home | clinical cancer research - a personalized preclinical … · 2013. 12. 9. · cancer...

Cancer Therapy: Preclinical

A Personalized Preclinical Model to Evaluate the MetastaticPotential of Patient-Derived Colon Cancer Initiating Cells

Isabel Puig1, Irene Chicote1, Stephan P. Tenbaum1, Oriol Arqu�es1, Jos�e Ra�ul Herance5, Juan D. Gispert5,Jos�e Jimenez2, Stefania Landolfi6, Karina Caci6, Helena Allende6, Leire Mendizabal3, Debora Moreno7,Ram�on Charco8, Eloy Espín9, Aleix Prat4, Maria Elena Elez7, Guillem Argil�es7, Ana Vivancos3,Josep Tabernero7, Santiago Rojas5, and H�ector G. Palmer1

AbstractPurpose: Within the aim of advancing precision oncology, we have generated a collection of patient-

derived xenografts (PDX) characterized at the molecular level, and a preclinical model of colon cancer

metastasis to evaluate drug-response and tumor progression.

Experimental Design: We derived cells from 32 primary colorectal carcinomas and eight liver

metastases and generated PDX annotated for their clinical data, gene expression, mutational, and

histopathological traits. Six models were injected orthotopically into the cecum wall of NOD-SCID

mice in order to evaluate metastasis. Three of them were treated with chemotherapy (oxaliplatin) and

three with API2 to target AKT activity. Tumor growth and metastasis progression were analyzed by

positron emission tomography (PET).

Results: Patient-derived cells generated tumor xenografts that recapitulated the same histopathological

and genetic features as the original patients’ carcinomas.We showan87.5% tumor take rate that is one of the

highest described for implanted cells derived from colorectal cancer patients. Cecal injection generated

primary carcinomas and distant metastases. Oxaliplatin treatment prevented metastasis and API2 reduced

tumor growth as evaluated by PET.

Conclusions: Our improved protocol for cancer cell engraftment has allowed us to build a rapidly

expanding collection of colorectal PDX, annotated for their clinical data, gene expression, mutational, and

histopathological statuses. We have also established a mouse model for metastatic colon cancer with

patient-derived cells in order to monitor tumor growth, metastasis evolution, and response to treatment by

PET. Our PDX models could become the best preclinical approach through which to validate new

biomarkers or investigate the metastatic potential and drug-response of individual patients. Clin Cancer

Res; 19(24); 6787–801. �2013 AACR.

IntroductionColorectal cancer is the second leading cause of death

from cancer worldwide (1). Although surgical resectioncombined with adjuvant therapy is mostly effective at theearly stages of the disease, both subsequent relapse and

diagnosis at late stage with metastasis are frequent andresponsible for the majority of patient deaths. At theseadvanced stages, resistance to conventional therapies arefrequent and treatments are therefore quite ineffective (2). Anew generation of target-directed drugs has being designedto overcome such resistance. However, a better understand-ing of the mechanisms driving drug-response and metasta-sis is crucial in order to better guide treatment decisions andimprove patient outcomes.

Although the combination of 5-fluorouracil (5-FU) withoxaliplatin (FOLFOX) or irinotecan (FOLFIRI) constitutesthe basis of current treatments for metastatic colorectalcancer, the use of biologics directed to block some alteredoncogenic pathways has also proven beneficial for patientswith advanced colorectal cancer. Cetuximab is an antibodythat specifically blocks epidermal growth factor receptor(EGFR) oncogenic signaling in cancer cells. It significantlyimproves the response of patients with advanced coloncancer to conventional chemotherapy increasing their over-all survival (3, 4). Patients that present K-RAS–activating

Authors' Affiliations: 1Translational Program, Stem Cells and CancerLaboratory; 2Molecular Oncology Group; 3Genomics Cancer Group; and4Translational Genomics Group, Vall d'Hebron Institute of Oncology(VHIO); 5ParcdeRecercaBiom�edicadeBarcelona (PRBB),Centre d'ImatgeMolecular (CRC) Corporaci�o Sanit�aria; Departments of 6Pathology, 7Med-ical Oncology, and 8HBP Surgery and Transplantation, Vall d'HebronUniversity Hospital, Universidad Aut�onoma de Barcelona; and 9GeneralSurgery Service, Vall d'Hebron University Hospital, Barcelona, Spain

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

CorrespondingAuthor:H�ectorG.Palmer, PasseigVall d'Hebron119-129,Barcelona 08035, Spain. Phone: 34-93-2746000, ext. 4925; Fax: 34-93-2746708; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-12-1740

�2013 American Association for Cancer Research.

ClinicalCancer

Research

www.aacrjournals.org 6787

on January 3, 2021. © 2013 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst October 29, 2013; DOI: 10.1158/1078-0432.CCR-12-1740

http://clincancerres.aacrjournals.org/

mutations are refractory to Cetuximab, indicating the rel-evance of genotyping tumors in order to select the mostappropriate personalized treatment (5, 6).

It is clear that a stepwise accumulation of particulargenetic alterations is a driving force in tumor progressionand may also define the resistance or sensitivity to specifictarget-directed drugs. The establishment of improvedpreclinical models that recapitulate the human diseasepreserving its intratumoral cell heterogeneity and histo-pathological and genetic alterations has become essentialfor testing the efficacy of new target-directed therapiesand validating biomarkers of drug response. Few collec-tions of subcutaneous patient-derived xenografts (PDX)from colorectal cancer patients have been described todate (7, 8). They recapitulate the major histologicalfeatures and genetic alterations of the original patient’scarcinoma and reproduce the KRAS-dependent responseto anti-EGFR drugs.

Most in vivo models developed thus far with patient-derived cells involve their transplantation to immunode-ficient mice either by subcutaneous injection or into thekidney capsule (9, 10). The accessibility of subcutaneoustumors constitutes a great advantage in both monitoringtumor growth and assessing the effects of therapeutic inter-vention. However, a major disadvantage is that the subcu-taneous microenvironment differs greatly from that of thecolon. Interactions between the host environment and thetumor graft determine tumor cell expression profiles, levelsof growth factors and nutrients, as well as tumor angiogen-esis and metastatic behavior (11). Consequently, thesemodels do not recapitulate the advanced stages of coloncancer because mice do not develop metastasis. The closestmodels to human disease involve the injection of coloncancer cell lines into the cecum wall or the transplantation

of pieces of tumoral tissue derived from patients in thecolon serosa of nude mice. Concerning the latter, cells canmetastasize to the lymph nodes, liver, and peritoneum, butthey do not generate lung metastasis (12). Finally, freezingtissue pieces reduces cell viability, making the generation ofa patient-derived cells collection for long-term studiesdifficult.

Orthotopic injection of cancer cell lines can also recapit-ulate the metastatic dissemination to the main tissuesaffected in patients with advanced colorectal including lungmetastasis (13, 14). However, although cell lines are able toself-renew, they lose their pluripotency, generating veryhomogeneous tumors which do not recapitulate the cellheterogeneity characteristic of human colon carcinomas(15). This is an important factor that may partly explainthe lack of correlation between the in vivo response of celllines to antitumoral drugs and the resistance of patients toequivalent treatments. The discrepancies are even moreimportant when cell lines are injected subcutaneously,observing in many cases a positive response to antitumoralagents in preclinical models that clearly fail in clinical trialswhere patients’ disease progresses (15).

Several positron emission tomography (PET) studieshave been reported in mouse models of human cancer(16, 17). The most frequently employed radiotracer was18F-fluorodeoxyglucose (18F-FDG) followed by 18F-fluor-othymidine (18F-FLT; ref. 18). Both have been used toidentify the tumoral lesion, follow their growth and eval-uate the effectiveness of new treatments in vivo (19, 20). Toour knowledge, there have been no other previouslyreported PET studies inmice inoculated orthotopically withpatient-derived colon cancer cells. We also aimed to estab-lish which is the most advantageous radiotracer to evaluatecolon cancer in mice.

In summary, currently available models of colon cancerhave produced incomplete results, misguiding oncologistsand pharmaceutical companies when important decisionsare taken based on such preclinical data. This misleadinginformation can affect the initiation, design, or evolution ofclinical trials with new antitumoral drugs and consequentlythe future of patients affected by advanced colon cancer.This unacceptable scenario calls for an accurate preclinicalmodel that faithfully recapitulates metastatic colon cancerin order to evaluate the potential benefit of new drugs forpatients with advanced disease more precisely.

