors

with endocrine therapy: an approach to overcoming hormone resistance in breast

cancer. Clin Cancer Res 2003; 9:524S–532S.

35. Kurokawa H, Lenferink AE, Simpson JF, et al. Inhibition of HER2/neu (erbB-2)

and mitogen-activated protein kinases enhances tamoxifen action against

180 Zelnak et al.


HER2-overexpressing, tamoxifen-resistant breast cancer cells. Cancer Res 2000; 60:

5887–5894.

36. Oh AS, Lorant LA, Holloway JN, et al. Hyperactivation of MAPK induces loss of

ERalpha expression in breast cancer cells. Mol Endocrinol 2001; 15:1344–1359.

37. Kato S, Endoh H, Masuhiro Y, et al. Activation of the estrogen receptor through

phosphorylation by mitogen-activated protein kinase. Science 1995; 270:1491–1494.

38. Bunone G, Briand PA, Miksicek RJ, et al. Activation of the unliganded estrogen

receptor by EGF involves the MAP kinase pathway and direct phosphorylation.

Embo J 1996; 15:2174–2183.

39. Fry MJ. Phosphoinositide 3-kinase signalling in breast cancer: how big a role might

it play? Breast Cancer Res 2001; 3:304–312.

40. Castoria G, Migliaccio A, Bilancio A, et al. PI3-kinase in concert with Src promotes the

S-phase entry of oestradiol-stimulated MCF-7 cells. Embo J 2001; 20: 6050–6059.

41. Campbell RA, Bhat-Nakshatri P, Patel NM, et al. Phosphatidylinositol 3-kinase/

AKT-mediated activation of estrogen receptor alpha: a new model for anti-estrogen

resistance. J Biol Chem 2001; 276:9817–9824.

42. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus

progestin in healthy postmenopausal women: principal results From the Women’s

Health Initiative randomized controlled trial. JAMA 2002; 288:321–333.

43. Vogel CL, Shemano I, Schoenfelder J, et al. Multicenter phase II efficacy trial of

toremifene in tamoxifen-refractory patients with advanced breast cancer. J Clin

Oncol 1993; 11:345–350.

44. Johnston SR, Gumbrell LA, Evans TR, et al. A cancer research (UK) randomized

phase II study of idoxifene in patients with locally advanced/metastatic breast cancer

resistant to tamoxifen. Cancer Chemother Pharmacol 2004; 53:341–348.

45. Pyrhonen S, Ellmen J, Vuorinen J, et al. Meta-analysis of trials comparing toremifene

with tamoxifen and factors predicting outcome of antiestrogen therapy in post-

menopausal women with breast cancer. Breast Cancer Res Treat 1999; 56: 133–143.

46. Pagani O, Gelber S, Price K, et al. Toremifene and tamoxifen are equally effective

for early stage breast cancer: first results of International Breast Cancer Study Group.

Ann Oncol 2004; 15:1749–1759.

47. Wickerham DL, Costantino JP, Vogel V, et al. The study of tamoxifen and raloxifene

(STAR): initial findings from the NSABP P-2 breast cancer prevention study. ASCO

Annual Meeting Proceedings Part I. J Clin Oncol 2006; 24(suppl 18):LBA5.

48. Howell SJ, Johnston SR, Howell A. The use of selective estrogen receptor modu-

lators and selective estrogen receptor down-regulators in breast cancer. Best Pract

Res Clin Endocrinol Metab 2004; 18:47–66.

49. Osborne CK, Wakeling A, Nicholson RI. Fulvestrant: an oestrogen receptor antag-

onist with a novel mechanism of action. Br J Cancer 2004; 90(suppl 1):S2–S6.

50. Howell A, Robertson JF, Abram P, et al. Comparison of fulvestrant versus tamoxifen

for the treatment of advanced breast cancer in postmenopausal women previously

untreated with endocrine therapy: a multinational, double-blind, randomized trial.

J Clin Oncol 2004; 22:1605–1613.

51. Johnston SR, Dowsett M. Aromatase inhibitors for breast cancer: lessons from the

laboratory. Nat Rev Cancer 2003; 3:821–831.

52. Bonneterre J, Buzdar A, Nabholtz JM, et al. Anastrozole is superior to tamoxifen as

first-line therapy in hormone receptor positive advanced breast carcinoma. Cancer

2001; 92:2247–2258.



53. Mouridsen H, Gershanovich M, Sun Y, et al. Superior efficacy of letrozole versus

tamoxifen as first-line therapy for postmenopausal women with advanced breast

cancer: results of a phase III study of the International Letrozole Breast Cancer

Group. J Clin Oncol 2001; 19:2596–2606.

54. Paridaens R, Dirix L, Lohrisch C, et al. Mature results of a randomized phase II

multicenter study of exemestane versus tamoxifen as first-line hormone therapy for

postmenopausal women with metastatic breast cancer. Ann Oncol 2003; 14:1391–1398.

55. Baum M, Budzar AU, Cuzick J, et al. Anastrozole alone or in combination with

tamoxifen versus tamoxifen alone for adjuvant treatment of postmenopausal women

with early breast cancer: first results of the ATAC randomised trial. Lancet 2002;

359:2131–2139.

56. Thurlimann B, Keshaviah A, Coates AS, et al. A comparison of letrozole and

tamoxifen in postmenopausal women with early breast cancer. N Engl J Med 2005;

353:2747–2757.

57. Coombes RC, Kilburn LS, Snowdon CF, et al. Survival and safety of exemestane

versus tamoxifen after 2 to 3 years tamoxifen treatment (Intergroup Exenestane

Study): a randomized controlled trial. Lancet 2007; 369:559–570.

58. Kaufmann M, Jonat W, Hilfrich J, et al. Improved overall survival in postmenopausal

women with early breast cancer after anastrozole initiated after treatment with

tamoxifen compared with continued tamoxifen: the ARNO 95 Study. J Clin Oncol

2007; 25:2664–2670.

59. Buzdar A, Jonat W, Howell A, et al. Anastrozole, a potent and selective aromatase

inhibitor, versus megestrol acetate in postmenopausal women with advanced breast

cancer: results of overview analysis of two phase III trials. Arimidex Study Group.

J Clin Oncol 1996; 14:2000–2011.

60. Dombernowsky P, Smith I, Falkson G, et al. Letrozole, a new oral aromatase

inhibitor for advanced breast cancer: double-blind randomized trial showing a dose

effect and improved efficacy and tolerability compared with megestrol acetate. J Clin

Oncol 1998; 16:453–461.

61. Kao YC, Cam LL, Laughton CA, et al. Binding characteristics of seven inhibitors

of human aromatase: a site-directed mutagenesis study. Cancer Res 1991; 56:

3451–3460.

62. Sourdaine P, Parker MG, Telford J, et al. Analysis of the aromatase cytochrome P450

gene in human breast cancers. J Mol Endocrinol 1994; 13:331–337.

63. Masamura S, Santner SJ, Heitjan DF, et al. Estrogen deprivation causes estradiol

hypersensitivity in human breast cancer cells. J Clin Endocrinol Metab 1995;

80:2918–2925.

64. Santen R, Jeng MH, Wang JP, et al. Adaptive hypersensitivity to estradiol: potential

mechanism for secondary hormonal responses in breast cancer patients. J Steroid

Biochem Mol Biol 2001; 79:115–125.

65. Chan CM, Martin LA, Johnston SR, et al. Molecular changes associated with the

acquisition of oestrogen hypersensitivity in MCF-7 breast cancer cells on long-term

oestrogen deprivation. J Steroid Biochem Mol Biol 2002; 81:333–341.

66. Martin LA, Farmer I, Johnston SR, et al. Enhanced estrogen receptor (ER) alpha,

ERBB2, and MAPK signal transduction pathways operate during the adaptation

of MCF-7 cells to long term estrogen deprivation. J Biol Chem 2003; 278:

30458–30468.

182 Zelnak et al.


67. Jeng MH, Yue W, Eischeid A, et al. Role of MAP kinase in the enhanced cell

proliferation of long term estrogen deprived human breast cancer cells. Breast

Cancer Res Treat 2000; 62:167–175.

68. Shim WS, Conaway M, Masamura S, et al. Estradiol hypersensitivity and mitogen-

activated protein kinase expression in long-term estrogen deprived human breast

cancer cells in vivo. Endocrinology 2000; 141:396–405.

69. Stephen RL, Shaw LE, Larsen C, et al. Insulin-like growth factor receptor levels are

regulated by cell density and by long term estrogen deprivation in MCF7 human

breast cancer cells. J Biol Chem 2001; 276:40080–40086.

70. Arpino G, Weiss H, Lee AV, et al. Estrogen receptor-positive, progesterone receptor-

negative breast cancer: association with growth factor receptor expression and

tamoxifen resistance. J Natl Cancer Inst 2005; 97:1254–1261.

71. De Laurentiis M, Arpino G, Massarelli E, et al. A meta-analysis on the interaction

between HER-2 expression and response to endocrine treatment in advanced breast

cancer. Clin Cancer Res 2005; 11:4741–4748.

72. Ellis MJ, Tao Y, Young O, et al. Estrogen-independent proliferation is present in

estrogen-receptor HER2-positive primary breast cancer after neoadjuvant letrozole.

J Clin Oncol 2006; 24:3019–3025.

73. Smith IE, Dowsett M, Ebbs SR, et al. Neoadjuvant treatment of postmenopausal

breast cancer with anastrozole, tamoxifen, or both in combination: the Immediate

Preoperative Anastrozole, Tamoxifen, or Combined with Tamoxifen (IMPACT)

multicenter double-blind randomized trial. J Clin Oncol 2005; 23:5108–5116.

74. Viale G, Regan MM, Maiorano E, et al. Prognostic and predictive value of centrally

reviewed expression of estrogen and progesterone receptors in a randomized trial

comparing letrozole and tamoxifen adjuvant therapy for postmenopausal early stage

breast cancer: BIG 1-98. J Clin Oncol 2007; 25:3846–3852.

75. Dowsett M, Allred DC, for the TransATAC investigators. Relationship between

quantitative ER and PgR expression and HER2 status with recurrence in the ATAC

trial. SABCS 2006 (abstr 48).

76. Mackey JR, Kaufman B, Clemens M, et al. Trastuzumab prolongs progression-free

survival in hormone-dependent and HER2-positive metastatic breast cancer. SABCS

2006 (abstr 3).



J Clin Oncol 2002; 20:719–726.

78. Dowsett M, Cuzick J, Wale C, et al. Retrospective analysis of time to recurrence in

the ATAC trial according to hormone receptor status: an hypothesis-generating

study. J Clin Oncol 2005; 23:7512–7517.

79. Van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a

predictor of survival in breast cancer. N Eng J Med 2002; 347:1999–2009.



81. Paik S, Shak S, Tang G, et al. Expression of the 21 genes in the recurrence score

assay and prediction of clinical benefit from tamoxifen in NSABP study B-14 and

chemotherapy in NSABP study B-20. Breast Cancer Res Treat 2004; 88 (abstr 24).

82. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in

women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol

2006; 24:3796–3834.



13

TAILORx: Rationale for the Study Design

Joseph A. Sparano

Albert Einstein College of Medicine, Montefiore Medical Center,Bronx, New York, U.S.A.

INTRODUCTION

There have been substantial technical and analytical advances within the past

decade that have facilitated high-throughput analysis of clinical specimens, a

process that has been referred to as “molecular profiling.” This term, as applied

here, may refer to evaluating various “markers,” including genomic, proteomic,

and epigenomic expression patterns, or a combination of these patterns, in

clinical specimens. Technological advances have led to the ability to measure

thousands of markers with a variety of validated methods (1), and analytical

models have been developed that facilitate analysis of the voluminous amount of

data that are generated (2). The technology has also led to an ability to perform

“discovery-based research,” in which large volumes of data are generated and

analyzed without a specific hypothesis, in contrast to the traditional scientific

paradigm of “hypothesis-based research,” in which a limited number of genes or

proteins are based upon a specific hypothesis and rationale (3). Discovery-based

research and hypothesis-based research are not mutually exclusive; however,

profiling may also be used to test specific hypotheses that are based on sound,

scientific rationale. A series of studies have been reported over the past few

years that have utilized gene expression profiling to discover “molecular

markers,” which may identify (i) distinct molecular subtypes of breast cancer

185


(i.e., genotypic-phenotypic correlation), (ii) molecular signatures–associated

prognosis (i.e., prognostic factors), and (iii) molecular signatures that predict

the benefit from specific therapies (i.e., predictive factors) (4). When a

“molecular marker” is shown to perform more reliably than clinical features in

predicting prognosis or the benefit from a specific intervention, then there is

interest in further evaluating the utility of the marker in clinical practice (5).

PROGRAM FOR THE ASSESSMENT OF CLINICAL CANCER TESTS

In order to address the issue of integrating molecular diagnostic testing into

clinical practice, the U.S. National Cancer Institute initiated the program for the

assessment of clinical cancer tests (PACCT) (http://www.cancerdiagnosis.nci

.nih.gov/assessment/index.html). The goals of PACCT are to ensure translation

of new knowledge about cancer and new technologies to clinical practice and

development of more informative laboratory tools to help maximize the impact

of cancer treatments. This effort has led to the publication of the REMARK

(REporting recommendations for tumour MARKer prognostic studies)

guidelines (6) and to the TAILORx (Trial Assigning IndividuaLized Options

for Treatment) trial developed by the North American Breast Cancer Inter-

group (http://www.cancer.gov/clinicaltrials/digestpage/TAILORx). In review-

ing the rationale for design of the TAILORx trial in breast cancer, the following

questions were considered by the Breast Cancer Intergroup and are reviewed

herein:

l What therapeutic intervention should the marker be used to select?l What patient population should be evaluated?l Which marker should be evaluated?l What trial design would integrate current knowledge about the marker and

address gaps in our knowledge about the current marker and future potential

markers?

WHICH TREATMENT INTERVENTION?

Therapeutic options are used in an adjunctive manner for patients with early-

stage breast cancer after surgery, irradiation, endocrine therapy, chemotherapy,

and trastuzumab (Table 1). In contrast to irradiation, endocrine therapy, and

trastuzumab, there are no clinical factors that may be used to predict the benefit

from adjuvant chemotherapy. Chemotherapy is usually recommended on the

basis of the risk of recurrence (i.e., prognostic factors) and age (7). In contrast

to other therapies, the treatment effect of chemotherapy is modest, acute tox-

icities are frequent, and the cost is high (8). Therefore, evaluation of a marker

that has the potential for predicting benefit from chemotherapy would be

desirable.

186 Sparano


WHICH PATIENT POPULATION?

In 2006, approximately 124,000 women in the United States were diagnosed

with estrogen receptor (ER)–positive breast cancer associated with negative

axillary lymph nodes, accounting for about 50% of all breast cancer diagnosed

each year and nearly 9% of all new cases of cancer (9). Treatment recom-

mendations have generally been based on prognostic factors, with chemotherapy

recommended if the residual risk of recurrence exceeds approximately 5% to

10% despite adjuvant hormonal therapy (i.e., the tumor size exceeds 1 cm or

smaller tumors are associated with unfavorable histologic features) (7). By these

criteria, adjuvant chemotherapy could be recommended annually for up to

75,000 women younger than 70 years in the United States. Currently, it is esti-

mated that approximately 25% of all women, or about 30,000 annually, with early-

stage ER-positive breast cancer receive adjuvant chemotherapy. Its use has

increased substantially over the past 15 years (10) is highly dependent on age, and

is estimated to be given to 75% of those younger than 50 years, 30% between the

ages of 50 and 69 years, and about 5% of those 70 years or older.

Approximately 85% of women with early-stage ER-positive breast cancer

are therefore adequately treated with adjuvant hormonal therapy. Although

adding chemotherapy reduces the risk of recurrence on average by about 30%,

the absolute benefit for an individual patient is small, ranging from 1% to 5%

(11). The vast majority of patients with ER-positive breast cancer are therefore

overtreated with chemotherapy, since most would have been cured with hor-

monal therapy alone. Decision aids such as Adjuvant! Online (12), decision

boards (13), and other tools, are often useful to assist patients and caregivers with

information regarding absolute benefits that might be expected from chemo-

therapy (14). Although such decision aids may assist some patients in making a

more informed decision regarding whether to accept adjuvant chemotherapy,

when faced with a choice, many patients and their clinicians err on the side of

overtreatment because of the imprecise nature of predicting the treatment benefit

Table 1 Standard Adjuvant Therapies for Operable Breast Cancer

Treatment Selection

Percent

eligible

Treatment

effect (%)

Acute

toxicity

Relative

cost

Chemotherapy Recurrence risk

> 5–10%

Up to 90 25–35 High þþþ

Endocrine

therapy

ER- and/or PR-

positive tumor

70 50 Low þþ

Trastuzumab HER2/neu-positive

tumor

15 50 Low þþþþ

Radiation Breast-sparing surgery 40–60 90 Low þAbbreviations: ER, estrogen receptor; PR, progesterone receptor; HER2, human epidermal growth

factor receptor 2.

TAILORx 187


(15). Therefore, developing a clinical trial that employs a molecular marker to

define the chemotherapy benefit was judged by the Breast Cancer Intergroup to

be the preferred setting. Recent evidence demonstrating that the incremental

benefits associated with contemporary chemotherapy regimens had minimal

impact in high-risk ER-positive disease reinforces this approach (16).

WHICH MARKER?

A number of molecular diagnostic tests that provide more accurate prognostic

information that standard clinical features have been developed and externally

validated and are summarized in Table 2, including the Amsterdam 70-gene

profile (17,18), Rotterdam 76-gene profile (19,20), and a 21-gene profile (21,22).

Recently, these assays have been shown to have comparable discriminatory

value in early-stage breast cancer (23).

The 21-gene Oncotype DXTM (Genomic Health, Inc., Redwood City,

California U.S.) includes 16 tumor genes and 5 reference genes, with the result

expressed as a computed “recurrence score” (RS) (Fig. 1) (21). The genes

utilized in the assay were identified in several training set trials (24), followed by

a validation study that was performed in patients with ER-positive, node-nega-

tive breast cancer who received a five-year course of tamoxifen (TAM) in

National Surgical Adjuvant Breast and Bowel Project (NSABP) trial B-14 (21).

A higher RS was associated with an elevated risk of distant recurrence, whether

evaluated as a categorical or a continuous variable (Fig. 2). This assay was

selected for evaluation in the TAILORx trial for several scientific and practical

reasons that are summarized in Table 3 and discussed herein. In addition to the

prospective validation study illustrated by Figure 2, a population-based external

validation study performed in patients treated in the Kaiser Permanente health

care system provided further evidence for the robustness of the assay (22).

Moreover, subsequent studies in patients treated with TAM with or without

Table 2 Externally Validated Multigene Breast Cancer Markers

Population

Refs.

Number

of genes Tissue ER/PR Nodes Adjuvant therapy

Commercially

available

27,28 534 Frozen þ/– þ/– Preoperative chemo Not available

17,18 70 Frozen þ/– þ/– þ/– Chemo;

þ/– hormones

MammaPrint1

21,22 21 Paraffin

embedded

þ � TAM Oncotype DX

19,20 76 Frozen þ/– � No therapy Not available

Abbreviations: ER, estrogen receptor; PR, progesterone receptor; Chemo, chemotherapy; TAM,

tamoxifen.

188 Sparano


adjuvant chemotherapy in NSABP trial B20 indicated that only patients with an

elevated RS (RS > 30) derived benefit from adjuvant chemotherapy (25). Other

analyses also indicated that an RS also predicts local recurrence, which may

occur in up to 20% of patients with operable breast cancer and which may be

reduced by adjuvant chemotherapy (26). Furthermore, an RS more accurately

predicted recurrence than the Adjuvant! Online model, providing evidence that it

offers information that is complementary to Adjuvant! (12). Finally, although

Oncotoype DX includes a reverse transcriptase polymerase chain reaction

(RT-PCR) assay-based evaluation of ER expression, a tumor with high ER

content by RT-PCR may have either a low or a high RS, indicating that it reflects

a measure other than simply ER expression (25).

Figure 1 Oncotype DX RS: genes and algorithm. Abbreviation: RS, recurrence score.

Figure 2 Oncotype DX RS predicts distant recurrence as either a categorical variable

(left) or continuous variable (right). Abbreviation: RS, recurrence score.

TAILORx 189


There were also several practical reasons for selecting Oncotype DX. First

and foremost is the fact that it involves an RT-PCR-based technique that may be

employed on as little as three 10-mm sections of routinely processed and stored

formalin-fixed, paraffin-embedded tissue. In addition, the assay was approved by

Clinical Laboratory Improvement Amendments (CLIA) in the United States and

has received favorable local coverage decisions regarding reimbursement by

Medicare, indicating acceptance of the value of the assay by third-party payers.

Finally, selecting Oncotype DX for further evaluation by the Breast Cancer

Intergroup represents a logical extension of the successful public-private col-

laboration between NSABP and Genomic Health.

WHAT TRIAL DESIGN?

In designing a trial integrating a molecular diagnostic test into clinical decision

making, three factors must be considered. First, it is essential to incorporate

features already known about the prognostic and predictive value of the test into

the trial design. In this regard, it is clear that patients with a very low RS do very

well with hormonal therapy alone, and patients with a very high RS derive great

benefit from chemotherapy, thereby providing sufficient information to make

definitive treatment recommendations for patients who fall into these groups.

Second, the trial must be designed to address the clinical utility of the test in

circumstances where its utility is uncertain or result uninformative. This scenario

applies to patients with a tumor that has a mid-range RS, in which the risk of

relapse may be sufficiently high to recommend chemotherapy, but in which

chemotherapy may be not be beneficial. Third, it is important to design the trial

in a manner that offers an opportunity to evaluate other clinical cancer tests as

they evolve, obviating the need to repeat a large-scale clinical trial and providing

an opportunity to evaluate the test within a very short time frame.

Table 3 Scientific Rationale for Selecting Oncotype DX Assay for TAILORx

Rationale Ref.

Validation study demonstrating that an RS is a prognostic factor for risk of distant

recurrence in tamoxifen-treated patients, which outperforms clinical criteria in

the multivariate model

21

External population-based validation study demonstrating an RS predictive of

breast cancer death

22

A high RS predictive of chemotherapy benefit in tamoxifen-treated patients 25

An RS correlating with risk of local recurrence 26

An RS predicting the outcome more accurately than Adjuvant! Online 29

Concordance in the predictive value of Oncotype DX compared with other

multigene markers

23

Abbreviation: RS, recurrence score.

190 Sparano


The factors outlined in the previous paragraph were taken into account

while designing TAILORx. Only patients who meet established clinical criteria

for recommending adjuvant chemotherapy, who are medically suitable candi-

dates for chemotherapy, and who agree to have their treatment assigned or

randomized on the basis of the molecular test are eligible. After informed consent

and “preregistration,” a tumor specimen is forwarded to Genomic Health for the

Oncotype DX assay, with the result reported to the site within 10 to 14 days. After

the Oncotype DX RS is known, patients are then registered and have their treat-

ment assigned or randomized on the basis of the RS (Fig. 3), as described below:

l RS is 10 or less: Hormonal therapy alonel RS is 26 or higher: Chemotherapy plus hormonal therapyl RS is 11 to 25: Patients randomized to receive chemotherapy plus hor-

monal therapy (the standard treatment arm) versus hormonal therapy alone

(the experimental treatment arm)

Because the study includes only women who meet standard clinical criteria

for adjuvant chemotherapy and are medically suitable candidates for therapy,

chemotherapy plus hormonal therapy is considered the standard treatment arm

for patients with a mid-range RS (i.e., 11–25). Hormonal therapy is considered

the experimental treatment arm for the mid-range group. The primary objective

Figure 3 TAILORx schema. Abbreviation: RS, recurrence score.

TAILORx 191


of the trial is to determine whether hormonal therapy alone is not inferior to

chemotherapy plus hormonal therapy. The trial utilizes a “noninferiority” design

for the mid-range group and is adequately powered to detect a 3% or greater

difference between the two treatment arms.

The chemotherapy regimen and hormonal agents used are prescribed at the

discretion of the treating physician but must be consistent with one of several

standard options described in the protocol document. Patients may also enroll in

other Intergroup trials as long as the protocol treatment is consistent with the

TAILORx-assigned treatment arm (i.e., hormonal therapy or chemotherapy plus

hormonal therapy). In addition, patients who have already had the Oncotype DX

assay performed are eligible to enroll if the RS is 11 to 25.

At the time of registration and treatment assignment, patients are asked to

provide blood samples for banking of plasma and peripheral blood mononuclear

cells. Tissue specimens are collected by the Eastern Cooperative Oncology

Group (ECOG) Pathology Coordinating Office, with central testing and confir-

mation of ER and progesterone receptor (PR) expression, creation of tissue

microarrays, and RNA extraction for future studies.

The trial is available through the Cancer Trials Support Unit (www.ctsu.org)

and is being coordinated by ECOG. Other participants include NSABP and all

members of the North American Breast Intergroup, including the Southwest

Oncology Group (SWOG), Cancer and Acute Leukemia Group B (CALGB), North

Central Cancer Treatment Group (NCCTG), National Cancer Institute of Canada

(NCIC), and the American College of Surgeons Oncology Group (ACOSOG).

RATIONALE FOR THE RS RANGES IN TAILORx

The RS ranges used in TAILORx are different from those originally described

for the low- (<18), intermediate- (18–30), and high- (>30) risk groups as

originally reported for the assay. The range was adjusted in order to minimize

the potential for undertreatment in both the high-risk group (also called “sec-

ondary study group 2”) and the randomized group (also called the “primary study

group”). When the NSABP B20 data were analyzed using the RS ranges used in

TAILORx, the treatment effect of chemotherapy was similar to the original

analysis (Table 4). For the analyses described in Table 4, disease-free survival

(DFS) is defined as time to first local, regional, or distant recurrence, second

primary cancer, or death due to any cause, and distant recurrence-free survival

(DRFS) is defined as time to first distant recurrence or death from breast cancer

(contralateral breast cancers, other second primary cancers, and deaths due to

other cancers are censored, and local and regional recurrences are ignored). DFS

has been the primary endpoint evaluated in most breast cancer trials and is the

primary endpoint for TAILORx, whereas DRFS was the primary endpoint in

the NSABP trials evaluating RS. Although a trend favoring the addition of

chemotherapy becomes evident at an RS of approximately 11 when the risk of

relapse is analyzed in a linear fashion, the 95% confidence intervals completely

192 Sparano


Tab

le4

RSandResponse

toChem

otherapyin

B20Trial

(n¼

651)byRS

RS

Number

of

patients

TAM

10-yr

DRFS(%

)

TAM

þchem

o

10-yr

DRFS(%

)

HR

(95%

CI)

for

recurrence

by

additionof

chem

oP-value

TAM

10-yr

DFS(%

)

TAM

þchem

o

10-yr

DFS(%

)

HR

(95%

CI)for

recurrence

by

additionofchem

oP-value

<11

177(27%)

98

95

1.788(0.360,8.868)

0.471

77

85

0.605(0.317,1.153)

0.124

11–25

279(43%)

95

94

0.755(0.313,1.824)

0.531

81

76

1.106(0.671,1.823)

0.691

>25

195(30%)

63

88

0.285(0.148,0.551)

<0.0001

53

75

0.446(0.270,0.738)

0.0012

Abbreviations:RS,recurrence

score;HR,hazardratio;CI,confidence

interval;DRFS,distantrelapse-freesurvival;DFS,disease-freesurvival;TAM,tamoxifen;

chem

o,chem

otherapy(w

hichincluded

cyclophospham

ide,

methotrexate,

5-fluorouracilormethotrexate/5-flurourucil).

TAILORx 193


overlap in the 11 to 25 RS range (Fig. 4). Finally, a high RS has also been

associated with an elevated risk of local relapse, which may also be reduced by

giving adjuvant chemotherapy (26). An RS of 11 is associated with a risk of both

local and distant relapse of about 10%, a threshold that has been typically used

for recommending adjuvant chemotherapy.

CONCLUSION

TAILORx represents a major step forward into the era of personalized medicine

for breast cancer. By integrating a molecular diagnostic test into clinical decision

making, patients and clinicians will be able to make more informed decisions

regarding the most appropriate treatment options for a subset of patients for

whom the test results in a clear treatment path and refine the utility of the assay

for those for whom the result may be uninformative. This trial will also serve as

an important resource for evaluating new molecular signatures and other tech-

nologies, such as proteomics, epigenomics, and pharmacogenomics, as these

technologies evolve.

REFERENCES

1. Hanash S. Integrated global profiling of cancer. Nat Rev Cancer 2004; 4:638–644.

2. Simon R. Diagnostic and prognostic prediction using gene expression profiles in

high-dimensional microarray data. Br J Cancer 2003; 89:1599–1604.

Figure 4 Treatment effect of chemotherapy by Oncotype DX RS. Abbreviation: RS,

recurrence score.

194 Sparano


3. Ransohoff DF. Evaluating discovery-based research: when biologic reasoning cannot

work. Gastroenterology 2004; 127:1028.

4. Sparano JA, Fazzari MJ, Childs G. Clinical application of molecular profiling in

breast cancer. Future Oncol 2005; 1:485–496.

5. Hayes DF, Bast RC, Desch CE, et al. Tumor marker utility grading system: a framework

to evaluate clinical utility of tumor markers. J Natl Cancer Inst 1996; 88: 1456–1466.

6. McShane LM, Altman DG, Sauerbrei W, et al. REporting recommendations for

tumour MARKer prognostic studies (REMARK). Eur J Cancer 2005; 41:1690–1696.

7. Carlson RW, Brown E, Burstein HJ, et al. NCCN Task Force Report: Adjuvant

Therapy for Breast Cancer. J Natl Compr Canc Netw 2006; 4(suppl 1):S1–S26.

8. Hassett MJ, O’Malley AJ, Pakes JR, et al. Frequency and cost of chemotherapy-

related serious adverse effects in a population sample of women with breast cancer.

J Natl Cancer Inst 2006; 98:1108–1117.

9. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2006. CA Cancer J Clin 2006;

56:106–130.

10. Harlan LC, Clegg LX, Abrams J, et al. Community-based use of chemotherapy and

hormonal therapy for early-stage breast cancer: 1987-2000. J Clin Oncol 2006; 24:

872–877.

11. Fisher B, Jeong JH, Dignam J, et al. Findings from recent National Surgical Adjuvant

Breast and Bowel Project adjuvant studies in stage I breast cancer. J Natl Cancer Inst

Monogr 2001; 30:62–66.

12. Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the prog-

nostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23:2716–2725.

13. Whelan T, Levine M, Willan A, et al. Effect of a decision aid on knowledge and

treatment decision making for breast cancer surgery: a randomized trial. JAMA

2004; 292:435–441.

14. Whelan TJ, Loprinzi C. Physician/patient decision aids for adjuvant therapy. J Clin

Oncol 2005; 23:1627–1630.

15. Simes RJ, Coates AS. Patient preferences for adjuvant chemotherapy of early breast

cancer: how much benefit is needed? J Natl Cancer Inst Monogr 2001; 30:146–152.

16. Berry DA, Cirrincione C, Henderson IC, et al. Estrogen-receptor status and outcomes

of modern chemotherapy for patients with node-positive breast cancer. JAMA 2006;

295:1658–1667.



18. Buyse M, Loi S, van ’t Veer LJ, et al. Validation and clinical utility of a 70-gene

prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst

2006; 98:1183–1192.

19. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant

metastasis of lymph-node-negative primary breast cancer. Lancet 2005; 365:671–679.



Oncol 2006; 24:1665–1671.



22. Habel LA, Shak S, Jacobs MK, et al. A population-based study of tumor gene

expression and risk of breast cancer death among lymph node-negative patients.

Breast Cancer Res 2006; 8:R25.

TAILORx 195


23. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based

predictors for breast cancer. N Engl J Med 2006; 355:560–569.

24. Cobleigh MA, Tabesh B, Bitterman P, et al. Tumor gene expression and prognosis in

breast cancer patients with 10 or more positive lymph nodes. Clin Cancer Res 2005;

11:8623–8631.



2006; 24(23):3726–3734.

26. Mamounas ETG, Bryant J, Paik S, et al. Association between the 21-gene recurrence

score assay (RS) and risk of loco-regional failure in node-negative, ER-positive

breast cancer: results from NSABP B-14 and NSABP B-20. Breast Cancer Res 2005;

94(suppl 1):S16 (abstr 29).



U S A 2001; 98:10869–10874.

28. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor sub-

types in independent gene expression data sets. Proc Natl Acad Sci U S A 2003;

100:8418–8423.

29. Bryant J. Oncotype DX correlates more closely with prognosis than Adjuvant

Online. Paper presented at: 9th International Conference on Primary Therapy of

Early Breast Cancer January 26-29, 2005; St. Gallen, Switzerland. Available at:

http://www.oncoconferences.ch/2005/program/overview.htm.

196 Sparano


14

Taking Prognostic Signatures to the Clinic

Michail Ignatiadis and Christos Sotiriou

Department of Medical Oncology, Translational Research Unit,Institut Jules Bordet, Universite Libre de Bruxelles, Brussels, Belgium

INTRODUCTION

The advent of microarray technology and the sequencing of the human genome

have enabled scientists to simultaneously interrogate the expression of thousands

of genes. These gene expression profiling studies have provided (i) a molecular

classification of breast cancer into clinically relevant subtypes, (ii) new tools to

predict disease recurrence and response to different treatments, and (iii) novel

insight into various oncogenic pathways and the process of metastatic progression.

Breast cancer prognosis has been extensively studied by gene expression

profiling, resulting in different signatures with little overlap in their constituent

genes. In a large meta-analysis of all the above studies, despite the disparity in

their gene lists, molecular subtyping and different prognostic signatures were

shown to carry similar prognostic information. In all these signatures, prolifer-

ation genes were the common driving force with regard to prognostication (1).

Gene expression signatures, despite their current limitations may further

refine breast cancer prognosis and prediction of benefit from systemic therapy,

beyond conventional clinicopathological prognostic and predictive factors (2). In

this chapter, we will not refer to gene expression signatures that predict response

to systemic therapy. We will focus only on genomic predictors related to prognosis

and the way these predictors are expected to impact patient management in the

future and accelerate the transition from the “empirical” to “tailored” oncology.

197


MOLECULAR CLASSIFICATION OF BREAST CANCERBY GENE EXPRESSION PROFILING

Clinicians have long recognized that the diagnosis of breast cancer includes tumor

types with different natural histories and responses to various treatments. Never-

theless, traditional histopathological characteristics have been unable to capture the

biological heterogeneity of breast tumors. The first studies employing microarray

technology to address this issue were performed by Perou and Sorlie (3–5). They

proposed a molecular classification of breast cancer with at least five subgroups on

the basis the expression patterns of 500 “intrinsic genes.” Three subgroups were

characterized by low to absent expression of the estrogen receptor (ER) and ER-

related genes, the basal-like, the HER2þ, and the normal breast-like subgroup. The

first subtype was characterized by high expression of genes found in myoepithelial

cells in the outer basal layer of a normal breast duct, like keratins 5/6 and 17, and

was therefore named basal-like. The HER2þ subgroup was characterized by high

expression of several genes of the ERBB2 amplicon. The normal breast-like sub-

group showed genes expressed by adipose and nonepithelial tissues and also

revealed a strong expression of basal epithelial genes. The other two subtypes

consisted of ER-positive tumors and expressed genes of the luminal epithelial cells

that are found in the inner layer of a normal breast duct, and, as a result, they were

termed luminal. The luminal A subgroup had the highest expression of the ERa and

the GATA-binding protein 3 gene, and the second subgroup (luminal B and C)

showed low to moderate expression of these genes. To assign each sample to one of

the five molecular subtypes (basal-like, HER2þ/ER�, luminal A, luminal B, nor-

mal breast-like), Hu et al. have created a single sample predictor by calculating the

mean expression profiles (centroids) of a new “intrinsic set of 300 genes” for each

subgroup (6). In a multivariate analysis, the basal-like, HER2þ/ER�, luminal A,

and luminal B subgroups provided important prognostic information beyond those

provided by traditional predictors like tumor size and grade.

Despite using a different microarray platform, Sotiriou et al. confirmed the

consistency and the prognostic relevance of the above molecular classification of

breast cancer in a different patient population (7). Again, ER status was the most

important discriminator of gene expression patterns. Interestingly, although

tumor grade was also strongly reflected in the transcriptome profile of the pri-

mary tumor, this was not the case for tumor size and axillary nodal invasion.

These results confirmed the important role of ER and tumor grade biology in

defining breast cancer molecular subtypes. In conclusion, microarray studies

with different technology platforms and in different patient populations treated

with heterogeneous treatments have been remarkably consistent in reproducing a

similar molecular classification of breast cancer.

However, it should be pointed out that the number of robust breast cancer

molecular subtypes is not certain, and there is no standardizedmethod for assigning

a molecular subtype to a new case (8). Nevertheless, the breast cancer molecular

classification has begun to alter the way clinical trials are designed. As an example,

198 Ignatiadis and Sotiriou


new trials have been designed to test the efficacy of targeted agents only in the

basal-like breast tumors.

GENE EXPRESSION SIGNATURES AS TOOLS FOR IMPROVINGPROGNOSIS IN BREAST CANCER

Besides the above studies aiming at providing a molecular taxonomy of breast

tumors, several independent groups, including our own, have conducted gene

expression profiling studies with the objective of improving upon traditional

prognostic markers used in the clinic for prediction of outcome. Two concep-

tually different, supervised approaches for prognostic marker discovery on a

genome-wide scale have been applied so far, the “top-down” approach and the

“bottom-up” or “hypothesis-driven” approach. The former derives from a

prognostic model simply by looking for gene expression patterns associated with

the clinical outcome without any a priori biological scenario, whereas the latter

approach first identifies gene expression profiles linked with a specific biological

phenotype and subsequently correlates these findings to survival.

Using the top-down approach, researchers from the Netherlands Cancer

Institute (NKI) first identified a 70-gene prognostic signature in a series of 78

systemically untreated node-negative breast cancer patients. This signature

(MammaPrint1 Agendia, Amsterdam, the Netherlands) was then validated on a

larger set of 295 young treated and untreated patients, with either node-negative

or -positive breast tumors, from the same institution (9,10). The Rotterdam group

later identified, using a training set of 115 breast cancer patients, a 76-gene

signature to predict the development of distant metastases in untreated node-

negative breast cancer patients (11). Interestingly, this group used the ER status

to divide the training set into two subgroups, and the genes selected from each

subgroup (60 and 16 genes for the ER-positive and ER-negative subgroups,

respectively) were combined to form a single signature. The 76-gene signature

(VDX2 array, Veridex LLC, Warren, New Jersey, U.S.) was recently validated in

180 lymph node–negative, untreated patients from multiple institutions (12). The

two groups have used different technology platforms (the NKI group used the

Agilent platform, Santa Clara, California, U.S. and the Rotterdam group used

the Affymetrix platform, Santa Clara, California, U.S.), and although there was

only a three-gene overlap between them, both signatures outperformed the

National Institutes of Health (NIH) (13) and St. Gallen criteria (14) in stratifying

patients according to risk of relapse. Indeed, while the prognostic signatures

correctly identified those patients who will relapse, they could also identify

many low-risk patients who could be spared from unnecessary chemotherapy.

However, the identification of high-risk patients could still be improved, since

half of the patients assigned to the poor-signature group, in fact, did not relapse.

A major challenge for gene expression profiling studies, especially those

with clear clinical implications, is independent validation. Consequently,

TRANSBIG, the translational research network of the Breast International Group

Taking Prognostic Signatures to the Clinic 199


(BIG), conducted an independent validation study of these two prognostic sig-

natures in a series of 302 untreated patients from five different centers and across

different statistical facilities (15,16). The two signatures were robust, and their

performance was reproducible and similar in this retrospective series. Interest-

ingly, both signatures were strong predictors of the development of metastasis

within five years of diagnosis (early metastasis), but their performance decreased

for the detection of late metastasis. This could be easily explained, since the

training set in both studies consisted of patients who have all relapsed within five

years. However, an intriguing question arises about the potential differences in

the biology and the oncogenic pathways that drive early and late metastasis.

Another group explored the correlation between the expression of 250

genes (selected from publicly available microarray data linked with breast cancer

biology) and the clinical outcome in 447 patients from three studies including the

tamoxifen-only group of the NASBP trial B-20 to build the Oncotype DX1

(Genomic Health Inc, Redwood City, California, U.S.) Recurrence Score1 (RS)

(17). This score (low, intermediate, and high) was based on the expression of a

final selected panel of 16 cancer-related genes and five reference genes, as

assessed by reverse transcriptase-polymerase chain reaction (RT-PCR) in par-

affin-embedded tumor tissue. Then, the Oncotype DX RS was validated for its

ability to quantify the likelihood of distance recurrence in the tamoxifen treated,

ER-positive, node-negative group of the NSABP trial B-14 (17). Using the RS,

around 50% of the patients in the above study were characterized as low risk

(low RS), with a 6.8% probability of distant recurrence at 10 years.

Using the second approach, the hypothesis-driven approach, Chang et al.

tried to identify the molecular similarities between cancer invasion/metastasis

and wound healing (18). They developed a gene signature to characterize the

response of fibroblasts to serum. When they applied these genes to primary

gene expression profiling data from multiple data sets, they observed that the

presence of a wound-healing phenotype in the primary tumor was related with

worse clinical outcome in various cancers, including breast cancer. Similar to the

70- and 76-gene signatures, the wound-response signature provided better risk

stratification than the NIH and St. Gallen criteria. Furthermore, it was able to

refine prognosis in the 295 patient data sets of van de Vijver et al. (10) beyond

existing clinicopathological risk factors and the 70-gene signature (19). In

conclusion, Chang et al., using a hypothesis-driven approach, apart from iden-

tifying a gene expression prognostic signature, have offered novel insight into

the molecular pathogenesis of metastasis by linking this process with wound

healing.

Another example of deriving a prognostic gene expression signature using

a hypothesis-driven approach is the study by our group (20) that focused on

histological grade, a well-established pathological parameter rooted in the cell

biology of breast cancer. Applying a supervised analysis, we developed a Gene

expression Grade Index (GGI) score based on 97 genes. These genes were

mainly involved in cell cycle regulation and proliferation and were consistently



differentially expressed between low- and high-grade breast carcinomas. Inter-

estingly, the GGI, which essentially quantifies the degree of similarity between

the tumor expression pattern of these 97 genes and tumor grade, was able to

reclassify patients with histological grade 2 tumors into two groups with distinct

clinical outcomes similar to those of histological grades 1 and 3, respectively

(20). This observation challenges the existence and clinical relevance of an

intermediate grade classification. When we explored the implications of the joint

distribution of ER status and GGI in predicting clinical outcome, we found that

almost all ER-negative tumors with poor clinical outcome were associated with a

high GGI score (high grade), whereas ER-positive tumors were associated with a

heterogeneous mixture of gene expression grade values. Recently, our group has

reported that GGI can be used to define two clinically relevant molecular sub-

types in ER-positive breast cancer (21). These two ER-positive molecular sub-

groups (high and low genomic grade) were highly comparable to the luminal A

and B classifications and were associated with distinct clinical outcome in

systemically untreated or tamoxifen-only-treated ER-positive breast cancer

populations involving more than 650 patients (21). GGI appeared to be the

strongest predictor of clinical outcome, highlighting the prognostic importance

of proliferation genes in ER-positive subgroups, as have been already reported

(17). Indeed, proliferation was one of the five components associated with the

highest weight in the Oncotype DX RS.

