technologies for high throughput single molecule...

TECHNOLOGIES FOR HIGH THROUGHPUT

SINGLE MOLECULE DNA SEQUENCING

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF BIOENGINEERING

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

Jerrod Joseph Schwartz

May 2009

UMI Number: 3364456

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality of the copy

submitted. Broken or indistinct print, colored or poor quality illustrations

and photographs, print bleed-through, substandard margins, and improper

alignment can adversely affect reproduction.

In the unlikely event that the author did not send a complete manuscript

and there are missing pages, these will be noted. Also, if unauthorized

copyright material had to be removed, a note will indicate the deletion.

®

UMI UMI Microform 3364456

Copyright 2009 by ProQuest LLC All rights reserved. This microform edition is protected against

unauthorized copying under Title 17, United States Code.

ProQuest LLC 789 East Eisenhower Parkway

P.O. Box 1346 Ann Arbor, Ml 48106-1346

I certify that I have read this dissertation and that, in my opinion, it

is fully adequate in scope and quality as a dissertation for the degree

of Doctor of Philosophy.

(Stephen R. Quake) Principal Advisor




(Zev Bryant)




(Mark Brongersma)

Approved for the Stanford University Committee on Graduate Studies.

<£L/. A?--,*'-"

m

Abstract

Next-generation DNA sequencing is rapidly accelerating biological research by per

mitting the inexpensive and routine analysis of genomes, transcriptomes, and interac-

tomes. Commercial instruments that sequence single DNA molecules are now capable

of generating 20-30 gigabases of sequence data per run, but technological advances

are required to further reduce costs, improve error rates, and increase throughput.

This dissertation focuses on developing the underlying technologies to address these

needs.

Single molecule sequencing-by-synthesis approaches employ a DNA polymerase

to sequentially incorporate fluorescently-labeled nucleotides into a surface-tethered

primer-template. One major bottleneck is the time required to image thousands of

fields of view after each nucleotide incorporation cycle. To maximize the amount

of data generated it is therefore critical to pack as many resolvable templates on

the surface as possible. Random deposition can at best achieve a density of «2

resolvable templates per square micron, so two new simple and scalable approaches

were developed using nanoparticle arrays and colloidal epitaxy to pattern surfaces at

up to 6-fold higher densities.

A comprehensive understanding of how DNA polymerases behave under different

conditions is also critical to optimize read length, coverage, and error rate. Single

iv

molecule measurements were used to make a detailed characterization of DNA repli

cation as a function of the template's secondary structure and the sequence context.

These data enable the measurement the intrinsic "speed limit" of DNA polymerase

for the first time by separating the burst synthesis rate from sequence-dependent

pausing.

Finally, the ability to use a thermophilic polymerase for single molecule sequencing

would offer a number of key advantages: improved enzyme heat stability, better ability

to incorporate nucleotide analogs, and the capacity to melt templates that are GC-

rich or have a high degree of secondary structure. To achieve this, colloidal lenses were

used to overcome the temperature limits of oil-immersion microscope objectives by

incorporating a focusing element in immediate proximity to an emitting fluorophore.

The optical system was completed by a low numerical aperture optic which can have a

long working distance and low light collection ability. As proof of principle, colloidal

lenses were used to measure real-time single molecule mesophilic and thermophilic

DNA polymerase kinetics at 23°C and 70°C using a 20X 0.5 NA air objective.

v

Acknowledgments

I thank my advisor Stephen Quake for his continuous support throughout my graduate

career. Steve is a brilliant scientist, a respected innovator, and a master motivator. I

appreciate his approach of leading by example and letting his passion for hard work

and quality results speak for themselves. Steve lets his students take ownership of

their own projects and gives them the freedom to innovate and work independently, all

the while stressing the importance of good scholarship, experimental virtuosity, and

critical thinking. I also thank my thesis committee members for thought-provoking

questions and insight: Zev Bryant, Mark Brongersma, WE Moerner, and KC Huang.

Most of this work was financially supported through a grant from the NIH NHGRI.

Additional funding was provided by the DARPA Center for Optofluidic Integration

and the Howard Hughes Medical Institute.

Many people have in some way contributed to my academic success, either by early

influence in my life or through more recent interactions. I will start by thanking both

of my parents, Janice and Gary, for helping foster my interest in science at a young

age. I also thank my brother Tyler and my sister Kimberly for their support over the

years. I thank Andre Marziali for introducing me to the field of "genome technology"

in 2001 and for being my constant advocate. I also thank Roger Donaldson for being

a great teacher, colleague, and friend over the last decade.

vi

I started my graduate career at the California Institute of Technology where Heun

Jin Lee, Brandon Birdwell, and Chris Lacenere introduced me to single molecule

DNA sequencing, for which I thank them. Chris was also great to have around

for his timeless wisdom and ability to keep me laughing. Brian Stoltz, Neil Garg,

and Carolyn Woodroofe synthesized the nucleotides used in Chapter 7. I also thank

Carl Hansen, Joshua Marcus, Mike Van Dam, Lin Zhu, Josh Klein, Ben Collins, and

Michael Torrice for their friendship during my time at Caltech.

At Stanford University I worked with Randy Stoltenberg, Ethan Townsend, and

Stavros Stavrakis. Randy and I collaborated on the nanoparticle array work in Chap

ter 3, Ethan wrote the microscope software used in most of this work, and Stavros

helped with the high temperature colloidal lensing experiments in Chapter 5. Over

the years I've had the opportunity to talk about science with a number of talented

Quake group members, including Frank Lee, Rafael Gomez-Sjoberg, Yinthai Chan,

Frederick Balagadde, Alan van Orden, Sebastian Maerkl, Yann Marcy, Piero Cas-

trataro, Yanyi Huang, Joshua Weinstein, Dmitry Pushkarev, Norma NefF, Mehmet

Fatih Yanik, Quy Tran, Paul Blainey, Doron Gerber, Mattias Meier, Jianbin Wang,

Aaron Streets, Richard White III, and Thomas Snyder. I also thank Mark Kwan,

Adam Grossman, Douglas Jones, and Craig Goergen for their friendship outside of

the lab. Finally I thank Sarah Macumber for her constant words of encouragement,

her regular reality checks, and her ability to always make me smile.