We have generated a PDX collection from primary coloncarcinomas and liver metastasis annotated for gene expres-sion,mutational status, histopathological, and clinical data.Our rapidly expanding collection is incorporating the maincolorectal tumor subtypes allowing us to test the efficacy oftarget-directed drugs. We have also developed a preclinicalmodel of colon cancer metastasis by injecting patient-derived colon cancer cells into the cecum wall of NOD-SCID mice and following tumor evolution by PET. Func-tional evaluation of treatment response could be performedwith cells derived from individual patients, providing pre-cise experimental data to oncologists upon selecting thebesttailored therapy.

Translational RelevanceMetastatic colorectal cancer patients present enhanced

resistance to conventional chemotherapy or target-directed drugs and a high rate of mortality. It is thereforecritical to establish preclinical models characterized atmolecular level, to test the efficacy of new therapies andfunctionally validate biomarkers of response. Translat-ing experimental results from these models to the clinicwill facilitate the identification of drug-sensitive patientsand better guide therapy selection.We have developed an improved procedure to gener-

ate patient-derived xenograftmodels of colorectal cancerthat retain the histological and genetic traits of theoriginal patients’ carcinomas. We have also establishedanorthotopicmodel of colorectalmetastasis by injectingtumor cells into the cecumwall of NOD-SCIDmice. Ourmodel allows us to monitor drug-response and meta-static capacity of colon cancer patient-derived cells andpotentially correlate the results with distinctive molec-ular biomarkers.

Puig et al.

Clin Cancer Res; 19(24) December 15, 2013 Clinical Cancer Research6788




Materials and MethodsTissue collection and patient informationWritten informed consent was signed by all patients. The

project was approved by the Research Ethics Committee ofthe Vall d’Hebron University Hospital, Barcelona, Spain(approval ID: PR(IR)79/2009). Human colon tissue sam-ples consisted of biopsies from nonnecrotic areas ofprimary adenocarcinomas or liver metastases correspond-ing to patients with colon and rectal cancer who under-went tumor resection. Sections for immunohistochemis-try were obtained from formalin-fixed paraffin-embedded(FFPE) tissue blocks. Detailed clinico-pathological infor-mation including tumor location, TNM (tumor, nodes,metastasis) status was compiled and genetic analysiscarried out for each patient. Histological diagnosis, eval-uated by the Pathology Service of Vall d’Hebron Univer-sity Hospital, was based on microscopic features of car-cinoma samples determining the tumor histotype, grade,and stage. The assessment of nuclear expression of mis-match-repair proteins (MLH1, MSH2, MSH6, PMS2) wasperformed by immunohistochemistry.

Gene expressionHematoxylin and eosin (H&E) stainingwas performed in

each FFPE tumor tissue. Areas enriched in tumor tissue wereidentified. A minimum of two 1-mm FFPE tumor tissuecores were collected. RNA was purified using the RocheHighPure FFPE Micro Kit, and �100 ng of total RNA wasused to measure expression of 292 selected genes using thenCounter platform from Nanostring Technologies (www.nanostring.com). In short, fluorescently labeled probes aredesigned for specific genes and allowed to hybridize totarget RNAs, and then captured and individual RNA mole-cules counted using color-coded probe pairs (21). Raw datawere log base 2 transformed and normalized using 5 house-keeping transcripts.

Mutational statusHuman tumor samples were genotyped as previously

described (22).Microsatellite instability was analyzed usingthe MSI-Analysis System (Promega).

Patient-derived cells isolation and cultureColon carcinoma tissues obtained upon surgery were

washed 3 times in cold PBS solution and incubatedovernight in DMEM/F12 (Gibco) containing a cocktailof antibiotics and antifungals (penicillin (250 U/mL),streptomycin (250 mg/mL), fungizone (10 mg/mL), kana-mycin (10 mg/mL), gentamycin (50 mg/mL), and nystatin(5 mg/mL; Sigma-Aldrich). Isolation of patient-derivedcells has been previously described (23, 24). Enzymaticdigestion was performed using collagenase (1.5 mg/mL;Sigma-Aldrich) and DNase I (20 mg/mL; Sigma-Aldrich)in a medium supplemented with a cocktail of antibioticsand antifungals (as described above) during 1 hour at37�C with intermittent pipetting every 15 minutes todisperse cells. The dissociated sample was then filtered

(100 mm pore size) and washed with fresh medium. Redblood cells were lysed by brief exposure to ammoniumchloride and the sample was washed again. Finally, cellswere used for subcutaneous or orthotopic injections inNOD-SCID mice.

Xenograft tumors in miceExperiments were conducted following the European

Union’s animal care directive (86/609/EEC) and wereapproved by the Ethical Committee of Animal Experimen-tation of the VHIR–the Vall d’Hebron Research Institute(ID: 40/08 CEEA and 47/08/10 CEEA). NOD-SCID (NOD.CB17-Prkdcscid/NcrCrl) were purchased from Charles RiverLaboratories. A total of 1 � 105 patient-derived cells sus-pended in PBS were mixed 1:1 with Matrigel (BD Biosci-ence) and injected subcutaneously into both flanks ofNOD-SCID mice. After 3 to 8 weeks, visible tumors weredetected. When the tumor grew to 1 cm3 in size, mice wereeuthanized and xenografts were processed to obtain a newcell suspension as previously described or fixed for histo-logical analysis.

For orthotopic transplantation, 1 � 106 patient-derivedcells suspended in 50mLof PBSwere injected into the cecumwall of NOD-SCID mice as has been previously reported(14). API-2 (Tocris Bioscience) at 1 mg/kg in PBS þ 2%dimethyl sulfoxide (DMSO) or oxaliplatin (Sigma) at 20mg/kg in PBS were injected intraperitoneally every secondday or twice a week respectively beginning the first day oftumor detection by palpation. Control mice were injectedwith the corresponding amount of vehicle (PBS þ 2%DMSOor PBS). In the case of API-2–treated animals, tumorgrowth was measured by micro-PET imaging. Tumorgrowth was assessed by palpation in oxaliplatin-treatedanimals.Whenmatching endpoint criteriamicewere eutha-nized and complete necropsies were performed. Primarycarcinomas in the cecumandmetastases in the liver, lung, orany other visible tissue affected were collected for histolog-ical analysis.

Immunohistochemistry and antibodiesAll immunostaining was performed on paraffin-

embedded tissues. Tissue blocks were sectioned, mountedon microscope slides, and heated at 56�C overnight.Paraffin was removed with xylene and tissues were seriallyrehydrated through descending ethanol concentrations towater. Sections were stained with H&E to assess cellularmorphology. For immunofluorescence, antigen retrievalwas performed by boiling the samples in a microwaveoven using 10 mmol/L sodium citrate buffer (pH 6).Slides were then washed twice in PBS and once in PBS-1% Tween-20 (Sigma-Aldrich) for 15 minutes. Tissuespecimens were blocked for 1 hour with PBS containing3% of bovine serum albumin. Slides were incubated withspecific primary antibodies at 4�C overnight: b-catenin1:100 (Abcam), caspase-3 1:100 (Cell Signaling Technol-ogy), chromogranin A 1:100 (Clone LK2H10; AbDSero-tec), cytokeratin 20 1:100 (Clone Ks 20.8; Dako), EpCAM1:100 (Clone E144; Abcam), Ki67 1:100 (DAKO

A Metastasis Model with Patient-Derived Colon Cancer Cells

www.aacrjournals.org Clin Cancer Res; 19(24) December 15, 2013 6789




Cytomation), MUC2 1:100 (Clone MOPC-21; BD Bio-science), and Villin1 1:100 (Lifespan Bioscience). Doubleimmunostaining was performed incubating slides withthe corresponding secondary antibodies (goat anti-mouseand goat anti-rabbit) conjugated to Alexa Fluor 488 andAlexa Fluor 555 (Invitrogen) at a dilution of 1:200 for1 hour at room temperature. Nuclei were stained withHoechst 33342 (5 mg/mL; Sigma-Aldrich). An OlympusFluoView FV1000 Confocal Microscope was used to visu-alize fluorescence and acquire images.

Positron emission tomographyMice were anesthetized in an induction cage using

isofluorane vaporized in O2 at a concentration of 4%.After the anesthesia induction, animals were placed in alateral position and received the radiotracer dose in oneof the tail veins. After dose injection, the animals werereturned to their cages for 2 hours’ radiotracer uptake. Inthe case of 11C-methionine and 11C-choline, the imagestarted immediately after radiotracer injection. DuringPET acquisition, mice were kept under anesthesia withisofluorane at 1% in O2 vaporized through orofacialmasks. Injected doses were (mean � SD) 271 � 88.8mCi for the 18F-FDG, 186.1 � 40.5 mCi for the 18F-FLT,294.2 � 106.9 mCi for the 11C-choline, and 216.8� 91.4mCi for 11C-methionine.