Four additional studies have used the hypothesis-driven approach to build

gene predictors of clinical outcome. Oh et al. used estrogen-induced genes,

identified by treating the MCF-7 cell line with 17b-estradiol to build a predictor

of clinical outcome on a training set of 65 patients with ER-positive breast

tumors (22). A total of 822 genes, mainly proliferation and antiapoptosis genes

could separate these tumors into two subgroups with different clinical outcomes.

This gene list was then mapped in three published datasets (4,5,19,23) and an

average expression index or “centroid” was created that could classify the ER-

positive tumors into two clinically relevant subgroups. In another study, Miller et

al. generated a 32-gene expression signature that distinguishes p53 mutant and

wild type tumors. The aim of this research was to provide a more accurate

assessment of the functional status of the p53 pathway than that provided by

mutation analysis using sequencing (24). It should be noted that none of the

signature genes were known transcriptional targets of p53, nor have they been

related to the p53 pathway. Instead, they were mainly proliferation-related genes,

various transcription factors, and ion transport genes. Once more, the prolifer-

ation genes were a prominent part of this signature. The signature outperformed

p53 mutation status, as identified by sequencing in prognosis and prediction of

response to hormonal therapy and chemotherapy. By combining the hypothesis-

driven approach with comparative cross-species functional genomics, Glinsky

et al. identified an 11-gene “death-from-cancer” signature (25). This signature

can identify a lethal subset of human cancers of various origins (including breast

cancer) with a high propensity to metastasize. These tumors display the



“phenotype” of an activated BMI-1 oncogenic pathway that is essential for the

self-renewal of normal stem cells. Again, proliferation- and cell cycle–regulating

genes comprise the most essential part of this signature. Along the same lines,

Liu et al. isolated and compared at the transcriptional-level tumorigenic CD44þ/

CD24�/low breast cancer cells derived from a small set of six breast cancer

patients with normal breast epithilium (26). They developed an “invasiveness”

gene signature (IGS) that was associated with the risk of death and metastasis not

only in breast cancer but also in other tumor types. Again, there was a small gene

overlap between IGS and other prognostic signatures, and, not surprisingly, its

prognostic power was significant only in ER-positive and intermediate-grade

tumors, a feature that had already been observed for other prognostic profiles.

CAN WE INTEGRATE THE DIFFERENT GENE EXPRESSION SIGNATURES?

The above gene expression signatures developed to improve breast cancer

prognostication have little overlap in their constituent genes. A representative

list of these signatures is depicted in Table 1. The inevitable question is whether

the “intrinsic subtypes” developed by unsupervised cluster analysis and the

other prognostic signatures have also little overlap in the prognostic information

they convey. Fan et al. recently compared the prognostic ability of five gene

expression–based models—intrinsic subtypes (6), 70-gene profile (9), wound

response (19), RS (17), and two-gene ratio (23)—in a single dataset of

295 patients and found that four of five predictors had significant agreement in

outcome predictions for individual patients (27). However, this study was limited

to only one dataset of 295 patients, examined only five gene expression signatures,

and did not investigate the biology that may drive prognosis among the five

signatures. To address this issue, a large meta-analysis of publicly available gene

expression and clinical data from 2833 patients (4–7,9,12,15,20,23,24,28–39) with

breast cancer was performed by the Swiss Institute of Bioinformatics in collab-

oration with our group (Table 1). In this meta-analysis, the concept of “co-

expression modules” (comprehensive list of genes with highly correlated

expression) was used to describe important biological processes in breast cancer,

like proliferation, ER and HER2 signaling, tumor invasion, and immune response.

In this study, we sought to depict the connection between these modules, the

previously reported molecular classification, several gene prognostic classifiers,

and the most established clinicopathological variables. A number of interesting

conclusions were drawn from this collaborative effort. First, the disparity of gene

lists of the various gene signatures may be attributed to (i) heterogeneity in patient

populations studied, (ii) the use of different microarray platforms with different

probe sets and different methods for data normalization, and (iii) sampling vari-

ation due to small sample size relative to the number of genes examined. Second,

using these co-expression modules, instead of the five molecular portraits reported

with the intrinsic genes, breast tumors were now grouped into three main sub-

groups namely ER�/HER�, HER2þ, and ERþ/HER2�. Third, the ER�/HER�,



Tab

le1

GeneExpressionSignaturesRelated

toBreastCancerPrognosis

Geneexpression

signatures

Biological

scenario

Microarrayplatform

Number

of

genes

inthe

signature

Independent

validation

Prospective

clinicalvalidation

Ref.

70-genesignature

Clinical

outcome

Agilent(oligonucleotides)

70

yes

Yes

(MIN

DACT

Trial)

(9,10,15)

76-genesignature

Clinical

outcome

Affymetrix

(oligonucleotides)

76

yes

No

(11,12,16)

Recurrence

Score

Clinical

outcome

RT-PCR

21

yes

Yes

(TAILORX

trial)

(17,37)

Wound-response

signature

Woundhealingandtumor

progression

CDNA

(custom

made)

512

yes

No

(18,19)

Genomic

grade

Histological

gradeandtumor

progression

Affymetrix

(oligonucleotides)

97

yes

No

(20)

P-53signature

Functional

statusofp-53

Affymetrix

(oligonucleotides)

32

yes

No

(24)

Death-from-cancer

signature

Bmi-1oncogenic

pathway

self-renew

al

Affymetrix

(oligonucleotides)

11

Yes

No

(25)

Invasivenessgene

signature

Tumorigenic

cancercells

CD44þ/

CD24�,

low

Affymetrix

(oligonucleotides)

186

Yes

No

(26)



HER2þ tumors were characterized by high proliferation, whereas the

ERþ/HER2� was divided in two subgroups, the ERþ/low proliferation and the

ERþ/high proliferation tumors resembling to luminal A and B subtypes,

respectively. Fourth, 10 prognostic signatures [RS (17), 70-gene signature (9),

76-gene signature (11), wound response (18), p53 signature (24), GGI (20),

NCH70 (33), CON52 (38), CCYC (39), ZCOX], despite the disparity in their

gene lists, showed similar prognostic performance (Fig. 1). It is interesting to

note that they all identified as “low risk” the ERþ subtype associated with low

proliferation. Fifth, proliferation was the common driving force responsible for

the performance of the various prognostic signatures (Fig. 1). Indeed, to further

investigate the role of proliferation genes in relation to the different prognostic

signatures, we divided each signature into two “partial signatures”—one with only

proliferation genes and the other with the complementary nonproliferation genes.

Interestingly, when only proliferation genes were used, the total performance was

not degraded. In contrast, the nonproliferation partial signatures typically show

degraded performance. Sixth, combining the signatures did not improve the per-

formance as expected from the high concordance in their classifications. Finally,

in the multivariate analysis, including the genomic predictor derived from co-

expressionmodules, the standard clinicopathological variables, namely tumor size

and nodal status, retained their prognostic value.

PROSPECTIVE VALIDATION OF TWO GENOMIC PREDICTORS

The initial promising results that breast cancer prognostication can be improved

using gene expression signatures need to be validated by well-designed pro-

spective clinical trials. Two gene expression predictors, the 70-gene expression

Figure 1 Despite the disparity in their gene lists, different prognostic signatures showed

similar prognostic performance. Proliferation is the common driving force responsible for

their prognostic performance.



signature, (MammaPrint) discussed earlier and the Oncotype DX RS have

reached the final step of prospective clinical testing. The BIG and the EORTC

have jointly launched the MINDACT (Microarray for Node-Negative Disease

may Avoid Chemotherapy) trial that directly compares two approaches for

deciding whether to offer adjuvant chemotherapy in node-negative breast cancer

patients, i.e., either using the prognostic information provided by standard

clinicopathological criteria (control arm) or by the 70-gene expression signature

(MammaPrint) (experimental arm). Six thousand women will participate, and

the primary endpoint is to prove that five-year disease-free survival in the

experimental arm will not be inferior to that of the control arm; at the same time,

it is estimated that 10% to 15% fewer women will be treated with chemotherapy

in the experimental arm.

A similarly ambitious initiative is taking place in the United States. The

Oncotype DX RS is being evaluated in the TAILORx [Trial Assigning

IndividuaLized Options for Treatment (Rx)] trial, which is part of the

National Cancer Institute (NCI) Program for the Assessment of Clinical

Cancer Tests (PACCT). This trial randomizes patients with intermediate RS

to receive either hormonal therapy alone or hormonal therapy plus chemo-

therapy. In it, patients with low RS will be treated with hormonal therapy

alone and those with high RS will receive chemotherapy plus hormonal

therapy.

These two trials will provide level I evidence that these gene expression

predictors provide prognostic information beyond standard clinicopathological

factors and are ready for use in clinical practice. Both trials will provide an

excellent opportunity to address issues related to tissue handling and shipping,

reproducibility, quality controls, and standardization of the microarray experi-

ments. Furthermore, they will provide a valuable database for future translational

research.

FUTURE DIRECTION–CONCLUSIONS

The real challenge, once the above trials prospectively validate the two genomic

predictors, will be to integrate these predictors into a comprehensive clinical

decision-making algorithm that will lead toward personalized medicine.

Parameters that need to be included in this algorithm include the following:

Traditional Clinicopathological Parameters

These include patient comorbidities, age, and primary tumor characteristics like

tumor size, tumor grade, nodal status, ER status, HER2/neu status, and lym-

phovascular invasion. Many of these parameters are incorporated in various

prognostic tools like the Nottingham Prognostic Index (40), Adjuvant! Online

(41), and the NIH (13) and St. Gallen (42) Guidelines.



Primary Tumor Gene Expression Signatures

From the above discussion, it is apparent that two genomic predictors are in the

late phase of clinical validation. However, apart from the question of who can

be spared unnecessary adjuvant chemotherapy (prognosis), a more important

question is which therapy works better in an individual patient (prediction).

Genomic predictors of response to multidrug regimens and newer targeted agents

have also been reported (43–49), but their description is beyond the scope of this

chapter. Other important technology platforms are being developed to analyze

epigenetic changes in DNA (50), the role of microRNAs (51) that will provide us

with a more comprehensive picture of the biology of the tumor and further refine

prognosis and prediction.

Micrometastatic Cells

The presence of cytokeratin positive tumor cells in the bone marrow of early breast

cancer patients has been correlated with worse outcome in a meta-analysis

involving more than 4500 patients (52). We have reported that the detection with a

real-time RT-PCR assay of Cytokeratin-19 (CK19) mRNA-positive circulating

tumor cells (CTCs) in patients with early breast cancer before (53,54) and after

(55,56) the initiation of adjuvant systemic treatment is associated with poor out-

come. However, the additional prognostic information of micrometastatic cells to

primary tumor gene expression profiling data has not yet been examined.Moreover,

there are pilot studies where HER2-positive CTCs and/or disseminated tumor cells

(DTCs) were successfully eliminated with trastuzumab (57), suggesting that CTCs

and/or DTCs may be tested also as tools to predict response to a particular drug.

Molecular Imaging

PET (positron emission tomography) scan using novel PET tracers that target

critical cellular processes like angiogenesis may provide important staging and

functional information and is under investigation in breast cancer (58).

Pharmacogenetics (Host)

The importance of pharmacogenetics in predicting clinical outcome in breast

cancer has recently been illustrated in studies focused on the polymorphism of

enzymes that metabolize tamoxifen (59,60). The study of the host is an important

piece of the clinical decision-making algorithm.

Proteomics

Proteomics provides the hope of discovering novel biological markers that can

be used for early detection, disease diagnosis, prognostication, and prediction of

response to therapy in early breast cancer (61).



Adjuvant Treatment

It is well documented that adjuvant treatment affects clinical outcome of breast

cancer patients (62). From that perspective, the positive adjuvant trials with the

aromatase inhibitors (63) and trastuzumab (64,65), the concept of dose-dense

adjuvant chemotherapy (66), and the emerging data about the use of taxanes

(67,68) and anthracyclines (69) in the adjuvant setting are rendering both the

treatment decision-making process and prediction of outcome increasingly more

complex.

Several efforts have been made to combine at least some of the above

parameters to predict the clinical outcome of breast cancer patients. For instance,

Pittman et al. have used a modeling approach based on classification trees to

combine clinical and genomic data (70). They nicely showed that by combining

multiple metagenes (defined as summary measures of the expression of subsets

of potentially related genes) with nodal status, they could create integrative

clinicogenomic models. These models provided improved prediction accuracy

compared with either clinical or genomic data alone. Another example of such an

effort is the Adjuvant! Online for breast cancer, Genomic Version 7.0, which

combines traditional clinical data, the RS, and information about adjuvant

treatment to provide an estimate of clinical outcome in patients with early breast

cancer (71).

However, the accumulation of highly complex and multilevel information

will ultimately impose specific data integration requirements. So far these dif-

ferent types of data are not only maintained in a variety of different data sources

but are also managed by different complex data management and analytical

systems. Additionally, they involve many institutions and numerous complex

workflows. To address all these critical issues, two initiatives have been

launched that intend to integrate all this information into one comprehensive

biomedical platform—the cancer Biomedical Informatics Grid, “caBIG” (https://

cabig.nci.nih.gov/) and the Advancing Clinico Genomic Trials, “ACGT” (http://

www.eu-acgt.org/) funded by the NCI and European Union, respectively.

In the future, medical oncologists should use a clinically validated treat-

ment decision-making algorithm that will take into account all this highly

complex and rapidly evolving, multilevel information. In this way, we will move

from the “empirical” to the new exciting era of “individual-tailored” oncology.

REFERENCES

1. Sotiriou C, Wirapati P, Kunkel S, et al. Meta-analysis of gene-expression profiles in

breast cancer: towards a unified understanding of breast cancer sub-typing and

prognosis signatures. Breast Cancer Res Treat. 2007; 106(suppl 1):1047(abstr).

2. Ignatiadis M, Desmedt C. Predicting risk of breast cancer recurrence using gene-

expression profiling. Pharmacogenomics 2007; 8(1):101–111.


Nature 2000; 406(6797):747–752.





U S A 2001; 98(19):10869–10874.

5. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes

in independent gene expression data sets. Proc Natl Acad Sci U S A 2003; 100(14):

8418–8423.

6. Hu Z, Fan C, Oh DS, et al. The molecular portraits of breast tumors are conserved

across microarray platforms. BMC Genomics 2006; 7:96.



U S A 2003; 100(18):10393–10398.

8. Andre F, Pusztai L. Molecular classification of breast cancer: implications for

selection of adjuvant chemotherapy. Nat Clin Pract Oncol 2006; 3(11):621–632.


clinical outcome of breast cancer. Nature 2002; 415(6871):530–536.


predictor of survival in breast cancer. N Engl J Med 2002; 347(25):1999–2009.

11. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant

metastasis of lymph-node-negative primary breast cancer. Lancet 2005; 365(9460):

671–679.



Oncol 2006; 24(11):1665–1671.

13. Eifel P, Axelson JA, Costa J, et al. National Institutes of Health Consensus Devel-

opment Conference Statement: adjuvant therapy for breast cancer, November 1–3,

2000. J Natl Cancer Inst 2001; 93(13):979–989.

14. Goldhirsch A, Wood WC, Gelber RD, et al. Meeting highlights: updated interna-

tional expert consensus on the primary therapy of early breast cancer. J Clin Oncol

2003; 21(17):3357–3365.

15. Buyse M, Loi S, van ’t Veer LJ, et al. Validation and clinical utility of a 70-gene

prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst

2006; 98(17):1183–1192.

16. Desmedt C, Piette F, Loi S, et al. Strong time-dependency of the 76-gene prognostic

signature for node-negative breast cancer patients in the TRANSBIG multi-centre

independent validation series. Clin Cancer Res 2007; 13:3207–3214.



18. Chang HY, Sneddon JB, Alizadeh AA, et al. Gene expression signature of fibroblast

serum response predicts human cancer progression: similarities between tumors and

wounds. PLoS Biol 2004; 2(2):E7.

19. Chang HY, Nuyten DS, Sneddon JB, et al. Robustness, scalability, and integration of

a wound-response gene expression signature in predicting breast cancer survival.

Proc Natl Acad Sci U S A 2005; 102(10):3738–3743.


understanding the molecular basis of histologic grade to improve prognosis. J Natl

Cancer Inst 2006; 98(4):262–272.



21. Loi S, Haibe-Kains B, Desmedt C, et al. Definition of clinically distinct molecular

subtypes in estrogen receptor-positive breast carcinomas through genomic grade.

J Clin Oncol 2007; 25(10):1239–1246.

22. Oh DS, Troester MA, Usary J, et al. Estrogen-regulated genes predict survival in

hormone receptor-positive breast cancers. J Clin Oncol 2006; 24(11):1656–1664.

23. Ma XJ, Wang Z, Ryan PD, et al. A two-gene expression ratio predicts clinical

outcome in breast cancer patients treated with tamoxifen. Cancer Cell 2004; 5(6):

607–616.

24. Miller LD, Smeds J, George J, et al. An expression signature for p53 status in human

breast cancer predicts mutation status, transcriptional effects, and patient survival.


25. Glinsky GV, Berezovska O, Glinskii AB. Microarray analysis identifies a death-

from-cancer signature predicting therapy failure in patients with multiple types of

cancer. J Clin Invest 2005; 115(6):1503–1521.

26. Liu R, Wang X, Chen GY, et al. The prognostic role of a gene signature from

tumorigenic breast-cancer cells. N Engl J Med 2007; 356(3):217–226.

27. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based pre-

dictors for breast cancer. N Engl J Med 2006; 355(6):560–569.

28. Calza S, Hall P, Auer G, et al. Intrinsic molecular signature of breast cancer in a

population-based cohort of 412 patients. Breast Cancer Res 2006; 8(4):R34.

29. Pawitan Y, Bjohle J, Amler L, et al. Gene expression profiling spares early breast

cancer patients from adjuvant therapy: derived and validated in two population-based

cohorts. Breast Cancer Res 2005; 7(6):R953–R964.


outcomes. Lancet 2003; 361(9369):1590–1596.

31. Huang E, Ishida S, Pittman J, et al. Gene expression phenotypic models that predict

the activity of oncogenic pathways. Nat Genet 2003; 34(2):226–230.

32. Korkola JE, DeVries S, Fridlyand J, et al. Differentiation of lobular versus ductal breast

carcinomas by expression microarray analysis. Cancer Res 2003; 63(21): 7167–7175.

33. Naderi A, Teschendorff AE, Barbosa-Morais NL, et al. A gene-expression signature

to predict survival in breast cancer across independent data sets. Oncogene 2007;

26(10):1507–1516.

34. Farmer P, Bonnefoi H, Becette V, et al. Identification of molecular apocrine breast

tumours by microarray analysis. Oncogene 2005; 24(29):4660–4671.

35. Richardson AL, Wang ZC, De NA, et al. X chromosomal abnormalities in basal-like

human breast cancer. Cancer Cell 2006; 9(2):121–132.

36. Espinosa E, Vara JA, Redondo A, et al. Breast cancer prognosis determined by gene

expression profiling: a quantitative reverse transcriptase polymerase chain reaction

study. J Clin Oncol 2005; 23(29):7278–7285.



2006; 24(23):3726–3734.

38. Teschendorff AE, Naderi A, Barbosa-Morais NL, et al. A consensus prognostic gene

expression classifier for ER positive breast cancer. Genome Biol 2006; 7(10):R101.

39. Whitfield ML, George LK, Grant GD, et al. Common markers of proliferation. Nat

Rev Cancer 2006; 6(2):99–106.



40. Galea MH, Blamey RW, Elston CE, et al. The Nottingham Prognostic Index in

primary breast cancer. Breast Cancer Res Treat 1992; 22(3):207–219.

41. Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the

prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23(12):

2716–2725.

42. Goldhirsch A, Glick JH, Gelber RD, et al. Meeting highlights: international expert

consensus on the primary therapy of early breast cancer 2005. Ann Oncol 2005;

16(10):1569–1583.






Oncol 2005; 23(6):1169–1177.

45. Jansen MP, Foekens JA, van Staveren IL, et al. Molecular classification of tamox-

ifen-resistant breast carcinomas by gene expression profiling. J Clin Oncol 2005;

23(4):732–740.



and cyclophosphamide in breast cancer. J Clin Oncol 2006; 24(26):4236–4244.

47. Bild AH, Yao G, Chang JT, et al. Oncogenic pathway signatures in human cancers as

a guide to targeted therapies. Nature 2006; 439(7074):353–357.


motherapeutics. Nat Med 2006; 12(11):1294–1300.

49. Solit DB, Garraway LA, Pratilas CA, et al. BRAF mutation predicts sensitivity to

MEK inhibition. Nature 2006; 439(7074):358–362.

50. Herman JG, Baylin SB. Gene silencing in cancer in association with promoter

hypermethylation. N Engl J Med 2003; 349(21):2042–2054.

51. Iorio MV, Ferracin M, Liu CG, et al. MicroRNA gene expression deregulation in

human breast cancer. Cancer Res 2005; 65(16):7065–7070.

52. Braun S, Vogl FD, Naume B, et al. A pooled analysis of bone marrow micro-

metastasis in breast cancer. N Engl J Med 2005; 353(8):793–802.

53. Xenidis N, Perraki M, Kafousi M, et al. Predictive and prognostic value of peripheral

blood cytokeratin-19 mRNA-positive cells detected by real-time polymerase chain

reaction in node-negative breast cancer patients. J Clin Oncol 2006; 24(23):3756–3762.

54. Ignatiadis M, Xenidis N, Perraki M, et al. Different prognostic value of Cytokeratin-

19 mRNA-positive Circulating Tumor Cells according to estrogen receptor status in

early breast cancer. J Clin Oncol 2007; 25(33):5194–5202.

55. Xenidis N, Vlachonikolis I, Mavroudis D, et al. Peripheral blood circulating cyto-

keratin-19 mRNA-positive cells after the completion of adjuvant chemotherapy in

patients with operable breast cancer. Ann Oncol 2003; 14(6):849–855.

56. Xenidis N, Mavroudis D, Apostolaki S, et al. Effect of adjuvant treatment on the

circulating CK19mRNA positive tumor cells in patients with early stage breast

cancer. Proc Am Soc Clin Oncol 2006; 24(No18S):20012 (abstr).

57. Bozionellou V, Mavroudis D, Perraki M, et al. Trastuzumab administration can

effectively target chemotherapy-resistant cytokeratin-19 messenger RNA-positive

tumor cells in the peripheral blood and bone marrow of patients with breast cancer.

Clin Cancer Res 2004; 10(24):8185–8194.



58. Quon A, Gambhir SS. FDG-PET and beyond: molecular breast cancer imaging.

J Clin Oncol 2005; 23(8):1664–1673.



Oncol 2005; 23(36):9312–9318.

60. Nowell S, Sweeney C, Winters M, et al. Association between sulfotransferase 1A1

genotype and survival of breast cancer patients receiving tamoxifen therapy. J Natl

Cancer Inst 2002; 94(21):1635–1640.

61. Somiari RI, Somiari S, Russell S, Shriver CD. Proteomics of breast carcinoma.

J Chromatogr B Analyt Technol Biomed Life Sci 2005; 815(1–2):215–225.

62. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of chemo-

therapy and hormonal therapy for early breast cancer on recurrence and 15-year

survival: an overview of the randomised trials. Lancet 2005; 365(9472):1687–1717.

63. Winer EP, Hudis C, Burstein HJ, et al. American Society of Clinical Oncology

technology assessment on the use of aromatase inhibitors as adjuvant therapy for

postmenopausal women with hormone receptor-positive breast cancer: status report

2004. J Clin Oncol 2005; 23(3):619–629.

64. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al. Trastuzumab after adjuvant

chemotherapy inHER2-positive breast cancer. N Engl JMed 2005; 353(16):1659–1672.

65. Romond EH, Perez EA, Bryant J, et al. Trastuzumab plus adjuvant chemotherapy for

operable HER2-positive breast cancer. N Engl J Med 2005; 353(16):1673–1684.

66. Citron ML, Berry DA, Cirrincione C, et al. Randomized trial of dose-dense versus

conventionally scheduled and sequential versus concurrent combination chemo-

therapy as postoperative adjuvant treatment of node-positive primary breast cancer:

first report of Intergroup Trial C9741/Cancer and Leukemia Group B Trial 9741.

J Clin Oncol 2003; 21(8):1431–1439.

67. Mamounas EP, Bryant J, Lembersky B, et al. Paclitaxel after doxorubicin plus

cyclophosphamide as adjuvant chemotherapy for node-positive breast cancer: results

from NSABP B-28. J Clin Oncol 2005; 23(16):3686–3696.

68. Martin M, Pienkowski T, Mackey J, et al. Adjuvant docetaxel for node-positive

breast cancer. N Engl J Med 2005; 352(22):2302–2313.

69. Piccart-Gebhart MJ. Anthracyclines and the tailoring of treatment for early breast

cancer. N Engl J Med 2006; 354(20):2177–2179.

70. Pittman J, Huang E, Dressman H, et al. Integrated modeling of clinical and gene

expression information for personalized prediction of disease outcomes. Proc Natl

Acad Sci U S A 2004; 101(22):8431–8436.

71. Adjuvant! For Breast Cancer, Genomic Version 7.0. http://www.adjuvantonline.com/

breastgenomic.jsp 2006.



15

Tissue is the Issue

Adekunle Raji

Eastern Cooperative Oncology Group Pathology Coordinating Office,Northwestern University, Evanston, Illinois, U.S.A.

INTRODUCTION

All eyes are on tissue. The biggest challenge facing biomedical discoveries today

is the issue with tissue. The era of personalized, individually tailored medicine has

begun; therefore, cells, tissue and its intra- and interenvironment, are the focus of

pharmacogenetics. The intracellular kinetic and intercellular dynamics are key

factors in the understanding of cell behavior before, during, and after therapy.

Undoubtedly the solid type of tissue carries the bulk of these issues. Other

fluid forms of tissue, however, do have their own share of these issues, of course.

Tissue will be defined here as biological material comprising a network of

characteristically diverse cells. In essence, blood, urine, surgical resections,

biopsies, marrow, lavage, hair, nail, and other specimen taken from a person will

be considered as tissue. Turn to any biomedical community member and most, if

not all, will have some sort of opinion about tissue. Between the ultimate giver

(patients) and ultimate user (laboratorians) lie the challenges. To the patient, care

team, and several entities, the tissue is the issue, and to the laboratories, tech-

nologists, and the interpreters of results, the issue is really the tissue.

In the past two years, several working groups were established not only to

understand current issues as well as future anticipated issues but also to address

them. These working groups are usually comprise various clinical and financial

institutions and experts from multiple disciplines, such as medical scientists,

basic scientists, surgeons, laboratories, pathologists, funding agencies, agent

213


manufacturers, patient advocates, statisticians, bioinformatics, research asso-

ciates, research administrators, medical staff, and clinical administrators. These

groups guide the interdisciplinary patient care team by addressing the pro-

curement, collection, preservation, acquisition, processing, transportation, and

storage of biological materials of high quality and usability that are adequate

for analysis and interpretation. Good news is that specimens with clinical data

are archived as far back as 40 years ago, the bad news is that they were

collected non-uniformaly; today’s good news is that the collection processing

and associated data are undergoing harmonization (uniformity). Contingency

plans are also being developed for future anticipated issues. The key areas of

the tissue issue are obviously procurement, collection, acquisition, processing,

preservation, storage, distribution, and bioinformatics that are successfully

conducted within stringent policies, procedures, rules, and regulations.

PROCUREMENT

A strategy for collecting any biological specimens from patients is needed to

ensure that the behavior of cells or tissue after it leaves the body of the patient is

consistent with what it was while in the patient’s body. Regardless of the

strategy, clear and simple instructions should be written and made available to

the procurement team comprising a preoperation nurse, phlebotomist, surgery

assistant, physician/surgeon, physician assistant, pathology assistant, research

coordinator, patient care nurse, internal specimen transporters, and medical/

research technologist. These instructions are typically included in the local

manual of operational procedures and/or clinical protocols and should provide

details of the type of collection devices such as vacutainers, syringes, punch size,

preservatives, buffers, reagents and the containers needed. The manual must also

include procedures for preparing reagents or buffers such as formalin, RNAlater,

and the optimal cutting temperature (OCT) compound. It must contain infor-

mation about the precollection requirements with regard to time, temperature,

and pressure for certain types of specimens (Table 1)

COLLECTION

There are several choices of collection devices, to the extent that they have

created a great deal of confusion in the medical field. It is not surprising to see

several new clinical trial protocols now restricting collection to the use of

preassembled collection devices, with instructions to assist the procurement

plan and ensure uniform collection, particularly in a multi-institutional study.

While it is possible to obtain serum from redtop, royal blue top, serum separator

top, and plain syringe, it is true that tissue can be fixed in paraformaldhyde,

formalin, acid alcohol, zinc formalin, and many other reagents. Each of these

fixatives presents its own advantages and disadvantages toward downstream

applications such as immunohistochemistry (IHC), in situ hybridization (ISH),

214 Raji


comparative genomic hybridization (CGH), polymerase chain reaction (PCR),

blotting (Western, Northern, Southern), enzyme-linked immunosorbent assay

(ELISA), radio immunoassay (RIA), enzyme immunoassay (EIA), and chem-

iluminescence assay (CLA). It is also true that plasma can be obtained from all

anticoagulated vacutainers as well as heparin-coated syringes. Protein and

nucleic acids can be obtained from biological samples regardless of collection

or preservation buffer. However, experience has shown a great deal of variation

in the yield and quality of sample products, as well as the analytical results

obtained from the samples. It is therefore important to select one device and one

alternate backup for the collection, preservation, and fixation that will provide

the best representation of the cellular content and its content behavior as it was

just before leaving the patient’s body. It is important to select a device and

preservative that are best for yield and quality, as well as most appropriate for

the next step in processing or analysis. Table 2 lists the collection devices,

preservatives, and fixatives that have been selected by large research groups

because of their overall performance success.

Fresh tissue. Nonfrozen and nonfixed tissue should be collected directly

into a labeled sterile screw-cap plastic container, with one of the

following buffers:

1. Phosphate buffered saline (PBS) for tissue culture and cell viability

studies.

2. Neutral buffered formalin (10%) for gradual fixation.

Table 1 Roles and Responsibility

Roles Responsibility

Physician/surgeon including

but not limited to pathologist

and intervention radiologist

Inform patient, order test, perform surgery/biopsy/

imaging/morphology/assay procedures.

Nurse Inform patient, determine timing of procedure/draw.

Draw blood via port.

Phlebotomist, nurse, medical

technologist

Draw blood into devices. Perform aspiration.

Physician/surgery assistant Guide physician and expedite preservation and delivery.

Transporter Ensure timely delivery from point of collection to next

critical step/location while preserving samples.

Research coordinators, project

manager, protocol specialist,

data manager, nurse

coordinator, clinical

research associates

Ensure all team members are familiar with their roles

and have adequate information to assist in

performing task. Liaison among team members as

well as facilitate collaboration.

Medical/research technologist Ensure availability of appropriate supplies for

collection and preservation. Effectively and

efficiently process specimens and analyze samples.

Tissue is the Issue 215


Tab

le2

CommonBloodCollectionTubes

Closure/stopper

color

Anticoagulantoradditive

Number

of

inversionsto

mix

Exam

plesforuse

Red/red

inglass

tube

None

None

Serum

forchem

istry,proteomics,im

munology,and

banking

Red/red

inplastic

tube

Clotactivator

5–10

Serum

forchem

istry,proteomics,im

munology,and

banking

Gold/red

marbledSST

Clotactivatorandseparationgel

5–10

Serum

forchem

istry,proteomics,im

munology,and

banking.Rem

ote

collection.

Lavender/lavender

(purple)

Sodium

EDTA

(dried),or

potassium

EDTA

(liquid)

5–10

Whole

bloodforhem

atology,certainwhole

blood

chem

istries,e.g.,hem

oglobin

A1corbloodtyping.

Proteomics,floatingDNA,genomic

DNA,cells.

Lightblue/lightblue

Liquid

3.2%

sodium

citrate,aor

3.8%

sodium

citrate

5–10

Plasm

aforcoagulation,Plateletpoorplasm

a.Growth

factors,fragileproteins,viable

cells.

Cellpreparation

tubeCPT

Liquid

3.2%

sodium

citrate,aor

3.8%

sodium

citrate.

Ficoland

gel

separator

5–10

Plasm

aforcoagulation,Plateletpoorplasm

a.Growth

factors,fragileproteins,viable

cells.Rem

ote

collectionofviable

mononuclearcells.

PaxgeneRNA/DNA

Preanalytic/Qiagen

nucleicacid

stabilizer

5–10

Genomic

expressions.

P100

Protein

stabilizer

5–10

Smallsize

protein

measurement.Proteomic

studies.

Tan

Sodium

heparin

(glass),or

potassium

EDTA

(plastic)

5–10

Wholebloodforlead,heavymetals,toxicologystudies.

216 Raji


Tab

le2

CommonBloodCollectionTubes

(Continued

)

Closure/stopper

color

Anticoagulantoradditive

Number

of

inversionsto

mix

Exam

plesforuse

Green/green

Lithium

heparin,or

5–10

Plasm

aorwhole

bloodforchem

istry,PK,andPD.

Sodium

heparin

5–10

Plasm

aforchem

istry

Lightgreen/green

marbled

Lithium

heparin

andseparationgel

5–10

Plasm

aforchem

istry

Gray/gray

Potassium

oxalate/sodium

fluoride,

orsodium

fluoride,

orlithium

iodoacetate,

orlithium

iodoacetate/lithium

heparin

5–10

Serum

orplasm

aforglucose

oralcoholrequiring

inhibitionofglycolysis,C-peptides.

Darkroyal

blue

Sodium

heparin,EDTA,orplain

5–10

Plasm

aorserum

fortraceelem

ents,toxicology,and

nutrients.

Paleyellow/brightyellow

Sodium

polyanetholesulfonate

5–10

Bloodcultures,microbiology,andinfectiousdiseases

markers.

Yellow

Acidcitratedextrose

(ACD)

5–10

Bloodbank,viable

cellsalternateto

CPTandCellSave

Orange/yellow

and

marbled

Clotactivator(thrombin)

5–10

Serum

forchem

istry,ELISA

Pink/pink

Potassium

EDTA

5–10

Molecular/viral

load

testing,bloodbank,hem

atology.

awww.bd.com

Abbreviations:SST,serum

separatortop;PK,pharmacokinetic;PD,pharmacodynam

ic.



3. RPMI (Roswell Park Memorial Institute) media for tissue culture, cell

viability studies, protein extraction, and multiphoton studies.

4. RNAlater for RNA extraction.

5. OCT bottom and top prior to freezing.

PROCESSING

Generally, processing methods that will ensure optimal quality and yield should be

selected for the processing of specimens and samples into next product aliquot.

Blood. It is important to perform all centrifugation in refrigerated and

temperature-controlled chambers. This is to avoid unnecessary heat

incubation of samples while spinning. In some instances, it will be

necessary to spin the sample at a slow speed followed by high speed to

eliminate platelets. Selection of complex processes to obtain delicate

and fragile by-products is important. Aliquoting into the smallest

useable quantity to avoid freeze-thaw-freeze cycle is also necessary.

For example, serum and plasma can be apportioned into 200 mL, DNAand RNA can be apportioned into 75 mg and 1 mg, respectively.

Fresh frozen tissue. OCT-protected frozen tissue should be cryosectioned

for hematoxylin and eosin staining to assess tumor area and percentage

as part of the annotation. The information will guide case selection as

well as steps in macro and micro dissection. Integrity of the sample must

be maintained through sterile processes to avoid degradation and/or

contamination. The tissue must be kept frozen until submersed into

extraction buffer. There are many choices of methods for the extraction

of cellular components such as RNA, DNA, and proteins.

Fixed tissue. Tissue biopsies and/or resection must undergo complete and

adequate fixation in a buffered environment. This is important for the

morphologic review as well as the evaluation of reference and novel

markers. Neutral buffered formalin has been widely used and is most

appropriate for current and future translational processes. As previously

mentioned, there are many choices of methods for the extraction and

isolation of DNA; however, there are only few methods for the

extraction of RNA and protein from fixed tissue. Care must be taken

in making selections to avoid loss of valuable tissue resource.

Sectioning should be performed with new blades and clean water.

Nonprecipitating formalin should be avoided.

Xylene rather than its substitutes should be used.

Manual rather than automated processes should be used for small,

starting samples.

Washes between steps in any and all processes must be complete to

avoid reagent carry over.

218 Raji


CURRENT NOVEL PROCESSING OF BLOOD AND FIXED TISSUE

Isolation of Mononuclear Cells

Monocytes, a type of white blood cells, are found both in bone marrow and blood.

The following procedure describes the isolation of monocytes in peripheral blood

(peripheral blood mononuclear cells, PBMC).

Steps

1. Centrifuge CPT tubes containing whole blood for 15 minutes at 258C at

2700 rpm.

2. Pipet out the top layer (plasma) and transfer into two labeled tubes and

freeze.

3. Pipet out the lymphocyte layer (the cloudy layer) with a transfer pipette,

placing the layer into a fresh 15-mL tube. Fill tube with room-temperature

PBS.

4. Centrifuge for 10 minutes at 1800 rpm.

5. Discard supernatant and resuspend pellet in 1 mL PBS.

6. Fill tube with PBS, inverting eight times to mix and wash.

7. Centrifuge for 15 minutes at 2000 rpm.

8. Discard supernatant.

9. Proceed to cell sorting or resuspend and freeze pellet in 1 mL of RPMI or

DMSO (dimethyl sulfoxide) for later sort.

Note: For blood collected in any other anticoagulated tube, dilute 1:1 with

RPMI, layer on Histopaque 1077, then proceed with step 1.

Specialized Sorting of Circulating Tumor/Endothelial Cells

1. Magnetic labeling (*45 min)

a. Centrifuge sample for 10 minutes at 300g at room temperature.

b. Carefully remove the supernatant completely.

c. Resuspend cell pellet in appropriate amount of buffer as specified by

the microbeads package insert. For fewer cells, use the same volume.

d. Add microbeads per package insert.

e. Incubate for 15 minutes at 68C to 128C. For less than 5 � 106 total cells,

use the same volume.

f. Wash cells by adding 10 to 20 times the labeling volume of buffer and

centrifuge at 300g for 10 minutes.

g. Remove supernatant completely (save the supernatant in appropriately

labeled 1 mL cryovial) and resuspend cell pellet in appropriate amount

of buffer (1 mL of buffer/108 total cells).

h. Proceed to magnetic separation.



2. Magnetic separation (*60 min)

a. Turn on the manual and/or automated cell sorter.

b. When the machine has finished initializing, perform a column exchange

to remove blank columns and insert magnetic columns. (To save

time, try to perform these functions while the sample and beads

are incubating.)

c. Wipe down probes with alcohol to disinfect.

d. Place the sample in the uptake position, and place empty 15-mL conical

tubes in the rack under the Neg1, Pos1, and Pos2 probes.

e. Choose “Separation” on the main menu.

f. Use the down arrows to select “possel_s” and press “start.”

g. For both the positive and negative cells, separate elution container

must be used (*15 min).

h. Prepare and/or submit samples for next step in processing (DNA extrac-

tion, thin prep slides, etc.).

Extraction of RNA from Peripheral Blood in PAXgene

1. Centrifuge the PAXgene Blood RNA Tube for 10 minutes at 3000g to

5000g using a swing-out rotor.

2. Remove the supernatant by decanting or pipetting. Add 4 mL RNase-free

water to the pellet, and close the tube using a fresh secondary Hemogard

closure.

3. Vortex until the pellet is visibly dissolved, and centrifuge for 10 minutes at

3000g to 5000g using a swing-out rotor. Remove and discard the entire

supernatant.

4. Add 350 mL Buffer BR1, and vortex until the pellet is visibly dissolved.

5. Pipet the sample into a 1.5-mL microcentrifuge tube. Add 300 mL Buffer

BR2 and 40 mL proteinase K. Mix by vortexing for 5 seconds, and incu-

bate for 10 minutes at 558C using a shaker/incubator at 400 to 1400 rpm.

After incubation, set the temperature of the shaker/incubator to 658C (for

step 20).

6. Pipet the lysate directly into a PAXgene Shredder spin column (lilac)

placed in a 2-mL processing tube, and centrifuge for 3 minutes at maxi-

mum speed (but not to exceed 20,000g).

7. Carefully transfer the entire supernatant of the flow-through fraction to a

fresh 1.5-mL microcentrifuge tube without disturbing the pellet in the

processing tube.

8. Add 350 mL ethanol (96–100%, purity grade p.a.). Mix by vortexing, and

centrifuge briefly (1–2 seconds at 500–1000g) to remove drops from the

inside of the tube lid.

9. Pipet 700 mL sample into the PAXgene RNA spin column (red) placed in a

2-mL processing tube and centrifuge for 1 minute at 8000 to 20,000g.

220 Raji


Place the spin column in a new 2-mL processing tube, and discard the old

processing tube containing flow-through.

10. Pipet the remaining sample into the PAXgene RNA spin column, and

centrifuge for 1 minute at 8000 to 20,000g. Place the spin column in a

new 2-mL processing tube, and discard the old processing tube containing

flow-through.

11. Pipet 350 mL Buffer BR3 into the PAXgene RNA spin column. Centrifuge

for 1 minute at 8000 to 20,000g. Place the spin column in a new 2-mL

processing tube and discard the old processing tube containing flow-through.

12. Add 10 mL DNase I stock solution to 70 mL Buffer RDD in a 1.5-mL

microcentrifuge tube. Mix by gently “flicking” the tube and centrifuge

briefly to collect residual liquid from the sides of the tube.

13. Pipet the DNase I incubation mix (80 mL) directly onto the PAXgene RNA

spin column membrane, and place on the bench top (20–308C.) for

15 minutes.

14. Pipet 350 mL Buffer BR3 into the PAXgene RNA spin column, and

centrifuge for 1 minute at 8000g to 20,000g. Place the spin column in a

new 2-mL processing tube and discard the old processing tube containing

flow-through.

15. Pipet 500 mL Buffer BR4 to the PAXgene RNA spin column, and

centrifuge for 1 minute at 8000g to 20,000g. Place the spin column in a

new 2-mL processing tube, and discard the old processing tube containing

flow-through.

16. Add another 500 mL Buffer BR4 to the PAXgene RNA spin column.

Centrifuge for 3 minutes at 8000g to 20,000g.

17. Discard the tube containing the flow-through, and place the PAXgene

RNA spin column in a new 2-mL processing tube. Centrifuge for 1 minute

at 8000g to 20,000g.

18. Discard the tube containing the flow-through. Place the PAXgene RNA

spin column in a 1.5-mL microcentrifuge tube, and pipet 20 mL Buffer

RNAse-free water directly onto the PAXgene RNA spin column mem-

brane. Centrifuge for 1 minute at 8000g to 20,000g to elute the RNA.

19. Repeat the elution step by adding the eluted RNA to the same PAXgene

RNA spin column membrane. Centrifuge for 1 minute at 8000g to 20,000g

to elute the RNA.

20. Incubate the eluate for 5 minutes at 658C. in the shaker/incubator without

shaking. After incubation, chill immediately on ice.

21. If the RNA samples will not be used immediately, store at�208C or�708C.

Laser Capture Microdissection

Laser capture microdissection (LCM) allows for the microscopic removal of

specific cells from paraffin or frozen tissue sections mounted on glass slides.



This procedure describes the isolation of a pure population of cells followed by

its nucleic acid or protein extraction.