Jerrod Schwartz

Stanford, California

May 2009

vii

Contents

Abstract iv

Acknowledgments vi

1 Introduction 1

1.1 Background 1

1.2 Thesis Organization 3

1.3 Scientific Contributions 4

2 Single Molecule D N A Sequencing 5

2.1 Introduction 5

2.2 Single Molecule DNA Sequencing . 7

2.2.1 Sequencing-by-Synthesis via Stepwise

Base Incorporation 7

2.2.2 Sequencing with Nanopores 8

2.2.3 Transmission Electron Microscopy for

DNA Sequencing 9

2.2.4 Sequencing-by-Synthesis in Real Time 10

2.2.5 Sequencing-by-Synthesis with Force Spectroscopy 10

viii

2.3 Progress of Single Molecule Sequencing 11

3 Single Molecule Surface Patterning 13

3.1 Introduction 13

3.2 Random Molecular Deposition 16

3.2.1 Nearest-Neighbor Analysis 16

3.2.2 Monte Carlo Simulation 18

3.3 Fitting the PSF 21


3.4 Patterned Nanoparticle Arrays 23

3.4.1 The Poisson Limit 23

3.4.2 Methods 27

3.4.3 Experimental Results 29

3.5 Single Molecule Colloidal Epitaxy 33


3.5.2 Mass Transport Considerations 35

3.5.3 Methods 38

3.5.4 Experimental Results 39

3.6 Future Work for Colloidal Epitaxy 45

3.7 Super-Resolution: Breaking the Diffraction Limit 46

3.8 Throughput Comparison 50

4 The "Speed Limit" of D N A Polymerase 52

4.1 Introduction 52

4.2 Primer Extension with Fluorescently Labeled Nucleotides 53

IX

4.2.1 Methods and Results 54

4.3 Strand Displacement Synthesis Through a DNA Hairpin 59

4.3.1 Hairpin Design 60

4.3.2 Hairpin FRET Calibration 64

4.3.3 Single Molecule Kinetics Experiment 68

4.3.4 Discussion and Results 70

4.4 The Energy Landscape and Kinetics of

Hairpin Refolding 83

4.5 Future work 87

5 Single Molecule Colloidal Lensing 89

5.1 Introduction 89

5.2 Theory of Colloidal Lensing 91

5.2.1 Geometric Optics 91

5.2.2 Maxwell's Equations 92

5.2.3 FDTD Simulation Methods 94

5.2.4 FDTD Simulation Results 95

5.3 Single Quantum Dots as Rotational Probes 96

5.4 Enhanced Fluorescence Effects 102

5.5 Single Molecule Imaging with Colloidal Lenses 104

5.6 Measuring Polymerase Kinetics with Colloidal Lenses 108

5.6.1 Escherichia coli Pol I(KF) Activity 108

5.6.2 Thermococcus 9°N-7 Therminator Activity 112

5.6.3 Replication Rates Measured with Colloidal Lenses 116

5.7 Future work 116

x

6 SPR Enhanced TIRF Microscopy 118

6.1 Introduction 118

6.2 Methods 120

6.3 Results 123

6.4 Future work 127

7 D N A Polymerases and Nucleotide Analogs 128

7.1 Introduction 128

7.2 Polymerase Structures 129

7.2.1 Crystal Structure Quality 131

7.2.2 Sequence Alignment 134

7.3 Accomodating Nucleotide Analogs 136

7.4 Custom Nucleotide Variants 147

7.5 Longer Linkers with Internal Esters 148

7.6 Prospects 152

Appendix

A C + + Code for Monte Carlo Simulations 154

B MATLAB Code for Image Processing and Analysis 158

C Meep Code for FDTD Simulations 162

Bibliography 169

XI

List of Tables

2.1 Single molecule sequencing metrics 12

3.1 Theoretical throughput of a super-resolution approach 49

3.2 Parameters that govern the bandwidth of single molecule imaging . . 50

4.1 DNA oligonucleotide sequences for hairpin ligation 63

4.2 Nucleotide combinations used for stepping through the hairpin. . . . 68

5.1 Photon statistics for Ti02-enhanced fluorescence 104

5.2 Photon collection statistics for Ti0 2 colloidal lensing 107

7.1 Summary of crystallographic and refinement data, Part I 132

7.2 Summary of crystallographic and refinement data, Part II 133

7.3 Sequence alignment of six polymerases 135

xii

List of Figures

2.1 Sequence information obtained from single DNA molecules 8

3.1 The Rayleigh criterion 14

3.2 Weibull distribution for NN analysis 19

3.3 Monte Carlo simulation for random deposition 20

3.4 Simulated image of random deposition 20

3.5 Monte Carlo simulation for fitting the PSF 22

3.6 Simulated image of a patterned array 25

3.7 Monte Carlo simulation for patterned surfaces 26

3.8 AFM image of Au nanoparticle array 30

3.9 Micelle density vs. BCP concentration 31

3.10 Nearest-neighbor distances for Au nanoparticle arrays 32

3.11 Fill-factor for semi-ordered arrays 32

3.12 Cartoon of colloidal epitaxy 34

3.13 Monte Carlo simulation of colloidal epitaxy 36

3.14 Simulated images of colloidal epitaxy 37

3.15 Restriction enzyme cleavage sites for colloidal epitaxy 40

3.16 Restriction enzymes are sensitive to colloid proximity 41

xiii

3.17 Brightfield images and power spectrum of colloids on surfaces . . . . 42

3.18 Colloidal epitaxy experimental results 43

3.19 Colloidal epitaxy NN distances and surface densities 44

3.20 Super-resolution Monte Carlo simulations 47

3.21 Bandwidth comparison of deposition methods 51

4.1 Sequences used for single turnover experiments 54

4.2 Spectra of bulk single nucleotide incorporation experiment 55

4.3 Sample split-field image of Cy3/Alexa647 single primer/templates . . 56

4.4 Sample two-color trajectory showing single nucleotide incorporation . 57

4.5 "Race track" template for multiple dNTP incorporations 58

4.6 Single molecule FRET trajectory showing nucleotide incorporation . . 59

4.7 The internal Cy3 position influenced FRET efficiency 61

4.8 Gel purification of the ligated hairpin 65

4.9 FRET can used to identify the polymerase position 66

4.10 Partially extended primers permit FRET distance calibration 69

4.11 DNA replication exhibited heterogenous pausing 74

4.12 Pause frequency as a function of polymerase position 76

4.13 Histograms of S/N ratios for pauses and extensions 78

4.14 Sequence-dependent pause lifetimes for Pol I(KF) and 029 79