Emission data were acquired for 30 minutes in a micro-PET R4 system (Concorde 175 Microsystems; Siemens).Data were corrected for nonuniformity, random coinci-dences, and radionuclide decay, but not for scatter orattenuation. 11C studies were reconstructed with a filteredback-projection algorithm into amatrix size of 128� 128�63, a voxel size of 0.85 � 0.85 mm, and slice thickness of1.21 mm. 18F studies were reconstructed with a OSEM-2Dalgorithm into a matrix size of 256 � 256 � 63 and a voxelsize of 0.42 � 0.42 � 1.21.

Images were visually inspected and when a tumor wasclearly identified, volumes of interest were then defined.Finally, in the images obtainedwith 18F-FDG, the volumeoftumoral tissue was calculated in order to follow the grownof the mass over time.

StatisticsDifferences in primary tumor or metastases parameter

classes were analyzed by Fisher and x2 tests. Differences inproliferation and apoptosis status were analyzed byunpaired t test with Welch correction. Differences in tumorvolume were analyzed by unpaired t test comparing themeans of untreated andAPI2-treated groups of values. In allcases, a P value lower than 0.05 was considered statisticallysignificant.

ResultsEngraftment of colorectal patient-derived cells

All patient-derived tumor tissues were disaggregated in asingle-cell suspension and a minimum of 1 � 105 viablecellswere subcutaneously injected into bothflanks ofNOD-

SCID mice. Consecutive cell purification and reinjectionwas performed from the initial PDX tumors to amplify thesamples in a second generation ofmice. The 40 PDX shown(Tables 1 and 2) are part of an expanding collection ofmodels stored as frozen single-cell suspensions, which canbe thawed and reimplanted in NOD-SCID mice withoutany reduction of their tumor take rate.

A total of 32 primary colorectal cancer carcinomas(Table 1) and 8 liver metastases (Table 2) were processedand implanted in NOD-SCID mice. A total of 27 primarycarcinomas and all 8 metastases engrafted and generateda PDX model, representing an overall 87.5% tumor takerate (Tables 1 and 2). This is one of the highest tumor takerates ever described for colorectal cancer patient-derivedxenografting (7–9). A total of 4 of the 5 failed engraft-ments corresponded to cells derived from patients withno lymph nodes affected (N0) presenting nonmucinousadenocarcinomas. Furthermore, the tumor take rate of thesuccessfully engrafted N0 tumors was significantly lowerthan Nþ tumors (P ¼ 0.0117). We also observed areduced engraftment capacity of tumors with lower dif-ferentiation grade G1. The only 2 G1 tumors processedfailed to grow in mice (P ¼ 0.0206). No other statisticallysignificant correlation was observed with the rest of thepatients’ clinicopathological characteristics.

Cells derived from all liver metastasis samples generatedPDX, indicating their enhanced engraftment potential aspreviously suggested (7, 25, 26). Liver-derived tumor cellsfrom advanced patients treated with adjuvant chemother-apy showed a lower implantation rate, although not astatistically significant difference (Table 2). Furthermore,metastasis-derived PDX showed a shorter xenograft latencytime (46.6 � 21.7 days) than those derived from primarycarcinomas (68 � 34 days).

Histological and molecular characterization of thePDX collection

The histopathology of all PDX (first and second passagein mice) presented in concordance with their respectivepatient’s original carcinoma (Supplementary Fig. S1A andS2B and data not shown). The collection included 8modelsderived from mucinous adenocarcinomas and 27 conven-tional adenocarcinomas. Samples P10 (primary mucinousadenocarcinoma) and P33 (liver metastasis) were derivedfrom the same patient who underwent surgery upon initialdiagnosis of colorectal cancer and then after relapse withliver metastasis. This patient is the only rectal carcinomacase in our PDX collection.

Because gene expression profiles can define a particulartumor subtype (27–29), we analyzed 33 of the 35 estab-lished PDX using the Nanostring platform with anidentifier of 292 genes (Fig. 1 and Supplementary TableS1). Sample preparation failed in the 2 remaining PDX.We observed a perfect clustering of all mucinous sepa-rated from the nonmucinous adenocarcinomas. P10 andP33 samples derived from the same patient clusteredtogether indicating their similarities at gene expressionlevel.

Puig et al.





All 35 PDX samples were also genotyped using Seque-nom technology to identify the most frequent mutationsin oncogenes and tumor suppressors (ref. 22; Fig. 1 andSupplementary Table S2). We detected mutations fre-

quent in colorectal tumors affecting KRAS, PIK3CA,APC, TP53, or BRAF (29). E542K or E455K mutationsin PIK3CA gene were frequent in PDX derived fromconventional carcinomas and absent from mucinous

Table 1. Patients' clinical characteristics and PDX implantation rates of primary tumors

All processed (n ¼ 32) Growing in NOD-SCID mice (n ¼ 27)

Parameters Class Number (%) Number (%) Implantation rate% Tumor latency (days)

Gender (P ¼ 0.7652)Female 18 (56.2) 15 (55.5) 78.8 57.1 � 16.7Male 14 (44.8) 12 (44.5) 82.3 78.1 � 42.7

Age (P ¼ 0.1219)�50 y and <60 5 (15.6) 3 (11.1) 62.5 60 � 15�60 y and <70 9 (28.1) 7 (25.9) 88.2 73.9 � 37.9�70 y and <80 10 (31.2) 10 (37) 91.3 62.9 � 28.4�80 y 8 (25) 7 (26) 68.4 72.9 � 46.2

pT-Primary tumor status (P ¼ 0.3573)pT1 2 (6.2) 2 (7.4) 100 67.5 � 31.8pT2 3 (9.4) 2 (7.4) 50 150 � 0pT3 11 (34.4) 9 (33.4) 83.3 69.1 � 30.9pT4 16 (50) 14 (51.8) 80.5 55.6 � 21

pN-Lymph node status (P ¼ 0.0117)pN0 14 (43.7) 10 (37) 64� 69 � 36.2pNþ 18 (56.3) 17 (63) 90.5 67.4 � 33.7

pM-Distant metastasis status (P ¼ 0.3125)Unknown 27 (84.4) 22 (81.5) 81 70.2 � 29.9pM0 3 (9.4) 3 (11.1) 60 75 � 65.4pM1 2 (6.2) 2 (7.4) 100 33.5 � 4.9

Differentiation grade (P ¼ 0.0206)G1 2 (6.2) 0 (0) 0� 0G2 11 (34.4) 10 (37) 76.2 77.9 � 43.1G3 19 (59.4) 17 (63) 82.6 62.2 � 27G4 n.d. n.d. n.d. n.d.

Type of tumor (P ¼ 1)Adenocarcinoma 24 (75) 20 (74) 81.2 70.8 � 36.5Mucinous adenocarcinoma 8 (25) 7 (26) 78.9 60 � 26

Primary tumor location (P ¼ 0.7491)Right colon 23 (71.9) 19 (70.3) 79.1 61.5 � 30.1Left colon 8 (25) 7 (26) 82.3 78.8 � 43.3Rectum 1 (3.1) 1 (3.7) 100 90

Treatment (P ¼ 0.3532)Surgery 16 (50) 13 (48.1) 74.2 77.3 � 38.2Surgery þ chemotherapy 16 (50) 14 (51.9) 86.1 59.4 � 28.2

Outcomes (P ¼ 0.4879)Alive 23 (71.9) 19 (70.4) 77.5 65.5 � 33.6Dead 9 (28.1) 8 (29.6) 88.9 74 � 36.4

Cells derived from primary colorectal tumors presented different implantation rate or tumor latency when injected subcutaneously inNOD-SCID mice. Histological differentiation grade (G1–G2, well/moderate; G3–G4, poor/undifferentiated). Tumor-node-metastasisstatus. pT1–pT4, invasive tumors (pT1, submucosa; pT2, tunica muscularis; pT3, subserosa; pT4, serosa or other organs). pN0, nomalignant lymphnodes; pNþ, at least 1 positive regional lymphnode. pM0, nodistantmetastasis; pM1, presenceof distantmetastasis;Unknown, at the stage of primary tumor evaluation, metastasis cannot be assessed. Cases included in this study were reported ashaving been staged according to 7th edition of the American Joint Committee on Cancer (AJCC) Staging Manual (2010). Asteriskindicates a significant difference (P < 0.05) between groups as quantified by Fisher test or x2 test.






adenocarcinomas. Mutation in PIK3CA at positionH1047R was only observed in samples 10 and 33 fromthe patient with rectal cancer.