1.1. Prepare paraffin section on glass slide.

a. Collect 5 mm paraffin section onto uncoated glass slide, using micro-

tome and water bath.

b. Allow to air-dry overnight at room temperature or at 428C for up to

8 hours.

c. Deparaffinize and lightly stain slide(s) by immersing in the following:

i. Xylene � 2: 5 minutes each

ii. 100% ethanol: 1 minute

iii. 95% ethanol: 1 minute

iv. 70% ethanol: 1 minute

v. Distilled water: 1 minute

vi. Harris hematoxylin: 1 dip (Fisher’s Protocol #245–678)

vii. Tap water: 2 minutes

viii. Distilled water: 1 minute

ix. 95% ethanol: 1 minute

x. Eosin Y: 1 dip (Fisher’s Protocol #314–631)

xi. 95% ethanol: 1 minute

xii. Allow to air-dry

or

1.2. Prepare frozen section on glass slide.

a. Collect 8 mm frozen section on an uncoated glass slide, using cryostat.

b. Store frozen until ready to postfix.

c. Postfix and lightly stain.

d. Remove a maximum of four slides from cryostat or freezer, place on

clean paper towel, i.e., KimWipe, and allow to thaw for approxi-

mately 30 seconds.

e. Place the slides in 75% ethanol for 30 seconds.

f. Transfer the slides to distilled water for 30 seconds.

g. Place slides on KimWipe, apply 100 mL of histogene staining

solution, and stain for approximately 20 seconds.

h. Place the slides in distilled water for 30 seconds.

i. Place the slides for 30 seconds in each of the following ethanols in

increasing concentration: 75%, then 95%, and then 100%.

j. Transfer the slides to xylene for 5 minutes.k. Place the slides on aKimWipe to dry in a ventilation hood for 5minutes.

2. Dissect area of tissue from slide (to be performed same day as steps 1.1

or 1.2).

a. Place slide on Pixcell II LCM [Arcturus (now part of Molecular

Devices), Mountain View, California, U.S.] and locate area of

interest using 4� or 10� objective.

222 Raji


b. Load CapSure HS caps.

c. Turn on laser key, push laser enable button, and locate laser spot on

screen.

d. Focus laser beam to bright, well-defined spot.

e. Test laser, observing melted plastic ring(s).

See following table for power and duration.

Laser spot size (mm) Power (mW) Duration

CapSure HS Small (7.5) 65–75 650–750 msMedium (15) 35–45 2.5–3.0 ms

Large (30) 45–55 6.0–7.0 ms

f. View captured cells on cap.

g. Move cap to capping station.

h. Transfer to ExtractSure device from CapSure HS cap.

3. Extract DNA from dissected tissue.

a. Pipet 155 mL of reconstitution buffer into one vial of proteinase K.

Gently vortex the tube to mix the reagents and immediately place

the tube on ice. Completely dissolve reagent, excessive mixing can

denature proteinase K.

b. Assemble the ExtractSure Sample Extraction Device onto the CapSure

HS LCM Cap.

c. Using clean forceps, remove the cap from the cap insertion tool and

place into the alignment tray (sample facing up).

d. Make sure that the cap snaps securely and lies flat on the bottom of the

opening of the alignment tray.

e. Position the ExtractSure Device over the cap using clean forceps. The

fill-port should be facing up.

f. Using forceps, push the ExtractSure Device down onto the cap until it

snaps securely into place. Use forceps to ensure that the cap is firmly

attached to the ExtractSure Device.

g. Pipet 10 mL of proteinase K extraction solution into the microchamber

formed by the assembled ExtractSure Device and CapSure HS LCM

Cap.

h. Using gloved hands, place a 0.5-mL thin-walled reaction tube over the

ExtractSure device.

i. Cover the assembly in the alignment tray with the incubation block and

incubate it overnight at 658C.j. After incubation, remove the assembled CapSure HS LCM Cap and

ExtractSure Device from the incubator, place the assembly into a

microcentrifuge and centrifuge for 1 minute at 4000g.



k. Separate the CapSure HS LCM Cap and ExtractSure Device. Close the

microcentrifuge tube containing the extract and heat to 958C for

10 minutes in a heating block to inactivate the proteinase K. Cool

the sample to room temperature.

Isolation of DNA, RNA, Protein Using TRIZOL Method

Isolation of RNA by TRI Reagent

Homogenization Homogenize fresh tissue samples (that have been stored in

RNAlater) in TRI Reagent (1 mL/50–100 mg tissue) using a glass-Teflon or

Polytron homogenizer. Sample volume should not exceed 10% of the volume of

TRI Reagent used for homogenization.

Phase separation

1. Store the homogenate for 5 minutes at room temperature to permit the

complete dissociation of nucleoprotein complexes.

2. Add 0.2 mL chloroform to the homogenate, cover the samples tightly, and

shake vigorously for 15 seconds.

3. Store the resulting mixture at room temperature for 2 to 15 minutes and

centrifuge at 12,000g for 15 minutes at 48C.4. Following centrifugation, the mixture separates into a lower red phenol-

chloroform phase, interphase, and the colorless upper aqueous phase. RNA

remains exclusively in the aqueous phase, whereas DNA and proteins are

in the interphase and organic phase. The volume of the aqueous phase is

about 60% of the volume of TRI Reagent used for homogenization.

RNA precipitation

1. Transfer the aqueous phase to a fresh tube.

2. Precipitate RNA from the aqueous phase by mixing with 0.5 mL of

isopropanol.

3. Store samples at room temperature for 5 to 10 minutes and centrifuge at

12,000g for eight minutes at 48C to 258C. RNA precipitate (often invisible

before centrifugation) forms a gel-like or white pellet on the side and

bottom of the tube.

RNA wash

1. Remove the supernatant and wash the RNA pellet (by vortexing) with 75%

ethanol (1 mL).

2. Centrifuge at 7500g for 5 minutes at 48C to 258C.

224 Raji


RNA solubilization

1. Remove the ethanol wash and briefly air-dry theRNApellet for 3 to 5minutes.

2. Dissolve RNA water by passing the solution a few times through a pipette

tip and incubating for 10 to 15 minutes at 55 to 608C.

Isolation of DNA by TRI Reagent

The DNA is isolated from the interphase and phenol phase separated from the

initial homogenate as described in the RNA isolation protocol.

DNA precipitation

1. Remove the remaining aqueous phase overlying the interphase. Precipitate

DNA from the interphase and organic phase with ethanol.

2. Add 0.3 mL of 100% ethanol and mix samples by inversion.

3. Store the samples at room temperature for 2 to 3 minutes and sediment

DNA by centrifugation at 2000g for 5 minutes at 48C to 258C.

DNA wash

1. Wash the DNA pellet twice in a solution containing 0.1M trisodium citrate

in 10% ethanol (1 mL).

2. Store the DNA pellet in the washing solution for 30 minutes at room temper-

aturewith periodicmixing andcentrifuge at 2000g for5minutes at 48C to258C.3. Suspend the DNA pellet in 75% ethanol (1.5–2 mL of 75% ethanol). Store

for 10 to 20 minutes at room temperature with periodic mixing and cen-

trifuge at 2000g for 5 minutes at 48C to 258C. This ethanol wash removes

pinkish color from the DNA pellet.

DNA solubilization

1. Remove the ethanol wash and briefly air-dry the DNA pellet by keeping

tubes open for 3 �to 5 minutes at room temperature.

2. Dissolve the DNA pellet in 8 mM NaOH by slowly passing through a

pipette. Add (0.3 mL) of 8 mM NaOH.

3. Centrifuge at 12,000g for 10 minutes and transfer the resulting supernatant

containing DNA to new tubes.

Isolation of Proteins by TRI Reagent

Protein precipitation

1. Aliquot a portion of the phenol-ethanol supernatant (0.2–0.5 mL, 1 volume)

into a microfuge tube.



2. Precipitate proteins by adding 3 volumes of acetone.

3. Mix by inversion for 10 to 15 seconds to obtain a homogeneous solution.

4. Store samples for 10 minutes at room temperature and sediment the protein

precipitate at 12,000g for 10 minutes at 48C.

Protein wash

1. Decant the phenol-ethanol supernatant and disperse the protein pellet in

0.5 mL of 0.3 M guanidine hydrochloride in 95% ethanol þ 2.5 %

glycerol.

2. Disperse the pellet using a pipette tip. After dispersing the pellet, add

another 0.5 mL aliquot of the guanidine hydrochloride/ethanol/glycerol

wash solution to the sample and store for 10 minutes at RT.

3. Sediment the protein at 8000g for 5 minutes.

4. Decant the wash solution and perform two more washes in 1 mL each of

the guanidine/ethanol/glycerol wash solution. Disperse the pellet by vor-

texing after each wash to efficiently remove residual phenol.

5. Wash in 1 mL of ethanol containing 2.5 % glycerol.

6. Sediment the protein at 8000g for 5 minutes.

7. Decant the alcohol, invert the tube, and dry the pellet for 7 to 10 minutes at

room temperature.

Protein solubilization

1. After briefly air-drying the protein pellet, add 1% SDS (0.2 mL) to the

protein pellet.

2. Gently disperse and solubilize the pellet for 15 to 20 minutes by flicking

the tube or pipetting as required.

Storage of RNA, DNA, and/or Protein

Store the solubilized proteins and DNA at �208C or colder, store RNA at �708Cor colder.

Tissue Microarray Creation

This procedure describes the use of the MTA-1 (Manual Tissue Arrayer) by

Beecher Instruments, Inc. (Sun Prairie, Wisconsin, U.S.) to create tissue micro-

array (TMA) blocks. TMA is a technology which facilitates research by permitting

the collection of biological specimens representing a great many individuals

without requiring a great deal of storage space. TMA eases comparison of the

histological characteristics of different patients’ pathologies by placing many

different tissue samples together on the same microscope slide for observation.

226 Raji


Also, TMA helps prevent the unnecessary exhaustion of irreplaceable pathological

material, since it permits the same diagnostic and research objectives to be ach-

ieved with far less tissue than was required for such objectives before TMA

existed.

A core of paraffin is removed from a “recipient” paraffin block (one

embedded without tissue) and the remaining empty space is filled with a core of

paraffin-embedded tissue from a “donor” block. This process can be repeated up

to 230 times for one TMA block, resulting in a block containing mapped pieces

from many different blocks, and when sectioned, produces a slide containing

mapped pieces from many different blocks. A corresponding map records each

core’s location in the grid and the information connected with it (patient ID

number, study arm, etc.). The array sections can be used for all histological

staining, including hematoxylin & eosin (H&E), IHC, and ISH.

1. Prepare the donor blocks and corresponding H&E slides.

a. Re-embed tissue, as needed, for more uniform wax consistency and

optimal tissue orientation.

b. Donor tissue should be at least 1.0 mm thick for best results.

c. Prepare a fresh H&E slide from each re-embedded potential donor block.

2. Pathologist characterizes each H&E slide for appropriateness of diagnosis

and percentage of tumor.

a. Examine each H&E slide microscopically and indicate, with marker,

the area of interest (AOI). This area corresponds to the area on the

block from which the core(s) will be taken.

3. Prepare the recipient (TMA) block.

a. Label a blank cassette, using the automated cassette labeler.

b. Determine the size of the TMA block and use:

M—the medium mold (30 mm � 23 mm) or

L—the large mold (35 mm � 23 mm)

c. Make a paraffin block, containing no tissue, using the labeled cassette.

d. Assure that the paraffin block surface is flat and parallel to the

underside of plastic cassette and that it contains no air bubbles.

4. Design the array.

a. Determine the diameter of the cores to be removed from the donor

blocks. Available core sizes include 0.6, 1.0, 1.5, and 2.0 mm. Core

size is often indicated by protocol, i.e., breast cancer studies require

the smallest core size—0.6 mm—to conserve tissue while removing

three cores from each donor block during the routine production

of three TMA blocks from each group of subjects.

b. Allow for enough space (3.0 mm) around the edge of the block to avoid

cracking paraffin.



c. Determine the number of columns and number of rows to use by noting

the following table, which describes the maximum number of cores per

TMA block, with respect to core size and mold size.

Maximum Number of Cores per TMA Block

Core size (mm)

Spacing between

cores (mm) Mold size

Maximum no.

of cores/spaces

0.6a yellow 1.5 LM 20 � 12 ¼ 240

1.0a green 2.0 LM 15 � 9 ¼ 135

1.5a black 2.5 LM 12 � 7 ¼ 84

2.0a white 3.0 LM 10 � 6 ¼ 60

aPunches are color coded.

5. Create a map, identifying each core’s position on the grid its donor block.

a. Enter information into Excel Program (TMA Mapping Tool, Attach-

ment 1) located on Server/Pathcore/TMA. TMA may be mapped using

alternate methods.

b. Include cores from tissue and cell line control blocks, in duplicate.

i. Tissue blocks, i.e., for breast protocols:

Fibroadenoma

Normal endometrium

Normal salivary gland

Normal testis

ii. Cell line blocks, i.e., for breast protocols:

MCF7

SKBR3

T47D

c. Each TMA block will be asymmetric to assure map orientation, i.e.:

2nd column blank,

4th row blank, and

bottom right space blank

6. Construct the TMA.

a. Install the pair of tissue punches, for desired core size, in MTA-1. Install

the slightly larger of the pair (identified in blue) on the right (for the

donor block). Install the slightly smaller of the pair (identified in red) on

the left for the (for the recipient or TMA block).

b. Insert the recipient (TMA) paraffin block into the MTA-1 block holder.

Secure in place by tightening two screws in block holder with MTA-1

allen wrench.

228 Raji


c. Remove the first core from TMA block.

i. Assure that recipient punch is swung into place.

ii. Align punch over site of first core, usually top left corner, by

turning micrometers.

iii. Zero micrometers by pressing the X and Y “ZERO/ABS” buttons.

iv. Gently lower punch into TMA block, using care not to punch too

deeply. Maximum punch depth can be adjusted with large vertical

screw on the left.

v. Raise punch.

vi. Remove paraffin core from punch, using punch’s stylet.

d. Remove core from first donor block.

i. Place the donor block bridge over the TMA block holder.

ii. Swing donor punch into place.

iii. Manually align donor block so that AOI is directly under the

donor punch.

iv. Lower and raise punch.

v. Remove core of tissue from punch using punch stylet.

e. Place donor core into TMA block.

i. Remove donor block bridge from over the TMA block holder,

noting location of delicate loose core of tissue.

ii. Using forceps, align donor core over hole created for it in the

TMA block.

iii. Lower donor core into hole and assure flat level block surface by

pressing evenly with clean glass slide.

f. Adjust the micrometer(s) to move the precision guide to the next X/Y

position. Follow the Spacing Between Cores (mm) Column in the

maximum number of cores per TMA block table above to determine

the appropriate increments.

g. Repeat this cycle (steps 6c–6f) to construct the whole array.

h. To punch holes on right half of TMA block, rotate it 1808 and replace

in recipient block holder, because the x axis micrometer should not be

extended beyond 9 mm.

i. Smooth and level surface of TMA block.

i. Remove the TMA block from the recipient block holder.

ii. Place the TMA block in oven at 378C, maximum temperature, for

30 minutes.

iii. Gently press a glass slide on the top of the TMA block, applying

even pressure to push all tissue cores on the block to same level.



7. Section TMA block.

i. Cut a section from the TMA block and stain the slide with H&E for

QA/QC review by the pathologist.

ii. Store TMA block and save TMA map for any future sectioning requests.

Label side of block

2nd column left blank.

4th row left blank

Bottom right corner left blank

230 Raji


16

Acquisition and Preservation of Tissue

for Microarray Analysis

Brian R. Untch and John A. Olson

Institute for Genome Sciences and Policy and the Department of Surgery,Duke University Medical Center, Durham, North Carolina, U.S.A.

INTRODUCTION

Increasingly, collection of normal and neoplastic human tissue is being incor-

porated in the design of clinical trials and in routine clinical settings under tissue

banking programs. The goal behind these activities is either to facilitate iden-

tification and development of robust predictive or prognostic biomarkers or to

enable basic investigation of tumor or host biology.

As experience in tissue banking has broadened, the role of rigorous collec-

tion protocols, tissue handling, and annotation has become increasingly appre-

ciated. To this end, several groups have published guiding principles and standard

operating procedures (SOPs) designed to advance the fields of tissue banking,

biomarker development, and tumor biology (1,2).

This chapter is designed to summarize key points in the process of informed

consent, tissue collection, tissue preservation and storage, as well as extraction of

macromolecules of interest.

INFORMED CONSENT

Tissue banking requires informed consent by patients donating their tissues. The

informed consent document should include standard topics such as study design,

study purpose, and patient risks/benefits. Privacy rights are paramount for

231


patients when their health information is stored in databases, perhaps indef-

initely. The Health Insurance Portability and Accountability Act and the Com-

mon Rule are federal regulations that require protection of patient identification.

Database information and patient identifiers (name, social security number,

institutional medical record ID) should remain separate. An arbitrary code sys-

tem should be created to uniquely label specimens and data. The informed

consent document should specify which investigator will keep the anonymous

identifiers and patient identifiers together. In the event that third-party sharing of

patient identifier data becomes necessary, written authorization must be obtained

from the patient. Also specified in the informed consent document is the

patients’ right to withdraw from the study or have their specimen removed from

the tissue databank. Contact information for these requests is provided to the

patient at the time of document signing.

A recent topic of debate has been whether giving consent for tissue

banking also allows for indiscriminate investigations related to and unrelated to

the initial study design. Implicit to the potential of tissue banking is the use of

specimens for areas of research that are unanticipated or currently unknown.

However, the use of specimens for research not specifically noted in the

informed consent document may violate religious and personal beliefs. The

inclusion of broad consent for any possible project has been proposed by some as

a solution (3). Others argue that this does not constitute informed consent, as the

possibilities for potential research are unpredictable (4,5). Our approach has been

to specifically describe research to be performed on collected specimens. This

prevents any misunderstanding between patient and researcher. This is not useful

for large repositories that collect tissue for future investigations.

TISSUE ACQUISITION

When tissues are devascularized and removed from the body, cellular metabo-

lism is altered as hypoxia and acidification occur and nutrient delivery ceases.

At a minimum, macromolecules in ischemic tissues are subject to degradation.

Further, cellular response to ischemia may lead to stress-specific changes in gene

expression. This may ultimately lead to changes in gene and protein expression

or degradation until the cellular machinery stops or tissue preservation occurs.

To accurately and reproducibly analyze gene and protein expression events

within cells, tissue must be preserved as quickly and uniformly as possible.

For all protocols where banking is superimposed upon clinical care, above

all else, banking strategies must not interfere with pathologic assessment of

either the tumor or specimen resection margins. When designing these protocols,

involvement of the pathologist who is custodian of the tissue record is critical.

This is not to say that banking cannot occur before specimen processing in

pathology, but that overall specimen integrity must be maintained during the

research specimen collection (e.g., core biopsy of lumpectomy specimens).

When a pathologist is immediately available, specimen banking after pathologic

232 Untch and Olson


gross evaluation is preferred. However, if a pathologist is not immediately

available, then tumor specimens must be collected before pathologic assessment

but without specimen disruption. Regardless of when during the processing of

the pathologic specimen banking occurs, representative samples of all banked

tissue should be reviewed by the pathologist to both ensure that critical diag-

nostic material is not lost and to document that research tissue samples are

indeed tumor tissues.

To further ensure that research tissue acquisition will not have an impact

on diagnosis, frozen specimens should be held and not analyzed until definitive

pathologic assessment of the entire remaining specimen has been completed. At

the time of definitive diagnosis of the surgical specimen, research specimens can

then be deposited in the tumor bank (Fig. 1). If analysis of the surgical specimen

does not demonstrate the abnormal pathology, then research specimens may be

Figure 1 Flow diagram demonstrating the sample handling of core biospy specimens in

patients suspected of having breast cancer. Research and clinical core biopsies should be

pathologically examined so that the patient’s diagnosis is not affected and that research

specimens contain the tissue of interest.

Acquisition and Preservation of Tissue for Microarray Analysis 233


used for frozen-section analysis. Should the frozen section of research specimens

reveal abnormal histology not present in the surgical specimen (e.g., neoplasia

is present in the research specimens and not in the surgical specimen), all

samples can be turned over to pathology for routine processing. While this may

result in a short delay as surgical pathology is reevaluating frozen research

specimens, this can be explained in a preoperative consent form.

In all cases, frozen research specimens should be held frozen until two

weeks after a definitive pathologic report is issued on the diagnostic specimens.

This hold period is desirable, as once DNA and RNA are extracted, no further

diagnostic information relevant to patient care will be available.

For clinical trials and research protocols, specimen collection needs to be

standardized so that tumor samples are treated in a similar manner to minimize

technical variability in gene and protein expression data. Several cooperative

groups have developed banking SOPs for collection of formalin-fixed specimens

and blood samples; however, few have until recently focused on frozen-tissue

collection. The American College of Surgeons Oncology Group (ACOSOG) has

developed effective procedures for collecting tumor tissue from breast cancer

patients in neoadjuvant studies with good success. The European Human Frozen

Tumor Tissue Bank has also developed similar SPOs that are followed across

institutions and countries in an effort to reduce variability (2). In the former case,

specialized collection kits have been privately developed and are in use by

ACOSOG investigators for multisite tissue and blood collection at the time of

diagnostic biopsy and again at surgical resection (Fig. 2). Technicians respon-

sible for tumor collection prepare reagents and containers before excision.

Figure 2 A stereotactic biopsy device to retrieve breast cancer samples from excised

lumpectomy specimens. (A, arrow) Excised lumpectomy specimens are placed in the

device in an oriented fashion, and a specimen radiogram is taken to identify the location of

the tumor. (B) Coordinates are taken from the radiogram and are used to guide core biopsies

taken from the tumor within an example lumpectomy specimen held in fixed orientation in

the device.

234 Untch and Olson


Samples are required to be snap frozen within five minutes of excision,

regardless of the method of tissue fixation (described subsequently). Rapid and

uniform handling of specimens from patient to preservative is essential.

TECHNIQUES OF CRYOPRESERVATION

RNA degradation by tissue RNases can be detrimental to RNA yield and quality,

with implications on microarray and quantitative PCR experiments. Traditional

practice has been to snap freeze samples in liquid nitrogen followed by per-

manent storage at a minimum of �808C. This prevents changes to the RNA, and

tumor banks using this method have been highly successful in producing high-

quality extracted RNA for microarray analysis. Given the complexity of sample

acquisition and multi-institutional collection, recent studies have tried to deter-

mine the importance of immediate ex vivo freezing. Micke et al. (6) used ex vivo

tumors at room temperature to determine the amount of time until RNA deg-

radation occurred. Distinct ribosomal peaks were identified up to 16 hours,

indicating that this tissue can remain stable for extended periods of time. Other

sample-handling issues, such as freeze-thawing cycles, have also been studied.

Jochumsen et al. (7) established that gene expression analysis is not altered in up

to three freeze-thaw cycles. Despite these reports, it has been our experience that

RNA degradation can be extremely variable, even when SOPs are employed

(Figs. 3,4). To ensure quality RNA, samples should be frozen as soon as possible

after being devascularized.

Figure 3 RNA degradation is variable between two breast tumor samples. Samples are

subjected to the same conditions over 120 minutes. Sample 2 degrades faster than sample 1,

demonstrating the stability variation.



Preservation of tissue morphology can be difficult in samples that are snap

frozen. These samples are also subject to desiccation or freezer burn. Samples

can be immersed in optimal cutting temperature (OCT) compound to preserve

tissue morphology and prevent sample damage. This method has been shown to

generate reproducible genomic DNA extraction results (8). Our group has not

observed differences in microarray data when using OCT (unpublished obser-

vation). However, Turbett et al. (9) reported inhibition of a polymerase chain

reaction (PCR) assay when compared with non-OCT compound–preserved

specimens. OCT compound can also make specimen thawing and dissection

more difficult. When specimens contain a mix of tissue (as in the case of

undissected tumor specimens with margins), separating whole tumor for RNA

extraction can be difficult, particularly when trying to avoid time-dependant

RNA degradation. Nonetheless, few studies have investigated OCT compound

and its potential to alter microarray results. A recommendation for or against its

use cannot be made at this time.

Because of the challenges of sample handling (pathology analysis) and the

potential for RNA degradation, RNA stabilization buffers have been developed.

Commercially available RNAlater (Ambion inc.) has been shown to maintain

RNA quality before snap freezing. Several studies have shown consistent

microarray results when storing samples in RNA stabilization buffers (10–12).

One such study found that samples could be stored up to one hour at 378C,without affecting the RNA or subsequent microarray data (13). These buffers

Figure 4 The effect of two-hour extended ischemia time on microarray data. The same

sample had RNA extracted at time 0 and after two hours and hybridized to the U95

Affymetrix platform. Raw expression values were log normalized and graphed. Data

demonstrate that at lower expression values greater variability exists between prolonged

ischemic and nonischemic tissue. There was also variability demonstrated in published

gene signatures describing prognosis. Source: From Refs. 26, 27, 31.

236 Untch and Olson


will be important for tissue acquisition protocols, particularly with regard to

multi-institutional trials that will require transfering tissue specimens to a central

tissue bank. This approach allows the samples to be shipped refrigerated rather

than frozen. The principal disadvantages are that proteins are degraded, and

subsequent sectioning for histologic verification of tumor identity is problematic.

PARAFFIN-EMBEDED TISSUE FOR MICROARRAY ANALYSIS

The discovery of paraffin wax is attributed to the noted hydrocarbon chemist,

Dr. Carl (Karl) Ludwig von Reichenbach (1788–1869). After being developed in

1830, pathology laboratories discovered its potential as a method for storing

and preserving specimens. Throughout the twentieth century, formalin-fixed,

paraffin-embedded tissue became a reliable way to archive tissue. Given the

large quantity of specimens available in paraffin, these tissue banks were identified

as a possible tissue source for array-based techniques. Numerous challenges have

been identified that complicate the macromolecular extraction process from these

samples.

Sample handling is an important step that can lead to variability observed

in microarray results. Ischemia time can profoundly affect protein, DNA, and

RNA extraction because of enzymatic degradation. As noted above, sample

collection is ideally performed by snap-freezing tissue in liquid nitrogen

immediately after being devascularized and removed from a patient. Sample

handling of formalin-fixed, paraffin-embedded tissue is variable, and this makes

application of the array data potentially unreliable.

Many investigators have attempted to develop methods for extraction of

these samples with mixed results. Given the cross-linking effects of formalin, it

was thought that protein extraction for proteomic analysis would be impossible.

However, attempts have been made to isolate proteins from paraffin, but results

have been mixed and more technical development is required (14).

Extraction of nucleic acids can be accomplished using commercially

available proteinase digestion buffers (15). This method also eliminates the risk

of Rnase activity in the sample as all previously fixed proteins are destroyed.

Formalin fixation does not appear to affect the yield of RNA, unless the tissue is

fixed for an extended period of time. This can result in cross-linking between

RNA and protein (16). As long as the methylol groups added to the bases during

fixation can be removed, the quality of the RNA is not perturbed (17) and results

have been comparable with frozen-tissue controls (18).

The most difficult problem encountered with these specimens has been that

the total DNA or RNA yield from these samples is inadequate and often

degraded (19). Yield of DNA or RNA is dependent on duration of protein

digestion, temperature, and pH (20). Most array platforms will require several

micrograms of RNA for hybridization. Amplification of the RNA has been

suggested as means to increase quantity, but these techniques can result in

considerable variability between samples (21). Further, amplification of partially



degraded samples can also result in unreliable data. Statistical adjustments for

nucleic acid amplification in microarray experiments are being developed and

may eventually prove to be useful.

As extraction, amplification, and statistical techniques improve, formalin-

fixed, paraffin-embedded tissue may provide an additional resource for researchers.

Nevertheless, the variability in sample handling history makes application of the

array data from these samples less than reliable. At this point in time, use of these

samples for application to clinical investigations is discouraged.

FINE NEEDLE ASPIRATION, CORE NEEDLE BIOPSY,AND SURGICAL SPECIMENS IN MICROARRAY ANALYSIS

Using micorarray analysis, numerous genomic signatures have been developed

that attempt to predict tumor phenotypes, including lymph node metastasis,

tumor growth patterns, postoperative margin positivity, chemosensitivity, and

oncogenic pathway activation. Most of these analyses have been based on

samples collected at the time of surgery. These large volume samples can yield

adequate RNA for array hybridization and also afford the opportunity to assess

multiple areas of the tumor and subcellular compartments via laser capture

microdissection. However, tumor size and growth is known to affect gene

expression, and successful application of the developed models to earlier stage,

preoperative biopsy specimens may not be possible.

Nonetheless, application of these predictive tests depends on obtaining

adequate tissue for microarray analysis before a definitive surgical intervention.

The preoperative diagnosis of breast cancer is typically achieved though two

methods of tumor biopsy: core needle biopsy (CNB) and fine needle aspiration

(FNA). For micoarray-based predictive testing, an optimal protocol would allow

these two modalities to collect tissue for pathologic analysis and for molecular

extraction.

FNA is performed by inserting a small gauge needle, typically under image

guidance, into a tumor or lesion. Tissue is removed through a negative-pressure

syringe. Tumor architecture is typically destroyed during this process, but the

resultant sample can be analyzed under the microscope for tumor cell charac-

teristics. Likewise, this small amount of tissue can be harvested for RNA and

microarray analysis. However, concern exists over adequate sample amount (22).

Several studies have shown that FNA can produce tissue in amounts appropriate

for pathology review and array applications (23,24).

CNB is performed under image or visual guidance in a fashion similar to

FNA. However, with CNB, a larger bore needle is used, and the resultant

specimen has maintained tissue architecture, allowing for a more accurate

pathologic analysis. This larger amount of tissue can also be used for array-based

studies. Several studies have addressed adequacy of tissue acquisition in core

238 Untch and Olson


biopsy samples (13,24). Studies from our laboratory indicate that core biopsy

specimens can be collected before surgery and can undergo successful RNA

extraction (unpublished observation). The average yield from core biopsy

specimens was 10.4 mg of RNA, enough for most array platforms. After

extraction, RNA samples underwent hybridization to the HU95 Affymetrix

microarray platform and then chemosenstivity and oncogenic pathway signatures

were successfully applied. This is consistent with the work of Ellis et al. (13),

who was able to extract RNA from core biopsy specimens and distinguish cancer

and noncancer phenotypes.

LASER CAPTURE MICRODISSECTION

Recently, focus has shifted away from whole-tumor microarray analysis to

analysis of individual cell types within a tumor. A tumor is composed of a

heterogeneous group of cells, malignant cells, premalignant cells, normal

parenchymal cells, and supportive cells. This cellular heterogeneity must be

taken into account when applying array-based techniques (25). Numerous suc-

cessful expression-profiling studies have been performed using whole-tumor

macromolecular extraction (26–28). These analyses thus incorporate a variety of

different cell types. Laser capture microdissection allows one to isolate histo-

logically distinct cell types and perform subsequent analysis. The technique

employs an infrared laser in conjunction with light microscopy. Prepared his-

tologic slides are covered with a transfer film. When cells of interest are iden-

tified, the laser is activated over the cells. This process causes the cells to attach

to the transfer film. The remainder of the sample remains, and the isolated cells

can undergo extraction procedures. Notably, this procedure can result in lower

yields of cells requiring amplification of nucleic acids for use in microarray

experiments (29,30). Laser capture microdissection has the potential to not only

improve our understanding of cellular deregulation but also to help determine

how other cells within the tumor contribute to the overall tumor phenotype.

Immune cells, vascular structures, and normal parenchymal cells within a tumor

may contribute significantly to the growth and metastatic potential. Expression

analysis of these individual cell types may help elucidate their roles.

SUMMARY

Acquisition and preservation of tissue for microarray experiments requires

understanding of informed consent, sample handling, macromolecular extrac-

tion, and sampling methods. Significant advances in these areas have been made

in recent years to improve the quality of microarray data. Appreciating these

aspects is paramount if the application of microarray techniques is to contribute

meaningfully to the care of the cancer patient.



REFERENCES

1. McShane LM, Altman DG, Sauerbrei W, et al. Reporting recommendations for tumor

marker prognostic studies (REMARK). J Natl Cancer Inst 2005; 97(16):1180–1184.

2. Mager SR, Oomen MN, Morente MM, et al. Standard operating procedure for the

collection of fresh frozen tissue samples. Eur J Cancer 2007; 43(5):828–834.

3. Hansson MG, Dillner J, Bartram CR, et al. Should donors be allowed to give broad

consent to future biobank research? Lancet Oncol 2006; 7(3):266–269.

4. Maschke KJ. Alternative consent approaches for biobank research. Lancet Oncol

2006; 7(3):193–194.

5. Rothstein MA. Expanding the ethical analysis of biobanks. J Law Med Ethics 2005;

33(1):89–101.

6. Micke P, Ohshima M, Tahmasebpoor S, et al. Biobanking of fresh frozen tissue:

RNA is stable in nonfixed surgical specimens. Lab Invest 2006; 86(2):202–211.

7. Jochumsen KM, Tan Q, Dahlgaard J, et al. RNA quality and gene expression analysis

of ovarian tumor tissue undergoing repeated thaw-freezing. Exp Mol Pathol 2007;

82(1):95–102.

8. Sun Y, Hegamyer G, Colburn NH. A simple method using PCR for direct sequencing

of genomic DNA from frozen tumor tissue embedded in optimal cutting temperature

compound. Biotechniques 1992; 12(5):639–640.

9. Turbett GR, Sellner LN. The use of optimal cutting temperature compound can inhibit

amplification by polymerase chain reaction. Diagn Mol Pathol 1997; 6(5):298–303.

10. Gotzer MA, Patti R, Geoerger G, et al. Biological stability of RNA isolated from

RNAlater-treated brain tumor and neuroblastoma xenografts. Med Pediatr Oncol

2000; 34(6):438–442.

11. Florell SR, Coffin CM, Holden JA, et al. Preservation of RNA for functional

genomic studies: a multidisciplinary tumor bank protocol. Mod Pathol 2001; 14(2):

116–128.

12. Mutter GL, Zahrieh D, Liu C, et al. Comparison of frozen and RNALater solid tissue

storage methods for use in RNA expression microarrays. BMC Genomics 2004;

5(1):88.

13. Ellis M, Davis N, Coop A, et al. Development and validation of a method for using

breast core needle biopsies for gene expression microarray analyses. Clin Cancer Res

2002; 8(5):1155–1166.

14. Becker KF, Schott C, Hipp S, et al. Quantitative protein analysis from formalin-fixed

tissues: implications for translational clinical research and nanoscale molecular

diagnosis. J Pathol 2007; 211(3):370–378.

15. Krafft AE, Duncan BW, Bijwaard KE, et al. Optimization of the isolation and

amplification of RNA from formalin-fixed, paraffin-embedded tissue: The Armed

Forces Institute of Pathology Experience and Literature Review. Mol Diagn 1997;

2(3):217–230.

16. Moller K, Rinke J, Ross A, et al. The use of formaldehyde in RNA-protein cross-

linking studies with ribosomal subunits from Escherichia coli. Eur J Biochem 1977;

76(1):175–187.

17. Lewis F, Maughan NJ, Smith V, et al. Unlocking the archive: gene expression in

paraffin-embedded tissue. J Pathol 2001; 195(1):66–71.

240 Untch and Olson


18. Specht K, Richter T, Muller U, et al. Quantitative gene expression analysis in

microdissected archival formalin-fixed and paraffin-embedded tumor tissue. Am J

Pathol 2001; 158(2):419–429.

19. Jackson DP, Lewis FA, Taylor GR. Tissue extraction of DNA and RNA and analysis

by the polymerase chain reaction. J Clin Pathol 1990; 43(6):499–504.

20. Gilbert MT, Haselkorn T, Bunce M, et al. The isolation of nucleic acids from fixed,

paraffin-embedded tissues-which methods are useful when? PLoS ONE 2007; 2(6):

E537.

21. Coombs NJ, Gough AC, Primrose JN. Optimisation of DNA and RNA extraction

from archival formalin-fixed tissue. Nucleic Acids Res 1999; 27(16):E12.

22. Centeno BA, Enkemann SA, Coppola D, et al. Classification of human tumors using

gene expression profiles obtained after microarray analysis of fine-needle aspiration

biopsy samples. Cancer 2005; 105(2):101–109.

23. Rapkiewicz A, Expina V, Zujewski JA. The needle in the haystack: application of

breast fine-needle aspirate samples to quantitative protein microarray technology.

Cancer 2007; 111(3):173–184.

24. Symmans WF, Ayers M, Clark EA, et al. Total RNA yield and microarray gene

expression profiles from fine-needle aspiration biopsy and core-needle biopsy

samples of breast carcinoma. Cancer 2003; 97(12):2960–2971.

25. Fuller AP, Palmer-Toy D, Erlander MG, et al. Laser capture microdissection and

advanced molecular analysis of human breast cancer. J Mammary Gland Biol

Neoplasia 2003; 8(3):335–345.


predictor of survival in breast cancer. N Engl J Med 2002; 347(25):1999–2009.



U S A 2001; 98(19):10869–10874.


Nature 2000; 406(6797):747–752.

29. Caretti E, Devarajan K, Coudry R, et al. Comparison of RNA amplification methods

and chip platforms for microarray analysis of samples processed by laser capture

microdissection. J Cell Biochem 2008; 103(2):556–563.

30. Schuetz CS, Bonin M, Clare SE, et al. Progression-specific genes identified by

expression profiling of matched ductal carcinomas in situ and invasive breast tumors,

combining laser capture microdissection and oligonucleotide microarray analysis.

Cancer Res 2006; 66(10):5278–5286.

31. Ramaswamy S, Ross KN, Lander ES, et al. A molecular signature of metastasis in

primary solid tumors. Nat Genet 2003; 33(1):49–54.



17

Tapping into the Formalin-Fixed

Paraffin-Embedded (FFPE) Tissue

Gold Mine for Individualization

of Breast Cancer Treatment

Mark Abramovitz

VM Institute of Research, Montreal, Quebec, Canada

Brian Leyland-Jones

Winship Cancer Institute, Emory University School of Medicine,Atlanta, Georgia, U.S.A.

INTRODUCTION

It is now clearly recognized that breast cancer is a heterogeneous disease that

requires personalized targeted therapy in order to have the greatest impact on

survival. This necessitates the development of biomarkers that can be used to

predict outcome and response to therapy. Unfortunately, very few biomarkers

have made their way from the research bench to the clinic. The main reason for

this is that in order to validate breast tumor markers, large sample sets from well-

defined patient groups are required, although this could be overcome by tapping

into large numbers of well-defined banked tumor specimens. The efforts that are

being made, though, toward the development of biomarkers in order to improve

the assessment of cancer prognosis and diagnosis are great, given the enormous

potential that they embody.

243


Formalin-fixed, paraffin-embedded (FFPE) tissue samples represent the

largest collection of well-annotated, clinical tumor samples that are readily

available for conducting retrospective studies and, therefore, represent a potential

gold mine from which important biomarkers are just waiting to be discovered,

given that the right tools can be brought to bear upon this invaluable resource.

However, until relatively recently, a major caveat regarding the use of these FFPE

specimens for genomewide studies was that due to the formalin fixation process,

the standard procedure for handling tumor tissue samples in pathology labs around

the world, RNA degradation occurred, greatly limiting their potential utility. In

fact, with regard to nucleic acid–based assays, FFPE specimens have been mainly

subjected to reverse transcribed-polymerase chain reaction (RT-PCR) analysis.

This chapter will describe the progress that has been made in greatly expanding the

use of FFPE tumor specimens for several genomic-based applications that will

greatly increase our ability to tap into this gold mine and extract the biomarker and

genomic data nuggets that will improve cancer prognosis, diagnosis, and treatment.

EXTRACTION OF RNA FROM FFPE TUMOR TISSUESAMPLES FOR GENE EXPRESSION ANALYSIS

FFPE tissue samples, a vast archive of pathologically well-characterized clinical

samples from randomized trials, are an enormous potential resource that will allow

translational scientists to more fully describe breast cancers at the molecular level

and will ultimately have important etiologic and clinical implications. Even

though the degradation of RNA that occurs because of the formalin fixation

process results in an RNA species with an average size of approximately 200 nt

(1), several groups have illustrated that it is feasible to extract and purify RNA

from such fixed tissue and perform gene expression profiling (2–9). With the

development of real-time quantitative (q) RT-PCR technology that has high

detection sensitivity and high dynamic range, it is now possible to detect even rare

messages in FFPE tissue and to examine the variation of expression over quite a

large dynamic range. Amplicons are designed specifically on small segments of

DNA [<100 base pairs (bp)] to achieve close to 100% efficiency for all amplicons,

regardless of their length and nucleic acid composition. The potential of this

technology becomes even more attractive in that it permits the analysis of thou-

sands of tissue samples available through existing banks and without the need to

collect the relatively complicated, freshly collected frozen tissue. Some studies

have addressed improvements to the process of isolating high-quality FFPE RNA

suitable for qRT-PCR or high-throughput gene expression profiling (10–13).

Several groups have exploited this technology to evaluate and validate

prognostic and predictive markers of human disease, the response to certain

cancer therapies and to predict recurrence. Initial work by Cronin et al. (2004)

(14) from Genomic Health, Inc., which led to the development of the Oncotype

DX clinical assay for breast cancer (see below), reported developing an RT-PCR

244 Abramovitz and Leyland-Jones


method with multianalyte capability for potential use in clinical research and

diagnostic assays. Two multianalyte assays, one with 48 genes and the other with

92 genes, were designed and tested in different experiments with preserved

human breast cancer tissue. In the first case, all of the genes profiled in FFPE

RNA yielded measurable values, and the overall profile was validated by its

similarity to that generated with well-preserved RNA from matched freshly

frozen tissue. In the second experiment using a 92-gene assay, only one of the

tested genes failed to yield a signal. Measured levels of estrogen receptor (ER),

progesterone receptor (PR), and human epidermal growth factor receptor 2

(HER2) mRNAs were concordant with the levels of the respective proteins as

measured by immunohistochemistry (IHC) at an independent clinical reference

laboratory. Approximately 90% concordance was obtained when RT-PCR

expression results for ER and PR were dichotomized into positive and negative

values and compared with ER- and PR-positive and ER- and PR-negative

assignments based on IHC. It is noteworthy that similar levels of concordance

were found in other comparative studies of IHC versus RT-PCR (15,16).

DEVELOPMENT OF AN RT-PCR-BASED BREAST CANCERCLINICAL DIAGNOSTIC TEST: ONCOTYPE DX ASSAY

A major development in this area has been the development of a multigene qRT-

PCR assay to predict the likelihood of breast cancer recurrence in node-negative,

ER-positive, tamoxifen-treated patients (17). The authors studied the likelihood

of distant recurrence in breast cancer patients who had involved lymph nodes and

ER-positive tumors that were poorly defined by clinical and histopathologic

measures. They tested whether a prospectively-defined 21-gene RT-PCR assay

of FFPE tumor tissue could predict the likelihood of distant recurrence in

patients in the National Surgical Adjuvant Breast and Bowel Project (NSABP)

B-14 clinical trial. Measured levels of gene expression for 16 cancer-related and

5 reference genes were used, by a prospectively defined algorithm, to calculate a

recurrence score (RS) and risk group (low, intermediate, or high) for each

patient. Adequate RT-PCR profiles were obtained in 668 of 675 tumor blocks.

The proportion of patients categorized as low, intermediate, or high risk by the

RT-PCR assay was 51%, 23%, and 27%, respectively. The Kaplan–Meier esti-

mates and 95% confidence intervals for the rates of distant recurrence at 10 years

were 6.8% (4.0%, 9.6%), 14.3% (8.3%, 20.3%), and 30.5% (23.6%, 37.4%),

respectively, for the low-, intermediate-, and high-risk groups; the rate for

the low-risk group was significantly lower than the rate for the high-risk group

(p < 0.001). In a multivariate Cox model, RS provides significant (p < 0.001)

predictive power that goes beyond age and tumor size. The RS was also pre-

dictive of overall survival (p < 0.001) and could be used as a continuous function

to predict distant recurrence for individual patients. Finally, the RS was pros-

pectively validated to predict the likelihood of distant recurrence in tamoxifen-

treated breast cancer patients with node-negative, ER-positive tumors (18,19).