4.15 "Speed limit" replication rates for Pol I(KF) and 029 82

4.16 Average trajectories for polymerase molecules that did not pause . . . 84

4.17 Pause durations and intensities locations during hairpin refolding . . 85

4.18 The observable energy landscape of cruciform transitions 86

4.19 Structure of the DNA cruciform intermediates 88

xiv

5.1 Colloidal lenses for single fluorophore detection 93

5.2 Power calculated with the FDTD method 97

5.3 Colloidal lenses as rotational probes 100

5.4 Static imaging of single quantum dots 101

5.5 Enhanced fluorescence effects of Ti02 colloids 103

5.6 Static imaging of single fluorophores with colloidal lenses 106

5.7 Colloidal lensing of DNA polymerase activity I l l

5.8 Replication rates measured with colloidal lenses 117

6.1 Au-coated surfaces are substrates for SAMs 122

6.2 XPS spectra of an Au-coated surface 124

6.3 XPS spectra of 11-amino-undecanethiol on an Au-coated surface . . . 125

6.4 Cy3-dUTP quenching on Au-coated surfaces 126

6.5 Pol I(KF) incorporated Cy3-dCTP on a SAM-Au surface 126

7.1 2D structures for R6G-dGTP and Cy5-dCTP 136

7.2 Crystal structure of Taq polymerase 137

7.3 Crystal structure of the active site of Taq and T7 138

7.4 The active site of T7 polymerase 139

7.5 Crystal structure of HIV-1 reverse transcriptase 141

7.6 Tgo and Vent polymerases 142

7.7 Pol I(KF) with Cy5-dCTP bound 144

7.8 dUTP-17-E-Cy5 with an internal ester 148

7.9 Primer and template sequences used for nucleotide screening 149

7.10 Validation of dUTP-10-Cy5, Part I 150

7.11 Validation of dUTP-10-Cy5, Part II 150

xv

7.12 Cy5 nucleotides can be incorporated by Pol I(KF) 151

7.13 Nucleotides with ester-containing linkers can be cleaved 152

xvi

Chapter 1

Introduction

1.1 Background

The ability to isolate and detect the behavior of a single molecule represents the ul

timate limit of spectroscopy. Bulk measurements use ensemble averaging to provide

general insight into the dynamics of complex systems, but important details are often

lost due to subtle sample heterogeneity. The first single molecule detection experi

ments were reported two decades ago in crystals at near absolute zero temperatures

[1, 2, 3]. Since then the field has expanded to include room temperature measure

ments, making it is possible to explore single molecule dynamics of a wide variety of

systems [4, 5, 6, 7, 8, 9, 10].

Among the wide variety of applications for single molecule techniques, single

molecule DNA sequencing has caught the attention of both academia and indus

try [11, 12, 13, 14, 15, 16, 17]. Scientists are interested in using high throughput, low

cost sequencing technology for a wide range of applications including unlocking the

1

CHAPTER 1. INTRODUCTION 2

cancer genome [18], building an atlas of genetic variation for humans, gaining a bet

ter understanding of the immune system and disease susceptibility, gene expression,

and epigenetics. Pharmaceutical companies are interested in the technology for many

of the same reasons, as they believe that this wealth of information will help them

develop better preventions and treatments for disease. As the cost of whole human

genome sequencing and targeted sequencing continues to drop, more individuals will

be inclined and able to learn about their genetic makeup. This will eventually lead

to personalized medicine, which represents a major departure from the traditional

"one-size-fits-all" pharmaceutical model of producing one drug to treat a common

disease in otherwise diverse patients.

Reagent cost has traditionally been a limiting factor for large scale sequencing

projects, but the ability to sequence individual molecules would reduce reagent con

sumption and require smaller initial quantities of precious DNA samples. Further

more, single molecule sequencing potentially offers an unprecedented degree of paral

lelism with millions of DNA templates being interrogated simultaneously. Dephasing

is another important issue that ensemble sequencing approaches have to worry about:

because not every reaction is 100% efficient, some molecules fall out of phase at each

step and contribute spurious signals in subsequent steps. Sample preparation for cur

rent sequencers is also cumbersome and often requires DNA amplification steps to

generate sufficient signal for detection. This point is particularly important as ampli

fication has a tendency to introduce sequence-dependent bias into a sample, whereby

some regions are over-represented compared to others.

The unifying theme of this thesis is the development of new technologies to improve

the throughput single molecule DNA sequencing platforms. However, much of the


work can potentially be applied to other areas of biology, nanofabrication, and single

molecule spectroscopy as well. The thesis covers five broad areas: strategic surface

deposition techniques, measuring DNA polymerase kinetics, single molecule colloidal

lensing, surface chemistries for enhanced single molecule signals, and novel nucleotides

for sequencing-by-synthesis.

1.2 Thesis Organization

This thesis is organized as follows. Chapter 2 gives a brief overview of the current

and future single molecule DNA sequencing platforms being developed, along with

some discussion on the advantages and disadvantages of each. The remaining chap

ters describe various technologies designed to improve a particular aspect of single

molecule DNA sequencing.