No other correlation was observed between mutationalstatus, gene expression clusters, and histopathologicalcharacteristics. Paired genotyping was performed on 12

Table 2. Patients' clinical characteristics and PDX implantation rates of liver metastases

All processed (n ¼ 8) Growing in NOD-SCID mice (n ¼ 8)

Parameters ClassNumber(%)

Number(%)

Implantationrate%

Tumorlatency (days)

Gender (P ¼ 0.668)Female 4 (50) 4 (50) 58.3 58 � 24.8Male 4 (50) 4 (50) 75 35.3 � 11.8

Age (P ¼ 0.1969)�50 y and <60 3 (37.5) 3 (37.5) 44.4 57.3 � 30.4�60 y and <70 2 (25) 2 (25) 83.3 52.5 � 10.6�70 y and <80 3 (37.5) 3 (37.5) 77.8 32 � 12.1�80 y n.d. n.d. n.d. n.d.

pT-Primary tumor status (P ¼ 0.1782)pT1 n.d. n.d. n.d. n.d.pT2 n.d. n.d. n.d. n.d.pT3 3 (37.5) 3 (37.5) 88.9 42 �19.7pT4 5 (62.5) 5 (62.5) 53.3 49.4 � 24.6

pN-Lymph node status (P ¼ 0.5362)pN0 1 (12.5) 1 (12.5) 100 45pNþ 7 (87.5) 7 (87.5) 63.6 46.9 � 23.5

pM-Distant metastasis status (P ¼ 0.1441)Unknown 4 (50) 4 (50) 75 51 � 25.1pM0 1 (12.5) 1 (12.5) 100 21pM1 3 (37.5) 3 (37.5) 44.4 48.5 � 4.9

Differentiation grade (P ¼ 0.1782)G1 1 (12.5) 1 (16.7) 33.3 52G2 5 (62.5) 5 (62.5) 80 49.2 � 27.2G3 2 (25) 2 (25) 50 37.5 � 10.6G4 n.d. n.d. n.d. n.d.

Type of tumor (P ¼ 0.5362)Adenocarcinoma 7 (87.5) 7 (87.5) 63.6 46.9 � 23.5Mucinous adenocarcinoma 1 (12.5) 1 (12.5) 100 45

Primary tumor location (P ¼ 0.7251)Right colon 1 (12.5) 1 (12.5) 50 30Left colon 2 (25) 2 (25) 66.7 46.2 � 26.5Rectum 5 (62.5) 5 (62.5) 71.4 56 � 5.7

Treatment (P ¼ 0.1304)Surgery 2 (25) 2 (25) 100 52.5 � 10.6Surgery þ chemotherapy 6 (75) 6 (75) 57.9 44.7 � 24.9

Outcomes (P ¼ 0.3625)Unknown 1 (12.5) 1 (12.5) 50 30Alive 6 (75) 6 (75) 64.7 53.7� 20.4Dead 1 (12.5) 1 (12.5) 100 21

Cells derived from livermetastastasis presenteddifferent implantation rateor tumor latencywhen injected subcutaneusly inNOD-SCIDmice. Histological differentiation grade (G1–G2, well/moderate; G3–G4, poor/undifferentiated). Tumor-node-metastasis status. pT1–pT4, invasive tumors (pT1, submucosa; pT2, tunicamuscularis; pT3, subserosa; pT4, serosaor other organs). pN0, nomalignant lymphnodes; pNþ, at least 1 positive regional lymphnode; pM0, nodistantmetastasis; pM1, presenceof distantmetastasis; Unknown, at thestage of primary tumor evaluation, metastasis cannot be assessed. Cases included in this study were reported as having been stagedaccording to 7th edition of the American Joint Committee on Cancer (AJCC) Staging Manual (2010). Asterisk indicates a significantdifference (P < 0.05) between groups as quantified by Fisher test or x2 test.

Puig et al.





original tumor samples obtained upon surgery and thosefrom the corresponding PDX. In all cases, we observed aperfect match between the mutations detected in theoriginal tumors and their paired PDX.

Metastatic potential of colorectal cancer patient-derived cellsWe derived tumor cells from colon carcinomas surgically

removed from 6 different patients (Table 3). Patient 1(sample P1) presented a pT4aN0 high-grade exophyticmucinous adenocarcinoma that invaded adjacent adiposetissue (reaching the serosa), but noneof the 20 lymphnodesanalyzed. Patient 1 received 6 months’ Capecitabine asadjuvant therapy following surgery. Patient 2 (sampleP2) developed a pT4aN1 high-grade conventional adeno-carcinoma that infiltrated the intestinal serosa and 2 of 29lymph nodes. Patient 3 (sample P3) presented a pT3N1alow-grade conventional adenocarcinoma that infiltratedserosa, adipose tissue, and 1 of 15 lymph nodes. Patient5 (sample P5) presented a high-grade, poorly differentiated,pT4aN0 mucinous carcinoma that invaded the adiposetissue but none of the 16 lymph nodes analyzed. Patient5 received De Gramont adjuvant chemotherapy for 6months starting after surgery. Patient 6 (sample P6) pre-sented a high-grade, pT4aN2b, ulceroinfiltrative mucinousadenocarcinoma that invaded the serosa and 7 of the 27lymphnodes analyzed. This patient relapsed 5months afterinitial surgery presenting infiltration in lymph nodes andmetastases in the liver and the adrenal gland. Patient 10 of

33 presented a low-grade, pT4aN0, primary mucinousadenocarcinoma in the rectum that invaded the serosa butnone of the 14 lymphnodes analyzed (sample P10). Patient10 of 33 received FOLFOX chemotherapy for 5 months butrelapsed 16 months after the initial surgery, developingpolylobulated liver metastasis (sample P33). Patient 2 diedbecause of surgical complications and patient 5 died 5months after diagnosis of relapse. The other 4 patients arestill alive and have not, up until now, developed distantmetastases, as observed by thoraco-abdominal computer-ized tomography.

A total of 1 � 106 cells derived from each patient wereinjected into the cecum wall of 10 NOD-SCID mice toevaluate their metastatic potential. A few mice died unex-pectedly a couple of days after injection because of surgicalcomplications.

All mice injected with cells derived from patient 1generated a primary adenocarcinoma in the cecum thatrecapitulated the same histology as the original patient’stumor and subcutaneous xenograft, demonstrating highcancer initiation potential and pluripotency capacity(Table 4 and Supplementary Fig. S1A). Equivalent resultswere obtained with cells derived from patients 2, 6, and33 with a 100% tumor take rate, whereas those frompatient 3 generated tumors in 3 of the 5 injected mice.Regarding the cells from patient 3, injection into thececum wall was performed just after thawing the cells,validating the fact that the freezing process still preservedmost of their tumorigenic potential. Finally, 8 of 9 mice

Figure 1. Hierarchical cluster based on gene expression of primary tumors andmetastases. Cells derived from each PDXwere analyzed for gene expression of293genesusing thenanostringplatform. Tumor sampleswereorderedbyhierarchical clustering usinguncenteredPearsoncorrelation distance andcompletelinkagewithCluster 3.0 software (43). The cluster treewas visualizedwithGenepattern tools (http://genepattern.broadinstitute.org) and represents the relativedistance similarities of samples. The same samples were also genotyped by Sequenom using a panel of frequent mutations in oncogenes and tumorsuppressor genes (22).






generated a tumor in the cecum when injected with cellsderived from patient 5.

Cells derived from the mucinous adenocarcinomas ofpatients 1 and 5 generated metastases in the abdominalcavity (carcinomatosis), lungs, and liver that were con-firmed by H&E staining and immunoflurescence forEpCAM and CK20 (Table 4, Fig. 2 and Supplementary Fig.S2). On the contrary, cells derived from patients 2 and 33did not generate metastases in any of the injected mice. Inthe case of patient 3, 1 mice out of 3 injected developedcarcinomatoses and lung metastasis. In the case of cellsderived frompatient 6, injectedmice died unexpectedly anddistant tissues could not be collected in order to determinethe presence of metastases.

The pluripotency of patient-derived cells was evaluatedin further detail by staining patients’ original adenocar-cinomas and the corresponding xenograft tumors forlineage differentiation markers. The 6 PDX models reca-pitulated the same differentiation heterogeneity as theoriginal carcinoma presenting particular proportions ofmucinosecretory (Mucin 2), absortive (Vilin 1), or enter-oendocryne (Chromogranin) cells (Supplementary Fig.S3A and S3B).