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 245


This methodology has been sufficiently successful, so that the Food and

Drug Administration (FDA) approved it in 2004 to become commercially avail-

able [Genomic Health, Inc. (GHI), Redwood City, California, U.S.]. Oncotype DX

is a diagnostic assay that quantifies the likelihood of breast cancer recurrence in

women with newly diagnosed stage I or II, node-negative, ER-positive breast

cancer who will be treated with tamoxifen. More importantly, this test will pave

the way for individualization of cancer therapy. While the St. Gallen classification

distinguished 7.9% of patients as low risk, the GHI RS identified 50.6% of

patients as low risk. Similarly, St. Gallen overestimated the high-risk patients at

58.8%, while the GHI test classified high-risk patients at only 27.1%. It is obvious

that St. Gallen underestimated the low-risk patients and overestimated the high-

risk ones. The advantage of the newly developed multigene RS is, therefore, that

low-risk patients can be saved from potentially toxic, high-dose chemotherapy.

RNA-BASED EXPRESSION PROFILING OFFFPE BREAST TUMOR SPECIMENS

Expression Profiling: The DASL Assay

Gene expression profiling has become an important aspect in prediction, prog-

nosis, and cancer modeling. Gene signatures resulting from expression profiling

studies have the potential to define cancer subtypes, predict the clinical outcome

(recurrence of disease) and response to specific therapies, and analyze oncogenic

pathways (20–23). Investigations of gene pathways and interactions indicated by

gene signatures that are truly predictive of the clinical endpoints are necessary to

understand the biology underlying this predictive value. More importantly, when

combined with clinical and demographic factors, multiple forms of molecular

(protein- and gene-based) data can provide information that has the potential to

identify unique characteristics of individuals and so lead to individualized

treatment strategies (24,25).

Although most array-based assays utilize RNA prepared from frozen

specimens, the newer cDNA-mediated annealing, selection, extension, and

ligation (DASL) assay (Illumina, Inc., San Diego, California, U.S.) platform was

specifically designed to profile small transcripts typically found in FFPE tissue

(26–31). The DASL platform is based on massively multiplexed RT-PCR

applied in a microarray format that allows for the determination of expression of

up to 512 genes (502 genes in the standard cancer panel available from Illumina)

using a minimal amount of total RNA (100 to *200 ng per assay) isolated

from 96 FFPE tumor tissue samples in a high-throughput format (29,30). In the

procedure, biotinylated random nonamers (biotin-d(N)9) and oligo d(T)18 are

used for cDNA synthesis, and therefore probes are designed so that they can

target any unique region of the gene without limiting the selection of the optimal

probe to the 30 ends of transcripts. In addition, because of the small size of the

targeted gene sequence (50 nucleotides), along with the use of random primers in



the cDNA synthesis, RNAs that are otherwise too degraded for conventional

microarray analysis can be readily detected. Sequence-specific query oligonu-

cleotides encompassing primer extension, ligation, and universal PCR in highly-

plexed reactions (1536-plex), two-color labeling, and redundant (*30-fold

redundancy of each bead type) feature representations are used to probe up to

three different sites on each cDNA (Fig. 1A) and, therefore, lend the assay the

necessary sensitivity and reproducibility for quantitative detection of differential

expression using RNA from FFPE tissue (29,30). In an initial study, a 231-gene

Figure 1 The DASL platform for profiling both RNA and miRNA. (A) The DASL assay.

FFPE tumor tissue samples, 5- to 10-mm shaved sections or from microscope slides, are

first subjected to deparaffinization and then RNA extraction. The extracted RNA is

fragmented with an average size of approximately 200 nucleotides. Biotinylated cDNA is

made from the fragmented RNA using random primers (9-mers). Complex oligonucleo-

tides that contain a gene-specific sequence and a unique address sequence are used to

selectively bind to the cDNA. Extension and then ligation follow, resulting in cDNA

products that are subsequently amplified using fluorescently labeled common primers.

Depicted are three gene-specific oligonucleotide probe sets hybridized to cDNAs from a

single gene. The amplified products contain 1536 unique addresses (3 addresses per gene

and therefore 512 genes) and are hybridized to the 50,000-bead array (each unique

complementary address bead at *30-fold redundancy). The address sequence directs

hybridization to their complements on the universal bead arrays. (B) The miRNA array.

cDNA derived from total RNA is recognized as a specific miRNA by MSO hybridization.

Primer extension imparts further target specificity. The product of this extension is used as

a template for universal PCR. Abbreviations: DASL, cDNA-mediated annealing, selec-

tion, extension, and ligation; FFPE, formalin-fixed, paraffin-embedded; PCR, polymerase

chain reaction.



cancer panel was used in the DASL assay in order to profile both breast and

colon cancer FFPE tumor samples. Cluster analysis was able to separate breast

from colon tissue types and, subsequently, divide each tissue sample set into

cancer versus normal (32). More recently, the DASL assay has been used to

identify RNA signatures in prostate cancer, including a 16-gene set that corre-

lates with prostate cancer relapse (31,33).

We have designed our own custom 512-gene breast cancer–related gene

panel so that it incorporates previously identified signature genes from various

breast cancer microarray studies that have been used in the classification of

breast tumors (34,35), in prognosis (MammaPrint) (36) and as predictors of

outcome to treatment (17,37–40), as well as a host of additional genes that have

been implicated as playing a role in a number of breast cancer–related processes,

including proliferation, angiogenesis, (41,42), metastasis (43,44), DNA repair,

apoptosis (45), thrombosis (46), and custom-selected genes taken from the most

recently published data on breast cancer (47–49). Additional cancer-related

genes include oncogenes, cell cycle genes, telomerase-related genes, amplified

genes, breast cancer stem cell genes, and senescence-related genes (50–52).

MiRNA Profiling

MiRNAs (miRNAs) are small noncoding RNA species that regulate the trans-

lation and stability of mRNA messages for hundreds of downstream target genes

via partial complementarity to short sequences in the 30-untranslated region

(30UTR) of mRNAs. The aberrant expression or mutation of miRNAs, a distinct

class of RNAs that control gene expression via translation repression or degra-

dation, has been noted in cancers and suggests a key role for miRNAs in

tumorogenesis. MiRNAs can act both as tumor suppressors and as oncogenes,

depending on the sets of downstream targets that they regulate (53). Studies have

recently implicated nearly 30 of the over 450 known human miRNAs in breast

cancer, as tumor suppressor genes or as oncogenes, with significant dereg-

ulation among only a handful (54). Others are correlated with specific tumor

types as identified by protein expression. For example, the mir-30 cluster is

downregulated in both ER-negative, PR-positive and ER-positive, PR-negative

tumors (54). Differential expression has also been noted by tumor stage or pro-

liferative indices. The tumor suppressor, p53, was recently shown to regulate

expression of mir-34b and mir-34c (55). One key set of tumor suppressor miRNAs

includes mir15a-16, which targets the translation of the suppressor of apoptosis,

Bcl-2 (56). These observations suggest that specific miRNAs are associated with

major breast cancer subgroups, including hormone-positive and hormone-negative

subgroups and the basal or triple negative subtype, and that the deregulated

miRNAs may differ between African-American and Caucasian women.

These low-molecular weight miRNAs can be extracted at the same time as

total RNA is being isolated from FFPE tissue samples. qRT-PCR can be used to

screen tumor samples for miRNA expression. The TaqManTM array human



miRNA panel from Applied Biosystems allows for the expression profiling of

365 miRNAs from one RNA sample. Illumina, Inc., has recently adapted its

DASL assay for high-throughput multiplexed miRNA expression profiling

(Fig. 1B). It has made a human miRNA panel that can be used to detect 470

miRNAs described in the miRBase database and an additional 265 potential

miRNAs that have been identified in an RNA-primed array-based Klenow

extension (RAKE) analysis study (57,58). With these arrays and tools now

available, our understanding of the role that miRNAs play in breast cancer will

be greatly enhanced.

Whole-Genome Expression Profiling

In order to conduct expression profiling using microarrays, it has been necessary

to obtain freshly frozen tissue as the starting material for RNA extraction. This

presents limitations with regard to obtaining enough samples, enough tissue per

sample, as well as the logistics of storing this type of material. Most recently,

new technologies have come into play that enable the use of RNA extracted from

FFPE to be used in genomewide gene expression profiling. The company

NuGEN Technologies, Inc. (San Carlos, California, U.S.) has developed an

amplification system that can take picogram amounts of total RNA to yield

micrograms of amplified cDNA that can be fragmented and labeled for same-day

hybridization to a variety of array formats, including Affymetrix, Agilent, and

Illumina microarrays (59). This type of technology will undoubtedly generate

significant microarray data from FFPE breast tumor samples in the near future

further adding to the utility of FFPE specimens.

DNA-BASED ARRAY PROFILING THAT EMPLOYS FFPE SAMPLES

Array CGH and SNP Profiling

DNA copy number alterations (CNAs), which are the result of genomic insta-

bility, can give rise to gene amplifications and gene deletions that play important

roles in tumor progression, and it has become apparent that an increased number

of CNAs correlate with a poor prognosis for the patient. Several genes that

undergo CNAs have been identified that act as either oncogenes or tumor sup-

pressors, which are crucial factors in the disease process. With regard to breast

cancer, array comparative genomic hybridization (CGH) has been used principally

to identify novel breast cancer-related genes, genes that can be used to classify

different types of breast cancer, genes that yield prognostic or predictive informa-

tion, or genes that may represent important therapeutic targets. Single nucleotide

polymorphisms (SNPs) have been linked to individual variations in breast cancer

susceptibility. Genetic variations in genes involved in ER and drug metabolism and

genes involved in cell cycle control as well as variations depending on ethnic

background have been found to contribute to breast cancer susceptibility.



Array CGH has been used in the classification of breast tumors and has

shown that subtypes can be discerned on the basis of genomic instability pat-

terns. In one study, array CGH, made up of 2464 genomic clones, was applied to

the analysis of CNAs in 62 sporadic invasive ductal carcinomas, and three breast

tumor subtypes were delineated based on genomic DNA CNAs (60). In another

study, 89 breast cancer tumors of the luminal A, luminal B, HER2, and basal-like

subtypes were also found to yield distinct CNA profiles, suggesting that genomic

instability mechanisms differ between subtypes (61). Most array CGH studies

have made use of DNA prepared from fresh or frozen tumor samples; however,

the ability to tap into archival FFPE clinical samples would greatly accelerate the

rate at which disease-related genomic change, including translocation break-

points, could be more easily identified. Recent studies have addressed this

important issue and shown that DNA can be prepared from FFPE tissue samples

and used successfully in array CGH (62–64). Van Beers and colleagues (2006)

have also developed a multiplex PCR predictor that can be used to assess the

quality of the DNA extracted from FFPE samples (65) prior to use in array CGH.

Recently, Oosting and co-workers (2007) (66) have compared the Affymetrix

GeneChip and Bac array CGH with the Illumina BeadArray platform, which is a

high-density SNP microarray that allows for the measurement of both genomic

copy number and loss of heterozygosity (LOH). The BeadArray platform (Fig. 2)

is amenable for use with fragmented DNA from FFPE tissue because of the fact

that only short, intact genomic segments of approximately 40 bp flank each SNP

on the array. The authors found that when paired comparisons of DNA prepared

from freshly frozen and FFPE tissue were conducted, the patterns of genomic

aberrations were basically identical.

The use of array CGH and SNP profiling in conjunction with the vast

archive of FFPE clinical breast cancer samples will allow for the identification of

chromosomal aberrations as well as novel variants that contribute to breast

cancer, and this will lead to uncovering novel breast cancer-related genes that

will be key to understanding the molecular mechanisms underlying the disease.

Methylation Profiling

Epigenetic changes, which involve very subtle molecular modifications, can

cause activation or suppression of genes that can lead to abnormalities resulting

in tumorigenesis (67), and these epigenetic changes have been found to occur in

many cancers, including breast, colon, prostate, and blood cancer. The epi-

genotype is affected by such factors as the environment, age, epigenotype of the

parents, and the genotype at the loci that regulate DNA methylation and chro-

matin, and these factors can cause aberrant methylation of the cytosine residue of

DNA, resulting in gene regulation.

Illumina has developed an array-based platform (Fig. 3) that can be used in

conjunction with its standard Cancer Methylation Panel, which spans 1505 CpG

loci selected from 807 genes with 71% of those selected genes containing at least



2 CpG sites. The Cancer Methylation Panel includes genes selected from various

functional classes, including oncogenes and tumor suppressor genes; genes

involved in cell cycle control, DNA repair, differentiation, and apoptosis; and

X-linked and imprinted genes (68). Although DNA prepared from FFPE tumor

samples has not yet been specifically validated for use in the assay, it would be

useful to be able to scan large numbers of FFPE samples for methylation pat-

terns, which could provide important clues regarding tumor type.

THE WAY FORWARD

The goal is to obtain as complete a molecular picture of a breast tumor as possible,

which can now be realistically undertaken using FFPE tissue specimens, truly

untapped gold now ready to be brought to the surface and subjected to a thor-

ough nucleic acid characterization. Profiling platforms can now be brought to

bear on these invaluable samples in order to build a comprehensive picture of the

tumor. In the scheme shown in Figure 4, FFPE samples, either sections or cores,

are used to extract DNA, RNA, and miRNA. These nucleic acids can now be

Figure 2 Whole-genome expression profiling on the BeadChip array. (A) Gene-specific

probes are attached to beads, which are then assembled into the arrays. For simplicity,

this diagram shows only one oligomer attached to the bead; actual beads have hun-

dreds of thousands of copies of the same sequence attached. (B) Sentrix Human-6 and

HumanRef-8 Expression BeadChips. More than 48,000 different bead types, each with

a unique sequence, are represented on the Human-6 Expression BeadChip, greater

than 24,000 on the HumanRef-8 Expression BeadChip. Each bead contains hundreds

of thousands of copies of covalently attached oligonucleotide probes. Beads are

assembled into greater than 1.6 million pits, each measuring 3 mm in diameter,

generating an average 30-fold redundancy for each sequence represented on the array.



Figure 3 The DNA methylation assay. (A) In order to perform a DNA methylation assay,

genomic DNA must first undergo bisulfite conversion. Cytosines are converted to uracils,

but methyl-cytosine is unreactive. (B) In order to take advantage of this differential

reactivity to bisulfite treatment, two pairs of complex oligonucleotide probe sets are

designed for each CpG site. The first probe pair is made up of an ASO and a LSO to

interrogate the methylated state of the CpG site and a corresponding ASO-LSO pair for the

unmethylated state. The ASO is used to determine whether a site is methylated or not and

also incorporates a universal PCR primer sequence, P1 or P2. P1 and P2 are fluorescently

labeled, each with a different dye, and associated with the “T” (unmethylated) allele or

the “C” (methylated) allele, respectively. The LSO is made up of three parts: a CpG

(Continued)



locus-specific sequence, an address sequence in the middle corresponding to a comple-

mentary address sequence on the bead array, and a universal PCR priming site (P3) at the

30-end. The pooled assay oligonucleotides are first annealed to bisulfite-converted

genomic DNA. An allele-specific primer extension step is then carried out using ASOs

that are extended only if their 30-base is complementary to their cognate CpG site in the

genomic DNA template. Allele-specific extension is followed by ligation of the extended

ASOs to their corresponding LSOs to create PCR templates. Common primers P1, P2, and

P3 are then used to amplify the ligated products, which are subsequently hybridized to a

microarray bearing the complementary address sequences. Abbreviations: ASO, allele-

specific oligonucleotide; LSO, locus-specific oligonucleotide; PCR, polymerase chain

reaction.

Figure 4 Analysis of FFPE specimens on multiple array platforms. FFPE tissue samples,

as sections on either slides, sections, or cores are available for extraction of nucleic acids.

DNA, RNA, and miRNA can be isolated and applied to various array platforms in order to

maximize the molecular information that can be obtained from a single tumor sample.

DNA can be subjected to array CGH, SNP analysis, LOH profiling, and methlylation

analysis. RNA can be used in conjunction with expression profiling platforms that include

the DASL assay and whole-genome expression profiling on different microarray plat-

forms. MiRNA can be used to conduct miRNA profiling. This ability to utilize FFPE

tumor tissues for detailed molecular characterization will undoubtedly have a profound

impact on our understanding of tumor biology. Abbreviations: FFPE, formalin-fixed,

paraffin-embedded; CGH, comparative genomic hybridization; SNP, single nucleotide

polymorphisms; LOH, loss of heterozygosity; DASL, cDNA-mediated annealing, selec-

tion, extension, and ligation.

<



subjected to a panoply of analysis platforms that can be used to detect CNAs,

analyze SNPs, and investigate methylation aberrations, regulated gene expression,

and signaling pathway changes, as well as miRNA alterations. Data derived from

tumor tissue analyses can be used to further our understanding of the different

subtypes, which could lead to new insights into breast cancer etiology, prediction

of disease outcome, response to therapy, and ultimately to better and personalized

patient management.

REFERENCES

1. Masuda N, Ohnishi T, Kawamoto S, et al. Analysis of chemical modification of RNA

from formalin-fixed samples and optimization of molecular biology applications for

such samples. Nucleic Acids Res 1999; 27:4436–4443.

2. Abrahamsen HN, Steiniche T, Nexo E, et al. Towards quantitative mRNA analysis in

paraffin-embedded tissues using real-time reverse transcriptase-polymerase chain

reaction: a methodological study on lymph nodes from melanoma patients. J Mol

Diagn 2003; 5:34–41.

3. Finke J, Fritzen R, Ternes P, et al. An improved strategy and a useful housekeeping

gene for RNA analysis from formalin-fixed, paraffin-embedded tissues by PCR.

Biotechniques 1993; 14:448–453.

4. Godfrey TE, Kim SH, Chavira M, et al. Quantitative mRNA expression analysis

from formalin-fixed, paraffin-embedded tissues using 50 nuclease quantitative

reverse transcription-polymerase chain reaction. J Mol Diagn 2000; 2:84–91.

5. Reichmuth C, Markus MA, Hillemanns M, et al. The diagnostic potential of the

chromosome translocation t(2;13) in rhabdomyosarcoma: a Pcr study of fresh-frozen

and paraffin-embedded tumour samples. J Pathol 1996; 180:50–57.

6. Rupp GM, Locker J. Purification and analysis of RNA from paraffin-embedded

tissues. Biotechniques 1988; 6:56–60.

7. Sheils OM, Sweeney EC. TSH receptor status of thyroid neoplasms–TaqMan RT-

PCR analysis of archival material. J Pathol 1999; 188:87–92.

8. Specht K, Richter T, Muller U, et al. Quantitative gene expression analysis in

microdissected archival formalin-fixed and paraffin-embedded tumor tissue. Am

J Pathol 2001; 158:419–429.

9. Stanta G, Bonin S. RNA quantitative analysis from fixed and paraffin-embedded

tissues: membrane hybridization and capillary electrophoresis. Biotechniques 1998;

24:271–276.

10. Abramovitz M, Leyland-Jones B. Application of array-based genomic and epi-

genomic technologies to unraveling the heterogeneous nature of breast tumors: on

the road to individualized treatment. Cancer Genomics Proteomics 2007; 4:135–145.

11. Chen J, Byrne GEJr., Lossos IS. Optimization of RNA extraction from formalin-

fixed, paraffin-embedded lymphoid tissues. Diagn Mol Pathol 2007; 16:61–72.

12. Chung JY, Braunschweig T, Hewitt SM. Optimization of recovery of RNA from

formalin-fixed, paraffin-embedded tissue. Diagn Mol Pathol 2006; 15:229–236.

13. Penland SK, Keku TO, Torrice C, et al. RNA expression analysis of formalin-fixed

paraffin-embedded tumors. Lab Invest 2007; 87:383–391.



14. Cronin M, Pho M, Dutta D, et al. Measurement of gene expression in archival

paraffin-embedded tissues: development and performance of a 92-gene reverse

transcriptase-polymerase chain reaction assay. Am J Pathol 2004; 164:35–42.

15. Chevillard S, Muller A, Levalois C, et al. Reverse transcription-polymerase chain

reaction (RT-PCR) assays of estrogen and progesterone receptors in breast cancer.

Breast Cancer Res Treat 1996; 41:81–89.

16. Knowlden JM, Gee JM, Bryant S, et al. Use of reverse transcription-polymerase

chain reaction methodology to detect estrogen-regulated gene expression in small

breast cancer specimens. Clin Cancer Res 1997; 3:2165–2172.





2006; 24:3726–3734.

19. Paik S. Development and clinical utility of a 21-gene recurrence score prognostic

assay in patients with early breast cancer treated with tamoxifen. Oncologist 2007;

12:631–635.

20. Bild AH, Yao G, Chang JT, et al. Oncogenic pathway signatures in human cancers as

a guide to targeted therapies. Nature 2006; 439:353–357.

21. Huang E, Ishida S, Pittman J, et al. Gene expression phenotypic models that predict

the activity of oncogenic pathways. Nat Genet 2003; 34:226–230.

22. Huang ES, Black EP, Dressman H, et al. Gene expression phenotypes of oncogenic

signaling pathways. Cell Cycle 2003; 2:415–417.

23. Ramaswamy S, Tamayo P, Rifkin R, et al. Multiclass cancer diagnosis using tumor

gene expression signatures. Proc Natl Acad Sci U S A 2001; 98:15149–15154.

24. Nevins JR, Huang ES, Dressman H, et al. Towards integrated clinico-genomic models

for personalized medicine: combining gene expression signatures and clinical factors

in breast cancer outcomes prediction. Hum Mol Genet 2003; 12(2):R153–R157.



26. Abramovitz M, Ordanic-Kodani M, Wang Y, et al. Optimization of RNA extraction

from FFPE tissues for expression profiling in the DASL assay. Biotechniques 2007

(in press).

27. Abramovitz M, Leyland-Jones B. A systems approach to clinical oncology: focus on

breast cancer. Proteome Sci 2006; 4:5.

28. Bibikova M, Yeakley JM, Chudin E, et al. Gene expression profiles in formalin-

fixed, paraffin-embedded tissues obtained with a novel assay for microarray analysis.

Clin Chem 2004; 50:2384–2386.



2004; 165:1799–1807.

30. Fan JB, Yeakley JM, Bibikova M, et al. A versatile assay for high-throughput gene

expression profiling on universal array matrices. Genome Res 2004; 14:878–885.

31. Li HR, Wang-Rodriguez J, Nair TM, et al. Two-dimensional transcriptome profiling:

identification of messenger RNA isoform signatures in prostate cancer from archived

paraffin-embedded cancer specimens. Cancer Res 2006; 66:4079–4088.





2004; 165:1799–1807.

33. Bibikova M, Chudin E, Arsanjani A, et al. Expression signatures that correlated with

Gleason score and relapse in prostate cancer. Genomics 2007; 89:666–672.



U S A 2001; 98:10869–10874.



U S A 2003; 100:10393–10398.








Oncol 2005; 23:1169–1177.

39. Jansen MP, Foekens JA, van Staveren IL, et al. Molecular classification of tamoxifen-

resistant breast carcinomas by gene expression profiling. J Clin Oncol 2005; 23:

732–740.

40. Pawitan Y, Bjohle J, Amler L, et al. Gene expression profiling spares early breast

cancer patients from adjuvant therapy: derived and validated in two population-based

cohorts. Breast Cancer Res 2005; 7:R953–R964.

41. Heuze-Vourc’h N, Liu M, Dalwadi H, et al. IL-20, an anti-angiogenic cytokine that

inhibits COX-2 expression. Biochem Biophys Res Commun 2005; 333:470–475.

42. Zhong H, Chiles K, Feldser D, et al. Modulation of hypoxia-inducible factor 1alpha

expression by the epidermal growth factor/phosphatidylinositol 3-kinase/PTEN/

AKT/FRAP pathway in human prostate cancer cells: implications for tumor angio-

genesis and therapeutics. Cancer Res 2000; 60:1541–1545.

43. Kang Y, Siegel PM, Shu W, et al. A multigenic program mediating breast cancer

metastasis to bone. Cancer Cell 2003; 3:537–549.

44. Minn AJ, Gupta GP, Siegel PM, et al. Genes that mediate breast cancer metastasis to

lung. Nature 2005; 436:518–524.

45. Miller LD, Smeds J, George J, et al. An expression signature for p53 status in human

breast cancer predicts mutation status, transcriptional effects, and patient survival.

Proc Natl Acad Sci U S A 2005; 102:13550–13555.

46. Kwaan HC, Parmar S, Wang J. Pathogenesis of increased risk of thrombosis in

cancer. Semin Thromb Hemost 2003; 29:283–290.

47. Gonzalez MA, Pinder SE, Callagy G, et al. Minichromosome maintenance protein 2

is a strong independent prognostic marker in breast cancer. J Clin Oncol 2003;

21:4306–4313.

48. McLean GW, Carragher NO, Avizienyte E, et al. The role of focal-adhesion kinase

in cancer: a new therapeutic opportunity. Nat Rev Cancer 2005; 5:505–515.

49. Zhao C, Yasui K, Lee CJ, et al. Elevated expression levels of NCOA3, TOP1, and

TFAP2C in breast tumors as predictors of poor prognosis. Cancer 2003; 98:18–23.



50. Brabletz T, Jung A, Spaderna S, et al. Opinion: migrating cancer stem cells - an

integrated concept of malignant tumour progression. Nat Rev Cancer 2005; 5:744–749.

51. Collado M, Gil J, Efeyan A, et al. Tumour biology: senescence in premalignant

tumours. Nature 2005; 436:642.

52. Dikmen ZG, Gellert GC, Jackson S, et al. In vivo inhibition of lung cancer by

GRN163L: a novel human telomerase inhibitor. Cancer Res 2005; 65:7866–7873.

53. Hammond SM. MicroRNAs as oncogenes. Curr Opin Genet Dev 2006; 16:4–9.

54. Iorio MV, Ferracin M, Liu CG, et al. MicroRNA gene expression deregulation in

human breast cancer. Cancer Res 2005; 65:7065–7070.

55. Corney DC, Flesken-Nikitin A, Godwin AK, et al. MicroRNA-34b and microRNA-

34c are targets of p53 and cooperate in control of cell proliferation and adhesion-

independent growth. Cancer Res 2007; 67:8433–8438.

56. Cimmino A, Calin GA, Fabbri M, et al. miR-15 and miR-16 induce apoptosis by

targeting BCL2. Proc Natl Acad Sci U S A 2005; 102:13944–13949.

57. Berezikov E, van Tetering G, Verheul M, et al. Many novel mammalian microRNA

candidates identified by extensive cloning and RAKE analysis. Genome Res 2006;

16:1289–1298.

58. Griffiths-Jones S, Grocock RJ, van Dongen S, et al. miRBase: microRNA sequences,

targets and gene nomenclature. Nucleic Acids Res 2006; 34:D140–D144.

59. Turner, L, de Feo, G, Heath, JD, et al. WT-OvationTM FFPE system: RNA ampli-

fication for gene expression profiling using RNA from formalin fixed, paraffin

embedded (FFPE) tissue. Los Angeles, CA: American Association of Cancer

Research (AACR) 20073 (abstr).

60. Fridlyand J, Snijders AM, Ylstra B, et al. Breast tumor copy number aberration

phenotypes and genomic instability. BMC Cancer 2006; 6:96.

61. BergamaschiA,KimYH,WangP, et al.Distinct patterns ofDNAcopynumberalteration

are associated with different clinicopathological features and gene-expression subtypes

of breast cancer. Genes Chromosomes Cancer 2006; 45:1033–1040.

62. Johnson NA, Hamoudi RA, Ichimura K, et al. Application of array CGH on archival

formalin-fixed paraffin-embedded tissues including small numbers of microdissected

cells. Lab Invest 2006; 86:968–978.

63. Joosse SA, van Beers EH, Nederlof PM. Automated array-CGH optimized for archival

formalin-fixed, paraffin-embedded tumor material. BMC Cancer 2007; 7:43.

64. Little SE, Vuononvirta R, Reis-Filho JS, et al. Array CGH using whole genome

amplification of fresh-frozen and formalin-fixed, paraffin-embedded tumor DNA.

Genomics 2006; 87:298–306.

65. van Beers EH, Joosse SA, Ligtenberg MJ, et al. A multiplex PCR predictor for aCGH

success of FFPE samples. Br J Cancer 2006; 94:333–337.

66. Oosting J, Lips EH, van Eijk R, et al. High-resolution copy number analysis of paraffin-

embedded archival tissue using SNP BeadArrays. Genome Res 2007; 17:368–376.

67. Esteller M. Cancer epigenomics: DNA methylomes and histone-modification maps.

Nat Rev Genet 2007; 8:286–298.

68. Bibikova M, Lin Z, Zhou L, et al. High-throughput DNA methylation profiling using

universal bead arrays. Genome Res 2006; 16:383–393.



18

Breast Cancer Stem Cells and Their

Niche: Lethal Seeds in Lethal Soil

Danuta Balicki

Department of Medicine and Research Centre, Centre Hospitalier del’Universite de Montreal; Department of Medicine, Universite de Montreal;

and Breast Cancer Prevention and Genetics and Breast Cancer RiskAssessment Clinic, Ville Marie Multidisciplinary Breast Cancer Centre,

Montreal, Quebec, Canada

Brian Leyland-Jones

Winship Cancer Institute, Emory University School of Medicine,Atlanta, Georgia, U.S.A.

Max S. Wicha

Comprehensive Cancer Centre, Department of Internal Medicine, Division ofHematology-Oncology, University of Michigan, Ann Arbor, Michigan, U.S.A.

INTRODUCTION

The quest for optimal breast cancer therapy has been thwarted by recurrent

treatment failures. As a result, the necessity has emerged to concomitantly rethink

the fundamental basis of breast cancer, and to ensure that it is adequately and

optimally targeted by therapeutic agents, and combinations of these agents.

Adding to the complexity of this disease is the heterogeneity of breast cancer, and

the realization that breast cancer is not a single disease, but a family of diseases.

The daunting challenge that we face with a newly diagnosed breast cancer patient

is one-to-one communication that tailors an individualized treatment strategy

259


based on the corresponding breast cancer pathophysiology. At the same time, we

are mandated to address the urgency of this therapy on the basis of the underlying

clinical assessment, the risks of metastases and death, as well as the benefits and

adverse effects associated with therapeutic regimens to ensure that an enlightened

decision among available therapeutic options can be mutually reached with an

individual patient. As our understanding of the pathophysiology of breast cancer is

incomplete, it is critical to channel the information that we glean from observations

that deviate from our current models into a unified paradigm shift where thera-

peutic targets are validated, and subsequently attacked and destroyed. Given the

complexity of breast cancer, the highest level of therapeutic success is most

likely to be achieved when the uniqueness of an individual patient’s breast

cancer is matched with individualized, combined, and targeted therapy.

CANCER STEM CELL HYPOTHESIS

The cancer stem cell hypothesis is a major player in the current paradigm shift in

oncology (1,2). Cancer is, in essence, a genetic disease where three groups of genes

are responsible for tumor progression, namely oncogenes, tumor suppressor genes,

and stability genes (3,4). Cancer is a multistage process involving the accumulation

of genetic and epigenetic changes in genes involved in the regulation of cell

growth, DNA repair, and metastasis. In the traditional stochastic model, the vast

majority of cancer cells can proliferate extensively to form new tumors. In contrast,

the cancer stem cell model suggests that only rare cancer stem cells have the

intrinsic ability to proliferate extensively to form new tumors (Fig. 1) (1). Clinical

observations indicate that breast cancer responses to therapy do not necessarily

correlate with patient survival. Tumor-initiating cells with the concomitant stem

cell properties of self-renewal and successive maturation of progenitor cells into

highly differentiated cells are referred to as “cancer stem cells” (Fig. 2).

The cancer stem cell hypothesis provides an explanation for our clinical

observations, including those which deviate from the stochastic model, by suggesting

that therapies that kill cancer cells but not cancer stem cells may lead to tumor

regression, but that the disease will subsequently recur as the cancer stem cells act as

lethal seeds to repopulate the tumor (5). Thus, tumor regression in itself may represent

an incomplete clinical end point, as it reflects cell death limited to differentiated tumor

cells, while sparing cancer stem cells (Fig. 3). The “dandelion hypothesis” has been

used to illustrate this pattern of clinical activity as it is analogous to cutting off a

dandelion (or other weed) at ground level, which may initially appear to produce the

desired effect, but only the elimination of the root will prevent the weed from

regrowing (6). Therefore, effective therapies should target cancer stem cells, with

limited effects on normal cells. In the presence of therapies that kill cancer stem

cells, tumors are predicted to lose their ability to generate new cancer cells,

subsequently degenerate, leading to complete tumor eradication.

Stem cells are multipotent with high proliferative potential and with

concomitant properties of self-renewal and differentiation into multilineage

260 Balicki et al.


cells. Recent findings suggest that stem cell properties are integral and funda-

mental to the formation and perpetuation of human cancers, including breast

cancer (7). The quiescence, self-renewal, and differentiation of stem cells are

regulated by both intrinsic and extrinsic mechanisms. Intrinsic mechanisms

include those governing the epigenetic state of stem cells, for example, hema-

topoietic stem cells are controlled by chromatin remodelers such as polycomb

group proteins (8,9). The extrinsic mechanisms include changes in stem cell

fate that are dictated by the environment, i.e., the stem cell niche. As our tissues

are organized into stem cell hierarchies ranging from stem cells with extensive

proliferative and self-renewal capacity to mature cells with little or no capacity

for cell division, it is likely that cancer cell populations might also be organized

in stem cell hierarchies, ranging from a small number of cells that are responsible

for fueling the uncontrolled growth of the tumor cells to a large array of dif-

ferentiated daughter cells (10). While the vast majority of cells in a tumor are

merely the nontumorigenic daughter cells of cancer stem cells, the highly pro-

liferative capacities of stem cells may drive the continued expansion of malig-

nant cells. As cellular proliferation decreases with differentiation, the long-lived

and slowly dividing stem and progenitor cells are likely candidates for the

accumulation of mutations associated with carcinogenesis (11,12).

Figure 1 A comparison of the stochastic model where most cancer cells can proliferate

extensively to give rise to new tumors with the cancer stem cell model where only a few

cancer stem cells can give rise to new tumors. Abbreviation: CSC, cancer stem cell.

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 261


The multilineage differentiation of cancer stem cells may contribute to the

heterogeneity of certain tumors, including breast cancer. The parallels between

normal stem cells and cancer stem cells suggest that some cancer stem cells

arise from mutated normal stem cells, and are responsible for the proliferative

capacity of cancers (Fig. 2) (7). The observation that stem and cancer cells share

certain signaling pathways that regulate their self-renewal suggests that normal

stem cells can, in fact, give rise to cancer cells (8). Mutated progenitor cells

which acquire self-renewal properties may also become cancer stem cells

(Fig. 2) (7,13). Epigenetic gene silencing and associated promoter CpG island

DNA hypermethylation may precede genetic changes in premalignant cells and

foster the accumulation of additional genetic and epigenetic hits that contribute

Figure 2 NSCs and CSCs have the capacity of self-renewal and differentiation into

progenitor cells. NSCs divide asymmetrically giving rise to a stem cell and a differ-

entiated progenitor cell. CSCs may lose asymmetric division, leading to selectively self-

renewal, thereby increasing the number of CSCs and the subsequent tumor burden.

Abbreviations: NSCs, normal stem cells; CSCs, cancer stem cells.

262 Balicki et al.


to the generation of cancer stem cells from stem or early progenitor cells (14).

The accumulation of mutations in cancer stem cells may disrupt the tight control

of normal stem cells, leading to the dysregulation of self-renewal, tumorigenesis,

loss of asymmetric division, and aberrant differentiation (Fig. 2) (12,15). The

fate of stem cells, including self-renewal and differentiation, may also be pro-

grammed by the stem cell niche, i.e., the cellular microenvironment providing

necessary support and stimuli to sustain self-renewal and other environmental

factors (11,13,16–19). In normal breast development, mammary stem cells dif-

ferentiate into myoepithelial, alveolar epithelial, and ductal epithelial cells (20).

Cancer stem cells may acquire additional features associated with tumor pro-

gression, metastases, and therapeutic failure, including genetic instability and

drug resistance (11). Mechanisms of therapeutic resistance include cellular

quiescence and the acquisition of protein expression responsible for the cellular

efflux of chemotherapeutic agents. The elimination of the cancer stem cell

compartment of a tumor may be necessary to achieve stable, long-lasting cancer

remissions. Therapeutic strategies designed to eradicate the cancer stem

Figure 3 Therapies which kill cancer cells, but not CSCs lead to an initial tumor

regression, but as the CSCs are not eliminated, the tumor will regenerate. In contrast,

therapies that target CSCs also eliminate the genesis of new tumor cells and result in

tumor regression. Abbreviation: CSCs, cancer stem cells.



compartment should selectively, or at least preferentially, target the self-renewal,

survival, and proliferative pathways specific to cancer stem cells, or induce

differentiation. Thus, advances in our knowledge of stem cell properties are

pivotal to the specific and effective targeting of these cells in therapeutic

strategies, including cancer therapies.

CURRENT STATUS OF BREAST STEM CELL RESEARCH

The human mammary gland is organized during development as a hierarchy of

stem and progenitor cells that are successively limited in their multilineage

potential and proliferative capacities. Stem cells represent approximately 1 in

250 epithelial cells of the normal human breast (17). The transplantation of

mammary cells into cleared murine fat pads is a functional assay used to identify

mammary epithelial cells with stem cell properties based on their ability to

generate tumors (21,22). The existence of a stem cell origin of murine mammary

development was established using transplantation of cells at limiting dilutions.

These experiments confirmed the clonal nature of the regenerated alveolar,

ductal, lobular, and more complex structures (22–24). The unequivocal dem-

onstration of the existence of a murine mammary stem cell occurred with the

reconstitution of a complete and functional mammary gland from a single lin-

eage negative (Lin�)/cluster of differentiation (CD)24þ/CD29þ cell, marked

with a b-galactosidase (LacZ) transgene (25).

Three types of murine mammary epithelial cell progenitors have been

identified thus far (22–24). The first is a bipotent progenitor with features of

both luminal and myoepithelial characteristics, the second has luminal features,

and the third has myoepithelial features. Analogous human studies of stem and

progenitor cells have been limited by the challenges associated with the devel-

opment of suitable in vivo xenotransplantation assays. Alternative strategies

have been undertaken to assay human breast stem cells and to optimize conditions

that support the in vitro growth and differentiation of primitive human epithelial

cells. These include the proliferation of nonadherent mammospheres in suspen-

sion from human breast cancer tissues (17,26) and murine mammary fat pads

(27). These mammospheres are enriched in stem and progenitor cells, as illus-

trated by their ability to differentiate along all three mammary epithelial lineages

and to clonally generate complex functional structures in reconstituted three-

dimensional culture systems (17). Additional studies to characterize stem and

progenitor cells include the identification of side populations using flow cytom-

etry, i.e., low Hoechst 33342 dye-retaining populations (28–30). Side populations

are enriched 30-fold in mammospheres derived from human tissues (17).

The use of mammosphere assays subsequently lead to the discovery that

the Notch signaling pathway plays a pivotal role in normal human mammary

development through its activity in both stem and progenitor cells, concurrently

affecting self-renewal and lineage-specific differentiation. As a result, it was

proposed that abnormal Notch signaling contributes to mammary carcinogenesis

264 Balicki et al.


by dysregulating the self-renewal of normal mammary stem cells (26). Addi-

tional pathways thought to be involved in stem cell self-renewal include

hedgehog (HH), Bmi-1, wingless (Wnt/b-catenin) and phosphatase and tensin

homolog deleted on chromosome ten (PTEN) (11,20,31).

Further studies in human breast stem/progenitor cells identified a subset of

human breast cancer cells using flow cytometry with the phenotype CD44þ/CD24�/Lin� that possess highly tumorigenic properties in a nonobese diabetic/

severe combined immunodeficiency (NOD/SCID) xenograft model (21). As few

as 200 cells with the epithelial-specific antigen (ESA)þ, CD44þ/CD24�/Lin�

phenotype consistently generated tumors in mice, whereas even 20,000 cells

with alternative phenotypes were ineffective. The isolation and serial passage of

the tumorigenic CD44þ/CD24�/Lin� cell population in NOD/SCID mice gave

rise to additional tumorigenic CD44þ/CD24�/Lin� cells, as well as phenotypi-

cally diverse nontumorigenic cells, that composed the bulk of the tumors,

thereby confirming the self-renewal and differentiating capacities of these cells.

While the unequivocal demonstration of stem cell characteristics necessitates

model systems capable of tumor generation from a single cell, a population of

CD44þ/CD24�/Lin� cells appears to exhibit properties of human breast stem/

progenitor cells in NOD/SCID mice.

GENETIC SIGNATURES AND POLYMORPHISMS WHICHMAY INFLUENCE BREAST CANCER DEVELOPMENTAND TREATMENT OUTCOMES

The study of molecular breast cancer portraits identifies patterns of molecular

breast cancer markers using gene expression microarrays, which subsequently

are useful in the selection of therapy. These portraits also seek to overcome

potential differences in drug sensitivity or target expression between tumor-

initiating cells and the more frequent nontumorigenic cells (32). The rationale for

the pursuit of new drug discovery based on molecular targets results from the

significant increment of the effectiveness of these drugs as compared with

conventional chemotherapy. A recent systematic analysis of the consensus of

human breast and colorectal cancers included the analysis of 13,023 genes in

11 breast and 11 colorectal cancers, and revealed that individual tumors accu-

mulate an average of approximately 90 mutated genes, but that only a subset of

these contribute to the neoplastic process (33). Using stringent criteria, 189 genes

(average of 11 per tumor) were found to be mutated at significant frequencies.

The vast majority of these genes were not known to be genetically altered in

tumors and are predicted to affect a wide range of cellular functions, including

transcription, adhesion, and invasion. The genetic landscape of breast and col-

orectal cancer defined in this study provides new targets for diagnostic and

therapeutic intervention.

Gene signatures of tumor invasiveness in tumorigenic CD44þCD24�/low

breast cancer cell populations correlate with overall and metastasis-free survival,



independently of established clinical and pathological variables (34).

CD44þCD24�/low breast cancer cells were also shown to increase from a median

of 4.8% to 14.8% after chemotherapy in paired breast cancer biopsies from

35 patients obtained before and after 12 weeks of neoadjuvant chemotherapy.

Microarray analyses of CD44þCD24�/low breast cancer cells using the Affymetrix

U133 GeneChip demonstrated an increased expression of self-renewal pathways.

An increase in mammosphere formation of residual tumors was also observed

postchemotherapy. Taken together, these results suggest that therapy resistant

cancer stem cells may exist in the neoadjuvant setting of breast cancer therapy, and

that breast cancer patients may benefit from the therapeutic targeting designed to

eradicate these cells (35).

In addition to chemoresistance, progenitor cells may also be resistant to

radiation. This radioresistance has been shown in an animal model to be

mediated at least in part by Wnt signaling, which has been implicated in stem

cell survival (36). Radioresistance was investigated by treating primary BALB/c

mouse mammary epithelial cells with clinically relevant doses of radiation. An

enrichment in normal progenitor cells [stem cell antigen (SCA) 1-positive and

side population progenitors] was identified.