Chapter 3 provides an in-depth look at the importance of strategic single molecule

surface patterning in high throughput applications. Two new approaches for pattern

ing single molecules on surfaces are presented, both of which offer improved resolvable

densities compared to random deposition. Chapter 4 discusses an approach for the

precision measurement of single DNA polymerase kinetics as a function of template

structure and sequence. DNA replication was followed in real-time through a hair

pin and a a form of sequence-specific pausing was characterized that was previously

thought to exist based on bulk experiments. Heterogeneous cruciform extrusion fol

lowing replication was also observed and the kinetics of this process were measured.

Chapter 5 presents the idea of using high index colloids as lenses to detect single flu-

orophores with long working distance, low NA microscope objectives. Having single

molecule sensitivity with a long working distance enabled high temperature single


molecule spectroscopy and the ability to measure the kinetics of a thermophilic DNA

polymerase. Chapter 6 describes preliminary work using thin metal films on surfaces

to improve the signal from nearby single fluorophores detected using through-the-

objective total internal reflection fluorescence microscopy. Single molecule sequencing

platforms often employ DNA polymerases to perform sequencing-by-synthesis reac

tions, and so Chapter 7 gives an overview of solved polymerase crystal structures and

discusses why they are able to incorporate unnatural substrates such as dye-labeled

nucleotides. In this Chapter a novel fluorescently-labeled nucleotide with a longer

and cleavable linker is also characterized. Appendices at the end of thesis provide

examples of some of the code used for image analysis, Monte Carlo simulations, and

finite difference time domain simulations.

1.3 Scientific Contributions

The work described in this thesis represents contributions in several fields. First, I

developed two new scalable approaches for patterning single molecules at densities up

to 6-fold higher than that achievable with random deposition (Chapter 3). This work

has been published [19] and one patent application is pending review [20], while a

second manuscript is in preparation [21] and a second provisional patent application

has been filed [22]. I also measured, for the first time, the intrinsic speed limit of DNA

polymerase by separating sequence-specific pausing from burst synthesis (Chapter 4).

This work has been submitted for publication [23]. Finally, I used colloidal lenses to

enable the first observation of single molecules at high temperature along with the

first real-time single molecule measurement of a thermophilic enzyme (Chapter 5).

This work is in preparation for submission [24].

Chapter 2

Single Molecule DNA Sequencing

2.1 Introduction

In 1977 Fred Sanger and colleagues reported using 2',3'-dideoxy nucleotides as chain

terminating inhibitors of DNA polymerase to perform DNA sequencing [25]. He was

awarded his second Nobel Prize in 1980 for his efforts, a fitting sequel to the Nobel

Prize he won in 1959 for determining the amino acid sequence of the protein insulin.

It wasn't until 1986 that this method was partially automated with fluorescent labels

[26], motivated in large part due to the increasing importance of sequence information

in molecular biology. Further streamlining of Sanger sequencing by Applied Biosys-

tems and Amersham Biosciences (now Life Technologies and General Electric Health

care, respectively) contributed to the release of the first draft of the human genome

in 2001 [27, 28]. Although this sequence was based on DNA from multiple anony

mous individuals, advances in sequencing technology over the last few years have

allowed people to have their own personal genomes sequenced [29, 30, 31, 32, 33].

Next-generation DNA sequencing will continue to accelerate biological research by

5

CHAPTER 2. SINGLE MOLECULE DNA SEQUENCING 6

enabling the routine and widespread analysis of genomes, transcriptomes, and inter-

actomes. Such endeavors used to only be attempted by production-sized teams of

researchers, but now it is possible for individual investigators to achieve this level of

throughput for reasonable cost.

The ability to quickly and rapidly resequence an individual human genome has

generated a lot of interest in both the scientific community and the general public.

The "1000 Genomes Project" is aiming to sequence the genomes of approximately

1200 people from around the world to develop a new map of the human genome. It

should provide a view of biologically relevant DNA variation unmatched by current

resources. This information will help us better understand cancer, genetic variation,

basic human biology, and disease, while setting the foundation for the development

of genome-based therapeutics and medicines. Consumer genetics companies are now

offering to genotype ^500,000 of an individual's single nucleotide polymorphisms

(SNPs) for a few hundred dollars, yet the challenge remains of how to correctly

interpret and act on the information.

Although the first draft of the human genome cost in excess of $100 million, the

personal genomes sequenced in the last few years have cost considerably less, in the

range of $250,000 to $1,000,000 [31, 32, 33]. One individual recently had his genome

sequenced using one of the platforms discussed in the next section for only $42,000. In

order to truly be affordable for the masses, however, further cost reductions are neces

sary. Single molecule DNA sequencing approaches have the potential of achieving the

NHGPJ's goal of a $1000 genome and are currently in various stages of development.

The leading contenders are discussed in the next section.


2.2 Single Molecule D N A Sequencing

The leading proposed and demonstrated single molecule sequencing technologies are

discussed below followed by a comparison of their demonstrated throughput and cost.

2.2.1 Sequencing-by-Synthesis via Stepwise

Base Incorporation

The ability to obtain sequence information from single DNA molecules was first

demonstrated by Stephen Quake's group in 2003 [12]. The approach was simple

in design and elegant in execution. Biotinylated DNA templates of known sequence

were annealed to a Cy3-labeled primer and attached to a streptavidin-coated glass

coverslip. The location of each Cy3-fluorophore, and therefore each primer/template,

was identified using a prism-based total internal reflection microscope. Next, the

Cy3 fluorophores were photobleached. Pol I(KF) and dUTP-Cy3 were then added

to the flowcell and single nucleotide incorporation took place on templates that had

a template adenine immediately adjacent to the primer (Figure 2.1). After washing

away unincorporated nucleotides, dark nucleotides (dATP and dGTP) were added to

extend the primer to the next template adenine or guanosine. The flowcell was again

washed, and dCTP-Cy5 was added and allowed to incorporate. Templates that incor

porated dCTP-Cy5 were identified using FRET between Cy3-Cy5. Cy5 fluorophores

were photobleached and this process was repeated to identify the order of appearance

of A and G in the template sequence.