Testing drug response in colorectal cancer PDXmodelsWefirst studied the effect of oxaliplatin chemotherapy on

tumor growth and metastasis using our orthotopic PDXmodel. We injected cells derived from patients 5, 6, or 33into the cecumwall of 10NOD-SCIDmice. A fewmice diedunexpectedly a couple of days after injection because ofsurgical complications. Twomonths after injection, half themice were treated every second day with oxaliplatin and theother half with vehicle until sacrificed. Only mice injectedwith cells from patient 5 and treated with vehicle generateddistant metastases in the lungs and liver (3 of 4 mice),whereas oxaliplatin preventedmetastases formation (4 of 4mice). Mice injected with cells from patient 33 did notgenerate distant metastasis irrespective of treatment. Final-ly, mice injectedwith cells frompatient 6 died unexpectedlyand tissues could not be collected for histological evalua-tion of distant metastasis.

To further evaluate the effect of treatment with oxalipla-tin, we quantified the presence of proliferative (Ki67) orapoptotic (cleaved caspase 3) cells on histological sectionsof primary xenograft tumors growing in the cecumofNOD-SCID mice (Fig. 3A). We observed a significant effect ofoxaliplatin on the proliferation of PDX from patient 33 anda trend in those derived from patient 5. In addition, apo-ptosis was increased in tumor xenografts frompatient 5 andno effect was observed in those from patient 33.

We also used our colorectal cancer metastasis modeland PET to evaluate the response to API2, a target-directeddrug that inhibits AKT activity. We first evaluated differentradiotracers in our mouse model to observe tumor growthand metastasis by PET (Supplementary Fig. S4). 18F-FLTpresented a very high uptake by normal intestinal muco-sa, especially in the colon. In addition, 18F-FLT is highlyexcreted into the urine producing a high radioactivity

Tab

le3.

Clinical

andhistop

atho

logica

lfea

turesof

patientswith

derived

orthotop

icmou

semod

els

Treatmen

t

Patient

sample

Age/se

xSite

Typ

eGrade

TNM

Drugs

Tim

e(m

onths

)Fo

llow-up

(months

)Relap

setime

(months

)M1

loca

lization

Outco

me

P1

75/M

Right

colon

Muc

inou

sG3

pT4

aN0

Cap

acetibine

632

0n.d.

Alive

P2

79/F

Leftco

lon

Con

ventiona

lG3

pT4

aN1

n.a.

–6(day

s)0

n.d.

Dea

dP3

78/F

Right

colon

Con

ventiona

lG2

pT3

N1a

n.a.

–28

0n.d.

Alive

P5

73/F

Right

colon

Muc

inou

sG3

pT4

aN0

DeGramon

t6

250

n.d.

Alive

P6

83/F

Right

colon

Muc

inou

sG3

pT4

aN2b

n.a.

–10

5Live

r,l.n

.,ad

.gland

Dea

dP33

72/M

(Rec

tum)L

iver

met.

Muc

inou

sG2

pT4

aN0

Folfo

x5

2016

Live

rAlive

Histologica

ldifferen

tiatio

ngrad

e(G1–

G2,

well/m

oderate;G3–

G4,

poo

r/un

differen

tiated).Tu

mor-nod

e-metas

tasisstatus

.pT3

–pT4

,inv

asivetumors(pT3

,sub

serosa

;pT4

a,se

rosa

);pN0,no

maligna

ntlymphno

des;pN1,1–

3lymphno

demetas

tasis;pN1a

,metas

tasisinon

eregion

allymphno

de;pN

2b,m

etas

tasisin7or

moreregion

allymphno

des

.Liver

met,liver

metas

tasis;

n.a.,n

otap

plicab

le;n

.d.,no

tdeterminate;

l.n.,lymphno

de;a

d.g

land

,adrena

lgland

.

Puig et al.





concentration at the level of the bladder. Its high uptake byintestinal mucosa and bladder made it impossible to iden-tify or localize the tumor. The radiotracers 11C-choline and11C-methionine provided equivalent results showingintense uptake in the liver and kidneys, and very low uptakein tumoral tissue. Such a low signal combinedwith the highuptake present in abdominal organs, compromised theprecise delimitation of the tumoral mass that was onlytherefore possible in a few cases.PET with 18F-FDG allowed us to observe tumors in all

cases that presented a detectable abdominal mass bymanual palpation (Supplementary Fig. S4). In healthyanimals, this radiotracer presents a very low uptake at thelevel of the abdomen. However, the urinary bladder accu-mulated significant levels of radioactivity. For this reason,a precise evaluation of tumors in close contact with thisorgan could be difficult. The signal provided by thisradiotracer was heterogeneously distributed or localizedin a rim region at the level of the external layer of thetumor. In contrast, the inner part of the tumor masspresented a very low uptake, even lower than that observedin nontumoral tissues (Supplementary Fig. S4). The con-ventional adenocarcinoma derived from patient 2 showeda higher uptake than mucinous tumors from patient 1,coinciding with their differences in cellularity (Supple-mentary Figs. S1A and S4). Large liver metastases orcarcinomatosis were detected by PET when located dis-tantly from the primary tumor growing in the cecum wall(Fig. 2A and Supplementary Fig. S4B).We then performed a longitudinal study by PET with 18F-

FDG to evaluate the effect of an AKT inhibitor, API-2, ontumor growth. Mice with primary carcinomas generatedfrom patient 1, reduced tumor growth upon long-termtreatment with API-2 every second day (Fig. 3B and C). Atthe end of the experiment, tumor signal was reduced in API-2 cohort compared with vehicle-treated mice. On the con-trary, no significant effect on tumors derived from resistantpatient 2, was observed upon API-2 treatment (Fig. 3D). A

nonstatistically significant reduction of proliferation andincrease in apoptosis was observed in the tumor xenograftsof API-2-treated mice injected with cells derived frompatient 1, as evaluated by histological inmunofluorence ofKi67 and cleaved caspase 3 respectively (data not shown).

DiscussionTo overcome resistance to conventional treatments,

numerous drugs blocking specific molecular targets havebeen developed over the last decade and are currently beingtested on patients with different tumor types. It is becomingevident that the identification of robust biomarkers topredict response to treatment is essential for the success ofclinical trials with target-directed drugs. Although some ofthese novel drugs are showing promising results inadvanced cancer, resistance is frequent in most metastaticcolorectal patients. There is therefore an urgent need forpreclinical models that permit the testing of the efficacy ofthis new generation of target-directed drugs and validationof biomarkers of response.

Historically, most preclinical studies have been based oncell line models in vitro and in vivo. However, the long-termgrowth of cell lines selects a homogenous population that isthe most efficient in proliferating in a particular culturecondition. One of the most important defects of suchmodels is that the response to antitumoral drugs is notrepresentative of what actually occurs in heterogeneoushuman carcinomas (15). It is well accepted that intratu-moral heterogeneity occurs with respect to a variety ofbiological, biochemical, and immunological properties(30). These properties determine the ability of particularcancer cell subpopulations to emerge from the primarytumor and establish metastatic growth within distantorgans (31). Furthermore, resistance to specific treatmentscan also be innate in some genetic subclones present inpolyclonal colorectal carcinomas (23).

The preservation of patients’ intratumoral heterogeneityat the cellular and genetic levels ismajorly improved in PDX

Table 4. Incidence of xenograft implantation and metastasis in orthotopic colorectal cancer PDX

Subcutaneous xenograft Cecum xenograft

No. cellsIncidence

Patient No. cells Incidence Passages Cecum Carc. Lung Liver

P1 1 � 105 33/33 9 1 � 106 8/8 7/8 4/8 5/8P2 1 � 105 41/41 8 1 � 106 6/6 0/6 0/6 0/6P3 1 � 105 13/13 4 1 � 106 (F) 3/5 1/3 1/3 0/3P5 1 � 105 54/54 10 1 � 106 8/9 8/8 1/8 2/8P6 1 � 105 7/7 2 1 � 106 5/5 n.d. n.d. n.d.P33 2.5 � 105 10/10 3 1 � 106 8/8 0/8 0/8 0/8

Abbreviation: F, frozen.Table indicating on left, number of cells subcutaneously injected, incidence of tumor initiation, and number of passages in NOD-SCIDmice; and on right, incidence of mice with metastasis upon injection into the cecum wall of cells derived from different colon cancerpatients.






models compared with preclinical mouse models based oncancer cell lines. Therefore, PDX are currently becoming thebest preclinical models to test drug response.