Specifically, radiation selectively enriched for progenitors in mammary

epithelial cells isolated from transgenic mice with activated Wnt/b-cateninsignaling but not for background-matched controls. Irradiated stem cell antigen

1-positive cells had a selective increase in active b-catenin and survivin

expression compared with SCA 1-negative cells. Radiation also induced enrich-

ment of side population progenitors cells of the MCF-7 human breast adeno-

carcinoma cell line. In comparison to differentiated cells, progenitor cells have

different cell survival properties that may facilitate the development of targeted

antiprogenitor cell therapies.

Closely intertwined with response prediction are an individual patient’s

genetic polymorphisms, which may increase breast cancer susceptibility and

influence the efficacy or safety of breast cancer treatments. The accumulation of

genetic factors that underlie breast cancer susceptibility or therapeutic escape

may trigger the transformation of a normal stem or progenitor cell into a cancer

stem cell. The variation of treatment responses between individuals also lies in

differences in DNA sequences that alter the expression or function of proteins

that are drug targets. These proteins are encoded by genes identified in early

studies that are linked to highly penetrant, single-gene traits. Germline poly-

morphisms which confer susceptibility to breast cancer are associated with the

highly penetrant breast cancer genes BRCA1 and BRCA2, p53, PTEN, ataxia

telangiectasia mutated (ATM), and candidate BRCA3 genes (37). In addition,

BRCA1 has been suggested to function as a breast stem cell regulator (38). Thus,

BRCA1 may contribute to a conceptual link between hereditary and sporadic

breast cancer. The loss of BRCA1 function may produce a block in cell dif-

ferentiation and an increase in the self-renewal of mammary stem cells, resulting

in the expansion of the stem cell pool. Taken together with its known function in

266 Balicki et al.


DNA repair, BRCA1may generate genetically unstable stem cells which are targets

for additional carcinogenic events and the aggressive triple negative “basal” stem

cell–like phenotype (39).

Low-penetrance genes that may influence breast cancer susceptibility

alone or in combination with environmental factors include genes in DNA repair,

metabolic pathways, and estrogen-related pathways including HRAS1 minis-

atellite, CHEK2, XRCC1, XRCC2, RAD51, XRCC3, and LIG4 (37). Metabolic

enzymes such as N-acetyltransferases (NAT) and gluthathione-S-transferases

(GST) are involved in the metabolic inactivation of carcinogens. Mutations in

these enzymes can result in increased toxicity and exposure to mutagenic sub-

stances, thereby increasing mutations, which contribute to carcinogenesis.

Superoxide dismutases (SODs) are metalloenzymes that protect cells from free

radicals, and their mutations can lead to increased breast cancer susceptibility

(37). Cytochrome p450s (CYP) metabolically activate carcinogens into chem-

ically reactive electrophiles. CYP1B1 is the most common form in the breast,

and its alanine and serine (ALA/SER) polymorphism significantly increases

breast cancer susceptibility (37). Additional targets of polymorphisms that

increase breast cancer susceptibility include COMT, CYP17, CYP19, PPAR-g,TGF-b, HSP70-1, HSP70-2, and HSP70-HOM (37). Germline polymorphisms

may also influence an individual’s susceptibility to the effects of breast cancer

treatments (37). For example, dihydropyrimidine dehydrogenase (DPD) and

thymidylate synthase (TS) are pivotal to the metabolism of 5-FU, and specific

polymorphisms of these enzymes lead to 5-FU toxicity. Paclitaxel is metabolized

by cytochrome P450, and specific CYP polymorphisms lead to defective metab-

olism. Additional breast cancer treatments with known polymorphisms, which

may effect drug efficacy or safety include tamoxifen (ER, CYP2D6, SULT1A1),

aromatase inhibitors (CYP19, CYP1A2, CYP2C9, CYP3A), cyclophosphamide

(GST), methotrexate (MTHFR), doxorubicin (GST), and epirubicin (UGT2B7)

(40). Future advances hinge on the more difficult challenge of elucidating mul-

tigene determinants of breast cancer susceptibility and drug responses through an

intersection of genomics and medicine, which has the potential to yield a new set

of molecular diagnostic tools that can be used to individualize and optimize drug

therapy (41). Additional refinements can be accomplished through coupling in

vitro drug sensitivity data with microarray data, to develop gene expression sig-

natures that predict sensitivity to individual chemotherapeutic drugs (42). Thus, as

we continue to fine-tune our stratification, we also approach our ultimate goal of

defining individual molecular breast cancer portraits.

STEM CELL SANCTUARIES AS A MECHANISMOF THERAPEUTIC RESISTANCE

Targeting of stem cells is particularly complex since this cellular population is

largely quiescent. Thus, therapeutic agents directed against cycling cells are

predictably ineffective in this population, which is essentially shielded from



these drugs. Cancers that respond to therapy initially may acquire drug resistance

during the course of treatment, while other cancers appear to be intrinsically

resistant. The cancer stem cell hypothesis suggests that in both of these cases, the

resting cancer stem cell, which is both the cancer-initiating cell and its source

of replenishment under selective pressure, has innate drug resistance by virtue of

its quiescent stem cell phenotype. Acquired drug resistance in more differ-

entiated cancer cells may occur through gene amplification or rearrangement,

thereby contributing to an aggressive phenotype. Through the accumulation of

mutations that distinguish cancer stem cells from their normal counterparts, these

cells may acquire additional features associated with tumor progression,

metastases, and therapeutic failure, including genetic instability, radioresistance,

and chemoresistance (11,36). In the case of gliomas, cancer stem cells contribute

to radioresistance through the preferential activation of the DNA damage

checkpoint response and an increase in DNA repair capacity (43). Radio-

resistance may also result from a selective increase of cell survival pathways in

progenitor cells, including b-catenin and survivin (36).

Stem cell niches regulate stem cell proliferation and cell fate decisions, and

they also play a protective role by shielding stem cells from environmental

insults. In this manner, vascular niches might protect brain cancer stem cells

from chemo- and radiotherapies, and the opportunity for these cells to reform a

tumor mass in spite of an initial clinical response (44). With the acquisition of

properties of tumor progression and metastases, cancer stem cells may travel to

sanctuary sites, such as across the blood-brain barrier, where it may be difficult

to eradicate them, even with targeted therapies.

The drug resistance of normal tissue stem cells is constitutively mediated

by multidrug resistance (MDR) transporters and detoxifying enzymes (45). DNA

repair mechanisms, resistance to apoptosis, and telomerase activity also con-

tribute to the stability of normal tissue stem cells. Among the mechanisms

responsible for stem cell drug resistance, there is mounting speculation that ABC

transporters repress the maturation and differentiation of stem cells (46,47).

However, since breast cancer resistance protein 1 (BCRP1) knockout mice

display normal hematopoiesis instead of accelerated hematopoietic maturation,

it is clear that multiple factors determine the maturation and differentiation of

stem cells (48). In addition, protection from hypoxia appears to be another

function of ABC transporters. As a result, it has been postulated that Bcrp

expression protects hematopoietic stem cells from hypoxic environments by

preventing porphyrin accumulation that causes mitochondrial death (49).

Thus, leading edge-targeted breast cancer therapies must overcome all of

the obstacles associated with cancer stem cell biology, including both the

intrinsic and extrinsic properties of resistance, which define these enigmatic cells

and the cloisters that keep them safe. From the cancer stem cell hypothesis, it

follows that systemic administration of an efficacious MDR reversal agent would

render tumor and normal tissue stem cells equally susceptible to the chemo-

therapeutic agents, offering no net gain in therapeutic index (45). Thus, targeting

268 Balicki et al.


the properties that distinguish cancer stem cells from their normal stem cell

counterparts is necessary.

CELLULAR BREAST CANCER STEM CELL TARGETS

The identification of cancer stem cell markers is necessary for the isolation

and purification of cancer stem cells as they are potential targets of novel drug

discovery and development. The functional targets that define stem cells are

their abilities to self-renew and differentiate using serial passages in xenotrans-

plantation models. The current breast cancer stem cell hypothesis suggests that

the most primitive human breast stem cell does not express the estrogen receptor

(ER) while more mature progenitors may or may not express ER depending on

their location in the differentiation pathway (12). A recent study using reverse

transcriptase polymerase chain reaction (RT-PCR) and immunohistochemical

analyses established that mouse mammary stem cells were negative for both ERaand the progesterone receptor (50). An independent study further supported this

result by showing that ER-positive cells retained very little stem cell activity (51).

Throughout much of mammary development during embryogenesis, the devel-

oping gland consists of a simple duct system, with a primitive mammary bud, the

growing terminal end of the duct, with no ERa expression until the 30th week of

gestation. These structures are thought to be made of primitive ductal and

myoepithelial progenitors. Later in gestation, ERa-positive cells are present in

the growing terminal ends of these ducts, followed by a surge of ERa-positivecells in this region after birth (12). In the developing mammary gland, insulin

growth factor-1 (IGF-1) is required to form ducts and terminal end buds, indi-

cating its critical role in stimulating the proliferation and differentiation of

primitive breast epithelium (12).

ER-positive progenitors arise after 30 weeks of gestation. They respond to

estrogen stimulation by proliferating and generating ductal epithelial cells and

alveolar cells that form the adult mammary tree. Maintenance of the myoepi-

thelial progenitor pool and the stem cell compartment may rely on paracrine

signaling initiated by the ER-positive progenitor cells. Breast carcinogenesis

results from mutations affecting stem and progenitor cells. Thus, the phenotype

of different subtypes of breast cancer is determined both by the cell of origin and

the particular mutations driving carcinogenesis. Thus, ER-negative tumors arise

from stem cells or early ER-negative progenitor cells, while ER-positive tumors

arise from either ER-negative or ER-positive progenitors (12).

On the basis of these observations, a breast tumor model has been proposed

consisting of three types of tumors. The first is an ER-negative tumor termed

type 1, which arises from the most primitive ER-negative stem or early pro-

genitor cells. Using classic histopathology and molecular profiling studies, these

tumors are poorly differentiated and display a basal phenotype with expression

markers of both luminal epithelial and myoepithelial cells. Clinically, these

tumors are highly aggressive and are associated with a poor prognosis. Type 2



tumors also arise from early ER-negative stem or early progenitor cells. How-

ever, this group includes stem cell mutations, which allow for the differentiation

of a specific subset of tumor cells into ER-positive cells, which may range from a

low percentage (6–10%) similar to that seen in normal tissue to a high percentage

(70–100%). Thus, these tumors are heterogeneous, with intermediate differentia-

tion and prognosis. Type 3 tumors arise from the transformation of ER-positive

progenitors. They are predicted to exclusively express luminal markers and have

the best prognosis as they contain differentiated cells and respond to antiestrogen

therapy (12). Additional breast stem cell markers include CD44, CD24, aldehyde

dehydrogenase (ALDH), Oct-4, Bmi-1, and Wnt/b-catenin.As discussed above, a small population of cancer cells characterized by

CD44 expression but low or undetectable levels of CD24 (CD44þCD24�/low)

have a high tumorigenic capacity when injected into immunodeficient mice (52).

In contrast, the rest of the cancer cells, called nontumorigenic breast cancer cells,

have little or no tumorigenic ability. Tumors in mice that originate from purified

tumorigenic breast cancer cells contain a mixture of both tumorigenic and

nontumorigenic breast cancer cells, which confirms that the CD44þCD24�/low

population shares the capacity for self-renewal with normal stem cells. Gene

expression profiling of tumorigenic breast cancer cells was utilized to develop

prognostic tools to assess survival in patients with breast cancer and to predict

clinical outcomes in breast cancer (34). An invasiveness gene signature of 186

genes was developed to predict the likelihood of a tumor to metastasize. This

invasiveness gene signature is associated with the risk of death and metastasis

not only in breast cancer but also in lung cancer, prostate cancer, and medul-

loblastoma. These results suggest the critical relevance of the CD44þCD24�/low

tumorigenic subclass of breast cancer cells as a pivotal contributor to the asso-

ciation of the invasiveness signature to clinical outcome. In addition, CD44, a

cell-surface receptor that mediates cell–cell contacts, promotes the engraftment

of leukemic stem cells in the bone marrow and may thus have an additional role

as a critical component of the cancer stem cell niche (53).

As a result of the growing interest of identifying new markers for identi-

fying and isolating stem cells, assays based on ALDH activity have been pro-

posed to be a promising alternative in both murine and human mammary stem

cells. ALDH is a family of cytosolic enzyme isoforms responsible for oxidizing

intracellular aldehydes, including vitamin A, to carboxylic acids. It is highly

expressed in hematopoietic progenitors, in intestinal crypt cells, as well as in

breast tumor cells (54–56). ALDH1 or (ALDH isoenzyme 1 or ALDH1A1) has

been shown to be responsible for the resistance observed in hematopoietic stem

cells and breast cancer tumor cells to the alkylating agent, cyclophosphamide

(56–58). A novel system has been designed to detect ALDH using a visible light

excitable fluorochrome named Aldefluor1 (BODIPY-conjugated amino-

acetaldehyde, BAAA) metabolized by both murine and human cells (59).

Fluorescence-activated cell sorter (FACS) of the normal breast cell population

revealed that 6% of this population is ALDH-positive (60). In breast cancer cell

270 Balicki et al.


lines, ALDH-positive cells were more frequently observed in basal-like than in

luminal cell lines. ALDH-positive cells form mammospheres at a high efficiency

and are tumorigenic in NOD/SCID mice (61).

A recent study showed that when acute myeloid leukemic (AML) patient

cells were sorted on the basis of Aldefluor staining, three distinct possibilities

were observed (62). High levels of aldehyde dehydrogenase (ALDHhi) cells from

a first group of patients were identified as normal nonleukemic stem cells, were

present in low number, and were able to reconstitute a complete hematopoietic

system when transplanted into NOD/SCID mice. However, a second group of

patients had ALDHhi cells displaying a higher side scattering pattern than normal

stem cells and these cells were also significantly more abundant. These cells

correspond to leukemic stem cells. Finally, a third group of patients showed no

ALDHhi cells. This study demonstrates that FACS analyses based on the first

ALDH isozyme (ALDH1 or ALDHA1) activity can be used to identify and

isolate leukemic stem cells (58). In contrast to the toxicity associated with

Hoechst 33342, ALDH is relatively mild and allows for cell sorting using viable

cell populations, enriched in stem cells. The availability of antibodies derived

against ALDH1 allows for the study of their expression in fixed tissue samples

using immunohistochemistry (Fig. 4). Immunohistochemistry of 577 human

breast carcinoma specimens from two different sources detected 24% and 30%

ALDH-positive cells, respectively. In this study, ALDH expression was strongly

correlated with tumor size, grade, ER status, human epidermal growth factor

receptor 2 (HER2) overexpression, and poor prognosis (60).

OCT-4 is an additional cancer stem cell marker. The OCT-4 oncogene

encodes a nuclear protein that belongs to a family of transcription factors con-

taining the POU DNA–binding domain that plays an important role in maintaining

the pluripotent state of embryonic stem cells (63). It may prevent expression of

genes activated during differentiation, and its expression is downregulated during

differentiation. OCT-4 is normally found in the pluripotent stem cells of pregas-

trulation embryos, including oocytes, early cleavage-stage embryos, and the inner

cell mass of the blastocyst, and knockout of OCT-4 causes early lethality in mice

because of the absence of an inner cell mass (63). Thus, these activities associated

with OCT-4 suggest that it plays a pivotal role in mammalian development and in

the self-renewal of embryonic stem cells. The expression of genes such as OCT-4

is potentially correlated with tumorigenesis and may affect some aspects of tumor

behavior, such as tumor recurrence or resistance to therapies. OCT-4 is expressed

in bladder cancer, as well as in CD44þ/CD24� breast cancer–initiating cells

(64,65). It may serve as a multifunctional factor involved in major biological

processes such as embryonic development, control of differentiation, and stem

cell–based carcinogenesis (64).

Bmi-1 is a cancer stem cell marker that belongs to the polycomb group

family of proteins, and was first identified as a c-myc cooperating oncogene in

murine lymphomagenesis. Bmi-1 is involved in the regulation of diverse bio-

logical processes involved in X chromosome inactivation, repression of cellular



senescence, carcinogenesis, and stem cell renewal (66). Bmi-1 oncoprotein

overexpression correlates with axillary lymph node metastases in invasive ductal

breast cancer (67). In addition, the Bmi-1 oncogene induces telomerase activity

and immortalizes human mammary epithelial cells (68). Bmi-1 regulates self-

renewal of normal and malignant human mammary stem cells, stimulates

mammosphere formation and colony size in vitro, and increases ductal/alveolar

development in humanized NOD/SCID mammary fatpads (20).

The final stem cell marker in this chapter is wingless (Wnt)/b-catenin. Thecanonical Wnt/b-catenin signaling pathway modulates the delicate balance

between stemness, differentiation, and tumorigenesis in several adult stem cell

Figure 4 ALDH1 immunostaining. Immunohistochemistry using anti-ALDH1 on tissue

microarray samples from two patients (A, B) with infiltrating ductal carcinoma of the breast,

both magnified 200�, with a scale bar of 100 mm. Abbreviation: ALDH1, aldehyde dehy-

drogenase isoenzyme 1.

272 Balicki et al.


niches such as the hair follicles in the skin, the mammary gland, hematopoiesis,

and the intestinal crypt (8,69). Differences in the levels of Wnt signaling activity

reflect tumor heterogeneity and are likely to account for distinct cellular activ-

ities such as proliferation and epithelial-mesenchymal transitions, which prompt

tumor growth and malignant behavior, respectively. This pivotal pathway is

highly conserved in evolution and is known to regulate cell-fate decisions, cell

proliferation, morphology, migration, apoptosis, differentiation, and stem cell

self-renewal. Recent evidence which suggests that Wnt signaling plays a role in

human breast cancer include the detection of elevated levels of nuclear and/or

cytoplasmic b-catenin using immunohistochemistry, and the overexpression or

downregulation of specific Wnt proteins. The Wnt pathway has also been

implicated in normal stem cell self-renewal in vivo, and there is evidence that

dysregulation of this pathway in the mammary gland and other organs may play

a key role in carcinogenesis (70).

The Wnt family of secreted proteins includes the well-characterized

canonical Wnt signaling pathway, in which Wnt ligands signal through the

stabilization of b-catenin. In this pathway, Wnt proteins bind to a family of

frizzled receptors in a complex with the low-density lipoprotein receptor–related

proteins 5 and 6 (LRP5/6) coreceptors to activate Dishevelled (Dsh). Sub-

sequently, Dsh inhibits the activity of the b-catenin destruction complex [ade-

nomatous polyposis coli (APC), axin, and glycogen synthase kinase-3b (GSK-

3b)], which phosphorylates b-catenin in the absence of the ligands. As a result,

b-catenin is stabilized and translocated to the nucleus, where it recruits trans-

activators to high mobility group (HMG)–box DNA-binding proteins of the

lymphoid-enhancer factor/T-cell factor (LEF/TCF) family. In the absence of

Wnt signaling, b-catenin remains in the cytoplasm, where it forms the b-catenindestruction complex. GSK-3b phosphorylates b-catenin, which targets the pro-

tein for ubiquitin-mediated degradation. When the Wnt pathway is activated,

GSK-3b is inhibited, blocking b-catenin phosphorylation and its subsequent

degradation (71). In addition, several b-catenin-independent Wnt signaling

pathways, known as noncanonical, have been shown to be crucial for different

aspects of vertebrate embryo development (71).

MOLECULAR BREAST CANCER STEM CELL TARGETS

Molecular signatures of stem cells are currently being identified. The mutations that

accumulate in cancer stem cells could inappropriately activateBmi-1,Wnt/b-catenin,Notch, sonic hedgehog (SHH), and PTEN pathways that are involved in the regu-

lation of self-renewal of normal stem cells and could potentially lead to cancer

(Figs. 5 and 6) (26,72–74). These molecular markers can be exploited to purify

cancer stem cells, which can then be characterized in the most intimate detail,

by the modern tools of gene expression profiling and informatics. The failure to

shut off the self-renewal pathways could confer self-renewal on cells and micro-

environments capable of evading homeostatic regulation, leading to tumorigenesis.



The intimate interrelationship of multiple pathways critical to tumorigenesis and

to normal stem cell self-renewal raises the specter that cancer therapies targeting

such pathways might ablate resident stem cells (75). Drugs that selectively target

the self-renewal pathways of tumors will be attractive candidates for cancer

therapies if they can effectively spare the normal bystander stem cells that use

self-renewal as part of their regular functional repertoire. Alternatively,

Figure 5 Stem cell self-renewal pathways include the HH, Notch, Bmi-1, Wnt, and

PTEN pathways. Dysregulation of the self-renewal pathways of NSCs and progenitor

cells may lead to CSCs and tumor formation. NSCs are ER negative and they differentiate

into ER-positive progenitor cells, which further differentiate into myoepithelial, alveolar

epithelial, and ductal epithelial cells. CSCs in their niche may lose asymmetric division,

leading to the clonal expansion of cancerous cells and tumor formation. Abbreviations:

NSCs, normal stem cells; CSCs, cancer stem cells; ER, estrogen receptor.

274 Balicki et al.


Figure 6 A model of the interrelationship of stem cell self-renewal pathways including a

conceptual link between sporadic and hereditary breast cancers. BRCA1 may inhibit stem

cell self-renewal, while HER2 overexpression may increase the stem cell pool through

regulation of self-renewal. In this model, there is bidirectional cross talk between the

Notch and HH pathways with subsequent regulation of the polycomb gene Bmi-1 and the

Wnt/b-catenin pathway. SHH, IHH, and DHH are HH ligands, while Gli1 and 2 are two

HH transcription factors that are positive mediator of HH signaling. Cylcopamine inhibits

the HH pathway. The DSL peptide activates Notch signaling, while the GSI blocks the

intramembrane cleavage of Notch required for signaling. PTEN affects Wnt signaling

through GSK-3b. Exisulind, bromoindirubin-30-oxime, and imatinib inhibit the Wnt/b-catenin pathway. Abbreviations: BRCA, breast cancer; HER, human epidermal growth

factor receptor 2; HH, hedgehog; SHH, sonic hedgehog; IHH, Indian hedgehog; DHH,

desert hedgehog; PTEN, phosphatase and tensin homolog deleted on chromosome ten;

GSI, gamma secretase inhibitor.



differentiating therapies may also be effective if they can efficiently shift the

balance of stem cell divisions toward differentiation rather than self-renewal and

stem cell expansion. In addition to Bmi-1, and Wnt/b-catenin discussed above,

Notch, SHH pathways, and PTEN will be reviewed below.

The Notch signaling pathway plays an important role in regulating pro-

liferation, survival, and differentiation of many cell types, including the regu-

lation of self-renewal of adult stem cells and differentiation of progenitor cells

along a particular lineage. These outcomes include increased survival or death,

proliferation or growth arrest, and commitment to or blockage of differentiation.

There are four mammalian Notch homologues, Notch 1 to 4, which interact with

the DSL ligands: Delta, Delta-like, Jagged1, and Jagged2 in vertebrates (71).

Notch 1 and Notch 4 homologues are involved in the normal development

of the mammary gland, and mouse mammary tumors are associated with their

mutated forms. Notch signaling is critical to normal human mammary devel-

opment where it acts on both stem cells and progenitor cells, affecting self-

renewal and lineage-specific differentiation. Notch signaling may also contribute

to mammary carcinogenesis by dysregulating the self-renewal pathways of

normal mammary stem cells. There is bidirectional cross talk between the Notch

and HH pathways with subsequent regulation of the polycomb gene Bmi-1, as

well as the Wnt signaling network (71). The DSL ligands activate Notch sig-

naling, while the gamma secretase inhibitor (GSI) blocks the intramembrane

cleavage of Notch required for signaling (71). The abnormal expression of Notch

receptors in epithelial metaplastic lesions and neoplastic lesions suggests that

Notch may act as a protooncogene (26). In spite of the simplicity of the Notch

signaling pathway, it yields varied outcomes in different contexts (76). These

different outcomes are mediated through a novel signaling pathway in which

Notch receptors on the cell surface are processed to give rise to a nuclear

transcriptional activation complex. Activated Notch modulates stem and pro-

genitor cell renewal through the upregulation of the cyclin-dependent kinase

inhibitors p21cip1 and p27kip1 (77). Notch signaling is also required for the

survival of niche cells (78). Notch receptor activation promotes the survival of

neural stem cells by inducing the expression of the specific target genes hairy and

enhancer of split 3 (Hes3) and SHH. The underlying mechanism occurs through

the rapid activation of cytoplasmic signals, including the serine/threonine kinase

AKT, the transcription factor STAT3, and mammalian targets of rapamycin

(mTOR) molecular targets of rapamycin (79).

The HH signaling pathway has important roles in embryonic development

and tumorigenesis (80). In a number of tumors, including prostatic and pan-

creatic cancers, HH signaling is abnormally upregulated, and effective signaling

inhibition can repress tumor growth and induce apoptosis (81–83). The core

components of the HH pathway are evolutionarily conserved from insects to

mammals. They include the secreted ligand HH, the membrane receptor Patched

(PTCH), the membrane signal transductor Smoothened (SMO), and the tran-

scription factor Ci/glioma (GLI). However, while in Drosophila, there is only

276 Balicki et al.


one secreted ligand (HH), one receptor (Ptc), and one transcription factor (Ci),

there are three secreted ligands, sonic HH (SHH )/desert HH (DHH)/Indian HH

(IHH), two receptors (PTCH1/2), and three transcription factors (GLI1/2/3) in

mammals (81). The HH signaling components PTCH1 and GLI1/2 are highly

expressed in normal mammary stem and progenitor cells, and activated HH sig-

naling is mediated by Bmi-1 (30). Addition of Hh ligands increases the expression

of the transcription factors Gli1 and Gli2, which are positive mediators of Hh

signaling (20). SHH can potentiate tumor cell proliferation. Components of the

SHH–GLI pathway are expressed in adult human prostate cancer, often with

enhanced levels in tumors versus normal prostatic epithelia. Cyclopamine and

anti-SHH antibodies inhibit the proliferation of GLI1–PSA primary prostate tumor

cultures (83). The SHH signaling pathway also regulates adult neuronal self-

renewal, and may play a role in brain tumor biology (4).

The final pathway discussed herein is PTEN, which is a phosphatase that

inhibits cellular proliferation by negative regulation of signaling through the

phosphatidylinositol-3-OH kinase (PI3K) pathway. The activation of PI3K leads

to phosphorylation and activation of the Akt protein, which in turn can poten-

tially phosphorylate a multitude of proteins (84). The PTEN–Akt pathway reg-

ulates stem cell activation by helping control nuclear localization of the Wnt

pathway effector b-catenin. PTEN inhibits, whereas Akt enhances, the nuclear

localization of b-catenin. Akt is thought to coordinate with Wnt signaling to

assist in the activation of b-catenin in intestinal stem cells, either through

phosphorylation of GSK-3b or by phosphorylation of b-catenin itself. The

phosphorylation by Akt inactivates GSK-3b, helping the Wnt signal to permit

b-catenin to escape proteasome-mediated degradation and facilitating its nuclear

translocation and activity in transcriptional regulation (85).

In addition to other downstream effectors, the highly branched PI3K

pathway activates the mammalian target of rapamycin (mTOR). Molecules such

as p21Cip1/Waf1 are found downstream to PTEN, and they influence stem cell

proliferation, without necessarily implicating self-renewal. FOXO regulatory

proteins may have the effects of PTEN on p21Cip1/Waf1 by influencing the

expression of a range of crucial genes for stem cell function, including cell

division, and tolerance of oxidative stress. Thus, the perturbation of PTEN may

thus result in nonrenewing cell divisions or death of stem cells (84). PTEN has

also been reported to be a guardian of genome integrity and of the maintenance

of chromosomal stability through the physical interaction with centromeres and

control of DNA repair. PTEN acts on chromatin and regulates the expression of

Rad51, which reduces the incidence of spontaneous double strand breaks (86).

PTEN is commonly deleted or otherwise inactivated in diverse cancers,

including hematopoietic malignancies, glioblastomas, endometrial carcinomas,

breast carcinomas, and prostate carcinomas (86). PTEN deletion or HER2

overexpression may modulate stem cell function leading to increased invasion

and tumorigenicity (87). PTEN deletions are implicated in the generation of

transplantable leukemia-initiating cells, and the depletion of normal hematopoietic



stem cells (88). PTEN normally maintains hematopoietic stem cells in a quiescent

state, and in the absence of PTEN, hematopoietic stem cells are driven into the cell

cycle, eventually leading to depletion of hematopoietic stem cell reserves. Thus,

PTEN activity identifies a mechanistic difference between the maintenance of

normal stem cells and cancer stem cells. These effects were mostly mediated by

mTOR as they are inhibited by rapamycin, which depletes leukemia-initiating

cells but also restores normal hematopoietic stem cell function (88). This mech-

anistic distinction shall enable the design of therapies that target the PTEN

pathway to combat leukemia with the goal of killing leukemic stem cells, without

damaging the normal stem cell pool (75). In addition to leukemia, this pathway

may be an attractive therapeutic target for treating a broad spectrum of tumors

carrying inactivated PTEN because mTOR activity can be suppressed by a number

of compounds including rapamycin (75). Genes that promote stem cell quiescence

might be particularly amenable to such targeted strategies because although nor-

mal stem cells thrive in a quiescent world, forcing cancer stem cells to adopt

quiescence could lead to their eradication (75).

TARGETS OF THE STEM CELL NICHE

While cancer stem cells may indeed be lethal seeds in themselves, they may

whither and die unless they are supported by environmental factors, which

perpetuate their survival and self-renewal. The stem cell niche provides this

microenvironment for cancer stem cells, and the combination of lethal seeds in

lethal soil is likely to lead to the most aggressive forms of breast cancer, with the

most fatal course. The stem cell niche determines cell fate decisions based on

surrounding cells, extracellular matrix components, and secreted local and sys-

temic factors (11). The 1889 proposal by Stephen Paget that metastasis depends

on cross talk between selected cancer cells (the “seeds”) and specific organ

microenvironments (the “soil”) still holds forth today, and may also apply to the

interrelationship between cancer stem cells and their niche (89–91). Cancer

metastasis requires the seeding, homing, and successful colonization of spe-

cialized cancer stem cells at distant sites (92). In this context, it is not surprising

that the inhibition of the homing factor CXCR4 has been shown to effectively

prevent both primary tumor formation as well as metastasis in animal models

(92). Stem cells of various tissues exist within protective niches that are com-

posed of differentiated cells which provide direct cell contacts and secreted

factors that maintain stem cells primarily in a quiescent state by providing cues

for the inhibition of proliferation (8,44). The mutations that distinguish cancer

stem cells from normal stem cells may enable them to escape niche control.

Alternatively, the niche provides proliferation or differentiation-inducing signals

at times when high numbers of progenitors are needed and can quickly give rise

to all committed cell lineages (8). As a result, the uncontrolled proliferation of

stem cells and tumorigenesis may be a consequence of the dysregulation of

extrinsic factors within the niche. In this case, the stem cell niche might represent

278 Balicki et al.


targets for cancer therapy (44). Given that postnatal estrogen exposure is related to

breast cancer risk, this effect could also apply in utero, where the developing fetus

is exposed to estrogen levels up to 10-fold higher than at any other time in the

individual’s lifetime. Under these circumstances, estrogens or other hormones may

create the fertile soil that is responsible for an increase in the responsiveness of

estrogen-responsive tissue to undergo oncogenesis later in life (93).

Epigenetic modulation of gene expression is highly abnormal in cancers,

which often show aberrant promoter CpG island hypermethylation and tran-

scriptional silencing of tumor suppressor genes and prodifferentiation factors.

The aberrant epigenetic landscape of the cancer cell is also characterized by a

massive genomic hypomethylation, an altered histone code for critical genes, and

a global loss of monoacetylated and trimethylated histone H4 (94). In embryonic

stem cells, these genes are maintained in a state that is ready for transcription

mediated by a bivalent promoter chromatin pattern consisting of the repressor,

histone H3 methylated at Lys27 (H3K27) by polycomb group proteins, plus the

activator, methylated H3K4. However, in contrast to embryonic stem cells,

embryonic carcinoma cells add two key repressors, dimethylated H3K9 and

trimethylated H3K9, which are both associated with DNA hypermethylation in

adult cancers (14). Polycomb group proteins reversibly repress genes required

for differentiation in embryonic stem cells. Stem cell polycomb group targets are

up to 12-fold more likely to have cancer-specific promoter DNA hyper-

methylation than nontargets, supporting a stem cell origin of cancer where

reversible gene repression is replaced by permanent silencing, locking the cell

into a perpetual state of self-renewal, thereby predisposing to subsequent

malignant transformation (95). Thus, cell chromatin patterns and transient

silencing of these important regulatory genes in stem or progenitor cells may

leave these genes vulnerable to aberrant DNA hypermethylation and heritable

gene silencing during tumor initiation and progression.

Vascular niches have been described in the case of neural stem cells, which

were found to be concentrated in regions that are rich in blood vessels (18).

These vessels shelter neural stem cells from apoptotic stimuli and maintain a

proper balance between self-renewal and differentiation. The endothelial cells

that line blood vessels are key components of the vascular niche as they secrete

factors that promote stem cell survival and self-renewal. As the cotransplantation

of tumor cells with endothelial cells leads to more rapid tumor formation, the

vascular niche may contribute to tumor initiation and development (44). In an

established tumor, cancer stem cells that find a vascular niche may continue to

self-renew, while the remaining cells may differentiate to form the bulk of the

tumor, but not to its long-term maintenance (18). Thus, the identification of

endothelial cell signals, which regulate the self-renewal of cancer stem cells,

may also contribute to our understanding of the interrelationship of cancer stem

cells and their niche. There is some evidence that there is a bidirectional rela-

tionship between cancer stem cells and their niche. For example, gliomas secrete

elevated levels of vascular endothelial growth factor (VEGF), which increases



endothelial cell migration and tube formation (43). In addition, VEGF promotes

neural stem cell survival (96). Thus, the effectiveness of antiangiogenic agents in

cancer therapy may relate to the prevention of new blood vessels that inhibit

tumor growth. In addition, antiangiogenic therapy may disrupt a vascular niche,

which is necessary for cancer stem cell renewal (18).

In the case of hematopoietic stem cell niches, studies have focused on the

endosteum, the inner surface of the bone that interfaces with the bone marrow.

Regulation of hematopoietic stem cell function is governed by factors secreted

by osteoblasts, osteoclasts, and stromal fibroblasts within the endosteum (8).

Imaging data suggests that hematopoietic stem cells reside near the endosteum,

and that this localization and maintenance is also usually adjacent to CXCL12-

secreting reticular cells, irrespective of whether the hematopoietic stem cells

are in vascular or endosteal locations (97,98). The stimulation via parathyroid

hormone–related protein receptors (PPR) or an increase in number of osteo-

blastic cells in the periosteum with high levels of the Notch ligand Jagged1,

enlarged the stem cell niche with a resultant increase in the number of hem-

atopoietic stem cells with evidence of Notch1 activation in vivo (99). The

activation of Notch1 has been shown to result in enhanced self-renewal of

hematopoietic stem cells possibly by inducing self-renewal genes including

Hes1 (8). In addition, the bone morphogenetic protein signal has an essential

role in inducing haematopoietic tissue during embryogenesis within the hem-

atopoietic stem cell niche (100). Mesenchymal stem cells have been shown to

improve engraftment of hematopoietic stem cells, and suppress graft-versus-

host disease after allogeneic hematopoietic stem cell transplantation. In addi-

tion, when tumor cells are injected into NOD/SCID mice in conjunction with

mesenchymal stem cells, their growth is much faster as compared to the group

receiving only tumor cells. To explain this observation, it was suggested that

mesenchymal stem cells have the ability to form a cancer stem cell niche in

which tumor cells can preserve their proliferative potential to sustain the

malignant process (101).

Activated, irradiated, mammary fibroblasts are critical to the micro-

environment of mammary stem cells as they are required for the successful

repopulation of cleared mammary fat pads in immunocompromised mice (102).

These epithelial-stromal interactions are also important in malignant transfor-

mation and likely to be important with cancer stem cells (11). Fibroblasts secrete

matrix metalloproteinases and protease inhibitors and mobilize and activate

growth factors such as tumor growth factor-b (TGF-b), which has been shown to

promote tumor progression and invasion (103). Carcinoma-associated fibroblasts

promote angiogenesis, as well as tumor growth, by secreting stromal cell-derived

factor-1 (SDF-1), which interacts with the chemokine receptor CXCR4 on tumor

cells and endothelial cells (104). As radiographic breast density correlates with

increased breast cancer risk, as well as increased fibroblasts and collagen

deposition, this modified niche may promote the transformation of initiated stem

cells into cancer stem cells (11,105).

280 Balicki et al.


FUTURE CHALLENGES: INDIVIDUALIZED THERAPY BASEDON BREAST CANCER STEM CELLS AND THEIR NICHE

In the last few years, there has been a significant evolution in breast cancer care, if

not a revolution (106). Whereas, we previously focused on tumor size, nodal

status, and certain prognostic features based on tumor pathology, the paradigm has

now shifted to include the underlying molecular pathways that drive canceri-

genesis, as well as the cellular targets of breast cancer, including cancer stem cells

and their niche. The intricate circuitry of pathways that define the hallmarks of

cancer and the corresponding complex molecular interactions are being translated

into therapeutic benefit (107,108). Individualized drivers are gradually being

factored into patient assessments in order to astutely and accurately determine the

clinical course of breast cancer patients on the basis of their unique and complex

cellular and molecular breast cancer portraits.

Breast cancer stem cell markers and the critical molecules in the pathways

that drive their self-renewal are necessary to complete the molecular portrait of

breast cancer. Our failure to eradicate most cancers may be as fundamental as a

misidentification of the target cancer stem cell (Fig. 3) (6). The fact that stem cells

rarely divide and have unique cellular properties, coupled with the observation that

they may have high levels of drug transporters to pump chemotherapy agents out

of the cell, has led many to believe that traditional chemo- or radiation therapies

are insufficient to clear these tumor-initiating cells from the body (92). Clinical

observations indicate that breast cancer responses to therapy do not necessarily

correlate with patient survival and are incomplete clinical endpoints. Instead, the

effectiveness of these treatments should be evaluated by decreased cancer recur-

rence and metastases as measures of therapeutic efficacy (92). As the molecular

pathways that govern normal and cancer stem cells overlap, the targeting of these

pathways could have harmful effects on the homeostatic function of normal stem

cells. Targeted therapies that specifically eliminate cancer stem cells could further

revolutionize the ways by which we treat cancer.

As the proliferation of cancer stem cells is important for the growth of the

primary tumor, as well as progression to metastatic disease, inhibition of the self-

renewal capacity of these cells is an obvious target for their elimination. Therapies

such as cyclopamine (targeting the HH signaling) and exisulind, bromoindirubin-30-oxime, and imatinib (targeting the Wnt/b-catenin pathways) were designed to

inhibit self-renewal pathways critical to cancer stem cells, and have achieved

varying levels of success (92). Another strategy to eliminate self-renewal is to force

cancer stem cells to differentiate (Fig. 6).

As cancer stem cells are currently being identified in a broad spectrum of

cancers (44,109–112), significant advances in their fundamental properties are

forthcoming, and will serve as a scaffold to expedite our understanding and iden-

tification of cancer stem cells within specific types of cancers, including breast

cancer. The existence of overlapping cancer stem cell pathways within different

cancer types will also stimulate the discovery and development of new drugs to



target these pathways, as they are likely to be effective and marketable in more than

merely a small subset of cancer patients.

The addition of molecular markers in our patient assessments acknowl-

edges the pivotal role that the corresponding molecular pathways play in fueling

tumorigenesis. This fine-tuning of patient prognostic factors will ensure that the

oncogenic drivers of each individual patient’s tumor pathophysiology are both

identified and tailored into the definition and optimization of their specific

therapeutic strategy, which will translate into their best possible clinical outcome

and overall survival. As a result, it is critical that our therapeutic armamentarium

target the specific fingerprint of cellular and molecular pathways defined for an

individual patient. Therapeutic resistance via quiescence and the sequestration of

cancer stem cells into sanctuaries that are impermeable to most therapeutic

agents must be taken into consideration in the design of individualized breast

cancer therapies. Therapeutic resistance may also result from an inadequate

depletion of key cellular and molecular elements within the cancer stem cell

niche. The elucidation of the secreted factors that set up the premetastasis and

metastatic niche could provide both diagnostic as well as therapeutic benefits.

Drugs such as Bevacizumab (Avastin1) that reduce intratumoral pressure

associated with leaky tumoral vasculature and thereby increase the introduction

of intravenous agents within the tumor bed may have the critical synergy nec-

essary in combinations of targeted agents to permeate cancer stem cell sanc-

tuaries and cancer stem cell niches (18,44,113,114). In addition, they may target

both the vascular niche and the associated cancer stem cells (44). Thus, it is not

surprising that clinical trials of Bevacizumab combined with the chemo-

therapeutic drug CPT-11 suggest that this combination is one of the most

effective treatments of glioblastoma (115). Other agents that stimulate cycling of

quiescent cancer stem cells may also be beneficial in combined therapies, which

include agents that eradicate cycling cells. In addition, individualized therapy

must also have the long-term objectives of predicting and preparing to eradicate

the molecular pathways that may circumvent those pathways that drove the

initial breast cancer pathophysiology, and were targeted by first-line targeted

therapies. This strategy also underlines the critical objective of eradicating every

single cancer stem cell and its niche in newly diagnosed breast cancer patients,

thereby striving to eliminate all possibilities that these lethal seeds in lethal soil

may ignite tumor recurrence in the future.

Therefore, the genesis of individualized breast cancer therapy including

the eradication of cancer stem cells has become a necessity to achieve optimal

breast cancer patient care. It is imperative to understand the biology of cancer

stem cells and their niche to optimize our cancer treatment strategies to prevent

the longevity of cancer stem cells from interfering with our own (18). The

associated costs of its future reality precipitate a call to rethink our healthcare

strategies with a focus on cancer prevention and therapeutic schedules which

deliver targeted agents when they are both necessary and most effective in our

quest to optimally treat an individual breast cancer patient with the ultimate goal

282 Balicki et al.


of a stable, long-term remission. In this setting, the identification and targeting of

breast cancer stem cells and their niche hold great promise in the development of

new therapeutic translational strategies using state-of-the-art individualized

therapies for breast cancer patients.

ACKNOWLEDGMENTS

Rhyna Salinas (CHUM) is gratefully acknowledged for her expert assistance in

the preparation of this chapter. The contributions made by Drs. Francine

Jolicoeur, Isabel Garcia de la Fuente, and Louis Gaboury (CHUM) to Figure 4

during their collaboration with D.B. are greatly appreciated. Drs. Connie Eaves,

Gabriela Dontu, Suling Liu, Christophe Ginestier, Hasan Korkaya, and Julie

Dutcher are thankfully acknowledged for insightful and inspirational dis-

cussions. Dr. Connie Eaves is also gratefully recognized for her invaluable role

as D.B.’s first stem cell research mentor and role model, who initially planted the

seed that she continues to nourish. D.B. is a scholar of Fonds de la recherche en

sante du Quebec (FRSQ) and principal investigator of the National Cancer

Institute of Canada’s Clinical Trials Group (NCIC CTG). This chapter was also

made possible through a Canadian Institutes for Health Research (CIHR)

operating grant to D.B..

REFERENCES

1. Wicha MS, Liu S, Dontu G. Cancer stem cells: an old idea—a paradigm shift.

Cancer Res 2006; 66(4):1883–1890.

2. Balicki D. Cancer stem cells and resistance to treatment. Oncol Exchange 2007;

6(1): 37–40.

3. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med

2004; 10(8):789–799.