Helicos Biosciences has since further developed this basic concept to allow for

consecutive identification of all four bases, read lengths over 30 bp, vastly improved


s Color of the illumination

, Incorporation events

-C»3

• C * 5

—?m v$i 3«i m mmwi c « 3 :

) 3 ' •

l I 1

i uu u

Figure 2.1: Sequence information obtained from single DNA molecules with FRET. The intensity trace from a single template is shown as a function of the presented nucleotides in (b), and the FRET efficiency is shown in (c). Adapted from [12].

parallelism, and phred quality scores up to 30 (99.9% accuracy). This platform has

been used to resequence the entire M13 genome [15] and is now available commercially.

The instrument is capable of generating up to 20-30 gigabases of sequence information

per run at a cost of « $14,000 per run. This instrument should help enable the routine

resequencing of human genomes and genomic changes in tumor samples.

2.2.2 Sequencing with Nanopores

Nanopore-based devices are capable of providing single molecule detection by elec-

trophoretically driving molecules in solution through a nano-scale pore [34]. The orig

inal hope for nanopore sequencing involved threading a single stranded DNA molecule


through the Staphylococcus aureus cn-haemolysin membrane protein pore under an ap

plied voltage. The ionic current passing through the pore can be recorded and, ideally,

each base would be serially identified by a characteristic decrease in current ampli

tude [35, 36, 37]. Individual bases can be identified on a static DNA molecule in a

nanopore [38] but the rate of DNA translocation is too fast for the required current

resolution unless the bases are chemically modified [39]. Recent work has focused

on using enzymes to reduce the rate of translocation [40, 41] or on the detection of

individual nucleoside monophosphates [17]. The latter approach might enable single

molecule sequencing by covalently coupling an exonuclease enzyme to the periphery

of the pore; as nucleotides are cleaved off a single strand of DNA, their identity can

be detected as they translocate through the pore. Although nanopore approaches

offer potentially the longest read lengths of all the single molecule methods, scaling

up the technology to sequence many strands in parallel remains a challenge.

2.2.3 Transmission Electron Microscopy for

DNA Sequencing

Directly "imaging" the sequence of DNA using transmission electron microscopy

(TEM) has been proposed by ZS Genetics Inc. [42] and others. The basic idea

is to synthesize DNA complementary to the target sequence using nucleotides labeled

with heavy elements with sufficient nuclear charge (e.g. iodine, bromine, osmium).

The labeled DNA can then be stretched on a substrate and the spatial locations of

the heavy elements can be identified via TEM. The potential for long read lengths

(5000-7000 bp) make this approach an exciting one to keep an eye on in the near

future.


2.2.4 Sequencing-by-Synthesis in Real Time

Many DNA polymerases are naturally highly processive enzymes with the ability to

synthesize hundreds or thousands of bases before dissociation. This has led some

groups to try to monitor DNA synthesis such that as each base is incorporated, its

identity can be read out optically in real-time. This requires that each dNTP is labeled

with a different fluorophore and that the nucleotides are present in solution at high

enough concentration so that synthesis proceeds at a reasonable rate. This can result

in significant background even with reduced excitation volume techniques like total

internal reflection (TIR). To overcome this limitation, zero-mode waveguides (ZMWs)

have been proposed a technology that will allow single molecule sensitivity even in the

presence of nanomolar or even micromolar concentrations of labeled dNTPs [13, 43].

Pacific Biosciences reported using 4>29 polymerase immobilized at the bottom of

ZMWs to sequence a 150 bp template with a raw error rate of 20% [16]. The com

pany places fluorophores on the 7-phosphate of the dNTPs so that incorporation by

polymerase naturally releases the dye. They plan to launch a commercial instrument

in the near future. Major challenges of this technology include improving the raw

error rate, improving the number of ZMWs containing single polymerase molecules,

and developing large and sensitive detectors to enable the monitoring of more ZMWs

simultaneously.

2.2.5 Sequencing-by-Synthesis with Force Spectroscopy

In 2006 Greenleaf and Block [44] demonstrated motion-based single molecule DNA

sequencing using force spectroscopy. A pair of optical traps was used to control

two polystyrene beads: one of the beads was tethered to a single RNA polymerase


molecule while the other was attached to the distal end of a primed DNA template.

As the RNA polymerase underwent transcriptional motion along the template, the

distance between the two beads increased and was measured with single base resolu

tion. Sequencing proceeded by adding a dNTP solution to the chamber with one of

the four dNTPs present at a very low concentration. When the polymerase encoun

tered a template base that was complementary to the limiting dNTP, a pause was

recorded. This process was repeated four times with a different limiting dNTP in

each case. Trajectories for RNAP for each mix were then superimposed and sequence

information was extrapolated on the temporal basis of pause locations.

While this is an interesting approach for single molecule sequencing, a number of

hurdles remain before it can be implemented on a large scale. Polymerase pausing

due to sequence-specific pausing [45] or misincorporation may result in significant

errors. The same template must also be sequenced four times, requiring that the

synthesized strand be melted off and a new primer annealed after each experiment.

Finally, the use of optical traps limits the number of molecules that can be observed

simultaneously to a range of 1-10.

2.3 Progress of Single Molecule Sequencing

Table 2.1 shows some of the key metrics for published single molecule sequencing

approaches. Over the last six years, single molecule DNA sequencing has shown

remarkable progress with over six orders of magnitude improvement in parallelism.

Longer read lengths remain an important area for improvement for all of the methods,

which will be necessary before they can be used for de novo whole genome sequenc

ing. The real-time sequencing approaches ([41] and [16]) currently offer the greatest


Table 2.1: Demonstrated single molecule sequencing metrics. Reference

[12] [44] [15] [16] [41] [46]

Year

2003 2006 2008 2008 2008 2009

Parallelism

100 1

280,000 449

1 500,000,000

Read Length

4 32 30 150 8

33

Throughput

0.5 bp/min 10 bp/min 25 bp/sec 500 bp/sec

2 bp/hr 25,000 bp/sec

Error Rate

4% 6%

0.5% 20% ??% 3.5%

potential for 1000+ bp reads. Chemistry improvements may lead to better raw error

rates, although these approaches make up for it with deep coverage. It will be exciting

to see if the parallelism and throughput of single molecule DNA sequencing platforms

can continue to advance at the current pace in the coming years. In the following

chapters, a number of new technologies are presented that might help further improve

the throughput of some of these platforms.