We have established a circuit to derive cancer cells justafter surgical removal of colorectal tumors. We first disag-gregate the patient’s tumor piece, prepare a suspension

of single cells, and then subcutaneously inject a minimumof 1 � 105 viable cells in NOD-SCID mice, all within lessthan 24 hours. This procedure has resulted in an 87.5%tumor take rate, which is higher than the average 60%previously described in colorectal cancer PDX collections(7–9). Most of the laboratories that have generated similar

Figure2. Patient-derivedcells preservemetastatic potential. A, representativemacroscopic views into theabdominal cavity of 2mice (Ms1andMs2) injected inthe cecum with colon cancer cells derived from patient 1 (P1) or derived from patient 2 (P2). Ce, cecum xenograft; M, metastasis; C, carcinomatosis.Dashed lines delineate primary xenograft tumor grown in the cecum wall. B, a panel of H&E staining of the corresponding carcinomatosis, lung, and livermetastasis generated in mice injected with cancer derived from P1 and P3. 1, tumoral area; 2, necrotic area; L, liver; T, tumor. Dashed line delineatemetastasis in liver parenchyma. Scale bar, 1mm inpictures in columns 1, 3, 5, and 200mm inmagnifications in columns 2, 4, and 6. C, lung and livermetastasisin mice injected with cells derived from P1 were immunostained for Cytokeratin 20 (CK20, red), and epithelial cell adhesion molecule (EpCAM, red).Representative confocal pictures are shown. Scale bar, 200 mm.

Puig et al.





Figure 3. Responseof orthotopic colorectal cancer PDX tochemotherapy andAPI2 treatment. A (top), cecumprimary xenograft tumors developed fromP5andP33 treated with oxaliplatin or vehicle were immunostained for Ki67 (red) and Caspase-3 (green). Representative confocal pictures are shown. Scale bar, 100mm; (bottom) column scatter plot showing the amount of Ki67 (proliferation) or caspase-3 (apoptosis) in cecum xenografts from P5 and P33 treated withoxaliplatin (red) or vehicle (green). Horizontal lines indicate arithmetic mean values, and error bars show the 95%CI. Asterisk indicates a significant difference(P < 0.05) between groups as quantified by unpaired t test withWelch correction. r.u., relative units. B, representative 18FDGPET images of mice injectedwithcells derived from P1, obtained at baseline and after 10, 30, and 60 days of vehicle or API-2 treatment. H, heart; B, urinary bladder. Dashed lines delineateprimary xenograft tumors growing in the cecum wall. C, plot representing the evolution of tumor xenograft volume growing in mice injected with cellsderived from P1 and treated during 48 days with API-2 or vehicle. D, plot showing the final tumor xenograft volume after 48 days of treatment with API-2 orvehicle inmice injectedwith cells derived from P1 (green bars) or P2 (red bars). Statistical significance was evaluated by unpaired t test withWelch correction.






collections of PDXmodels transplant a piece of tumor tissuein immunodeficient mice (7, 8). Intact human tumor tissueoften contains large necrotic areas thus the number of viablecells implanted is unknown and engraftment could becompromised. Such inaccuracy would lead to a more var-iable tumor initiation efficacy, latency time, and growthrate, making it more difficult to compare between differentexperiments or even individual mice, as well as complicat-ing the setup of robust experiments to test the activity ofantitumoral agents (11). Furthermore, freezing pieces oftumor tissue compromises cell viability more than proto-cols with single cells. In our models, we can freeze andtherefore perpetuate each patient sample for future testswhen required, better preserving their capacity to reinitiatean equivalent xenograft tumor.

In addition, different tumor pieces from the same patientcould be enriched for particular genetic or epigenetic sub-clones and their separated implantation would generatetumor xenografts with different biological properties. Toovercome this potential bias, we inject a suspension ofdisaggregated single cells ensuring a better representationof the original patients’ intratumoral heterogeneity.

Usingour cell suspensionprotocolweobserved a reducedtumor take rate from patients that presented no invadedlymph nodes (N0) or thosewith lower differentiation grade(G1). Similar correlations have been reported in othercolorectal cancer PDX collections, indicating that lessaggressive tumors have a reduced capacity to engraft inimmunodeficient mice (7, 25).

The use of single cells disaggregated from patient tumortissue, opens upmany experimental avenues such as sphereculturing to test drug-response in vitro, or to purify (FACS)and study the biological characteristics of different cellsubpopulations present in heterogeneous colorectal carci-nomas. We have also observed that cancer initiation poten-tial is preserved in sphere cultures of tumor cells derived formost of the patients with colorectal cancer. The PDX gen-erated from cultured cells also preserve pluripotencybecause they recapitulate the same cell heterogeneity andhistopathological traits as the original patients’ colorectalcarcinoma (data not shown).

At the genetic level, our protocol permits the generationof tumor xenografts that present the samemutational statusas the original patients’ carcinoma. It would therefore allowthe study of the correlation between drug-responseobserved in our mouse models with the mutations presentin each particular patient. Twelve original patient carcino-mas and PDX pairs were genotyped presenting the samemutational pattern. Only in 2 cases was the allele frequencyincreased for a particular oncogenic mutation in the PDXversus the original patient’s sample (data not shown).

We could detect the most frequent mutations in genesclassically altered in colorectal cancer such us KRAS,PIK3CA, BRAF, or APC. The analysis of APC is clearlyincomplete because the Sequenomplatformonly permittedtesting some of the most frequent single nucleotide muta-tions described. Sequencing all APC exons should be per-formed to identify any possible mutations affecting this

tumor suppressor gene, which is essential for colorectalcancer carcinogenesis.

The genetic data generated demonstrates that our PDXmodels faithfully represent the main genetic characteristicsof patients with colorectal cancer (7). The percentage ofcases mutated for each particular gene (e.g., KRAS) in ourcollection differs from those previously described in largercollections of patients with colorectal cancer (29). Suchdiscrepancy could be because of the fact that our collectionis enriched in advanced tumors and liver metastasis.

Using the Nanostring platform with a discrete panel ofgenes, PDX sampleswere also evaluated for gene expression.The gene expression study allowed the clustering of PDXwith similar profiles. All PDX derived frommucinous coloncarcinomas clustered together, similarly to the results pre-viously shown in larger collections of colorectal cancersamples (29). We could even observe that samples P10 andP33, corresponding respectively to the primary rectal tumorand the liver metastasis of the same patient, clusteredtogether and separately from the rest of the PDX analyzed.Similar studies with microarrays have demonstrated thatgene expression profiles from original patients’ carcinomasand different PDX passages in mice cluster. Such dataevidences that the control of gene expression patterns ismostly a tumor cell autonomous trait. Such reproducibilityand cell autonomous behavior in PDX models wouldfacilitate the future use of gene expression profiles as pow-erful biomarkers to predict drug response or tumorprogression.

In all cases, PDX models recapitulate the same histologyas the original patients’ carcinoma. Therefore, tumor archi-tecture seems to be a cell autonomous trait mainly inde-pendent of the accompanying stroma. Such characteristic ofcancer cells has also been reported in other PDX collectionsof colorectal cancer tumors where patients’ stroma isreplaced by an equivalent mouse stroma. Such capacity oftumor cells to educate the host stroma reinforces thestrength of PDX as cancermodels that faithfully recapitulatehuman disease. Thus, PDX models are the optimal preclin-ical approach to test target-directed therapies that could alsoaffect the tumor stromal component.

Becausemetastatic colorectal cancer is currently lethal forthe vast majority of patients, most therapeutic efforts arenow focused on testing new target-directed drugs toimprove their survival. It has therefore become crucial tostudy the antitumoral properties of such novel drugs inmodels that reproduce advanced human disease. Aiming toaddress such an urgent need, we have developed a mousemodel of colorectal cancer metastasis with patient-derivedcells that recapitulates human advanced disease with greatprecision. It could therefore become the gold standardpreclinical model to test new target-directed drugs or tovalidate potential biomarkers of tumor progression orresponse to treatment. We inject a suspension of viablesingle cells derived frompatients with colorectal cancer intothe cecum wall of inmunodeficiente NOD-SCID mice.Primary tumors that grow in the cecum present the samehistopathological features as the original patients’

Puig et al.