4. Massard C, Deutsch E, Soria JC. Tumour stem cell-targeted treatment: elimination

or differentiation. Ann Oncol 2006; 17(11):1620–1624.

5. Wicha MS. Cancer stem cells and metastasis: lethal seeds. Clin Cancer Res 2006;

12(19):5606–5607.

6. Jones RJ, Matsui WH, Smith BD. Cancer stem cells: are we missing the target?

J Natl Cancer Inst 2004; 96(8):583–585.

7. Jordan CT, Guzman ML, Noble M. Cancer stem cells. N Engl J Med 2006; 355(12):

1253–1261.

8. Rizo A, Vellenga E, de Haan G, et al. Signaling pathways in self-renewing hem-

atopoietic and leukemic stem cells: do all stem cells need a niche? Hum Mol Genet

2006; 15(spec no 2):R210–R219.

9. Schwartz YB, Pirrotta V. Polycomb silencing mechanisms and the management of

genomic programmes. Nat Rev Genet 2007; 8(1):9–22.

10. Bernstein A. Cancer stem cells: the centrality of translational research to cancer

control. CMAJ 2007; 176(1):29–30.



11. Dontu G, Liu S, Wicha MS. Stem cells in mammary development and carcinogenesis:

implications for prevention and treatment. Stem Cell Rev 2005; 1(3):207–213.

12. Dontu G, El Ashry D, Wicha MS. Breast cancer, stem/progenitor cells and the

estrogen receptor. Trends Endocrinol Metab 2004; 15(5):193–197.

13. Krivtsov AV, Twomey D, Feng Z, et al. Transformation from committed progenitor

to leukaemia stem cell initiated by MLL-AF9. Nature 2006; 442(7104):818–822.

14. Ohm JE, McGarvey KM, Yu X, et al. A stem cell-like chromatin pattern may pre-

dispose tumor suppressor genes to DNA hypermethylation and heritable silencing.

Nat Genet 2007; 39(2):237–242.

15. Clevers H. Stem cells, asymmetric division and cancer. Nat Genet 2005; 37(10):

1027–1028.

16. Scadden DT. The stem-cell niche as an entity of action. Nature 2006; 441(7097):

1075–1079.

17. Dontu G, Abdallah WM, Foley JM, et al. In vitro propagation and transcriptional

profiling of human mammary stem/progenitor cells. Genes Dev 2003; 17(10):

1253–1270.

18. Yang ZJ, Wechsler-Reya RJ. Hit ’em where they live: targeting the cancer stem cell

niche. Cancer Cell 2007; 11(1):3–5.

19. Moore KA, Lemischka IR. Stem cells and their niches. Science 2006; 311(5769):

1880–1885.

20. Liu S, Dontu G, Mantle ID, et al. Hedgehog signaling and Bmi-1 regulate self-

renewal of normal and malignant human mammary stem cells. Cancer Res 2006;

66(12):6063–6071.

21. Al-Hajj M, Wicha MS, ito-Hernandez A, et al. Prospective identification of


22. Stingl J, Eirew P, Ricketson I, et al. Purification and unique properties of mammary

epithelial stem cells. Nature 2006; 439(7079):993–997.

23. Smith GH. Experimental mammary epithelial morphogenesis in an in vivo model:

evidence for distinct cellular progenitors of the ductal and lobular phenotype. Breast

Cancer Res Treat 1996; 39(1):21–31.

24. Kordon EC, Smith GH. An entire functional mammary gland may comprise the

progeny from a single cell. Development 1998; 125(10):1921–1930.

25. Shackleton M, Vaillant F, Simpson KJ, et al. Generation of a functional mammary

gland from a single stem cell. Nature 2006; 439(7072):84–88.

26. Dontu G, Jackson KW, McNicholas E, et al. Role of Notch signaling in cell-fate

determination of human mammary stem/progenitor cells. Breast Cancer Res 2004;

6(6):R605–R615.

27. Youn BS, Sen A, Behie LA, et al. Scale-up of breast cancer stem cell aggregate

cultures to suspension bioreactors. Biotechnol Prog 2006; 22(3):801–810.

28. Hadnagy A, Gaboury L, Beaulieu R, et al. SP analysis may be used to identify

cancer stem cell populations. Exp Cell Res 2006; 312(19):3701–3710.

29. Goodell MA, Brose K, Paradis G, et al. Isolation and functional properties of

murine hematopoietic stem cells that are replicating in vivo. J Exp Med 1996;

183(4):1797–1806.

30. Balicki D, Beaulieu B. Cancer stem cell side populations. In: Bapat SA, ed. Cancer

Stem Cells. Hoboken: John Wiley and sons, Inc., 2008 (in press).

31. Dontu G, Al-Hajj M, Abdallah WM, et al. Stem cells in normal breast development

and breast cancer. Cell Prolif 2003; 36(suppl 1):59–72.

284 Balicki et al.


32. Al-Hajj M, Becker MW, Wicha M, et al. Therapeutic implications of cancer stem

cells. Curr Opin Genet Dev 2004; 14(1):43–47.

33. Sjoblom T, Jones S, Wood LD, et al. The consensus coding sequences of human

breast and colorectal cancers. Science 2006; 314(5797):268–274.

34. Liu R, Wang X, Chen GY, et al. The prognostic role of a gene signature from

tumorigenic breast-cancer cells. N Engl J Med 2007; 356(3):217–226.

35. Chang JC, Li X, Wong H, et al. Journal of Clinical Oncology, 2007 ASCO Annual

Meeting Proceedings Part I. Vol. 25, No. 18S (June 20 Supplement), 2007:9S.

36. Woodward WA, Chen MS, Behbod F, et al. WNT/beta-catenin mediates radiation

resistance of mouse mammary progenitor cells. Proc Natl Acad Sci U S A 2007;

104(2):618–623.

37. Lymberis SC, Parhar PK, Katsoulakis E, et al. Pharmacogenomics and breast

cancer. Pharmacogenomics 2004; 5(1):31–55.

38. Foulkes WD. BRCA1 functions as a breast stem cell regulator. J Med Genet 2004;

41(1):1–5.

39. Liu S, Ginestier S, Dontu G, et al. BRCA1 regulates human mammary stem cell

self-renewal and differentiation. In: Proceedings of the 98th Annual Meeting of the

American Association for Cancer Research, Los Angeles, CA, 14–18 April 2007.

Philadelphia, PA: AACR; 2007:1346–1347. Abstract nr 5700.

40. Stearns V, Davidson NE, Flockhart DA. Pharmacogenetics in the treatment of

breast cancer. Pharmacogenomics J 2004; 4(3):143–153.

41. Evans WE, Relling MV. Moving towards individualized medicine with pharma-

cogenomics. Nature 2004; 429(6990):464–468.

42. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of

chemotherapeutics. Nat Med 2006; 12(11):1294–1300.

43. Bao S, Wu Q, McLendon RE, et al. Glioma stem cells promote radioresistance

by preferential activation of the DNA damage response. Nature 2006; 444(7120):

756–760.

44. Calabrese C, Poppleton H, Kocak M, et al. A perivascular niche for brain tumor

stem cells. Cancer Cell 2007; 11(1):69–82.

45. Donnenberg VS, Donnenberg AD. Multiple drug resistance in cancer revisited: the

cancer stem cell hypothesis. J Clin Pharmacol 2005; 45(8):872–877.

46. Kim M, Turnquist H, Jackson J, et al. The multidrug resistance transporter ABCG2

(breast cancer resistance protein 1) effluxes Hoechst 33342 and is overexpressed in

hematopoietic stem cells. Clin Cancer Res 2002; 8(1):22–28.

47. Zhou S, Schuetz JD, Bunting KD, et al. The ABC transporter Bcrp1/ABCG2 is

expressed in a wide variety of stem cells and is a molecular determinant of the side-

population phenotype. Nat Med 2001; 7(9):1028–1034.

48. Zhou S, Morris JJ, Barnes Y, et al. Bcrp1 gene expression is required for normal

numbers of side population stem cells in mice, and confers relative protection to

mitoxantrone in hematopoietic cells in vivo. Proc Natl Acad Sci U S A 2002;

99(19):12339–12344.

49. Krishnamurthy P, Ross DD, Nakanishi T, et al. The stem cell marker Bcrp/ABCG2

enhances hypoxic cell survival through interactions with heme. J Biol Chem 2004;

279(23):24218–24225.

50. Asselin-Labat ML, Shackleton M, Stingl J, et al. Steroid hormone receptor status of

mouse mammary stem cells. J Natl Cancer Inst 2006; 98(14):1011–1014.



51. Sleeman KE, Kendrick H, Robertson D, et al. Dissociation of estrogen receptor

expression and in vivo stem cell activity in the mammary gland. J Cell Biol 2007;

176(1):19–26.

52. Al Hajj M, Wicha MS, Benito-Hernandez A, et al. Prospective identification of


53. Williams DA, Cancelas JA. Leukaemia: niche retreats for stem cells. Nature 2006;

444(7121):827–828.

54. Kastan MB, Schlaffer E, Russo JE, et al. Direct demonstration of elevated aldehyde

dehydrogenase in human hematopoietic progenitor cells. Blood 1990; 75(10):

1947–1950.

55. Russo JE, Hilton J, Colvin OM. The role of aldehyde dehydrogenase isozymes in

cellular resistance to the alkylating agent cyclophosphamide. Prog Clin Biol Res

1989; 290:65–79.

56. Sladek NE, Kollander R, Sreerama L, et al. Cellular levels of aldehyde dehydrogenases

(ALDH1A1andALDH3A1)as predictors of therapeutic responses to cyclophosphamide-

based chemotherapy of breast cancer: a retrospective study. Rational individualization

of oxazaphosphorine-based cancer chemotherapeutic regimens. Cancer Chemother

Pharmacol 2002; 49(4):309–321.

57. Gordon MY, Goldman JM, Gordon-Smith EC. 4-Hydroperoxycyclophosphamide

inhibits proliferation by human granulocyte-macrophage colony-forming cells (GM-

CFC) but spares more primitive progenitor cells. Leuk Res 1985; 9(8):1017–1021.

58. Sladek NE. Human aldehyde dehydrogenases: potential pathological, pharmaco-

logical, and toxicological impact. J Biochem Mol Toxicol 2003; 17(1):7–23.

59. Storms RW, Trujillo AP, Springer JB, et al. Isolation of primitive human hema-

topoietic progenitors on the basis of aldehyde dehydrogenase activity. Proc Natl

Acad Sci U S A 1999; 96(16):9118–9123.

60. Ginestier C, Hur M, Charafe-Jauffret E, et al. ALDH1 is a marker of normal and

breast cancer stem cells and predicts poor clinical outcome. Targeted Oncol 2007;

2(suppl 1):S9.

61. Charafe-Jauffret E, Ginestier C, Hur MH, et al. Aldefluor-positive populations in

breast cell lines: evidence for cancer stem cells properties and gene expression profile

(GEP). In: Proceedings of the 98th Annual Meeting of the American Association for

Cancer Research, Los Angeles, CA, 14–18 April 2007. Philadelphia, PA: AACR,

2007:310. Abstract 1307.

62. Pearce DJ, Taussig D, Simpson C, et al. Characterization of cells with a high

aldehyde dehydrogenase activity from cord blood and acute myeloid leukemia

samples. Stem Cells 2005; 23(6):752–760.

63. Lee J, Kim HK, Rho JY, et al. The human OCT-4 isoforms differ in their ability to

confer self-renewal. J Biol Chem 2006; 281(44):33554–33565.

64. Atlasi Y, Mowla SJ, Ziaee SA, et al. OCT-4, an embryonic stem cell marker, is

highly expressed in bladder cancer. Int J Cancer 2007; 120(7):1598–1602.

65. Ponti D, Costa A, Zaffaroni N, et al. Isolation and in vitro propagation of tumori-

genic breast cancer cells with stem/progenitor cell properties. Cancer Res 2005;

65(13):5506–5511.

66. Kang MK, Kim RH, Kim SJ, et al. Elevated Bmi-1 expression is associated with

dysplastic cell transformation during oral carcinogenesis and is required for cancer

cell replication and survival. Br J Cancer 2007; 96(1):126–133.

286 Balicki et al.


67. Kim JH, Yoon SY, Jeong SH, et al. Overexpression of Bmi-1 oncoprotein correlates

with axillary lymph node metastases in invasive ductal breast cancer. Breast 2004;

13(5):383–388.

68. Dimri GP, Martinez JL, Jacobs JJ, et al. The Bmi-1 oncogene induces telomerase

activity and immortalizes human mammary epithelial cells. Cancer Res 2002;

62(16):4736–4745.

69. Fodde R, Brabletz T. Wnt/beta-catenin signaling in cancer stemness and malignant

behavior. Curr Opin Cell Biol 2007; 19(2):150–8.

70. Dontu G, Wicha MS. Survival of mammary stem cells in suspension culture:

implications for stem cell biology and neoplasia. J Mammary Gland Biol Neoplasia

2005; 10(1):75–86.

71. Liu S, Dontu G, Wicha MS. Mammary stem cells, self-renewal pathways, and

carcinogenesis. Breast Cancer Res 2005; 7(3):86–95.

72. Warner JK, Wang JC, Takenaka K, et al. Direct evidence for cooperating genetic

events in the leukemic transformation of normal human hematopoietic cells. Leukemia

2005; 19(10):1794–1805.

73. He XC, Zhang J, Tong WG, et al. BMP signaling inhibits intestinal stem cell self-

renewal through suppression of Wnt-beta-catenin signaling. Nat Genet 2004;

36(10):1117–1121.

74. Adolphe C, Narang M, Ellis T, et al. An in vivo comparative study of sonic, desert

and Indian hedgehog reveals that hedgehog pathway activity regulates epidermal

stem cell homeostasis. Development 2004; 131(20):5009–5019.

75. Rossi DJ, Weissman IL. Pten, tumorigenesis, and stem cell self-renewal. Cell 2006;

125(2):229–231.

76. Roy M, Pear WS, Aster JC. The multifaceted role of Notch in cancer. Curr Opin

Genet Dev 2007; 17(1):52–59.

77. Bullock TE, Wen B, Marley SB, et al. Potential of CD34 in the regulation of

symmetrical and asymmetrical divisions by hematopoietic progenitor cells. Stem

Cells 2006; 25(4):844–851.

78. Song X, Call GB, Kirilly D, et al. Notch signaling controls germline stem cell niche

formation in the Drosophila ovary. Development 2007; 134(6):1071–1080.

79. Androutsellis-Theotokis A, Leker RR, Soldner F, et al. Notch signalling regulates

stem cell numbers in vitro and in vivo. Nature 2006; 442(7104):823–826.

80. Kalderon D. The mechanism of hedgehog signal transduction. Biochem Soc Trans

2005; 33(pt 6):1509–1512.

81. Shimokawa T, Svard J, Heby-Henricson K, et al. Distinct roles of first exon variants

of the tumor-suppressor Patched1 in Hedgehog signaling. Oncogene 2007; 26(34):

4889–4896; [Epub 2007 February 19].

82. Sanchez P, Hernandez AM, Stecca B, et al. Inhibition of prostate cancer proliferation

by interference with Sonic Hedgehog-GLI1 signaling. Proc Natl Acad Sci U S A

2004; 101(34):12561–12566.

83. Ohuchida K, Mizumoto K, Fujita H, et al. Sonic hedgehog is an early devel-

opmental marker of intraductal papillary mucinous neoplasms: clinical implications

of mRNA levels in pancreatic juice. J Pathol 2006; 210(1):42–48.

84. Janzen V, Scadden DT. Stem cells: good, bad and reformable. Nature 2006;

441(7092):418–419.

85. He XC, Yin T, Grindley JC, et al. PTEN-deficient intestinal stem cells initiate

intestinal polyposis. Nat Genet 2007; 39(2):189–198.



86. Shen WH, Balajee AS, Wang J, et al. Essential role for nuclear PTEN in main-

taining chromosomal integrity. Cell 2007; 128(1):157–170.

87. Korkaya H, Iovino F, Wicha MS. PTEN and HER2 regulate self-renewal and

invasion of human mammary stem cells. In: Proceedings of the 98th Annual

Meeting of the American Association for Cancer Research, Los Angeles, CA, 14–18

April 2007. Philadelphia, PA: AACR, 2007:305. Abstract nr 1287.

88. Yilmaz OH, Valdez R, Theisen BK, et al. Pten dependence distinguishes haemato-

poietic stem cells from leukaemia-initiating cells. Nature 2006; 441(7092):475–482.

89. Fidler IJ. The pathogenesis of cancer metastasis: the ‘seed and soil’ hypothesis

revisited. Nat Rev Cancer 2003; 3(6):453–458.

90. Fidler IJ, Yano S, Zhang RD, et al. The seed and soil hypothesis: vascularisation

and brain metastases. Lancet Oncol 2002; 3(1):53–57.

91. Fidler IJ. Seed and soil revisited: contribution of the organ microenvironment to

cancer metastasis. Surg Oncol Clin N Am 2001; 10(2):257–269, vii–viiii.

92. Li F, Tiede B, Massague J, et al. Beyond tumorigenesis: cancer stem cells in

metastasis. Cell Res 2007; 17(1):3–14.

93. Savarese TM, Low HP, Baik I, et al. Normal breast stem cells, malignant breast

stem cells, and the perinatal origin of breast cancer. Stem Cell Rev 2006; 2(2):

103–110.

94. Esteller M. The necessity of a human epigenome project. Carcinogenesis 2006;

27(6):1121–1125.

95. Widschwendter M, Fiegl H, Egle D, et al. Epigenetic stem cell signature in cancer.

Nat Genet 2007; 39(2):157–158.

96. Wada T, Haigh JJ, Ema M, et al. Vascular endothelial growth factor directly inhibits

primitive neural stem cell survival but promotes definitive neural stem cell survival.

J Neurosci 2006; 26(25):6803–6812.

97. Sugiyama T, Kohara H, Noda M, et al. Maintenance of the hematopoietic stem cell

pool by CXCL12-CXCR4 chemokine signaling in bone marrow stromal cell niches.

Immunity 2006; 25(6):977–988.

98. Kiel MJ, Morrison SJ. Maintaining hematopoietic stem cells in the vascular niche.

Immunity 2006; 25(6):862–864.

99. Calvi LM, Adams GB, Weibrecht KW, et al. Osteoblastic cells regulate the

haematopoietic stem cell niche. Nature 2003; 425(6960):841–846.

100. Zhang J, Niu C, Ye L, et al. Identification of the haematopoietic stem cell niche and

control of the niche size. Nature 2003; 425(6960):836–841.

101. Ramasamy R, Lam EW, Soeiro I, et al. Mesenchymal stem cells inhibit proliferation

and apoptosis of tumor cells: impact on in vivo tumor growth. Leukemia 2007;

21(2):304–310.

102. Kuperwasser C, Chavarria T, Wu M, et al. Reconstruction of functionally normal

and malignant human breast tissues in mice. Proc Natl Acad Sci U S A 2004;

101(14):4966–4971.

103. Bhowmick NA, Chytil A, Plieth D, et al. TGF-beta signaling in fibroblasts

modulates the oncogenic potential of adjacent epithelia. Science 2004; 303(5659):

848–851.

104. Orimo A, Gupta PB, Sgroi DC, et al. Stromal fibroblasts present in invasive human

breast carcinomas promote tumor growth and angiogenesis through elevated SDF-1/

CXCL12 secretion. Cell 2005; 121(3):335–348.

288 Balicki et al.


105. Boyd NF, Guo H, Martin LJ, et al. Mammographic density and the risk and

detection of breast cancer. N Engl J Med 2007; 356(3):227–236.

106. Hortobagyi GN. Trastuzumab in the treatment of breast cancer. N Engl J Med 2005;

353(16):1734–1736.

107. Chabner BA. In defense of cell-line screening. J Natl Cancer Inst 1990; 82(13):

1083–1085.

108. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000; 100(1):57–70.

109. Li C, Heidt DG, Dalerba P, et al. Identification of pancreatic cancer stem cells.

Cancer Res 2007; 67(3):1030–1037.

110. Dirks PB. Brain tumour stem cells: the undercurrents of human brain cancer and

their relationship to neural stem cells. Philos Trans R Soc Lond B Biol Sci 2008;

363(1489):139–52.

111. Piccirillo SG, Reynolds BA, Zanetti N, et al. Bone morphogenetic proteins inhibit

the tumorigenic potential of human brain tumour-initiating cells. Nature 2006;

444(7120):761–765.

112. Prince ME, Sivanandan R, Kaczorowski A, et al. Identification of a subpopulation

of cells with cancer stem cell properties in head and neck squamous cell carcinoma.


113. Jain RK, Forbes NS. Can engineered bacteria help control cancer? Proc Natl Acad

Sci U S A 2001; 98(26):14748–14750.

114. Heldin CH, Rubin K, Pietras K, et al. High interstitial fluid pressure—an obstacle in

cancer therapy. Nat Rev Cancer 2004; 4(10):806–813.

115. Reardon DA, Wen PY. Therapeutic advances in the treatment of glioblastoma:

rationale and potential role of targeted agents. Oncologist 2006; 11(2):152–164.



19

Molecular Imaging in Individualized

Cancer Management

David M. Schuster

Division of Nuclear Medicine and Molecular Imaging, Departmentof Radiology, Emory University, Atlanta, Georgia, U.S.A.

Diego R. Martin

Department of Radiology, Emory University, Atlanta, Georgia, U.S.A.

INTRODUCTION

X-rays were discovered by Karl Roentgen in 1895, and with that discovery came

the ability to see structures within the living human body without having to directly

sample the tissue. Imaging, particularly cross-sectional techniques, has revolu-

tionized the practice of medicine. These techniques employ ionizing radiation, such

as computed tomography (CT) scanning, single-photon emission computed

tomography (SPECT), and positron emission tomography (PET), as well as non-

ionizing energy, such as ultrasound and magnetic resonance imaging (MRI). Some

of the most exciting and practical major advances are unfolding within theMRI and

PET technologies, specifically with regard to molecular imaging.

Molecular imaging is defined as the in vivo characterization and mea-

surement of biological processes at cellular and molecular levels. By definition,

molecular imaging is not restricted to any one tool; it may involve optical,

ultrasound, gamma ray, positron, or magnetic imaging (1). Since the 1960s, bone

scanning has played a major role in the management of breast cancer. However,

in the past decade, the role of radionuclide molecular imaging has expanded

291


significantly in the clinical management of breast cancer because of PET,

mammoscintigraphy, and sentinel lymph node techniques. Molecular imaging is

also instrumental in drug development, gene therapy, and basic science research

of breast cancer. Mammography and ultrasound have been and continue to be

very important tools in the diagnosis and management of breast cancer, but other

modalities, such as MRI, have enhanced our ability to detect occult disease,

especially in those with high genetic risk, and to follow breast carcinoma during

and after therapy (2,3).

In this chapter, we will examine MRI and PET with regard to molecular

imaging as applied to individualized breast cancer management; this chapter is

not intended to be a comprehensive review of molecular imaging methods. We

have specifically chosen MRI and PET because we believe the integration of

both technologies will lead to further progress in our ability to interrogate dis-

ease processes in a manner analogous to the classical pathologist, but in a

noninvasive fashion. This ability will continue to expand as technology pro-

gresses. But in essence, the future is already here.

RADIONUCLIDE TECHNIQUES:CURRENT CONCEPTS AND APPLICATIONS

The advantage of radionuclide techniques over other imaging approaches is their

ability to label any chemical species with an isotope of choice and to detect that

radiotracer with high sensitivity. When certain radionuclides decay, gamma rays

(photons) are given off, which are detected with special crystal detectors. The

gamma emissions of SPECT or PET radiotracers are energetic enough to travel

through tissues and thus enable imaging of the whole body. Carbon, nitrogen,

hydrogen, and oxygen have positron-emitting isotopes and are ideally suited for

designing probes to image natural biochemical and molecular process in vivo.

Probes can either home in on targets on cell surfaces, such as antigens and

receptors, or intracellular structures such as internalized receptors.

Single-photon imaging is the basis of simple techniques such as planar

bone scanning or more sophisticated cross-sectional approaches such as SPECT.

High-energy dual-photon imaging is utilized for PET, permitting better resolu-

tion and sensitivity than single-photon techniques. Yet, each method has its

place. For example, sentinel lymph node procedures, employing single-photon

radiotracers, are rapidly becoming the standard of care in breast cancer lymph

node staging (4). For all practical purposes modern PET scanning is now

accomplished as part of a PET-CT device, which fuses the functional imaging of

PET with the anatomic detail of CT.

In 1984, Beaney and coworkers reported (5) on the use of [15O]H2O PET

to study blood flow, the oxygen extraction ratio, and oxygen utilization in breast

cancer. The earliest studies on the use of 18F-fluorodeoxyglucose (FDG) in the

diagnosis of breast cancer were conducted in 1989 (6,7). Figure 1 is an example

of 18F-FDG PET uptake in a primary breast mass and axillary lymph nodes.

292 Schuster and Martin


Fluorodeoxyglucose is actively transported into the cell, where it is irre-

versibly phosphorylated by hexokinase, trapping it within the cell, where it does

not proceed further down the glycolytic pathway. There is a correlation between

FDG uptake and glucose transporter-1 (Glut-1) expression, the mitotic activity

index, percent necrosis (likely reflecting the proliferation rate), tumor cells/

volume, hexokinase-1 expression, the number of lymphocytes (but not macro-

phages), microvessel density, and possibly the expression of the p53 gene (8–12).

PET is currently of limited value in routine screening for breast cancer (13)

and cannot take the place of fine sectioning and immunohistochemical lymph

node evaluation, but it has an important role to play in a select group of prob-

lematic diagnostic situations and in staging patients with high-risk disease (14–17).

PET can provide important information by detecting locoregional lymph node

involvement as well as distant metastases. Because of its high positive predictive

value, FDG PET may obviate a sentinel lymph node procedure and axillary

dissection.

The use of neoadjuvant chemotherapy has increased the rate of breast

conservation surgery (18). Clinical response does not necessarily correlate with

pathologic response. FDG PET has been shown to aid in monitoring response to

chemotherapy and has prognostic benefit (19). Nonresponders and those

developing progressive disease or distant metastases can be identified earlier,

and this information may prove useful in changing therapies and avoiding side

effects of chemotherapy that is not effective. Baseline PET combined with PET

after the first course of chemotherapy is highly accurate in this regard (16). A

meta-analysis of the FDG PET literature (20) reports an overall 81% sensitivity,

96% specificity, and 92% accuracy for monitoring response to therapy. Elevated

FDG uptake was found to be an independent predictor of worse prognosis and

decreased relapse-free survival in breast cancer (10). Figure 2 illustrates a dramatic

response of metastatic breast cancer to fulvestrant (Faslodex) therapy.

FDG PET has been shown to locate unsuspected metastases with high

sensitivity and specificity (21,22). Detection of early recurrence may have

Figure 1 (A) Axial CT, (B) attenuation correctedPET, and (C) fused image in a50-year-old

female with a 4-cm primary right breast cancer (arrow) and malignant axillary lymph

nodes (arrowhead ). Abbreviations: CT, computed tomography; PET, positron emission

tomography.

Molecular Imaging in Individualized Cancer Management 293


important survival benefits, prompting the use of new therapies as well as curative

or palliative surgery. FDG PET is considered of great efficacy in the evaluation of

patients with suspected recurrent breast cancer, especially to differentiate true

recurrence from postsurgical and radiation sequelae. The ability of PET to detect

suspected clinical recurrence and in patients with elevated tumor markers has been

demonstrated in a number of studies (23–32).

MRI: CURRENT CONCEPTS AND APPLICATIONS

In 1973, Paul Lauterbur discovered that introduction of magnetic gradients

superimposed over a magnetic field makes it possible to create two-dimensional

images of structures, with the first experiments showing the ability to differen-

tiate between tubes filled with regular or heavy water (33). No other imaging

method can achieve this differentiation. This work was the foundation for a

revolutionary tissue-imaging technique that generates images on the basis of

molecular differences between or within tissues.

To understand the added value of MRI to clinical sciences, it is useful to

compare it with CT. CT can produce images of high spatial resolution, which are

interpreted on the basis of anatomy. A distinct feature of MRI is that anatomi-

cally correct images are produced, but these images are based on molecular

events that produce much higher soft tissue contrast between tissues and between

Figure 2 (A) Anterior view from a maximal intensity projection of an FDG-PET scan

showing extensive breast cancer pleural implants in the left chest and (B) after one dose of

fulvestrant (Faslodex) after which the implants resolved. Note normal cardiac (arrow) and

renal (arrowhead ) uptake is better seen in B than in A. The patient was status post

unrelated remote right nephrectomy. Abbreviations: FDG, 18F-fluorodeoxyglucose;

PET, positron emission tomography.



abnormal and normal tissues. It has been shown for most clinically important

applications, including the brain and the solid organs of the body, that soft tissue

contrast resolution is more important than spatial resolution. Even on contrast-

enhanced CT, the ability to generate soft tissue contrast resolution is less than

that of MRI, and this difference may contribute to the relative advantage that

MRI provides for lesion detection (Fig. 3).

Unlike CT, which relies on the degree of X-ray absorption to differentiate

tissues, MRI contrast differences are based on molecular interactions between

protons located in water molecules within and between cells, neighboring

molecules that influence mobility, such as charged glycoproteins and mem-

branes, and species that can alter the magnetic field at the molecular level, such

as iron and lipid. There are essentially an infinite variety of formulations of MRI

techniques that can exploit these interactions to generate different contrast pat-

terns. Similar to the concept of a pathologist using different histochemical stains

to visualize molecular species, the expert MR imager will utilize an optimized

set of MRI sequences to visualize the characteristics that may be unique to

specific pathologies. Though the initial applications of MRI had been for

imaging diseases of the brain, MRI is now able to image faster and with con-

comitant improvement in image quality. This has been critical for improving the

capacity of MRI for rapidly imaging the chest, abdomen, and pelvis, where

motion from respiration or cardiac contractions may deteriorate images (34,35).

Breast MRI (Fig. 4) is a recent additional imaging method for the evalu-

ation of primary breast carcinoma and is useful for early detection and staging

(36–40). MRI has been found to provide maximal sensitivity, as compared to

traditional mammography and ultrasound, but with comparatively lower spe-

cificity. The fundamental components to the current state of the art of breast MRI

is to use the combination of time-resolved gadolinium uptake analysis in com-

bination with the evaluation of the contours of a mass using high-contrast and

Figure 3 Liver cross-sectional imaging showing multiple hypervascular breast metas-

tases. (A) On contrast-enhanced CT, the liver lesions are not visualized. (B) On contrast-

enhanced MRI, the lesions are conspicuous with innumerable tumors scattered throughout

the liver (6 are shown: B, black arrows). Abbreviations: CT, computed tomography; MRI,

magnetic resonance imaging.



high-resolution imaging. MRI techniques have benefited from improvements in

surface coils specifically dedicated to imaging both breasts simultaneously.

The most important sequence used for routine breast MRI is the three-

dimensional gradient echo (3D GRE) technique. Uniform and complete sup-

pression of the breast lipid signal has been a key area of focus for improvement

of these sequences. With time-resolved imaging after an MRI contrast injection

(Fig. 4), malignant focal breast carcinomas typically show enhancement that

Figure 4 (A) Right breast carcinoma shown on sagittal plane dynamically enhanced

gadolinium MRI at the indicated time intervals. Subsequently, (B) a region of interest is

applied within the tumor, and a time-intensity curve is generated. (C) A lymph node

showing a similar enhancement pattern (arrow) is shown on a more lateral slice through

the right axilla on a representative 3-min sagittal image. (D) The enhancing primary breast

carcinoma and the metastatic lymph node in the right axilla can be seen on an axial MRI

image of both breasts (arrows), which correlates with findings on (E) PET-CT (arrows),

confirming the presence of metastasis in the node. Note that the position of the breast is

different on the MRI compared with the PET-CT due to prone MRI versus supine PET

positioning. Abbreviations: MRI, magnetic resonance imaging; PET, positron emission

tomography; CT, computed tomography.



remains constant over several minutes or may diminish. Benign tumors such as

fibromas may show differential contrast uptake, with progressive increase in

contrast accumulation within the benign mass. However, the specificity for these

differentiating features is relatively moderate. A more recent approach has

incorporated lessons learned previously from conventional X-ray mammog-

raphy; the margins of the tumor may show patterns specific to malignancy. Data

are developing to support that the analysis of tumor margin morphology will

improve MRI specificity for the diagnosis of breast carcinoma and separate this

condition from benign tumors. Here, further development of newer MRI strat-

egies, such as perfusion studies or spectroscopy, or a possible combination with

metabolic imaging provided by PET, may help improve this apparent current

limitation.

Lymphatic spread may result in local lymph node involvement. Findings

on MRI may be relatively specific when the lymph nodes are found to be over

1 cm in diameter and show contrast enhancement that is similar in time course

and intensity to those seen in the primary breast mass. However, for smaller

lymph nodes the specificity may be significantly lower. Combined imaging with

FDG-PET and MRI may help aid the specificity of the MRI study and the

identification of the most concerning lymph nodes (Fig. 4).

Currently accepted applications for MRI include screening for high-risk

women, evaluation of dense fibroglandular breast tissue, and surveillance after

surgical alteration or breast implant placement (36–40). Generalized use of MRI

for evaluation of all women in the screening age group has not been widely

accepted.

With regard to hematogenous tumor spread, the liver is the organ most

commonly involved in metastatic diseases, usually from breast, colon, pancreas,

kidney, and carcinoid cancers, or melanoma. However, most liver tumors are

benign, and benign liver lesions are common, including cysts, bile duct hamar-

tomas, hemangiomas, adenomas, and focal nodular hyperplasia. The specificity of

a noninvasive test is the litmus test for utility in tumor evaluation. We have shown

that although MRI is slightly more sensitive than CT for tumor detection, the most

important distinction is that MRI is highly specific, approaching 98%, while CT is

significantly less specific, approaching 80% (41–44). The ability to image rapidly

and include the evaluation of the entire body is undergoing development. The

potential for assessing metastatic burden will be a primary potential role for this

capability.

FUTURE DEVELOPMENTS

The resolution of PET scanners is currently limited in comparison with ana-

tomic imaging, but continued technological improvements are expected

(45,46). Sensitivity for initial diagnosis will likely improve as device resolution

advances and as special focused field PET-mammography imaging systems are

developed (47–49). Small field-of-view cameras not only offer better sentinel



lymph node and primary tumor detection because of increased spatial resolution

but can also be mounted on maneuverable arms and used in an intraoperative

setting, helping the surgeon realize a tumor-free margin (50).

MR spectroscopic analysis has the potential to elucidate pathophysio-

logical processes and may become useful for the assessment of primary or

metastatic neoplasms, or the assessment of tumor response to therapy. MRI

strategies are providing new diagnostic methods of evaluating dynamic patho-

physiological processes using time-resolved imaging. These methods include the

use of an injected contrast agent that can be visualized as it perfuses and then

redistributes within the tissues, much as a radiotracer is sometimes used for

nuclear imaging techniques (51,52). The advantages of MRI for this application

include the capacity for high temporal and spatial resolution and the ability to

combine this component of an examination with other elements of MRI to

provide comprehensive evaluation during a single study. This approach is being

used to detect and characterize diseases, including neoplasms, inflammation and

infection, and fibrotic processes (53–57).

Apoptosis, or programmed cell death, is an elementary process of life that

maintains homeostasis. Apoptosis is physiologic and is different from necrosis,

which is pathologic, often eliciting an inflammatory reaction. Successful che-

motherapy increases apoptosis and enhances the uptake of annexin in the tumor.

As the tumor becomes necrotic, the uptake of annexin reduces. PET and SPECT

imaging with radiolabeled annexin have been used to assess apoptosis (58–61).

Molecular process that occur within breast cancer cells can be imaged by

using a radiotracer based on amino acid metabolism, such as methionine, which

can serve as an indirect measure of protein synthesis, or an amine, such as

choline, which can reflect membrane biosynthesis. 11C-methionine and11C-choline have been used in assessing breast cancer and metastases (62,63).

DNA synthesis within breast cancer has been examined with thymidine ana-

logues such as 30-deoxy-30-[18F]fluorothymidine (FLT), which has been

employed to image intratumoral proliferation (64,65). [15O]H2O PET can also be

useful for tumor perfusion (65). These radiotracers can serve to characterize

tumor as well as measure therapeutic response.

Estrogen and progestin receptors offer another avenue to image breast

cancer. Estrogen receptors (ERs) have been successfully targeted using 16a-[18F]fluoro-17b-estradiol (FES), and early clinical studies have shown them to be as

sensitive as FDG imaging and probably more specific (66). In one study of

patients with ER-positive breast cancer, the patients also underwent FES PET.

PET was able to predict global response to hormonal therapy in human epi-

dermal growth factor receptor 2 (HER2)-negative patients better than pathology

(24% vs. 47%) (67,68). Radiolabeled chemotherapeutic agents such as [18F]

fluorotamoxifen may be useful to detect ER-positive metastases and can also

help in following response to treatment (69). 18F-labeled paclitaxel has also been

shown in a mouse model with a breast carcinoma xenograft to predict tumor

response to paclitaxel chemotherapy (70).



Peptides that localize to receptors are small-sized molecules and clear

rapidly via the kidney, making these peptides attractive imaging probes. Pep-

tides, such as bombesin (BN), labeled with single-photon and dual-photon

radionuclides are being studied both for imaging and for therapeutic purposes

(71). Gastrin-releasing peptide (GRP) receptors, a subtype of BN receptors, are

overexpressed in breast carcinoma. Octreoscan, a single-photon radiotracer

somatostatin analogue, has also been used to study breast cancer (72).

Several gene therapy approaches have been considered for the manage-

ment of breast cancer, such as ablation of oncogenic products, restoration of ER

expression, alteration of genes that are involved in apoptosis, and activation of

tumor suppressor genes. Reporter genes have been used to study promoter or

regulatory elements involved in gene expression conventionally studied by tissue

biopsy and immunohistochemistry. Now noninvasive in vivo methods are

available in the form of radionuclide techniques or optical imaging. The ability

to image deeper structures in larger animal models and in humans is the main

advantage of radionuclide technique over optical imaging. For example, 8-[18F]

Fluoropenciclovir has been studied as a reporter probe for herpes simplex virus

type 1 thymidine kinase (HSV1-tk) gene expression (73). Other possibilities still

in very early development include specially designed, small radiolabelled anti-

sense oligodeoxynucleotides that may help image gene expression (20). PET

imaging with an appropriate PET reporter gene and probe combination may

become a powerful tool for molecular imaging of human gene expression (74).

HER2/neu overexpression in breast cancer is associated with poor response

to chemotherapy and an unfavorable prognosis. Phase I clinical trials of E1A

gene therapy targeting HER2/neu-overexpressed breast cancer has shown

increased apoptosis (75). A novel PET radiotracer, (68)Ga-DOTA-F(ab0)(2)-herceptin, has been developed, which has the potential to follow tumor

response to the heat shock protein 90 (Hsp90) inhibitor 17-allylamino-17-

demethoxygeldanamycin (17AAG) (76). In an animal model, downregulation of

HER2 occurs independently of glycolytic response as measured by 18F-FDG-

PET. This method could be used to rapidly screen tumor response to a variety of

HER2 inhibitors and choose the correct agent within hours rather than wait for

weeks or months to determine if there is a response. Multidrug resistance as

manifested by MDR1 P-glycoprotein transport activity also has the potential to

be studied with SPECT and PET radiotracers (77).

Angiogenesis, the outgrowth of new capillaries, is an important component

in tumor formation and spread. Imaging of angiogenesis for diagnosis and fol-

lowing therapy can be performed with PET and SPECT radiotracers targeting

integrin, VEGF, and other biomarkers (78). For example, (64)Cu-DOTA-VEGF

(121) PET has been successful in small animals to specifically image in vivo

expression of VEGF (79).

Tumor hypoxia may result in resistance to therapy. [18F]Fluoromisonida-

zole-3-fluoro-1-(20-nitro-10-imidazolyl)-2-propanol ([18F]MISO) and Cu-diacetyl-

bis(N4-methylthiosemicarbazone) (Cu-ATSM) radiotracers for PET and blood



oxygen level–dependent (BOLD) MRI are some of the techniques being utilized

for in vivo imaging but require further study (80,81).

Finally, nanoparticles technology may be utilized to achieve diagnostic

imaging with both PET and MRI and enable combined modality radionuclide

therapy with intracellular drug delivery (82). The feasibility of magnetic

frequency–activated thermal necrosis with antibody-linked iron oxide nano-

particles conjugated with the single-photon radiotracer Indium-111 has been

recently demonstrated in breast cancer xenografts (83).

An understanding of pathology, together with the capabilities of MRI,

facilitates the use of a particular MRI feature for the evaluation of a range of

diseases that affect different organ systems. The application of MRI time-

resolved perfusion imaging for tumor ranges from primary diagnosis to the

determination of the degree of tumor neoangiogenesis. The ability to quantify

tumor angiogenesis may provide a useful surrogate marker of tumor respon-

siveness prior to initiating therapy or indicate tumor response even prior to any

change in the overall volume of tumor tissue (Fig. 4). MRI may be useful for the

specific evaluation of chemotherapeutic antiangiogenic agents such as anti-

VEGF (84,85). The possibility of using MRI molecular imaging capabilities for

directing therapy is being explored, including the detection of reporter gene

transcription, the use of MRI tissue thermometry to guide therapeutic ultrasound

for noninvasive tumor ablation, and the use of functional brain MRI to plan the

surgical resection of tumors, including metastases to the brain (51,52,86–91).

Most importantly, in a manner analogous to the fusion of PET and CT

scanning, the fusion of PET and MRI scanning holds even greater promise in

providing a one-stop shop for molecular function and exquisite anatomic detail

but with a markedly reduced radiation dose. There have been a number of studies

examining the utility of MRI and FDG PET. For example, Walter et al. studied

42 lesions in 40 patients preoperatively (92). The sensitivity of FDG PET was

63%, and the specificity was 91%; the sensitivity of MRI was 89%, and the

specificity was 74%. Thus, both modalities can play complementary roles, with

overall improved accuracy when used in conjunction. Other papers examining

the complementary strengths and weaknesses of both modalities have emerged

(93–95). A recent intriguing study in a small group of patients examines the

utility of MRI with ultrasmall superparamagnetic iron oxide particles in con-

junction with PET, achieving 100% sensitivity and specificity on a per-patient

basis for local lymph node staging when both modalities are employed (96).

Further technical development is anticipated to facilitate combining MRI

and PET (97). We have already started accumulating evidence of the advantages

in this pairing of technologies, as compared with using the combination of

CT and PET. Figure 5 is an example from our institution of how PET and MRI

data can be combined utilizing special software registration. While the above

PET-MRI studies have been performed on separate devices, necessitating

moving and scheduling the patients between two physically separate units, a

combined PET-MRI scanner in a single gantry is now a reality (45,46,97,98).



OUR VISION

Molecular imaging will continue to have an evolving role in diagnosis, staging,

personalization of therapy, monitoring response to therapy, and detecting

recurrence of breast carcinoma, both local and widespread. A patient will arrive

in the imaging suite and have the entire disease process interrogated on a spe-

cially modified PET-MRI system, which will utilize dynamic MRI techniques

and MR spectroscopy as well as novel PET radiotracers. Complex processes

such as glycolysis, tumor perfusion, cell proliferation, protein synthesis, genetic

expression, angiogenesis, and various other profiling will be examined. There is

potential for additional information and improvements over current invasive

tissue sampling and histopathology, which, for example, may not be truly rep-

resentative of the heterogeneity present within a primary tumor or metastases.