Chapter 3

High Density Single Molecule

Surface Patterning

3.1 Introduction

With the application of spectroscopy to studies that require the interrogation of many

molecules in parallel, there is growing interest in developing methods for strategically

patterning different molecules on surfaces at small fixed length scales. A number

of approaches have been developed to control the deposition of DNA molecules on

surfaces, including molecular combing by capillary flow [47, 48], casting solutions on

a surface pre-patterned with polydimethylsiloxane (PDMS) [49, 50], spin stretching

[51], and using a PDMS stamp inked with DNA for printing on mica [52]. Ordered

DNA arrays have been generated by drop projection [53, 54], microcontact printing

[55], deposition on surfaces pre-patterned by e-beam lithography [56], and dip-pen

nanolithography [57, 58]. These approaches typically either yield high densities over

small areas or low densities over large areas and are not well suited for single molecule

13

CHAPTER 3. SINGLE MOLECULE SURFACE PATTERNING 14

Figure 3.1: The Rayleigh criterion for two adjacent Airy disks is met when the center of one Airy disk is exactly aligned with the first minimum of the other. At this point, the two molecules are said to just be resolvable.

deposition. Nanopatterning of DNA has been demonstrated on large surfaces of

anodic porous alumina [59] and through the use of micron-sized beads to deposit

nanoscale DNA spots [60], but neither of these approaches has yet been demonstrated

for single molecule deposition. Optical traps have been used to deposit single colloidal

particles in a designed pattern [61] but this approach has not yet scaled to deposit

millions of features in a dense ordered array.

In order to pattern single molecules at the highest possible resolvable density, the

approach employed has to take into account the optical setup and the fluorophore be

ing imaged. The spatial resolution of a diffraction-limited microscope is traditionally

determined by the Rayleigh criterion [62]. The Rayleigh criterion is satisfied when

the center of the Airy disk of one point spread function (PSF) is superimposed on the

first minimum of the other (see Figure 3.1 and Equation 3.1). The diffraction-limited

resolution d,R depends on the numerical aperture (NA) of the microscope objective:

1.22A 2ra sin 9

(3.1)


where A is the wavelength of collected photons. The NA is a measure of how good

the objective is at collecting light and is defined as NA = nsin#, where n is the

refractive index of the interface and 6 is the half-collection angle. Due to limitations

on the values of 6, A, and n, the resolution limit of a light microscope using visible

light is on the order of RS 250 nm.

Another important consideration for wide-field single molecule imaging has to do

with the digital sampling of an analog signal, as in the case of detecting a collection

of point spread functions on a charge-coupled device (CCD). To address this funda

mental issue, Harold Nyquist developed a theorem in 1928 that states that in order

to reproduce an analog signal, the digital sampling rate must be at least twice the

frequency of the original signal [63]. This means that the detector should sample

each PSF with at least 2.5-3 pixels. The tradeoff of sampling over more pixels is a

decrease in the signal-to-noise per pixel and few PSFs can be detected simultaneously.

The majority of the images described in this chapter were collected with a 60X 1.45

NA objective on a CCD with 85 nm square pixels. For Cy3 fluorophores with peak

emission at A = 570 nm, the diffraction-limited resolution of the optical setup was

d,R = 240 nm, and it is therefore within the sampling limits of the Nyquist criterion.

The imaging throughput of any wide-field single molecule technique can be gen

erally described by Equation 3.2:

bp/sec - 0 f i m a g i n g area\ ( 1 ^ /3 2) \ L Or a r e a J y tmove i ''image /

In this equation 9 is the surface "fill factor", or the number of resolvable single

molecules per diffraction limited region (DLR), imaging area is the physical size of

the field of view, PSF area is the area occupied by a single molecule, tmove is the time


required to move the stage to a new field, and timage is the time required to take an

image. Any of these parameters can be optimized to improve throughput, and the

stage/image times have already been optimized by to some extent. However, there is

a need for easy, scalable approaches to maximize 6. The theoretical density limit for

this problem is the circle packing density limit whereby « 91% of the surface area

is occupied with resolvable molecules. While achieving this density over large areas

remains the holy grail of single molecule surface patterning, there is still much room

for improvement over simple random deposition. The following sections compare

random single molecule deposition to viable alternatives with improved fill factors.

3.2 Random Molecular Deposition

3.2.1 Nearest-Neighbor Analysis

Random single molecule deposition is typically achieved by incubating a solution of

target molecules over a reactive surface for some time to allow the molecules to bind

at random spatial locations. This is followed by a washing step to remove unbound

species. Molecules are free to bind to the surface wherever there are reactive groups

and the surface density is only limited by the incubation time, the binding affinity,

the target concentration, and possibly molecule-molecule interactions on the surface.

Nearest-neighbor (NN) analysis can be used to describe the distance from a point

to its closest neighbor for a random distribution of points. Consider a point population

of size N randomly situated in a region of area S. The NN distances for the population


follows a Weibull distribution [64]:

0 t = 0 / ( * , * ) = < (3-3)

2nSte-n5t2 t > 0

where / is the fraction of points with a nearest-neighbor distance t and 5 = N/S

is the point density. Figure 3.2a shows this distribution for a variety of different

S. For simplicity, the diffraction-limited resolution is defined to be at t = 1. Note

that throughout this chapter, equations and parameters will be defined unitless to

maintain broad applicability. The "resolvable density" is therefore given by point

density multiplied by the fraction of points with NN distances > 1, or

rt=oo I

S / /(<$, t)dt = 6 -e-*6*

= Se'wS

t=oo^

t = l

To find the maximum resolvable density, take the derivative with respect to 5:

i-A*'-*6) = s7*(r°)+*-*'

= e-*d(l-n5)

This function goes to zero when 5 = 7r_1, SO the maximum resolvable density occurs

when 8 = 7r_1 = 0.3183. For a surface at this density, the NN distribution can


therefore be written as:

/(*) = 2te-'2 (3.4)

To find the fraction of points with a NN distance > 1, we have to find the total area

under the curve and the area under the curve to the right of t = 1 (see Figure 3.2b).