carcinoma. We have observed peritoneal carcinomatoses,liver and lung metastases. Lung metastasis was observed in23% of the injected mice, whereas the liver was affected in27% of all animals. Such incidences are similar to thoseobserved in patients presenting 10% to 20% lungmetastasisand 20% to 70% liver metastasis (1).We injected a suspension of single cells instead of attach-

ing a piece of tumor tissue to cecum wall. Tumor tissuepieces attached to the serosal side of the cecum couldpresent a reduced capacity to reach vascular and lymphaticsystems located in the submucosa. That would explain thefailure of thismodel to generate lungmetastasis. Contrarily,cancer cell injection directly under the intestinal mucosawould facilitate their access to the abundant vasculature thatfeeds the cecum submucosal layer, prompting cell dissem-ination and distant metastasis to both the lung and liver.Mice orthotopically injected with cells derived from patient1 developed numerous distant metastases and presentedtumoral cells detached from the primary xenograft tumor inthe cecum that were invading local vasculature (data notshown).We observed that cells derived from primary mucinous

adenocarcinoma of patients 1 and 5 have a high metastaticpotential. It is generally recognized that mucinous tumorsof the colon have a worse prognosis than nonmucinouscarcinomas. The mucinous adenocarcinomas have a higherpenetration rate, increased lymph node invasion, less pro-tective lymphocyte infiltration in tumor margins, highermetastatic potential, and lower 5-year survival rate (32, 33).Furthermore, also in accordance with our results, in vivostudies in experimental animals demonstrated increasedtumorigenicity and metastatic potential of tumors derivedfrom cell lines that produced large amounts of mucin (34–36).Curiously, cells derived from the liver metastasis sam-

ple P33 did not generate distant metastases when injectedinto the cecum of NOD-SCID mice. It is possible thattumor cells growing in the patient’s liver acquired atransient capacity to escape the primary tumor but havelost such potential after homing in the new tissue. Thistransient behavior has already being observed in othermetastasis models (37).Although we observed some similarities between the

evolution of PDX mouse models and the clinical progres-sion of the corresponding patients, a large-scale study willbe required to prove that our orthotopic PDX mousemodels completely recapitulate colorectal cancer progres-sion in patients. It would be essential to reproduce inparallel in the mouse models equivalent adjuvant che-motherapy treatments as those received by patients. Ourresults with 6 patients with colorectal cancer provide thefirst evidence regarding the potential usefulness of ourorthotopic PDXmouse model as a preclinical approach topredict disease progression of patients or drug responseand hopefully help to better tailor treatment in thecoming future.We observed that oxaliplatin treatment was able to pre-

vent metastasis in our mouse models injected with cells

derived from patient 5. Proliferation was reduced andapoptosis increased in the primary cecal tumors of micetreatedwith oxaliplatin chemotherapy. Curiously, althoughpatient 5 was not treated with oxaliplatin she followed a DeGramont adjuvant chemotherapy (5-fluorouracil þ folicacid) andhas not relapsedwithmetastasis during the last 25months after surgery (Table 3). Our results are similar toprevious reports with subcutaneous PDX responding tochemotherapy (7), but our model also permits to evaluatethe antimetastatic effect of treatment contrarily to subcuta-neous PDX which do not produce metastases.

To test the validity of our model for investigating theactivity of target-directed drugs, we treated cecum-injectedmice with API2, an AKT inhibitor with proven antitumoralactivity (38, 39). API2 reduced tumor growth of mucinous(patient 1) but not conventional (patient 2) adenocarcino-mas.We recently described nuclear b-catenin as responsiblefor this differential response to treatment. High nuclearb-catenin content in patient 2 would confer resistance toFOXO3a-induced apoptosis promoted by PI3K or Akt inhi-bitors, whereas mucinous carcinomas from patient 1 thataccumulate low nuclear b-catenin amounts are sensitive(22). The evaluation of tumor response to API2 treatmentwas based in PET measurements.

Conventional noninvasive imaging methods, includingPET, that are routinely used in the clinic, have been adjustedto monitor tumor initiation, progression, and response totherapy in mouse models of cancer (40, 41). However, weare the first to use PET to monitor tumor xenografts fromcolon cancer patient-derived cells growing in the cecum ofimmunodeficient mice. From our studies, we can concludethat PET with 18F-FDG permits the evaluation of primarytumor localization, growth, and development of distantmetastases or quantify response to treatment in our mousemodel of advanced colorectal cancer with patient-derivedcells. Indeed, PET is used in the clinic to detect the presenceof metastasis or the response to treatments in patients withcolon cancer (42).

PDX models open a promising avenue for precisiononcology because functional assays could be performedwith cells derived from patients with colon cancer uponsurgery. For instance, patients at early stages of disease thatundergo surgery to remove a primary carcinoma and thatcould relapse and develop metastasis in the future, couldpotentially benefit from these assays to test the metastaticcapacity of their cells derived from the primary carcinoma.At the same time, assays to evaluate the response of xeno-grafted tumors derived from a particular patient to antitu-moral drugs could be performed just upon initial surgery orwhen relapse occurs. Such assays would provide valuableinformation about functional resistance or sensibility toapproved biologics such as anti-EGFR drugs, or experimen-tal drugs such as PI3K, AKT, MEK, or BRAF inhibitorscurrently tested in early-stage clinical trials. This functionalinformation is perfectly complementary to the particularmutation status of each patient with colon cancer, betterguiding therefore oncologists to select the best targetedtherapy.






ConclusionsWepresent an improved procedure to generate colorectal

cancer PDXwith a high tumor take rate that has allowed thegeneration of a rapidly expanding collection of models forfurther preclinical studies. The gene expression profiling,genotyping, and histological description of each PDXwould allow the study of the strength of new biomarkersof tumorprogressionor response to treatment, aswell as testthe efficacy of new target-directed drugs.

We have also generated a mouse model of metastaticcolorectal cancer with patient-derived cells that allowsevaluating their metastatic potential and response to treat-ment. Using this approach we have generated a model thatalso recapitulates advanced human disease at the geneticand histopathological level.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: I. Puig, J.R. Herance, J. Jimenez, J. Tabernero,S. Rojas, H.G. PalmerDevelopment of methodology: I. Puig, I. Chicote, S.P. Tenbaum, J.R.Herance, J. Jimenez, K. Caci, J. Tabernero, S. Rojas, H.G. Palmer

Acquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): I. Puig, I. Chicote, J.R. Herance, J.D. Gispert,S. Landolfi, K. Caci, L. Mendizabal, R. Charco, A. Prat, M. Elena Elez,G. Argiles, A. Vivancos, J. Tabernero, S. RojasAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): I. Puig, S.P. Tenbaum, O. Arques,J.D. Gispert, J. Jimenez, S. Landolfi, K. Caci, A. Prat, J. Tabernero, S. Rojas,H.G. PalmerWriting, review, and/or revision of themanuscript: I. Puig, J.R. Herance,J.D. Gispert, J. Jimenez, S. Landolfi, E. Espin, M. Elena Elez, G. Argiles,J. Tabernero, S. Rojas, H.G. PalmerAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): I. Chicote, S. Landolfi, D. Moreno,R. CharcoStudy supervision: J. Jimenez, J. Tabernero, S. Rojas, H.G. Palmer

AcknowledgmentsExperiments were supported by a VHIO starting grant and grants from

Fondode Investigaciones Sanitarias–Instituto de SaludCarlos III (ISCIII; FIS-PI081356, FIS-PI11/00917 andRD12/0036/0012). I. Puigwas fundedby theFundaci�on CientUfica de la Asociaci�on Espa~nola Contra el Cancer (AECC),S.P. Tenbaum was supported by a Fundaci�o Olga Torres Fellowship, and H.G. Palmer was supported by the Miguel Servet Program, ISCIII.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received May 30, 2012; revised August 14, 2013; accepted September 25,2013; published OnlineFirst October 29, 2013.

References1. JemalA,BrayF,CenterMM, Ferlay J,WardE, FormanD.Global cancer

statistics. CA Cancer J Clin 2011;61:69–90.2. Janne PA, Mayer RJ. Chemoprevention of colorectal cancer. N Engl J

Med 2000;342:1960–8.3. Van Cutsem E, Kohne CH, Hitre E, Zaluski J, Chang Chien CR,

Makhson A, et al. Cetuximab and chemotherapy as initial treatmentfor metastatic colorectal cancer. N Engl J Med 2009;360:1408–17.

4. Van Cutsem E, Kohne CH, Lang I, Folprecht G, Nowacki MP, CascinuS, et al. Cetuximab plus irinotecan, fluorouracil, and leucovorin as first-line treatment for metastatic colorectal cancer: updated analysis ofoverall survival according to tumor KRAS andBRAFmutation status. JClin Oncol 2011;29:2011–9.

5. NormannoN, Tejpar S,Morgillo F, De LucaA, VanCutsemE,CiardielloF. Implications for KRAS status and EGFR-targeted therapies inmetastatic CRC. Nat Rev Clin Oncol 2009;6:519–27.

6. Walther A, Johnstone E, Swanton C, Midgley R, Tomlinson I, Kerr D.Genetic prognostic and predictive markers in colorectal cancer. NatRev Cancer 2009;9:489–99.

7. JulienS,Merino-TrigoA, Lacroix L, PocardM,GoereD,Mariani P, et al.Characterization of a large panel of patient-derived tumor xenograftsrepresenting the clinical heterogeneity of human colorectal cancer.Clin Cancer Res 2012;18:5314–28.

8. Bertotti A, Migliardi G, Galimi F, Sassi F, Torti D, Isella C, et al. Amolecularly annotated platform of patient-derived xenografts ("xeno-patients") identifies HER2 as an effective therapeutic target in cetux-imab-resistant colorectal cancer. Cancer Discov 2011;1:508–23.