This process will not simply serve as a baseline to monitor therapy but

will help individualize therapy since all the parameters can be entered into a

matrix and therapy chosen on the basis of data derived from a cumulative

outcomes database, including surgical, chemotherapeutic, radiotherapy, and

genetic therapy options. Drug-target matching and optimal dosing can be

assessed utilizing the powerful molecular techniques described above. These

techniques will also facilitate targeted biopsies, which will improve genomic

and proteomic profiling and permit a more tailored radiotherapy. Designer

therapeutic and imaging agents may also be synthesized on the basis of a

patient’s particular profile.

Figure 5 Breast cancer patient (from Fig. 4) with axial MRI, PET, and CT images through

the pelvis. (A) Shows the MRI and PET images before and after deformable registration of

the PET to the MRI. (B) The original CT (upper left) and PET-CT (lower left), compared

with the MRI (upper right) and deformable registered PET-MRI (lower right). Increased

FDG activity in the perirectal space on PET-CT (arrow) is of uncertain origin; however, the

activity from the PET scan clearly corresponds to a benign fibroid on the MRI, which is not

visible on CT. Benign fibroids may occasionally take up FDG in premenopausal women.

The fusion of PET and MRI increased diagnostic accuracy in this patient. Abbreviations:

MRI, magnetic resonance imaging; PET, positron emission tomography; CT, computed

tomography; FDG, 18F-fluorodeoxyglucose.



These same methods may then be used to quickly determine treatment

response as measured by glycolysis, amino acid transport, tumor perfusion,

proliferation, and apoptosis, among other metabolic indices, and also monitor for

resistant or mutating clones and for recurrence. Chemotherapeutic and genetic

therapies may then be altered or doses optimized on the basis of rapid molecular

imaging feedback rather than maximizing therapy based only on toxicity effects.

The possibility of therapy using nanoparticle delivery of killing doses of

radiotherapy combined with chemotherapy may also be possible. The same

nanoparticles could also carry MR and PET imaging tracers.

The concept of the “one magic bullet” for the generic enemy of cancer has

fallen by the wayside. A host of ordinary tools optimized for a particular patient

and his or her own cancer can be combined in a unique fashion with the help of

the “magic window” afforded by molecular imaging. Though defeating the

cancer is a lofty goal, these tools may also enable patients to live with cancer and

manage the entire disease process holistically in cooperation with their medical

caregivers. The true cost-benefit ratios of therapies can then be ascertained. We

believe the technology already exists, but further research, social, and financial

investment, as well as societal vision is required for this dream to flower into

everyday reality.

REFERENCES

1. Basilion JP. Current and future technologies for breast cancer imaging. Breast

Cancer Res 2001; 3(1):14–16.

2. Trecate G, Vergnaghi D, Manoukian S, et al. MRI in the early detection of breast

cancer in women with high genetic risk. Tumori 2006; 92(6):517–523.

3. Bartella L, Smith CS, Dershaw DD, et al. Imaging breast cancer. Radiol Clin North

Am 2007; 45(1):45–67.

4. Krynyckyi BR, Kim CK, Goyenechea MR, et al. Clinical breast lymphoscintigraphy:

optimal techniques for performing studies, image atlas, and analysis of images.

Radiographics 2004; 24(1):121–145 (discussion 139–145).

5. Beaney RP, Lammertsma AA, Jones T, et al. Positron emission tomography for in-

vivo measurement of regional blood flow, oxygen utilisation, and blood volume in

patients with breast carcinoma. Lancet 1984; 1(8369):131–134.

6. Minn H, Soini I. [18F]fluorodeoxyglucose scintigraphy in diagnosis and follow up of

treatment in advanced breast cancer. Eur J Nucl Med 1989; 15(2):61–66.

7. Kubota K, Matsuzawa T, Amemiya A, et al. Imaging of breast cancer with [18F]

fluorodeoxyglucose and positron emission tomography. J Comput Assist Tomogr

1989; 13(6):1097–1098.

8. Bos R, van Der Hoeven JJ, van Der Wall E, et al. Biologic correlates of (18)

fluorodeoxyglucose uptake in human breast cancer measured by positron emission

tomography. J Clin Oncol 2002; 20(2):379–387.

9. Brown RS, Wahl RL. Overexpression of Glut-1 glucose transporter in human breast

cancer. An immunohistochemical study. Cancer 1993; 72(10):2979–2985.



10. Oshida M, Uno K, Suzuki M, et al. Predicting the prognoses of breast carcinoma

patients with positron emission tomography using 2-deoxy-2-fluoro[18F]-D-glucose.

Cancer 1998; 82(11):2227–2234.

11. Crippa F, Seregni E, Agresti R, et al. Association between [18F]fluorodeoxyglucose

uptake and postoperative histopathology, hormone receptor status, thymidine label-

ling index and p53 in primary breast cancer: a preliminary observation. Eur J Nucl

Med 1998; 25(10):1429–1434.

12. Smith TA, Sharma RI, Thompson AM, et al. Tumor 18F-FDG incorporation is

enhanced by attenuation of P53 function in breast cancer cells in vitro. J Nucl Med

2006; 47(9):1525–1530.

13. Samson DJ, Flamm CR, Pisano ED, et al. Should FDG PET be used to decide

whether a patient with an abnormal mammogram or breast finding at physical

examination should undergo biopsy? Acad Radiol 2002; 9(7):773–783.

14. Noh DY, Yun IJ, Kang HS, et al. Detection of cancer in augmented breasts by

positron emission tomography. Eur J Surg 1999; 165(9):847–851.

15. Wahl RL, Helvie MA, Chang AE, et al. Detection of breast cancer in women after

augmentation mammoplasty using fluorine-18-fluorodeoxyglucose-PET. J Nucl Med

1994; 35(5):872–875.

16. Zimny M, Siggelkow W. Positron emission tomography scanning in gynecologic and

breast cancers. Curr Opin Obstet Gynecol 2003; 15(1):69–75.

17. Czernin J. FDG-PET in breast cancer: a different view of its clinical usefulness. Mol

Imaging Biol 2002; 4(1):35–45.

18. Rose C, Dose J, Avril N. Positron emission tomography for the diagnosis of breast

cancer. Nucl Med Commun 2002; 23(7):613–618.

19. Kostakoglu L, Goldsmith SJ. 18F-FDG PET evaluation of the response to therapy for

lymphoma and for breast, lung, and colorectal carcinoma. J Nucl Med 2003; 44(2):

224–239.

20. Gambhir SS, Czernin J, Schwimmer J, et al. A tabulated summary of the FDG PET

literature. J Nucl Med 2001; 42(suppl 5):S1–S93.

21. Avril N, Dose J, Janicke F, et al. Assessment of axillary lymph node involvement in

breast cancer patients with positron emission tomography using radiolabeled 2-

(fluorine-18)-fluoro-2-deoxy-D-glucose. J Natl Cancer Inst 1996; 88(17):1204–1209.

22. Dose J, Bleckmann C, Bachmann S, et al. Comparison of fluorodeoxyglucose positron

emission tomography and “conventional diagnostic procedures” for the detection of

distant metastases in breast cancer patients. Nucl Med Commun 2002; 23(9):857–864.

23. Kim TS, Moon WK, Lee DS, et al. Fluorodeoxyglucose positron emission tomog-

raphy for detection of recurrent or metastatic breast cancer. World J Surg 2001; 25(7):

829–834.

24. Eubank WB, Mankoff DA, Vesselle HJ, et al. Detection of locoregional and distant

recurrences in breast cancer patients by using FDG PET. Radiographics 2002; 22(1):

5–17.

25. Lonneux M, Borbath II, Berliere M, et al. The place of whole-body PET FDG for the

diagnosis of distant recurrence of breast cancer. Clin Positron Imaging 2000; 3(2):

45–49.

26. Siggelkow W, Zimny M, Faridi A, et al. The value of positron emission tomography

in the follow-up for breast cancer. Anticancer Res 2003; 23(2C):1859–1867.



27. Moon DH, Maddahi J, Silverman DH, et al. Accuracy of whole-body fluorine-

18-FDG PET for the detection of recurrent or metastatic breast carcinoma. J Nucl

Med 1998; 39(3):431–435.

28. Grahek D, Montravers F, Kerrou K, et al. [18F]FDG in recurrent breast cancer:

diagnostic performances, clinical impact and relevance of induced changes in

management. Eur J Nucl Med Mol Imaging 2004; 31(2):179–188.

29. Suarez M, Perez-Castejon MJ, Jimenez A, et al. Early diagnosis of recurrent breast

cancer with FDG-PET in patients with progressive elevation of serum tumor

markers. Q J Nucl Med 2002; 46(2):113–121.

30. Liu CS, Shen YY, Lin CC, et al. Clinical impact of [(18)F]FDG-PET in patients with

suspected recurrent breast cancer based on asymptomatically elevated tumor marker

serum levels: a preliminary report. Jpn J Clin Oncol 2002; 32(7):244–247.

31. Pecking AP, Mechelany-Corone C, Bertrand-Kermorgant F, et al. Detection of occult

disease in breast cancer using fluorodeoxyglucose camera-based positron emission

tomography. Clin Breast Cancer 2001; 2(3):229–234.

32. Maldonado A, Sancho F, Calvo FA, et al. The role of FDG-PET in diagnosing

suspected tumor recurrence based on increased levels of plasma markers. Clin

Positron Imaging 1998; 1(4):246.

33. Pearson H. Magnetic pioneers net Nobel for putting medicine in the picture. Nature

2003; 425(6958):547.

34. Martin DR, Danrad R, Hussain SM. MR imaging of the liver. Radiol Clin North Am

2005; 43(5):861–886, viii.

35. Martin DR, Friel HT, Danrad R, et al. Approach to abdominal imaging at 1.5 Tesla and

optimization at 3 Tesla. Magn Reson Imaging Clin N Am 2005; 13(2):241–254, v–vi.

36. Deurloo EE, Peterse JL, Rutgers EJ, et al. Additional breast lesions in patients eligible

for breast-conserving therapy by MRI: impact on preoperative management and

potential benefit of computerised analysis. Eur J Cancer 2005; 41(10):1393–1401.

37. Lalonde L, David J, Trop I. Magnetic resonance imaging of the breast: current

indications. Can Assoc Radiol J 2005; 56(5):301–308.

38. Lehman CD, Blume JD, Weatherall P, et al. Screening women at high risk for breast

cancer with mammography and magnetic resonance imaging. Cancer 2005; 103(9):

1898–1905.

39. Liberman L. Breast MR imaging in assessing extent of disease. Magn Reson Imaging

Clin N Am 2006; 14(3):339–349, vi.

40. Warner E, Causer PA. MRI surveillance for hereditary breast-cancer risk. Lancet

2005; 365(9473):1747–1749.

41. Eisenberg LB, Warshauer DM, Woosley JT, et al. CT and MRI of hepatic focal

nodular hyperplasia with peripheral steatosis. J Comput Assist Tomogr 1995; 19(3):

498–500.

42. Martin DR, Semelka RC. Imaging of benign and malignant focal liver lesions. Magn

Reson Imaging Clin N Am 2001; 9(4):785–802, vi–vii.

43. Semelka RC, Martin DR, Balci C, et al. Focal liver lesions: comparison of dual-

phase CT and multisequence multiplanar MR imaging including dynamic gadoli-

nium enhancement. J Magn Reson Imaging 2001; 13(3):397–401.

44. Lee VS, Lavelle MT, Krinsky GA, et al. Volumetric MR imaging of the liver and

applications. Magn Reson Imaging ClinN Am 2001; 9(4):697–716, v–vi.

45. Lucignani G. Time-of-flight PET and PET/MRI: recurrent dreams or actual realities?

Eur J Nucl Med Mol Imaging 2006; 33(8):969–971.



46. Cherry SR. The 2006 Henry N. Wagner Lecture: Of mice and men (and positrons)–

advances in PET imaging technology. J Nucl Med 2006; 47(11):1735–1745.

47. Raylman RR, Majewski S, Weisenberger AG, et al. Positron emission mammog-

raphy-guided breast biopsy. J Nucl Med 2001; 42(6):960–966.

48. Murthy K, Aznar M, Thompson CJ, et al. Results of preliminary clinical trials of the

positron emission mammography system PEM-I: a dedicated breast imaging system

producing glucose metabolic images using FDG. J Nucl Med 2000; 41(11):1851–1858.

49. Adler LP, Weinberg IN, Bradbury MS, et al. Method for combined FDG-PET and

radiographic imaging of primary breast cancers. Breast J 2003; 9(3):163–166.

50. Aarsvold JN, Mintzer RA, Yasillo NJ, et al. A miniature gamma camera. Ann N Y

Acad Sci 1994; 720:192–205.

51. Curran SD, Muellner AU, Schwartz LH. Imaging response assessment in oncology.

Cancer Imaging 2006; 6:S126–S130.

52. Kiessling F, Morgenstern B, Zhang C. Contrast agents and applications to assess tumor

angiogenesis in vivo by magnetic resonance imaging. Curr Med Chem 2007; 14(1):

77–91.

53. Ansari M, Araoz PA, Gerard SK, et al. Comparison of late enhancement cardio-

vascular magnetic resonance and thallium SPECT in patients with coronary disease

and left ventricular dysfunction. J Cardiovasc Magn Reson 2004; 6(2):549–556.

54. Goldfarb JW, Shinnar M. Free-breathing delayed hyperenhanced imaging of the

myocardium: a clinical application of real-time navigator echo imaging. J Magn

Reson Imaging 2006; 24(1):66–71.

55. Saeed M, Weber O, Lee R, et al. Discrimination of myocardial acute and chronic

(scar) infarctions on delayed contrast enhanced magnetic resonance imaging with

intravascular magnetic resonance contrast media. J Am Coll Cardiol 21 2006; 48(10):

1961–1968.

56. Schneider G, Ahlhelm F, Seidel R, et al. Contrast-enhanced cardiovascular magnetic

resonance imaging. Top Magn Reson Imaging 2003; 14(5):386–402.

57. Martin DR, Semelka RC. Magnetic resonance imaging of the liver: review of

techniques and approach to common diseases. Semin Ultrasound CT MR 2005; 26(3):

116–131.

58. Yagle KJ, Eary JF, Tait JF, et al. Evaluation of 18F-annexin V as a PET imaging

agent in an animal model of apoptosis. J Nucl Med 2005; 46(4):658–666.

59. Murakami Y, Takamatsu H, Taki J, et al. 18F-labelled annexin V: a PET tracer for

apoptosis imaging. Eur J Nucl Med Mol Imaging 2003.

60. Dekker B, Keen H, Shaw D, et al. Functional comparison of annexin V analogues

labeled indirectly and directly with iodine-124. Nucl Med Biol 2005; 32(4):403–413.

61. Keen HG, Dekker BA, Disley L, et al. Imaging apoptosis in vivo using 124I-annexin

V and PET. Nucl Med Biol 2005; 32(4):395–402.

62. Inoue T, Kim EE, Wong FC, et al. Comparison of fluorine-18-fluorodeoxyglucose

and carbon-11-methionine PET in detection of malignant tumors. J Nucl Med 1996;

37(9):1472–1476.

63. Leskinen-Kallio S, Nagren K, Lehikoinen P, et al. Carbon-11-methionine and PET is

an effective method to image head and neck cancer. J Nucl Med 1992; 33(5):691–695.

64. Smyczek-Gargya B, Fersis N, Dittmann H, et al. PET with [18F]fluorothymidine for

imaging of primary breast cancer: a pilot study. Eur J Nucl Med Mol Imaging 2004;

31(5):720–724.



65. Buck AK, Schirrmeister H, Mattfeldt T, et al. Biological characterisation of breast

cancer by means of PET. Eur J Nucl Med Mol Imaging 2004; 31(suppl 1):S80–S87.

66. Dehdashti F, Mortimer JE, Siegel BA, et al. Positron tomographic assessment of

estrogen receptors in breast cancer: comparison with FDG-PET and in vitro receptor

assays. J Nucl Med 1995; 36(10):1766–1774.

67. Linden HM, Stekhova S, Link JM, et al. HER2 expression and uptake of 18F-fluo-

roestradiol predict response of breast cancer to hormonal therapy. J Nucl Med 2004;

45:85–86.

68. Shields AF. Positron emission tomography measurement of tumor metabolism and

growth: its expanding role in oncology. Mol Imaging Biol 2006; 8(3):141–150.

69. Inoue T, Kim EE, Wallace S, et al. Preliminary study of cardiac accumulation of F-18

fluorotamoxifen in patients with breast cancer. Clin Imaging 1997; 21(5):332–336.

70. Hsueh WA, Kesner AL, Gangloff A, et al. Predicting chemotherapy response to

paclitaxel with 18F-Fluoropaclitaxel and PET. J Nucl Med 2006; 47(12):1995–1999.

71. Zhang H, Schuhmacher J, Waser B, et al. DOTA-PESIN, a DOTA-conjugated

bombesin derivative designed for the imaging and targeted radionuclide treatment of

bombesin receptor-positive tumours. Eur J Nucl Med Mol Imaging 2007; 34(8):

1198–1208.

72. van Eijck CH, Kwekkeboom DJ, Krenning EP. Somatostatin receptors and breast

cancer. Q J Nucl Med 1998; 42(1):18–25.

73. Iyer M, Barrio JR, Namavari M, et al. 8-[18F]Fluoropenciclovir: an improved

reporter probe for imaging HSV1-tk reporter gene expression in vivo using PET.

J Nucl Med 2001; 42(1):96–105.

74. Penuelas I, Boan J, Marti-Climent JM, et al. Positron emission tomography and gene

therapy: basic concepts and experimental approaches for in vivo gene expression

imaging. Mol Imaging Biol 2004; 6(4):225–238.

75. Hung MC, Hortobagyi GN, Ueno NT. Development of clinical trial of E1A gene

therapy targeting HER-2/neu-overexpressing breast and ovarian cancer. Adv Exp

Med Biol 2000; 465:171–180.

76. Smith-Jones PM, Solit D, Afroze F, et al. Early tumor response to Hsp90 therapy using

HER2 PET: comparison with 18F-FDG PET. J Nucl Med 2006; 47(5):793–796.

77. Sharma V, Prior JL, Belinsky MG, et al. Characterization of a 67Ga/68Ga radio-

pharmaceutical for SPECT and PET of MDR1 P-glycoprotein transport activity in

vivo: validation in multidrug-resistant tumors and at the blood-brain barrier. J Nucl

Med 2005; 46(2):354–364.

78. Choe YS, Lee KH. Targeted in vivo imaging of angiogenesis: present status and

perspectives. Curr Pharm Des 2007; 13(1):17–31.

79. Cai W, Chen K, Mohamedali KA, et al. PET of vascular endothelial growth factor

receptor expression. J Nucl Med 2006; 47(12):2048–2056.

80. Padhani AR, Krohn KA, Lewis JS, et al. Imaging oxygenation of human tumours.

Eur Radiol 2007; 17(4):861–872.

81. O’Donoghue JA, Zanzonico P, Pugachev A, et al. Assessment of regional tumor

hypoxia using 18F-fluoromisonidazole and 64Cu(II)-diacetyl-bis(N4-methylthiose-

micarbazone) positron emission tomography: comparative study featuring microPET

imaging, Po2 probe measurement, autoradiography, and fluorescent microscopy in

the R3327-AT and FaDu rat tumor models. Int J Radiat Oncol Biol Phys 2005; 61(5):

1493–1502.



82. DeNardo SJ, Denardo GL. Targeted radionuclide therapy for solid tumors: an

overview. Int J Radiat Oncol Biol Phys 2006; 66(suppl 2):S89–S95.

83. Denardo SJ, Denardo GL, Natarajan A, et al. Thermal dosimetry predictive of

efficacy of 111In-ChL6 nanoparticle AMF-induced thermoablative therapy for

human breast cancer in mice. J Nucl Med 2007; 48(3):437–444.

84. Kanematsu M, Semelka RC, Leonardou P, et al. Angiogenesis in hepatocellular

nodules: correlation of MR imaging and vascular endothelial growth factor. J Magn

Reson Imaging 2004; 20(3):426–434.

85. Kanematsu M, Semelka RC, Osada S, et al. Magnetic resonance imaging and

expression of vascular endothelial growth factor in hepatocellular nodules in cir-

rhosis and hepatocellular carcinomas. Top Magn Reson Imaging 2005; 16(1):67–75.

86. Andrew ER. Nuclear magnetic resonance and the brain. Brain Topogr 1992; 5(2):

129–133.

87. Cai W, Rao J, Gambhir SS, et al. How molecular imaging is speeding up anti-

angiogenic drug development. Mol Cancer Ther 2006; 5(11):2624–2633.

88. Charles-Edwards EM, deSouza NM. Diffusion-weighted magnetic resonance imag-

ing and its application to cancer. Cancer Imaging 2006; 6:135–143.

89. Chenevert TL, Sundgren PC, Ross BD. Diffusion imaging: insight to cell status and

cytoarchitecture. Neuroimaging Clin N Am 2006; 16(4):619–632, viii–ix.

90. Lee JH, Huh YM, Jun YW, et al. Artificially engineered magnetic nanoparticles for

ultra-sensitive molecular imaging. Nat Med 2007; 13(1):95–99.

91. Lee JH, Jun YW, Yeon SI, et al. Dual-mode nanoparticle probes for high-perfor-

mance magnetic resonance and fluorescence imaging of neuroblastoma. Angew

Chem Int Ed Engl 2006; 45(48):8160–8162.

92. Walter C, Scheidhauer K, Scharl A, et al. Clinical and diagnostic value of preop-

erative MR mammography and FDG-PET in suspicious breast lesions. Eur Radiol

2003; 13(7):1651–1656.

93. Antoch G, Vogt FM, Freudenberg LS, et al. Whole-body dual-modality PET/CT and

whole-body MRI for tumor staging in oncology. JAMA 2003; 290(24):3199–3206.

94. Schmidt GP, Schoenberg SO, Schmid R, et al. Screening for bone metastases: whole-

body MRI using a 32-channel system versus dual-modality PET-CT. Eur Radiol

2006; 17(4):939–949.

95. Brix G, Henze M, Knopp MV, et al. Comparison of pharmacokinetic MRI and [18F]

fluorodeoxyglucose PET in the diagnosis of breast cancer: initial experience. Eur

Radiol 2001; 11(10):2058–2070.

96. Stadnik TW, Everaert H, Makkat S, et al. Breast imaging. Preoperative breast cancer

staging: comparison of USPIO-enhanced MR imaging and 18F-fluorodeoxyglucose

(FDC) positron emission tomography (PET) imaging for axillary lymph node stag-

ing: initial findings. Eur Radiol 2006; 16(10):2153–2160.

97. Schmidt GP, Haug AR, Schoenberg SO, et al. Whole-body MRI and PET-CT in the

management of cancer patients. Eur Radiol 2006; 16(6):1216–1225.

98. Catana C, Wu Y, Judenhofer MS, et al. Simultaneous acquisition of multislice PET

and MR images: initial results with a MR-compatible PET scanner. J Nucl Med

2006; 47(12):1968–1976.



20

Cancer Nanotechnology for Molecular

Profiling and Individualized Therapy

May Dongmei Wang

Departments of Biomedical Engineering and Electricaland Computer Engineering, Georgia Institute of Technology

and Emory University, Atlanta, Georgia, U.S.A.

Jonathan W. Simons

The Prostate Cancer Foundation, Santa Monica, California, U.S.A.

Shuming Nie

Departments of Biomedical Engineering and Chemistryand the Winship Cancer Institute, Emory University andGeorgia Institute of Technology, Atlanta, Georgia, U.S.A.

INTRODUCTION

Cancer nanotechnology is an interdisciplinary area of research in science, engi-

neering, and medicine with broad applications in molecular profiling and individ-

ualized therapy (1–4). The basic rationale is that nanometer-sized particles such as

biodegradable micelles, semiconductor quantum dots (QDs), and iron oxide nano-

crystals have functional or structural properties that are not available from either

molecular or bulk materials (5–15). When linked with biotargeting ligands such as

monoclonal antibodies, peptides, or small molecules, these nanoparticles are used to

target malignant tumors with high affinity and specificity. In the “mesoscopic” size

range of 5 to 100 nm diameter, nanoparticles also have large surface areas and

functional groups for conjugating to multiple diagnostic (e.g., optical, radioisotopic,

309


or magnetic) and therapeutic (e.g., anticancer) agents. The emergence of nano-

technology has opened new opportunities for personalized oncology in which cancer

detection, diagnosis, and therapy are tailored to each individual’s molecular profile,

and also for predictive oncology inwhich genetic ormolecular information is used to

predict tumor development, progression, and clinical outcome.

Recent advances have led to the development of functional nanoparticles

that are covalently linked to biological molecules such as peptides, proteins,

nucleic acids, or small-molecule ligands (1–4). Medical applications have also

appeared, such as the use of superparamagnetic iron oxide nanoparticles as a

contrast agent for lymph node prostate cancer detection (16), and the use of

polymeric nanoparticles for targeted gene delivery to tumor vasculatures (17). In

particular, the U.S. Food and Drug Administration (FDA) recently approved

AbraxaneTM, an albumin-paclitaxel (TaxolTM) nanoparticle for the treatment of

metastatic breast cancer (18). A phase I clinical trial determined that the max-

imum tolerated dose (MTD) of single-agent albumin-bound paclitaxel every

three weeks was 300 mg/m2 in patients with solid tumors (breast cancer and

melanoma). A second phase I trial demonstrated five patient responses among

39 pretreated patients with advanced solid tumors, including one response in a

patient with non–small cell lung cancer (NSCLC), three responses in patients

with ovarian cancer, and one in breast cancer. The dose-limiting toxicity was

myelosuppression, the MTD was 270 mg/m2, and premedication was not

required. Subsequent use of Abraxane in both phase II and phase III trials proved

that this new formulation was considerably superior to Taxol. The use of

nanoparticles for drug delivery and targeting is one of the most exciting and

clinically important applications of cancer nanotechnology.

CANCER BIOMARKERS AND NANOTECHNOLOGY

Cancer markers are broadly defined as altered or mutant genes, RNA, proteins,

lipids, carbohydrates, small metabolite molecules, and altered expression of

those that are correlated with a biological behavior or a clinical outcome (19–22).

Most cancer biomarkers are discovered by molecular profiling studies on the

basis of an association or correlation between a molecular signature and cancer

behavior. In the cases of both breast and prostate cancers, a major progression

step is the appearance of so-called “lethal phenotypes” (causing patient death)

such as bone metastatic, hormone-independent, and radiation- and chemotherapy-

resistant phenotypes. It has been hypothesized that each of these aggressive

behaviors or phenotypes could be understood and predicted by a defining set of

biomarkers (20). By critically defining the inter-relationships between these

biomarkers, it could be possible to diagnose and determine the prognosis of a

cancer on the basis of a patient’s molecular profile, leading to personalized and

predictive medicine.

One example is the drug trastuzumab (HerceptinTM; Genentech/Roche,

California, U.S./Basel, Switzerland), a monoclonal antibody designed to target

310 Wang et al.


amplified and overexpressed ERBB2 (also known as HER2) tyrosine kinase

receptor found in only approximately 25% to 30% of breast cancers. FDA

approval of trastuzumab was predicated on the availability of a test to detect

ERBB2 overexpression. Both an immunohistochemistry (IHC) assay for the

expressed protein (HercepTestTM; Dako, Glostrup, Denmark) and a nucleic acid–

based fluorescence in situ hybridization (FISH) test (PathVysion; Abbott Lab-

oratories, Illinois, U.S.) have been approved as in vitro diagnostics to guide

trastuzumab treatment decisions. In another example, the clinical response of

lung cancer patients to the epidermal growth factor receptor (EGFR) tyrosine

kinase inhibitor gefitinib (IressaTM; AstraZeneca, London, U.K.) is associated

with a small number of genetic mutations (23,24). Thus, a molecular diagnostic

test could be used to identify patients that are most likely to respond to this drug.

Despite these advances, critical studies that can clearly link biomarkers

with cancer behavior remain a significant challenge. One difficulty is that most

cancer tumors (especially prostate and breast cancer) are highly heterogeneous,

containing a mixture of benign, cancerous, and stromal cells. Current tech-

nologies for molecular profiling including reverse-transcriptase polymerase

chain reaction (RT-PCR), gene chips, protein chips, two-dimensional gel elec-

trophoresis, biomolecular mass spectrometry (e.g., MALDI-MS, i.e., matrix

assisted laser desorption ionization mass spectrometry, ES-MS, i.e., electro spray

mass spectroscopy, and SELDI-MS, i.e., surface enhanced laser desorption/

ionization) are not designed to handle this type of heterogeneous samples

(25,26). Furthermore, a limitation shared by all these technologies is that they

require destructive preparation of cells or tissue specimens into a homogeneous

solution, leading to a loss of valuable three-dimensional cellular and tissue

morphological information associated with the original tumor. The development

of nanotechnology, especially bioconjugated nanoparticles, provides an essential

link by which biomarkers could be functionally correlated with cancer behavior

(Fig. 1).

NANOTYPING

Semiconductor QDs are tiny light-emitting particles on the nanometer scale, and

are emerging as a new class of fluorescent labels for biology and medicine

(8–15). In comparison with organic dyes and fluorescent proteins, QDs have

unique optical and electronic properties such as size-tunable light emission,

superior signal brightness, resistance to photobleaching, and simultaneous exci-

tation of multiple fluorescence colors (Fig. 2). These properties are most promising

for improving the sensitivity and multiplexing capabilities of molecular histo-

pathology and disease diagnosis. Recent advances have led to highly bright and

stable QD probes that are well suited for profiling genetic and protein biomarkers

in intact cells and clinical tissue specimens (2–4). In contrast to in vivo imaging

applications where the potential toxicity of cadmium-containing QDs is a major

concern, immunohistological staining is performed on in vitro or ex vivo clinical

Cancer Nanotechnology 311


patient samples. As a result, the use of multicolor QD probes in IHC is likely one

of the most important and clinically relevant applications in the near term (27).

Indeed, significant opportunities exist at the interface between QD nano-

technology and signature biomarkers for cancer diagnosis and individualized

therapy. In particular, QD nanoparticle probes have been used to quantify a panel

of biomarkers on intact cancer cells and tissue specimens, allowing a correlation

of traditional histopathology and molecular signatures for the same material

(2–4,27). A single nanoparticle is large enough for conjugation to multiple

ligands, leading to enhanced binding affinity and exquisite specificity through a

“multivalency” effect (28). These features are especially important for the

analysis of cancer biomarkers that are present at low concentrations or in small

numbers of cells.

Quantitative biomarker information can be obtained by using a spec-

trometer attached to the fluorescence microscope. It is however very important to

use a common protein such as b actin or glyceraldehyde 3-phosphate dehydro-

genase (GAPDH) as an “internal control.” That is, one of the QD–Ab (antibody)

conjugates should be designed to measure the product of a housekeeping gene

that is expressed at relatively constant levels in all cells. The use of an internal

control holds great promise for overcoming a number of major problems in

biomarker quantification, such as differences in the probe brightness, variations

in probe binding efficiency, uneven light illumination, and detector responses.

Figure 1 Size-tunable optical properties of semiconductor QDs for multiplexed molec-

ular profiling of cancer. (A) Fluorescence emission spectra of different sizes of CdSe QDs.

(B) Relative sizes of the five sizes of CdSe QDs corresponding to the above spectrum.

From left to right the diameters are: 2.1 nm, 2.5 nm, 2.9 nm, 4.7 nm, and 7.5 nm.

Abbreviations: CdSe, cadium selenide; QDs, quantum dots.

312 Wang et al.


The majority of available tumor specimens are archived, formalin-fixed paraffin-

embedded (FFPE) tissues that might be several decades old. As the clinical

outcomes of these tissues are already known, these specimens are well suited for

examining the relationship between molecular profile and clinical outcome in

retrospective studies.

It is also critically important to validate the QD staining data with other

available techniques. For this purpose, O’Regan and coworkers (2) have

obtained QD molecular profiling data from standard human breast cell speci-

mens and have compared the corresponding biomarker data with traditional IHC

and FISH techniques. Briefly, slides from formaldehyde-fixed paraffin cell

blocks were stained in accordance with standard pathological protocols for three

breast cancer biomarkers—estrogen receptor (ER), progesterone receptor (PR),

and HER2. This panel of protein biomarkers was selected because of its clinical

significance in human breast cancer diagnosis and treatment (29–32). The tra-

ditional IHC results were analyzed by two independent observers and scored

with a standard scale from 0 (no visible staining in the nucleus or membrane) to

Figure 2 Nanotechnology linking biomarkers with cancer behavior and clinical out-

come. The schematic diagram shows multiplexed detection and quantification of cancer

biomarkers on intact cells or tissues with multicolor nanoparticle probes. (Left) The

images show cancer cells labeled with quantum dots, and (right) the drawings suggest

how surface and intracellular biomarkers could be quantified by wavelength-resolved

spectroscopy or spectral imaging.



3þ (strong and complete membrane or nuclear staining in > 10% of malignant

stained cells). For a comparative analysis of QD profiling with traditional IHC, it

is necessary to normalize the absolute fluorescence intensities of QD–Ab signals,

so that relative percentage values are calculated from the maximum signal

strength.

The results reveal that a 3þ score for ER, PR, or HER2 by traditional IHC

corresponds to 85% to 100% relative expression of the antigen by QD–Ab mea-

surement, and that 1þ or 2þ scores by traditional IHC correspond to 11% to 48%

expression as determined by QD quantification. It should be noted that classifi-

cation of antigens expressed at low levels (1þ or 2þ) is subjective, requiring

experience and often resulting in considerable interobserver variations. In contrast,

quantitative QD measurements allow accurate determination of tumor antigens at

low levels. For example, PR expression in MCF-7 cells and ER expression in BT-

474 cells are both classified as 1þ by traditional IHC, but quantitative QD

measurements indicate major differences in PR expression (16.8%) and ER

expression (47.7%) in these two cell lines. This indicates that the quantitative

nature of QD-based molecular profiling could simplify and standardize catego-

rization of antigens that are expressed at low levels. This is of fundamental

importance in the management of breast cancer, as the likely benefit of hormonal

therapies and trastuzumab depends directly on not just the presence but also the

quantity of hormone or HER2 receptors (33–35) (Fig. 3).

Wang and coworkers (27) have developed an integrated image processing

and bioinformatics software tool called Q-IHC for quantitative analysis of bio-

marker expression and distribution in IHC images. In comparison with previous

image processing software for automated feature extraction and quantitative

Figure 3 Targeting the HER 2/Neu antigen on breast cancer cells by using anti-

bodyconjugated quantum dots. The image shows individual breast cancer cells with

nuclear counterstaining.

314 Wang et al.


analysis (36,37), this software system is capable of handling imaging data from

both traditional and QD-based IHC. To measure the distribution of labeled

antigens, multiple slides of IHC imaging data are acquired to capture selected

tissue structures. After image acquisition, an image-processing module carries

out automatic boundary identification, semiautomatic image segmentation, and

color-based tissue classification on the basis of biomarker staining. Then, an

image analysis module quantifies the various biomarker features into numerical

values. These values become distinct features and are used for comparison with

clinical diagnosis. After validation by a physician, the quantitative data and rules

describing biomarker features are stored in a database. This semiautomatic

image processing and quantification system is designed to provide molecular

profiling data that are more objective, more consistent, and more reproducible

than completely manual or automated quantification methods. The software tools

process image files from slide scanners in Matlab, which is a collection of

various engineering processing tools. In addition, a user-friendly graphical

interface allows users to give input and feedback to improve the system quality.

NANOTHERAPEUTICS

Most current anticancer agents do not greatly differentiate between cancerous

and normal cells, leading to systemic toxicity and adverse effects. Consequently,

systemic applications of these drugs often cause severe side effects in other tissues

(such as bone marrow suppression, cardiomyopathy, and neurotoxicity), which

greatly limit the maximal allowable dose of the drug. In addition, rapid elimination

and widespread distribution into nontargeted organs and tissues require adminis-

tration of a drug in large quantities, which is not economical and often complicated

because of nonspecific toxicity. Nanotechnology offers a more targeted approach

and could thus provide significant benefits to cancer patients.

Rapid vascularization in fast-growing cancerous tissues is known to result

in leaky, defective architecture and impaired lymphatic drainage. This structure

allows an enhanced permeation and retention (EPR) effect (38–42), as a result of

which nanoparticles accumulate at the tumor site. For such passive targeting

mechanism to work, the size and surface properties of drug delivery nano-

particles must be controlled to avoid uptake by the reticuloendothelial system

(RES) (43). To maximize circulation time and targeting ability, the optimal size

should be less than 100 nm in diameter and the surface should be hydrophilic to

circumvent clearance by macrophages. A hydrophilic surface of the nano-

particles safeguards against plasma protein adsorption, and it can be achieved

through hydrophilic polymer coatings such as polyethylene glycol (PEG),

poloxamines, poloxamers, polysaccharides or through the use of branched or

block amphiphilic copolymers (44–47). The covalent linkage of amphiphilic

copolymers (polylactic acid, polycaprolactone, and polycyanonacrylate chemi-

cally coupled to PEG) is generally preferred, as it avoids aggregation and ligand

desorption when in contact with blood components.



An alternative passive targeting strategy is to utilize the unique tumor

environment in a scheme called tumor-activated prodrug therapy. The drug is

conjugated to a tumor-specific molecule and remains inactive until it reaches the

target (48). Overexpression of the matrix metalloproteinase (MMP), MMP-2, in

melanoma has been shown in a number of preclinical as well as clinical

investigations. Mansour et al. (49) reported a water-soluble maleimide derivative

of doxorubicin (DOX) incorporating an MMP-2 specific peptide sequence (Gly-

Pro-Leu-Gly-Ile-Ala-Gly-Gln) that rapidly and selectively binds to the cysteine-

34 position of circulating albumin. The albumin-DOX conjugate is efficiently

and specifically cleaved by MMP-2, releasing a DOX tetrapeptide (Ile-Ala-Gly-

Gln-DOX) and subsequently DOX. pH and redox potential have been also

explored as drug-release triggers at the tumor site (50). Another passive targeting

method is the direct local delivery of anticancer agents to tumors. This approach

has the obvious advantage of excluding the drug from the systemic circulation.

However, administration can be highly invasive, as it involves injections or

surgical procedures. For some tumors such as lung cancers that are difficult to

access, the technique is nearly impossible to use.

Active targeting is usually achieved by conjugating to the nanoparticle a

targeting component that provides preferential accumulation of nanoparticles in

the tumor-bearing organ, in the tumor itself, individual cancer cells, or intra-

cellular organelles inside cancer cells. This approach is based on specific

interactions such as lectin-carbohydrate, ligand-receptor, and antibody-antigen

(51). Lectin-carbohydrate is one of the classic examples of targeted drug delivery

(52). Lectins are proteins of nonimmunological origin, which are capable of

recognizing and binding to glycoproteins expressed on cell surfaces. Lectin

interactions with certain carbohydrates are very specific. Carbohydrate moieties

can be used to target drug delivery systems to lectins (direct lectin targeting), and

lectins can be used as targeting moieties to target cell surface carbohydrates

(reverse lectin targeting). However, drug delivery systems based on lectin-

carbohydrate have mainly been developed to target whole organs (53), which can

pose harm to normal cells. Therefore, in most cases, the targeting moiety is

directed toward specific receptors or antigens expressed on the plasma mem-

brane or elsewhere at the tumor site.

The cell surface receptor for folate is inaccessible from the circulation to

healthy cells because of its location on the apical membrane of polarized epi-

thelia, but it is overexpressed on the surface of various cancers like ovary, brain,

kidney, breast, and lung malignancies (54,55). Surface plasmon resonance

studies revealed that folate-conjugated PEGylated cyanoacrylate nanoparticles

had a 10-fold higher affinity for the folate receptor than the free folate ones did

(56). Folate receptors are often organized in clusters and bind preferably to the

multivalent forms of the ligand. Furthermore, confocal microscopy demonstrated

selective uptake and endocytosis of folate-conjugated nanoparticles by tumor

cells bearing folate receptors. Interest in exploiting folate receptor targeting

in cancer therapy and diagnosis has rapidly increased, as attested by many

316 Wang et al.


conjugated systems including proteins, liposomes, imaging agents, and neutron

activation compounds (54,55).

For enhanced tumor-specific targeting, the differences between cancerous

cells and normal cells may be exploited. By virtue of their small size, nano-

particles entail a high surface area that not only paves a way for more efficient

drug release but also a better strategy for functionalization. There is a growing

body of knowledge on unique cancer markers due to recent advances in pro-

teomics and genomics. They form the basis of complex interactions between

bioconjugated nanoparticles and cancer cells. Carrier design and targeting

strategies may vary according to the type, developmental stage, and location of

cancer (57). There is much synergy between imaging and nanotechnology in

biomedical applications. Many of the principles used to target delivery of drugs

to cancer may also be applied to target imaging and diagnostic agents to enhance

detection sensitivity in medical imaging. With engineered multifunctional

nanoparticles, the full in vivo potential of cancer nanotechnology in targeted

drug delivery and imaging can be realized.

BIONANOINFORMATICS

For the long-term goal of personalized and predictive oncology, it is important to

integrate biotechnology and nanotechnology with information technology

(Fig. 4). A major step in this direction is the development of the Cancer Bio-

informatics Grid (caBIG), the “World Wide Web for Cancer Research” spear-

headed by the U.S. National Cancer Institute (NCI) (58). Through data sharing

and standardization, this national initiative aims at improving the infrastructures

of health care information technology, to facilitate the process from clinical data

Figure 4 Schematic diagram showing the integration of biotechnology, nanotechnology,

and information technology for personalized oncology.



collection to computational data mining, and to accelerate the use of bio-

molecular markers for personalized treatment. The caBIG organizational struc-

ture is divided into working groups, each focused on different aspects of the

interoperability problem specific to cancer research. Two overarching work

groups, “Architecture” and “Vocabularies and Common Data Elements” have

begun to produce detailed specifications that will guide the design of inter-

operable bioinformatics tools. In practice, caBIG provides a common platform

from which to launch even more ambitious integrated solutions. Bioinformatics

laboratories can adopt the basic infrastructure, databases, and functionality

provided by caBIG working groups to fill gaps in knowledge and to carry out

information management responsibilities. In addition to these shared tools and

architectures, caBIG is the perfect test bed for a suite of standard tools that can

be understood, verified, and used by clinicians around the world. Another major

effort is Cancer Bioinformatics Infrastructure Objects (caBIO), an application

programming interface (API) that was first developed as Java objects and was

later extended to support various programming platforms. These objects repre-

sent the most fundamental concepts in bioinformatics research such as genes,

ontology, and sequences. Also under development are network programs such as

caBIONet based on Microsoft’s .NET technology for data analysis and inte-

gration (59).

OUTLOOK

Looking into the future, there are a number of research themes or directions that

are particularly promising but require concerted effort for success. The first

direction is the design and development of nanoparticles with monofunctions,

dual functions, three functions, or multiple functions. For cancer and other

medical applications, important functions include imaging (single or dual-

modality), therapy (single drug or combination of two or more drugs), and

targeting (one or more ligands). With each added function, nanoparticles could

be designed to have novel properties and applications. For example, binary

nanoparticles with two functions could be developed for molecular imaging,

targeted therapy, or for simultaneous imaging and therapy (but without target-

ing). Bioconjugated QDs with both targeting and imaging functions will be used

for targeted tumor imaging and molecular profiling applications. Conversely,

ternary nanoparticles with three functions could be designed for simultaneous

imaging and therapy with targeting, targeted dual-modality imaging, or for tar-

geted dual-drug therapy. Quaternary nanoparticles with four functions can be

conceptualized in future to have the abilities of tumor targeting, dual-drug

therapy, and imaging. The second direction is nanoparticle molecular profiling

(nanotyping) for clinical oncology; i.e., the use of bioconjugated nanoparticle

probes to predict cancer behavior, clinical outcome, treatment response, and

individualize therapy. This should start with retrospective studies of archived

specimens because the patient outcome is already known for these specimens.