/ •OO /-OO

/ f(t)dt = / 2te" Jo Jo

dt

= -€-*

= 1

/

oo />oo

f{t)dt = I 2te

tZdt

oo

= —e

1

This result states that e _ 1 = 0.3679 = 36.8% of the points have NN distances > 1

and that (1 — e_1) = 0.6321 = 63.2% of the points have NN distances < 1. One can

therefore expect that at the ideal density of 5 = n"1, only 9 = (7re)~ = 11.7% of the

available surface area will contain points with the closest NN > 1 unit away.

3.2.2 Monte Carlo Simulation

Monte Carlo simulations were run to confirm the distribution of NN distances for

a random set of points on a surface. Using custom software written in C++ (see

Appendix A for example code), a random XY coordinate was assigned to each of


2.0

1.5

) 1.0

0.5

0

B 00

J2te~'2dt i .

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

NN distance (r/a.u.) NN distance (r/a.u.)

(a) Weibull distribution for various 6. (b) Weibull distribution for 5 = TT~1.

Figure 3.2: (a) As the point density increases, the distribution shifts towards smaller NN values as intuitively expected, (b) For the 8 = n~l case, only the fraction of the population with t > 1 are resolvable.

N different molecules in a region containing S diffraction limited regions (DLRs)

at a density of 8 = N/S. The distance between each point and every other point

was calculated and points with a NN distance < 1 were flagged as unresolvable. To

eliminate boundary effects, points located within 2 units of any edge were not scored.

Of the remaining points, those with a NN distance > 1 were counted to give m. The

'fill factor" 9 is then defined as the fraction of the available DLRs containing points

with a NN distance > 1, so 9 = m/S. Simulations were run in a 75 x 75 unit area

with a point population ranging from 1 to 10000 at intervals of 100. The values of 9

and 8 were calculated for each simulation and the results are shown in Figure 3.3. A

sample field of view at the peak of from Figure 3.3 is shown in Figure 3.4.

As expected, the simulations show that the relationship between 9 and 8 approx

imates a Weibull distribution. The maximum value of 6 — 0.119 « (7re)~ when 8 =

0.32 « 7T_1, in good agreement with the theoretical predictions. Fitting the Weibull


0.14

0.12

0.10

0.06

0.04

0.02

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

6

Figure 3.3: Monte Carlo simulation for random deposition. Blue x's are the result of individual simulations for random deposition at a given 8. The red curve shows is the Weibull distribution fit from Equation 3.5

0 10 20 30 40

Figure 3.4: Monte Carlo simulation of randomly deposited molecules on a surface at a density of 6 = 7r_1. Red colored circles represent unresolvable molecules; green circles are resolvable molecules; white space is wasted space.


equation to this data gives Equation 3.5 which is plotted in red in Figure 3.3.

9RND(S) = Se'*6 (3.5)

With only « 12% of the available surface area containing resolvable molecules, and

nearly twice that containing unresolvable molecules, a random deposition approach

clearly leaves much room for improvement.

3.3 Fitting the PSF

There has been considerable effort spent to devise novel approaches of resolving more

than one molecule per DLR using various super-resolution techniques [65, 66, 67, 68,

69, 70, 71, 72, 73]. Being able to break the diffraction limit would clearly increase the

amount of useable surface area for single molecule arrays, and some of the approaches

are capable of resolving two molecules separated by as few as 10-20nm [68, 73, 72]. The

main drawbacks of these approaches are increased image acquisition and processing

time, and this is discussed in detail in section 3.7. Assuming time isn't an issue,

the potential 6 for being able to resolve up to 2 molecules per DLR was determined

by Monte Carlo simulations. This method was termed "Fitting the PSF" after the

groups who pioneered the idea [70, 68].


An algorithm was implemented in silico to determine the resolvable density improve

ment potentially achievable by fitting the PSF. The Monte Carlo simulations for

random deposition were repeated with a single modification: each point can have at


0.35

0.30

0.25

©0.20

0.15

0.10

0.05

0 (

5

Figure 3.5: Monte Carlo simulations for fitting the PSF. Blue x's show the fraction of points resolvable without the need for PSF fitting; red circles show the fraction of points only resolvable with PSF fitting, and green circles are the sum of the two.

most one NN with the property 0.03 < t < 1. A lower limit was imposed because two

molecules within «10 nm of each other cannot be resolved even with super-resolution

techniques. For these simulations, two molecules can therefore be resolved within

a single DLR while all other neighbors must be at least 1 unit away. Any DLRs

containing three or more points resulted in all responsible molecules being flagged as

unresolvable. The results of this simulation are shown in Figure 3.5. A least-squares

fit of the points only resolvable with PSF fitting to a Weibull distribution gives:

BP8F{S) = 0.40 (-£-) ' e ( - ( ^ ) 1 0 1 ) (3.6)

D 0.2 0.6 1 1.4 1.8


The fraction of resolvable molecules as a function of surface density is simply the sum

of equations 3.6 and 3.5:

0RDN+PSF(S) = 6e~*s + 0.40 C^j eH^)1 '"1) (3.7)

With this approach, the optimal "fill factor" improves to 6RDN+PSF ~ 0.27 when

S ~ 0.5. While this is more than double the resolvable density of random deposition,

it would likely be difficult to reliably implement over a large surface area due to the

large image acquisition times.