9. Ricci-Vitiani L, Lombardi DG, Pilozzi E, BiffoniM, TodaroM, PeschleC,et al. Identification and expansion of human colon-cancer-initiatingcells. Nature 2007;445:111–5.

10. O'Brien CA, Pollett A, Gallinger S, Dick JE. A human colon cancer cellcapable of initiating tumour growth in immunodeficient mice. Nature2007;445:106–10.

11. PocardM, Tsukui H, SalmonRJ, Dutrillaux B, PouponMF. Efficiency oforthotopic xenograft models for human colon cancers. In vivo (Athens,Greece) 1996;10:463–9.

12. Rashidi B, Gamagami R, Sasson A, Sun FX, Geller J, Moossa AR, et al.An orthotopic mouse model of remetastasis of human colon cancerliver metastasis. Clin Cancer Res 2000;6:2556–61.

13. Bikle DD, Munson S, Mancianti ML. Limited tissue distribution of theintestinal brush border myosin I protein. Gastroenterology 1991;100:395–402.

14. Cespedes MV, Espina C, Garcia-Cabezas MA, Trias M, Boluda A,Gomez del Pulgar MT, et al. Orthotopic microinjection of human coloncancer cells in nude mice induces tumor foci in all clinically relevantmetastatic sites. Am J Pathol 2007;170:1077–85.

15. Diehn M, Cho RW, Lobo NA, Kalisky T, Dorie MJ, Kulp AN, et al.Association of reactive oxygen species levels and radioresistance incancer stem cells. Nature 2009;458:780–3.

16. AideN, Poulain L, BriandM,Dutoit S, Allouche S, Labiche A, et al. Earlyevaluation of the effects of chemotherapywith longitudinal FDG small-animal PET in human testicular cancer xenografts: early flare responsedoes not reflect refractory disease. Eur J Nucl Med Mol Imaging2009;36:396–405.

17. Brepoels L, De Saint-Hubert M, Stroobants S, Verhoef G, Balzarini J,Mortelmans L, et al. Dose-response relationship in cyclophospha-mide-treated B-cell lymphoma xenografts monitored with [18F]FDGPET. Eur J Nucl Med Mol Imaging 2010;37:1688–95.

18. Allen NE, Appleby PN, Roddam AW, Tjonneland A, Johnsen NF,Overvad K, et al. Plasma selenium concentration and prostate cancerrisk: results from the European Prospective Investigation into Cancerand Nutrition (EPIC). Am J Clin Nutr 2008;88:1567–75.

19. Waldherr C,Mellinghoff IK, Tran C, Halpern BS, Rozengurt N, Safaei A,et al. Monitoring antiproliferative responses to kinase inhibitor therapyin mice with 30-deoxy-30-18F-fluorothymidine PET. J Nucl Med2005;46:114–20.

20. Brepoels L, Stroobants S, Verhoef G, De Groot T, Mortelmans L, DeWolf-Peeters C. (18)F-FDG and (18)F-FLT uptake early after cyclo-phosphamide and mTOR inhibition in an experimental lymphomamodel. J Nucl Med 2009;50:1102–9.

21. Geiss GK, Bumgarner RE, Birditt B, Dahl T, Dowidar N, Dunaway DL,et al. Direct multiplexed measurement of gene expression with color-coded probe pairs. Nat Biotechnol 2008;26:317–25.

22. Tenbaum SP, Ordonez-Moran P, Puig I, Chicote I, Arques O, LandolfiS, et al. b-catenin confers resistance to PI3K and AKT inhibitors andsubverts FOXO3a to promote metastasis in colon cancer. Nat Med2012;18:892–901.

Puig et al.





23. Kreso A, O'Brien CA. Colon cancer stem cells. Curr Protoc Stem CellBiol 2008;Chapter 3:Unit 3 1.

24. Todaro M, Alea MP, Di Stefano AB, Cammareri P, Vermeulen L, IovinoF, et al. Colon cancer stem cells dictate tumor growth and resist celldeath by production of interleukin-4. Cell Stem Cell 2007;1:389–402.

25. Marangoni E, Vincent-Salomon A, Auger N, Degeorges A, Assayag F,de Cremoux P, et al. A new model of patient tumor-derived breastcancer xenografts for preclinical assays. Clin Cancer Res 2007;13:3989–98.

26. Nemati F, Sastre-Garau X, Laurent C, Couturier J, Mariani P, Desjar-dins L, et al. Establishment and characterization of a panel of humanuveal melanoma xenografts derived from primary and/or metastatictumors. Clin Cancer Res 2010;16:2352–62.

27. Sadanandam A, Lyssiotis CA, Homicsko K, Collisson EA, Gibb WJ,Wullschleger S, et al. A colorectal cancer classification system thatassociates cellular phenotype and responses to therapy. Nat Med2013;19:619–25.

28. DeSousaEMF,WangX, JansenM,Fessler E, TrinhA, deRooij LP, et al.Poor-prognosis colon cancer is defined by a molecularly distinctsubtype and develops from serrated precursor lesions. Nat Med2013;19:614–8.

29. Network TCGA. Comprehensive molecular characterization of humancolon and rectal cancer. Nature 2012;487:330–7.

30. Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, GronroosE, et al. Intratumor heterogeneity and branched evolution revealed bymultiregion sequencing. N Engl J Med 2012;366:883–92.

31. Killion JJ, Fidler IJ. The biology of tumor metastasis. Semin Oncol1989;16:106–15.

32. Sasaki O, Atkin WS, Jass JR. Mucinous carcinoma of the rectum.Histopathology 1987;11:259–72.

33. Yamamoto S, Mochizuki H, Hase K, Yamamoto T, Ohkusa Y,Yokoyama S, et al. Assessment of clinicopathologic features ofcolorectal mucinous adenocarcinoma. Am J Surg 1993;166:257–61.

34. Bresalier RS, Niv Y, Byrd JC, DuhQY, ToribaraNW,Rockwell RW, et al.Mucin production by human colonic carcinoma cells correlates withtheir metastatic potential in animal models of colon cancermetastasis.J Clin Invest 1991;87:1037–45.

35. Schwartz B, Bresalier RS, Kim YS. The role of mucin in colon-cancermetastasis. Int J Cancer 1992;52:60–5.

36. Niv Y. Mucin and colorectal cancer. Isr Med Assoc J 2000;2:775–7.37. Monteiro J, Fodde R. Cancer stemness and metastasis: therapeu-

tic consequences and perspectives. Eur J Cancer 2010;46:1198–203.

38. Yang L, Dan HC, Sun M, Liu Q, Sun XM, Feldman RI, et al. Akt/proteinkinase B signaling inhibitor-2, a selective small molecule inhibitor ofAkt signalingwith antitumor activity in cancer cells overexpressingAkt.Cancer Res 2004;64:4394–9.

39. Karst AM, Dai DL, Cheng JQ, Li G. Role of p53 up-regulatedmodulatorof apoptosis and phosphorylated Akt in melanoma cell growth, apo-ptosis, and patient survival. Cancer Res 2006;66:9221–6.

40. Lewis JS, Achilefu S, Garbow JR, Laforest R, Welch MJ. Small animalimaging. current technology andperspectives for oncological imaging.Eur J Cancer 2002;38:2173–88.

41. Weissleder R. Scaling down imaging: molecular mapping of cancer inmice. Nat Rev 2002;2:11–8.

42. Bomanji JB, Costa DC, Ell PJ. Clinical role of positron emissiontomography in oncology. Lancet Oncol 2001;2:157–64.

43. de Hoon MJ, Imoto S, Nolan J, Miyano S. Open source clusteringsoftware. Bioinformatics 2004;20:1453–4.






2013;19:6787-6801. Published OnlineFirst October 29, 2013.Clin Cancer Res Isabel Puig, Irene Chicote, Stephan P. Tenbaum, et al. Potential of Patient-Derived Colon Cancer Initiating CellsA Personalized Preclinical Model to Evaluate the Metastatic

Updated version

10.1158/1078-0432.CCR-12-1740doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2013/10/21/1078-0432.CCR-12-1740.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/19/24/6787.full#ref-list-1

This article cites 42 articles, 11 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/19/24/6787.full#related-urls

This article has been cited by 6 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/19/24/6787To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-12-1740

http://clincancerres.aacrjournals.org/content/suppl/2013/10/21/1078-0432.CCR-12-1740.DC1

http://clincancerres.aacrjournals.org/content/19/24/6787.full#ref-list-1

http://clincancerres.aacrjournals.org/content/19/24/6787.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/19/24/6787


home | clinical cancer research - a personalized preclinical … · 2013. 12. 9. · cancer...

Documents