318 Wang et al.


The key hypotheses to be tested are that nanotyping a panel of tumor markers

will allow more accurate correlations than single tumor markers, and that the

combination of nanotyping tumor gene expression and host stroma are both

important in defining the aggressive phenotypes of cancer as well as determining

the response of early-stage disease to treatment (chemotherapy, radiation, or

surgery). The third important direction is to study nanoparticle distribution,

excretion, metabolism, and pharmacodynamics in in vivo animal models. These

investigations will be very important in the development of nanoparticles for

clinical applications in cancer imaging or therapy.

In conclusion, nanotechnology is emerging as an enabling technology for

personalized oncology in which cancer detection, diagnosis, and therapy are

tailored to each individual’s tumor molecular profile and also for predictive

oncology in which genetic or molecular markers are used to predict disease

development, progression, and clinical outcomes (60,61).

ACKNOWLEDGEMENTS

We are grateful to Dr. Leland Chung, Dr. Ruth O’Regan, and Dr. Dong Shin for

insightful discussions. Our research in cancer nanotechnology was supported by

National Institutes of Health (NIH) grants (R01 CA108468-01 and U54CA119338)

and the Georgia Cancer Coalition Distinguished Cancer Scholars Program (to May

Dongmei Wang and Shuming Nie).

REFERENCES

1. Ferrari M. Cancer nanotechnology: Opportunities and challenges. Nat Rev Cancer

2005; 5:161–171.

2. Yezhelyev M, Gao XH, Xing Y, et al. Emerging use of nanoparticles in diagnosis

and treatment of breast cancer. Lancet Oncol 2006; 7:657–767.

3. Sinha R, Kim GK, Nie SM, et al. Nanotechnology in cancer therapeutics: bio-

conjugated nanoparticles for drug delivery. Mol Cancer Ther 2006; 5:1909–1917.

4. Xing Y, Kim GJ, Nie SM, et al. Nanotechnology Applications in Cancer. Annu Rev

Biomed Eng 2007; 9:257–288 (review).

5. Henglein A. Small-particle research - physicochemical properties of extremely small

colloidal metal and semiconductor particles. Chem Rev 1989; 89:1861–1873.

6. Schmid G. Large clusters and colloids - metals in the embryonic state. Chem Rev

1992; 92:1709–1727.

7. Niemeyer CM. Nanoparticles, proteins, and nucleic acids: biotechnology meets

materials science. Angew Chem Int Ed Engl 2001; 40:4128–4158.

8. Alivisatos AP. The use of nanocrystals in biological detection. Nat Biotechnol 2004;

22:47–52.

9. Alivisatos AP. Semiconductor clusters, nanocrystals, and quantum dots. Science

1996; 271:933–937.

10. Alivisatos AP, Gu WW, Larabell C. Quantum dots as cellular probes. Annu Rev

Biomed Eng 2005; 7:55–76.



11. Pinaud F, Michalet X, Bentolila LA, et al. Advances in fluorescence imaging with

quantum dot bio-probes. Biomaterials 2006; 27:1679–1687.

12. Michalet X, Pinaud FF, Bentolila LA, et al. Quantum dots for live cells, in vivo

imaging, and diagnostics. Science 2005; 307:538–544.

13. Gao XH, Yang LL, Petros JA, et al. In vivo molecular and cellular imaging with

quantum dots. Curr Opin Biotechnol 2005; 16: 63–72.

14. Smith AM, Gao X, Nie S. Quantum dot nanocrystals for in vivo molecular and

cellular imaging. Photochem Photobiol 2004; 80:377–3785.

15. Chan WC, Maxwell DJ, Gao X, et al. Luminescent quantum dots for multiplexed

biological detection and imaging. Curr Opin Biotechnol 2002; 13:40–46.

16. Harisinghani MG, Barentsz J, Hahn PF, et al. Noninvasive detection of clinically

occult lymph-node metastases in prostate cancer. N Engl J Med 2003; 348:2491–2495.

17. Hood JD, Bednarski M, Frausto R, et al. Tumor regression by targeted gene delivery

to the neovasculature. Science 2002; 296:2404–2407.

18. Robinson DM, Keating GM. Albumin-bound paclitaxel in metastatic breast cancer.

Drugs 2006; 66:941–948.

19. Hahn WC, Weinberg RA. Modelling the molecular circuitry of cancer. Nat Rev

Cancer 2002; 2:331–341.

20. Liotta L, Petricoin E. Molecular profiling of human cancer. Nat Rev Genet 2000;

1:48–56.

21. Petricoin EF, Zoon KC, Kohn EC, et al. Clinical proteomics: translating benchside

promise into bedside reality. Nat Rev Drug Discov 2002; 1:683–695.

22. Menon U, Jacobs IJ. Recent developments in ovarian cancer screening. Curr Opin

Obstet Gynecol 2000; 12:39–42.

23. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth

factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib.

N Engl J Med 2004; 350:2129–2139.

24. Paez JG, Janne PA, Lee JC, et al. EGFR mutations in lung cancer: correlation with

clinical response to gefitinib therapy. Science 2004; 304:1497–1500.

25. Michener CM, Ardekani AM, Petricoin EF, et al. Genomics and proteomics:

application of novel technology to early detection and prevention of cancer. Cancer

Detect Prev 2002; 26:249–255.

26. Srinivas PR, Verma M, Zhao Y, et al. Proteomics for cancer biomarker discovery.

Clin Chem 2002; 48:1160–1169.

27. Xing Y, Chaudry Q, Shen C, et al. Bioconjugated quantum dots for multiplexed and

quantitative immunohistochemistry. Nat Protoc 2007; 2:1152–1165.

28. Akerman ME, Chan WC, Laakkonen P, et al. Nanocrystal targeting in vivo. Proc

Natl Acad Sci U S A 2002; 99:12617–12621.

29. Umemura S, Osamura RY. Utility of immunohistochemistry in breast cancer prac-

tice. Breast Cancer 2004; 11:334–338.

30. Thomson TA, Hayes MM, Spinelli JJ, et al. HER-2/neu in breast cancer: interobserver

variability and performance of immunohistochemistry with 4 antibodies compared

with fluorescent in situ hybridization. Mod Pathol 2001; 14:1079–1086.



J Clin Oncol 2002; 20:719–726.

32. Elledge RM, Green S, Pugh R, et al. Estrogen receptor (ER) and progesterone

receptor (PR), by ligand-binding assay compared with ER, PR and pS2, by

320 Wang et al.


immunohistochemistry in predicting response to tamoxifen in metastatic breast

cancer: a Southwest Oncology Group Study. Int J Cancer 2000; 89:111–117.

33. Marquez DC, Pietras RJ. Membrane-associated binding sites for estrogen contribute

to growth regulation of human breast cancer cells. Oncogene 2001; 20:5420–5430.

34. Hicks DG, Tubbs RR. Assessment of the HER2 status in breast cancer by fluores-

cence in situ hybridization: a technical review with interpretive guidelines. Hum

Pathol 2005; 36:250–261.

35. Konecny G, Pauletti G, Pegram M, et al. Quantitative association between HER-2/

neu and steroid hormone receptors in hormone receptor-positive primary breast

cancer. J Natl Cancer Inst 2003; 95:142–153.

36. Matkowskyj KA, Schonfeld D, Benya RV. Quantitative immunohistochemistry by

measuring cumulative signal strength using commercially available software Pho-

toshop and Matlab. J Histochem Cytochem 2000; 48:303–311.

37. Matkowskyj KA, Schonfeld D, Benya RV. Quantitative immunohistochemistry by

measuring cumulative signal strength accurately measures receptor number. J His-

tochem Cytochem 2003; 51:205–214.

38. Matsumura Y, Maeda H. A new concept for macromolecular therapeutics in cancer

chemotherapy: mechanism of tumoritropic accumulation of proteins and the anti-

tumor agent smancs. Cancer Res 1986; 46:6387–6392.

39. Duncan R. The dawning era of polymer therapeutics. Nat Rev Drug Discov 2003;

2:347–360.

40. Jain RK. Transport of molecules, particles, and cells in solid tumors. Annu Rev

Biomed Eng 1999; 1:241–263.

41. Jain RK. Delivery of molecular medicine to solid tumors: lessons from in vivo

imaging of gene expression and function. J Control Release 2001; 74:7–25.

42. Jain RK. Understanding barriers to drug delivery: High resolution in vivo imaging is

key. Clin Cancer Res 1999; 5:1605–1606.

43. Gref R, Minamitake Y, Peracchia MT, et al. Biodegradable Long-Circulating

Polymeric Nanospheres. Science 1994; 263:1600–1603.

44. Ringsdorf H. Structure and properties of pharmacologically active polymers. J Polymer

Sci Polymer Symp 1975; 51:135–153.

45. Davis FF. The origin of pegnology. Adv Drug Deliv Rev 2002; 54:457–458.

46. Moghimi SM, Hunter AC. Poloxamers and poloxamines in nanoparticle engineering

and experimental medicine. Trends Biotechnol 2000; 18:412–420.

47. Park EK, Lee SB, Lee YM. Preparation and characterization of methoxy poly(eth-

ylene glycol)/poly(epsilon-caprolactone) amphiphilic block copolymeric nanospheres

for tumor-specific folate-mediated targeting of anticancer drugs. Biomaterials 2005; 26:

1053–1061.

48. Chari RV. Targeted delivery of chemotherapeutics: tumor-activated prodrug therapy.

Adv Drug Deliv Rev 1998; 31:89–104.

49. Mansour AM, Drevs J, Esser N, et al. A new approach for the treatment of malignant

melanoma: Enhanced antitumor efficacy of an albumin-binding doxorubicin prodrug

that is cleaved by matrix metalloproteinase 2. Cancer Res 2003; 63:4062–4066.

50. Guo X, Szoka FC. Chemical approaches to triggerable lipid vesicles for drug and

gene delivery. Acc Chem Res 2003; 36:335–341.

51. Allen TM. Ligand-targeted therapeutics in anticancer therapy. Nat Rev Cancer 2002;

2:750–763.



52. Kannagi R, Izawa M, Koike T, et al. Carbohydrate-mediated cell adhesion in cancer

metastasis and angiogenesis. Cancer Sci 2004; 95:377–384.

53. Yamazaki N, Kojima S, Bovin NV, et al. Endogenous lectins as targets for drug

delivery. Adv Drug Deliv Rev 2000; 43:225–244.

54. Leamon CP, Reddy JA. Folate-targeted chemotherapy. Adv Drug Deliv Rev 2004;

56:1127–1141.

55. Leamon CP, Low PS. Folate-mediated targeting: from diagnostics to drug and gene

delivery. Drug Discov Today 2001; 6: 44–51.

56. Stella B, Arpicco S, Peracchia MT, et al. Design of folic acid-conjugated nano-

particles for drug targeting. J Pharm Sci 2000; 89:1452–1464.

57. Vicent MJ, Duncan R. Polymer conjugates: nanosized medicines for treating cancer.

Trends Biotechnol 2006; 24:39–47.

58. Covitz PA, Hartel F, Schaefer C, et al. caCORE: a common infrastructure for cancer

informatics. Bioinformatics 2003; 19:2404–2412.

59. Kraj P, McIndoe RA. caBIONet: a .NET wrapper to access and process genomic data

stored at the National Cancer Institute’s Center for Bioinformatics databases. Bio-

informatics 2005; 21:3456–3458.

60. Weinshilboum R, Wang LW. Pharmacogenomics: Bench to bedside. Nat Rev Drug

Discov 2004; 3:739–748.

61. Evans WE, Relling MV. Moving towards individualized medicine with pharmaco-

genomics. Nature 2004; 429:464–468.

322 Wang et al.

[gajendra][D:/informa_Publishing/H5293_Leyland_Jones_112035/z_production/z_3B2_3D_files/978-1-4200-5293-0_Index_O.3d] [11/4/08/15:24:22] [323–334]

Index

ABCB1 polymorphisms, 19–20

AbraxaneTM, 310

Acute myeloid leukemic (AML) cells, 271

Adenosine triphosphate (ATP), 52

Adjuvant chemotherapy, 100–101, 207

Adjuvant! Online, 187, 189, 205, 207

Adjuvant on Line (AOL), vs. RS assay,

103–104

AdnaGen system, 157

Adriamycin. See Doxorubicin

Advancing Clinico Genomic Trials

(ACGT), 207

16a-[18F] fluoro-17b-estradiol (FES), 298Affymetrix GeneChip, 250, 266

Affymetrix microarray, 239

AIB1, 171

Akt, 173, 277

Aldefluor1, 270

Aldehyde dehydrogenase (ALDH), 270–271

ALDH. See Aldehyde dehydrogenase

(ALDH)

17-allylamino-17-demethoxygeldanamycin

(17AAG), 299

American College of Surgeons Oncology

Group (ACOSOG), 234

American Joint Committee on Cancer

(AJCC) classification system,

159–160

Aminoglutethimide, 174

Amonafide, 17

Amphiphilic copolymers, covalent linkage

of, 315

Amplicons, 244

Anastrozole, 32, 174

chemical structure of, 175

vs. tamoxifen, 177

Angiogenesis

PET and SPECT imaging of, 299

Animal model, 299

Anthracyclines

CBR and, 28

oxidative stress and, 28–29

Antiapoptosis genes, 201, 202

Anticytokeratin antibodies, 153, 154

Antiestrogens, 168–174

Anti-VEGF, 300

API. See Application programming

interface (API)

Apoptosis, 298

Application programming interface

(API), 318

Aromatase inhibitors (AI), 32, 137

analysis of gene expression with,

177–178

chemical structures of, 175

mechanisms of resistance, 175–176

steroidal and nonsteroidal, 174

vs. tamoxifen, 176–177

323


Array CGH, 250

ATAC trial, 175–177

AutoMACSTM system, 153–155

Automatic magnetic cell sorting system.

See AutoMACSTM system

Avidin-biotin complex peroxidase

technique, 154

BAC. See Bacterial artificial

chromosomes (BAC)

Bacterial artificial chromosomes (BAC),

64–65

b-catenin, 273Bcl-2 tumor supressor, 248

BCRP1 (breast cancer resistance

protein 1), 268

17b-estradiol (E2), 168, 201Bioinformatics laboratories, 318

Biomarkers, 51. See also Cancer

biomarkers

Bionanoinformatics, 317–318

Biopsies, gene expression profiling in,

91–92

Biotinylated random nonamers, 246

Blood collection tubes, 216–217

Blood tisssue, 218

Bmi-1, 271–272

BMI-1 oncogenic pathway, 202

BOLD (blood oxygen level-dependent)

MRI, 300

Bone marrow

cancer stem cells in, 152

cytokeratin positive tumor cells in, 206

micrometastasis, 154

tumor cells, 154

Bone mineral density, 178

Bone scanning, 291, 292

BRCA1, 5, 50–51, 266, 267

BRCA2, 5

Breast cancer

genetic susceptibility, 5

luminal, 198

molecular classification of, 3, 198–199

patients information

storage and protection of, 231–232

treatment with genetic

polymorphism, 267

Breast cancer cells

circulating. See Circulating tumor cells

subgroups of, 202–204

Breast Cancer Intergroup, 186, 188, 190

Breast International Group (BIG) 1-98

study, 175

Breast MRI, 295–296

Breast tumor model, 269

caBIG. See Cancer Bioinformatics Grid

(caBIG)

caBIONet, 318

Cancer Bioinformatics Infrastructure

Objects (caBIO) API, 318

Cancer biomarkers, 310–311

Cancer Biomedical Informatics Grid

(caBIG), 207

development of, 317

organizational structure of, 318

Cancer cells vs. normal cells, 317

Cancer metastasis, 278

Cancer Methylation Panel, 250, 251

Cancerous tissues, vascularization in, 315

Cancer stem cells. See Stem cells

Cancer Trials Support Unit, 192

CapSure HS LCM Cap, 223, 224

Carbohydrate moieties, 316

Carbonyl reductase (CBR), 28

Carcinogenesis, 269

CAT. See Catalase

Catalase (CAT), 2911C-choline, 298

CD24� breast cancer cells, 202

CD44+ breast cancer cells, 202

CD44+CD24�/loo breast cancer cells,

265, 266, 270

CD44 expression, 270

cDNA synthesis, 246, 247

CellProfileTM kit, 153

CellSearchTM system, 153

for prognostic and predictive value of

CTCs, 158

CellSpotterTM Analyzer, 153

c-erbB-2, 161

CGH. See Comparative genomic

hybridization (CGH)

Chemotherapeutic agents, 263

324 Index


Chemotherapy, 159

adjuvant. See Adjuvant chemotherapy

cytotoxic, 161

evaluation of marker, 186

neoadjuvant. See Neoadjuvant

chemotherapy

residual clinical risk and, 102–103

Chromatin immunoprecipitation (ChIP), 53

Chromogenic in situ hybridization

(CISH), 139

Circulating female hormone, 168

Circulating tumor cells (CTCs)

baseline and follow-up, 159

CD326-positive, 154, 155

clinical application of, 160–161

detection, 153–157

clinical benefits of, 161

future uses of, 162

immunostains of, 154

magnetic labeling and separation of,

219–220

prognostic value of, 158

for staging of MBC, 159–160

CISH. See Chromogenic in situ

hybridization (CISH)

Classification of breast tumors, 250

Clinical decision making, 190

Clinical Laboratory Improvement

Amendments (CLIA), 190

Clinical trials

for antiestrogens, 168

designing, 190–192

protocols, 214, 228

for treatment with MAPK-specific

inhibitors, 172

Clinicogenomic models, 207

Cloned probe arrays, in CGH, 65–66

Cluster analysis, 24811C-methionine, 298

CNA, Copy number abnormalities (CNA)

CNB. See Core needle biopsy (CNB)

Coexpression modules, 202

Combined imaging, with FDG-PET and

MRI, 297

Common Rule, 232

Comparative genomic hybridization

(CGH), 62, 249

BAC, 64–65

[Comparative genomic hybridization

(CGH)]

cloned probe arrays in, 65–66

DNA in, 69–71

metaphase chromosomes in, 65

MIP, 66–69

oligonucleotide-based arrays, 65–66

Complete pathological response (pCR),

101–102

Computed tomography (CT) scanning, 294

Contingency plans, 214

Copy number abnormalities (CNA), 61, 249

assessment by CGH. See Comparative

genomic hybridization (CGH)

gene deregulation by, 62–63

Coregulators, 171

Core needle biopsy (CNB), 83, 238–239

Cox model, 245

Cox regression, 159

CpG dinucleotides, 45–46

Cryopreservation techniques, 235–237

CTCs. See Circulating tumor cells (CTCs)

Cu-ATSM, 299

Cu-DOTA-VEGF, 299

CXCR4 chemokine receptor, 278, 280

Cycling cells, 267, 282

Cyclophosphamide

CYP3A4 in, 27

CYP2B6 in, 26–27

and GST, 27–28

CYP. See Cytochrome p450s (CYP)

CYP3A4, in cyclophosphamide, 27

CYP1B1, 267

CYP2B6, in cyclophosphamide, 26–27

CYP2C8

in paclitaxel, 31

CYP2D6

genotyping, 170

in TAM, 32

Cytochrome p450s (CYP), 267

Cytology, 51–52

Cytotoxic agents, in neoadjuvant

chemotherapy, 114–115

Cytotoxic therapies, 109

Dandelion hypothesis, 260

DASL assay, 246

Index 325


Database information, and patient

identifiers, 232

Daunorubicin, 28

DCIS. See Ductal carcinoma in situ

(DCIS)

Deacetylation, of histones, 47

Decision aids, 187

Decision boards, 187

DES. See Diethylstilbestrol (DES)

DFS. See Disease-free survival

3D GRE technique, 296–297

Diethylstilbestrol (DES), vs.

tamoxifen, 168

Dihydropyrimidine dehydrogenase

(DPD), 18–19

Discovery-based research, 185

Disease-free survival (DFS), 32–36, 109

defined, 192

GST in, 33

MnSOD, 36

MTHFR in, 33

SULT1A1, 36

UGT2B15, 36

Disseminated tumor cells (DTC), 206

Distant recurrence-free survival (DRFS)

defined, 192

DNA, in CGH, 69–71

DNA hypermethylation, 279

DNA isolation, 225–226

DNA methylation, 45

DNMT in, 46–47

inhibitors, 53–54, 55

DNA methyltransferases (DNMT), 46–47

DNA microarray analysis, 107–108

DNA synthesis, 298

40,6-doamidino-2-phenylindole

(DAPI), 153

Docetaxel, 19–20

Doxorubicin (DOX), 28, 316

DPD. See Dihydropyrimidine

dehydrogenase (DPD)

Drosophila, 276

Drugs

chemotherapy, 159

development

chemotherapy, 159

for CTC, 161

predictive markers in, 141–143

DSL ligands, 276

Ductal carcinoma in situ (DCIS), 174

E2. See 17b-estradiol (E2)Eastern Cooperative Oncology Group

(ECOG), 192

EasySepTM system, 155–157

EGFR. See Epidermal growth factor

receptor (EGFR)

Embryogenesis, 269

Embryonic stem cells, 271, 279

Endocrine therapy, 167–168

with antiestrogens, 168–174

with aromatase inhibitors,

174–178

Endothelial cells, 279, 280

magnetic labeling and separation of,

219–220

Endoxifen. See 4-hydroxy-N-desmethyl

tamoxifen

Enzymes, metabolic, 267

EpCAM. See Epithelial cell adhesion

molecule-1 (EpCAM)

Epidermal growth factor receptor (EGFR),

140–141, 171, 172

Epigenetics

DNA methylation. See DNA

methylation

histones. See Histones

Epithelial cell adhesion molecule-1

(EpCAM), 153

Epithelial stem cells, 263

ER. See Estrogen receptors (ER)

ERa, 136–138, 168gene transcription, 169

ERa-positive breast cancer, 168, 173

ERa-positive cells, 269

ERBB2, 62, 311

ER-negative breast cancer, 168, 170

ER-negative tumor cells, 269–270

ER-positive breast cancer, 187, 246

ER-positive tumors, 198, 245

Estrogen receptors (ER), 51, 108, 298

ERa, 136–138, 168gene transcription, 169

gene expression of, 55–56

mutations, 170

326 Index


European Human Frozen Tumor Tissue

Bank, 234

European Union, 207

Exemestane, 174


ExtractSure device, 223, 224

FDG PET, 293–294

FES. See 16a-[18F] fluoro-17b-estradiol(FES)

FES PET, 29818F-fluorodeoxyglucose (FDG), 292

[18F] fluorotamoxifen, 298

FFPE. See Formalin-fixed

paraffin-embedded (FFPE)

FFPE tissues, 313

Fibroblasts, 280

Fibromas, 297

Ficoll-PaqueTM, density gradient

procedure, 156

Fine needle aspiration (FNA), 83, 238

Fixed tissue, 21818F-labeled paclitaxel, 298

Flow cytometry, 264

Fluorescence-activated cell sorter

(FACS), 270, 271

Fluorescence microscope, 312

Fluorescent in situ hybridization

(FISH), 311

for HER2/neu testing, 138–139

Fluorodeoxyglucose, 293

5-Fluorouracil (5FU)

and DPD, 18–19

DPD and, 30

MTHFR and, 30, 31

TS and, 31

TS inhibition by, 19

[18F]MISO ([18F]fluoromisonidazole-

3-fluoro-1-(20-nitro-10-imidazolyl)-

2-propanol), 299

FNA. See Fine needle aspiration (FNA)

Folate receptors, 316

Food and Drug Administration

(FDA), 153, 174, 246,

311, 3111

Formalin, 215, 218

cross-linking effects of, 237

Formalin-fixed paraffin-embedded

(FFPE) tissue

extraction of RNA from tissue samples

of, 98–99, 244–245

RNA-based expression profiling of, 246

Forward genetics approach, 16

FOXO regulatory proteins, 277

Freeze-thawing cycles, 235

Fulvestrant, 174

Ga-DOTA-F(ab0)(2)-herceptin, 299Gamma rays, 292

Gamma secretase inhibitor (GSI), 276

Gefitinib, 172, 311

Gene expression

analysis, 177–178, 244–245

profiling. See Gene expression profiling

70-gene expression, 108, 124–126, 205.

See also Gene expression signature

in MINDACT trial, 130–132

prognostic value of, 126–127

in RASTER trial, 130

in TRANSBIG study, 127–129

Gene expression analysis, 177–178,

244–245

Gene expression Grade Index (GGI),

200–201

Gene expression profiling, 79–81, 97–98,

198–199, 246, 270

and chemotherapy, 100–101, 111–114

in neoadjuvant setting, 101–102

for classification of breast cancer,

198–199

and docetaxel, 112–113

oncogenic pathway analysis by, 116–118

pCR prediction, 101–102

RNA expression, 98–100

RS in. See Recurrence score (RS) assay

RT-PCR for, 99

Gene expression signatures, 265–267. See

also Gene expression profiling

for improving prognosis, 199–202

integration with, 202–204

Gene methylation

BRCA1, 50–51

estrogen receptors (ER), 51

RASSF1A, 49–50

Index 327


21-gene profile. See Oncotype DXTM

76-gene profile, 129

Genetic polymorphisms, 266–267

Genetic susceptibility, 5

Gene transcription, 19

Genomic alterations

prognostic value of, 158

Genomic Health, 190, 191

Genomic Health, Inc. (GHI), 246

Genomic hypomethylation, 279

Genomic predictors, 204–205

Germline polymorphisms, 267

Gestation, 269

GHI. See Genomic Health, Inc. (GHI)

Gliomas, 268

Glutathione-S-transferases (GST), 16, 267

and cyclophosphamide, 27–28

in DFS, 33

Glycophorin A, 156

Graphical interface, user-friendly, 315

GRP (gastrin-releasing peptide)

receptors, 299

GSK-3b, 273

Harmonization, 214

HAT. See Histone acetyltransferases

(HAT)

Health Insurance Portability and

Accountability Act, 232

Hedgehog (HH), 265

ligands, 276–277

signaling pathway, 276–277

transcription factors, Gli1 and

Gli2, 277

H & E (hematoxylin & eosin), 227

Hematopoietic cells, human, 156

Heparin-coated syringes, 215

HER2. See Human epidermal growth

factor receptor 2 (HER2)

HER2+, 198

HER3, 171

HER4, 171

HER2/neu, 138–139, 171–172

Heterochromatin, 47

HH. See Hedgehog (HH)

Histone acetyltransferases (HAT), 47

Histone deacetylases (HDAC), 55

Histones

deacetylation, 47

HAT, 47

HDAC. See Histone deacetylases

(HDAC)

methylation of, 278

modifiers, 48

posttranslational modification of, 47

ChIP for, 53

H3K9 repressors, 278

Homogenization, 224

Hormonal therapy, 191

adjuvant, 187

Hormone receptors, 161

Hormone replacement therapy

(HRT), 173

Human epidermal growth factor receptor

2 (HER2). See also HER2/neu

gene amplification in CTCs, 158

Hybridization. See Comparative genomic

hybridization (CGH)

4-hydroxy-N-desmethyl tamoxifen,

20, 170

Hypothesis-based research, 185

IHC. See Immunohistochemistry (IHC)

Illumina, Inc., 249

Illumina BeadArray, 250

Illumina microarrays, 249

Image file processing, 315

Immunohistochemistry (IHC), 135–136,

227, 311, 313

for ERa, 136–138for HER2/neu, 138

for PR, 136–138

Immunohistological staining, 311–312

Indium-111, 300

In situ hybridization (ISH), 227

International Stage IV Stratification

Study, 160

International Union Against Cancer

(UICC), 159

Invasiveness gene signature (IGS), 202

Iressa. See Gefitinib

Isoniazid, 16–17

328 Index


Java objects, 318

Kaiser Permanente health care system,

188

Kaplan-Meier estimator, 245

Krishnamurthy, Savitri, 154

Laser capture microdissection (LCM),

2–3, 221–224, 239

Lauterbur, Paul, 294

LCM. See Laser capture microdissection

(LCM)

Lectin-carbohydrate, 316

Lectins, 316

Lethal phenotypes, 310

Letrozole, 32, 174


vs. tamoxifen, 176, 177

Ligand-binding domain (LBD), 170

Liver tumors, 297

Lobular carcinoma in situ (LCIS), 174

Loss of heterozygosity (LOH), 66, 250

Luminal epithelial cells, genes of, 198

Lymph nodes, 297

microarrays for, 3–5

recurrence of breast cancer in, 245

Macromolecular extraction, 237, 239

Magnetic labeling and separation, of

tumor tissue, 219–220

MammaPrint. See 70-gene expression

Mammary gland, 264

Mammography, 292

Manganese superoxide dismutase

(MnSOD), 29

in DFS, 36

MAPK. See Mitogen-activated protein

kinase (MAPK)

MassARRAY, 52

Matlab slide scanners, 315

Matrix metalloproteinase (MMP), 316

MBC. See Metastatic breast cancer

(MBC)

MCF-7 breast cancer cells, 170, 172

Medicare, 190

Mercaptopurine, 17

Metaphase chromosomes, in CGH, 65

Metastatic breast cancer (MBC)

development of, 151

diagnosis of, 158–159

stage IV, 160

HER2-positive, 177, 178

staging, 159–160

Methotrexate (MTX), 18, 30, 31

Methylation

detection of, 51–53

of DNA. See DNA methylation

of genes. See Gene methylation

profiling, 250–251

therapy, 53–55

Methyltetrahydrofolate reductase

(MTHFR), 18

in DFS, 33

and 5-Fluorouracil (5FU), 30, 31

Microarray for Node-Negative Disease

may Avoid Chemotherapy

(MINDACT) trial, 130–132, 205

Microarray Quality Control Project

(MAQC), 82

Microarrays

for lymph node status, 3–5

with OCT, 236

with paraffin-embeded tissue, 236–237

RNA expression, 2, 4

with RNA stabilization buffers,

237–238

technical reproducibility of, 82–83

using CNB, 238–239

using FNA, 238

using LCM, 239

Micrometastatic carcinoma, 154

Micrometastatic cells, 206

microRNAs, 206

Microsoft Excel, 228

MIP. See Molecular inversion probe (MIP)

miRNA profiling, 248–249

Mitogen-activated protein kinase

(MAPK), 172

MMP. See Matrix metalloproteinase

(MMP)

MMP-2, 316

Index 329


MnSOD. See Manganese superoxide

dismutase (MnSOD)

Molecular classification, of breast

cancer, 3

Molecular diagnostic test, 186, 188,

190, 311

Molecular epidemiologic studies, 32–36

Molecular imaging, 291–292. See also

Positron emission tomography

(PET)

Molecular inversion probe (MIP), 66–69

PCR in, 67

Molecular markers, 79, 160, 188–190

defined, 185–186

development process of, 81–82

in stem cells, 265, 273

Molecular profiling

defined, 185

developmental technologies for, 311

Molecular signatures, 186

Monoclonal antibodies, 153, 156

Monocytes, 219

MPO. See Myeloperoxidase (MPO)

MRI, 294–297

applications for, 297

mRNAs, 245

MTA-1 (Manual Tissue Arrayer), 226

MTHFR. See Methyltetrahydrofolate

reductase (MTHFR)

mTOR (mammalian targets of

rapamycin), 276

MTX. See Methotrexate

Mutations

estrogen receptors, 170

Myeloperoxidase (MPO), 29

N-acetyltransferases, 16, 267

Nanoparticles. See also Nanotechnology

bioconjugated, 311, 317

design and development of, 318

molecular profiling. See Nanotyping

superparamagnetic iron oxide, 310

Nanotechnology, 311

for cancer detection, diagnosis, and

therapy, 319

QD, 312

Nanotherapeutics, 315–317

Nanotyping, 311–315, 318–319

National Cancer Institute (NCI), 186,

207, 317

National Institutes of Health (NIH),

199, 200

cancer guidelines, 205

National Surgical Adjuvant Breast and

Bowel Project (NSABP), 99–100,

188, 190, 192, 245

N-CoR, 171

Neoadjuvant chemotherapy, 5–7, 78,

109–111

cytotoxic agents in, 114–115

predictive markers of response to,

101–102

response predictors of, 78–79

use of, 293

Netherlands Cancer Institute (NKI), 199

.NET technology, 318

Noncanonical pathways. See Wnt

signaling

Nonnucleoside inhibitors, 53, 54

Non-small cell lung cancer (NSCLC), 310

Nontumorigenic cells, 265

North American Breast Intergroup, 192

Notch receptors, 276

Notch signaling pathway, 264, 276

Nottingham Prognostic Index, 205

NSABP. See National Surgical Adjuvant

Breast and Bowel Project

(NSABP)

Nucleic acids

amplification, 238

collection, 215

extraction, 237

profilation, 251

Nucleoside inhibitors, 53, 54

NuGEN Technologies, Inc., 249

OCT. See Optimal cutting temperature

(OCT)

OCT-4, 271

Octreoscan, 299

[15O]H2O PET, 292

4OHT (4-hydroxytamoxifen), 171

Oligonucleotide-based arrays,

in CGH, 65–66

330 Index


Oligonucleotides, 247

Oncogenes, 62, 251

Oncogenic pathways, deregulation of,

116–118

cytotoxic agents, 118

Oncologists, 207

Oncotype DX assay, 188, 190, 192, 194,

200, 205, 246

Optimal cutting temperature (OCT),

218, 236

p53

signature genes, 201

tumor suppressor, 248

PACCT. See Program for the assessment

of clinical cancer tests (PACCT)

Paclitaxel, 19–20, 267

CYP2C8 in, 31

Paget, Stephen, 278

Papanicolaou (Pap)-stained cytospin, 154

Paraffin, 222, 227

Paraffin-embedded tissue, 237–238

Paraffin wax, discovery of, 237

Partial signatures, 204

Pathologic report, 234

Pathologist, 232, 233

Patient identifiers, 232

PAXgene, 220–221

p21Cip1/Waf1, 277

PCR. See Polymerase chain reaction

(PCR); Polymerase chain reaction

(PCR), in MIP

pCR. See Complete pathological response

(pCR)

Peptides, 299

Peripheral blood

cancer stem cells in, 152

extraction of RNA from, 220–221

isolation of monocytes in, 219

separation from mononuclear cells, 156

Peripheral blood mononuclear cells

(PBMC), 219

PET. See Positron emission tomography

(PET)

PET-MRI scanner, 300

PET scanners, 297

Pharmacodynamics, 13

Pharmacoepidemiology, 14

Pharmacogenetics, 14, 206

molecular epidemiologic studies, 32–36

Pharmacogenomic predictors, 79–80

published studies, 86–90

validation studies of, 84–85

Pharmacokinetics, 13

Pharmacology, 13–14

Phosphatase and tensin homolog deleted

on chromosome ten (PTEN),

265, 277

Phosphate buffered saline (PBS), 215

Phosphatidylinositol-3-OH kinase (PI3K),

173, 277

Polycomb group (PcG) proteins, 46,

261, 279

Polymerase chain reaction (PCR), 236

detection of tumor cells, 158

in MIP, 67

Positron emission tomography (PET),

206, 292

Posttranslational modification,

of histones, 47

PR. See Progesterone receptor (PR)

Predictive markers

defined, 136

in drug development, 141–143

ERa, 136–138HER2/neu, 138–139

PR, 136–138

Preoperative chemotherapy. See

Neoadjuvant chemotherapy

Preoperative systemic chemotherapy

(PST). See Neoadjuvant

chemotherapy

Preregistration, 191

Preservatives, 214, 215

Primary study group, 192

Privacy rights, 231

Progenitor cells, mutated, 262, 266

Progesterone receptor (PR), 136–138,

176–177

Prognosis, in breast cancer

improving, 199–202

Prognostic marker, 199

Program for the assessment of clinical

cancer tests (PACCT), 186

Proliferation genes, 201, 202, 204

Index 331


Prostate cancer, RNA signatures in, 248

Proteinase digestion buffers, 237

Protein biomarkers, 313. See also Cancer

biomarkers

Protein collection, 215

Protein isolation, 225–226

Proteomics, 206

PTEN. See Phosphatase and tensin

homolog deleted on chromosome

ten (PTEN)

PTEN pathway, 277–278

Pyrosequencing, 52–53

QD-Ab, 312

QDs. See Quantum dots (QDs)

Q-IHC tools, 314

qRT-PCR (realtime quantitative-PCR)

for screening tumor samples for

miRNA expression, 248

Quantification system, 315

Quantum dots (QDs)

multicolor, 312

nanotechnology, 312

semiconductor, 311

Radiation, resistance to

(Radioresistance), 266

Radionuclide techniques, 292–294

vs. optical imaging, 299

RAKE (RNA-primed array-based Klenow

extension) analysis study, 249

Randomization, 192

Rapamycin, molecular targets of. See

MTOR (mammalian targets of

rapamycin)

RASSF1A, 49–50

RASTER trial, 130

Rationally designed inhibitors, 53,

54, 55

Reactive oxygen species (ROS), 27

and mitochondrial permeability, 29

Recurrence score (RS), 100, 200, 245

and adjuvant chemotherapy, 100–101

clinical utility of, 103–104

GHI, 246

for node positive patients, 104

[Recurrence score (RS)]

and pCR, 101

to predict recurrence of breast

cancer, 245

vs. AOL, 103–104

REMARK guidelines, 186

Reporter genes, 299

Response predictors, of neoadjuvant

chemotherapy, 5–7, 78–79

molecular markers as, 79

Reticuloendothelial system (RES), 315

Reverse transcriptase-polymerase chain

reaction (RT-PCR), 99, 244

assay-based evaluation of ER

expression, 189

and gene signature, 200

RNA

amplification, 237–238

degradation, 235, 247

expression microarray, 2, 4

hybridization, 238, 239

isolation, 224–225

RNAlater, 218, 236

RNA stabilization buffers, 236–237

RosetteSepTM method, 156

Rotterdam group, 199

RPMI (Roswell Park Memorial Institute)

media, 218

RS. See Recurrence score (RS) assay

RT-PCR assay, 245

Sample handling, 237

Selective ER downregulators (SERDs),

173, 174

Selective ER modulators (SERMs), 173

Semiautomatic image processing, 315

Single nucleotide polymorphism (SNP),

14–15, 249

profiling, 250

Single-photon imaging, 292

SMRT, 171

SNP. See Single nucleotide polymorphism

(SNP)

Sonic hedgehog (SHH), 276, 277

SRC-1, 171

St. Gallen classification, 246

St. Gallen criteria, 199, 200

332 Index


St. Gallen guidelines, 205

Standard operating procedures (SOPs)

for collection of tissue samples, 234

for RNA degradation, 235

STAR (Study of Tamoxifen and

Raloxifene) trial, 173

STAT3 transcription factor, 276

Stem cell lines, derivation of, 266

Stem cell niche, 261, 263, 268

molecular targets of, 278–280

Stem cell regulation, 280

Stem cells

cellular proliferation, 261

clinical research on, 264–265

differentiation of, 261, 262

division of, 276

drug resistance of, 268

embryonic, 271, 279

functional targets, 269–273

gene expression profiling of, 270

hematopoietic, 278, 280

hypothesis, 260–264, 268

intestinal, 277

leukemic, 278

mesenchymal, 280

molecular markers, 265, 273

mutations, 263, 270, 278

nontumorigenic daughter cells of, 261

Stem cell sanctuaries, 267–269

Stem cell survival, neural, 279, 280

Stochastic model, 260, 261

SULT1A1

in DFS, 36

in TAM, 32

Superoxide dismutases (SODs), 267

Swiss Institute of Bioinformatics, 202

TAM. See Tamoxifen (TAM)

Tamoxifen (TAM), 20, 31–32, 56, 137

adjuvant, 169

CYP2D6 in, 32

metabolizers of, 170

SULT1A1 in, 32

UGT2B15 in, 32

vs. aromatase inhibitors, 176–177

vs. DES, 168

TaqManTM array, 248–249

Taxanes, 19–20

CYP2C8 in, 31

Thiopurine methyltransferase (TPMT),

17–18

Thymidylate synthase (TS), 19

Tissue

acquisition of, 232–235

biopsies, 218

collection, 214–218

control blocks, 228

defined, 213

processing, 218

procurement, 214

Tissue banking, 231–232

Tissue microarray (TMA), 226–227

blocks, 228, 229, 230

mapping, 227, 228

TMA. See Tissue microarray (TMA)

TRANSBIG study, 127–129, 199–200

Transcriptional regulation, 277

Trastuzumab, 138, 172, 310–311

Trial Assigning Individualized Options for

Treatment (TAILORx) trial, 178, 186

designing of, 191

recurrence score ranges in, 192–194

TRI reagent

for isolation of DNA, 225–226

for isolation of RNA, 224–225

for tissue microarray creation, 226–230

TRIZOL method, 224–230

TS. See Thymidylate synthase (TS)

Tumor-activated prodrug therapy, 316

Tumor growth factor-b (TGF-b), 280Tumor hypoxia, 299

Tumorigenesis, 250, 273, 274

Tumors

detection of, 158

invasion, 202

regression, 260

size and growth, 238, 239

Tumor suppressor genes, 251

Tumor tissue. See Tissue

UGT2B15

in DFS, 36

in TAM, 32

Ultrasound, 292

Index 333


Vacutainers, use of, 215

Vascular endothelial growth factor

(VEGF), 279, 280

Vascularization, 315

VEGF. See Vascular endothelial growth

factor (VEGF)

White blood cells. See Monocytes

Whole-genome expression profiling, 249

Wnt/b-catenin, 272

Wnt signaling, 266, 273

World Wide Web for Cancer Research,

317

Wound-healing, gene phenotype, 200

WWOX expression, 56

Xenograft model, 265

Xenotransplantation, 264, 269

X-ray mammography, 297

Xylene, 218

334 Index

H8637

Pharmacogenetics of Breast Cancer

Towards the Individualization of Therapy

Edited by

Brian Leyland-Jones

Leyland-Jones

Pharmacogenetics of Breast C

ancerTow

ards the Individualization of T

herapy

Translational Medicine Series 7

about the book…

Pharmacogenetics is becoming increasingly relevant in the diagnosis, treatment, and recovery of cancer patients. A major problem facing oncologists is the out-standing varied efficacy of treatment. Promising advances in pharmacogenetics have allowed the development of effective agents which will enable personalized cancer chemotherapy to become routine for the clinical practice.

Written by experts in the field and combining information that is unable to be found in a single source volume, Pharmocogenetics of Breast Cancer:• combines a complete overview of pharmacogenetics and how it relates to

breast oncology for diagnosis, treatment, and the recovery of patients using an individualized therapy model

• is the first single source reference demonstrating the application of pharmacogenetics in the care of breast cancer patients

• enables physicians a coherent interpretation of the emerging science of pharmacogenetics, aiding them to incorporate individual therapies in their own practice

• gives practical guidance on various forms of specimen collection, tissue selection, and handling procedures

Oncology

about the editor...

BRIAN LEYLAND-JONES is Associate Vice President and Director, Emory Winship Cancer Institute, Emory University, Atlanta, Georgia. Dr. Leyland-Jones’ main research interests are pharmacodynamics, pharmacokinetics, and pharmacogen-etics in oncological clinical trials; translation of preclinical models into the clinic; biomarker endpoints in Phase I/II clinical trials; and screening and mechanistic studies of novel targeted and chemotherapeutic anti-cancer agents. He has authored more than 125 peer-reviewed articles and book contributions, 150 abstracts, and 29 patents.

Printed in the United States of America

of breast cancer - pharmaceutical review · pdf filepharmacogenetics in the care of breast...

Documents