3.4 Pat terned Nanoparticle Arrays

Instead of using a super-resolution technique to improve the resolvable density, one

obvious alternative is to pre-pattern the surface with small reactive features spaced

out by at least a diffraction limit. Each feature is capable of binding more than one

molecule but single molecules bound to two adjacent features are resolvable. Ordered

arrays for single molecule imaging have been fabricated using e-beam lithography [74],

albeit not at a diffraction-limited pitch. Furthermore, e-beam and cleanroom steps

remain expensive and time consuming to manufacture surfaces in large numbers. In

this section ordered arrays are discussed and a novel method for fabricating a semi-

ordered array without the need for e-beam lithography is presented.

3.4.1 The Poisson Limit

Consider a surface with one small feature capable of specifically binding a molecule of

interest. If there are n molecules each with a probability p of binding, the probability


of getting exactly k molecules bound follows a binomial distribution:

Pr{K = k) = m^Wpt(1 -"rt (3'8)

The probability of getting exactly 1 molecule (k = 1) is given by equation 3.8:

Pr(K = 1) = np(l - p)"-1 (3.9)

Now consider the case where there are m squares and the probability of binding

p = 1/m. If S(n, m) is the expected number of squares containing exactly 1 molecule,

then:

S{n, m) = mnp{\ - p ) n - 1 (3.10)

Simplifying and taking the natural logarithm:

log(S) = log(n) + (n - 1) log(l - p) (3.11)

Approximating the second term using a Taylor series gives:

log(S) w log(n) - np (3.12)

S = ne-np = ne-n/m (3.13)


Figure 3.6: Simulated image of a patterned array. Monte Carlo simulation of single molecules deposited on a patterned array at a density of 1 molecule per DLR. White spots are features without a molecule; red spots are features with more than one molecule; and green spots are features with exactly one molecule.

To find the case where S reaches a maximum value for fixed m, take the derivative

of Equation 3.13 with respect to n:

— = e~nP _ —e-nlm

dn m

= e~np(l--) m

The derivative goes to zero when n = m, ie. when the number of deposited molecules

exactly equals the number of features. At this point, Smax = iVe-1. A simulated

surface at this density is shown in Figure 3.6.

The binomial distribution (Equation 3.8) approaches the Poisson distribution as

the number of molecules n is large while p is small. The fraction of squares containing


0

0.5 1 1.5 2 2.5

Figure 3.7: Monte Carlo simulations of single molecule deposition on a patterned surface. Blue squares show the results of individual trials for various 8. The red curve is a fit to a Poisson distribution.

k molecules is given by:

9[5, k) = 6ke

k\ (3.14)

The fraction of squares containing exactly one molecule is thus given by:

0(6) = Se -S (3.15)

To find the conditions with the highest fraction of squares containing k = 1 one

molecule, take the derivative of Equation 3.15 with respect to 6 and set it equal to

zero: d0 dS

e-s(l - S2) = 0. (3.16)

The only non-negative solution to this equation occurs when 6 = 1. At this point

the fill factor 9array = e_ 1 w 0.37. Monte Carlo simulations were run to confirm


this result. Points were randomly assigned to integer coordinates on a square array.

Points deposited at different coordinates were resolvable from each other; two or more

points with the same coordinates were unresolvable. The results of this simulation

are shown in Figure 3.7. From this figure it is apparent that 6max « 0.36 at 5 = 1, in

agreement with our theoretical predictions.

3.4.2 Methods

Instead of fabricating an ordered array with e-beam lithography, a novel approach

has been developed that is completely scalable over a large area and involves simple

solution processing. The method involves creating chemically modifiable nanoparticle

arrays, such as Au, AI2O3 or TiC"2 (1-5 nm in diameter), on surfaces with interparticle

distances of « 280 nm.

The ability of blockcopolymers (BCPs) to self-assemble on the nanometer scale

makes them attractive templates for patterning dense nanostructures. BCPs allow

simultaneous control over interparticle spacing and nanoparticle size. Moreover, this

process can be performed on transparent glass substrates compatible with fluores

cence microscopy. The key advantages of this BCP templating method are the low

cost and high throughput of semi-ordered nanoparticle arrays when compared to tra

ditional lithographic patterning techniques. Not only have BCPs been used as a mask

material to form such structures in a subtractive fashion, but the amphiphilic ver

sions are capable of additive patterning by localizing inorganic precursors within their

hydrophilic regions. A variety of metal [75] and metal oxide [76] nanoparticles have

been formed by the latter method. Whereas most BCP patterning studies focus on

highly ordered and dense features, little work has been done studying the low density


patterning capability of BCPs. To this end, a method was developed for forming thin

micellar films of BCP from solution as templates for ordered gold nanoparticle arrays

with interparticle spacings > 200 nm.

Surface patterning proceeded via spin coating of a block copolymer solution that

was infused with inorganic precursors (HAuCL4, TiCL4, or A1C13). The precursors

selectively load in the cores of the spherical micelles formed by the block copolymers

in solution. The physical dimensions of these micelles cause them to pack in well-

defined geometry on a surface when a solution is spin cast, resulting in each micelle

core having a specific minimum distance with its nearest neighbors. This minimum

spacing is controlled by the molecular weight of the block copolymer, the concen

tration of the block copolymer micelle solution, and the spin speed. Removal of the

polymer matrix from the surface by oxygen plasma cleaning causes the inorganic

precursors in the micelle cores to coalesce and form nanometer sized particles (Au,

TiC>2, AI2O3 respectively) where the micelle cores once stood. The soft pattern gen

erated by self-assembly of the micelle solution on the surface during spin coating is

thereby transferred into a hard pattern consisting of metal or metal-oxide nanoparti-

cles. These particles can then modified to enable attachment of fluorescently labeled

molecules. For Au nanoparticles, a gold-thiol interaction may be employed for direct

attachment; the metal oxide particles can be modified with aminated phosphonic

acids to enable amide coupling reactions.

The following BCPs were purchased from Polymer Source Inc. and used as re

ceived: Poly(styrene-b-2-vinyl pyridine) (PSP2VP)172k-42k, PS-P2VP 464k-24k, and

PS-P4VP 557k-75k. Anhydrous toluene (99.8%) was purchased from Acros Organ-

ics and hydrogen tetrachloroaurate (III) hydrate (99.999%) was purchased from Alfa


Aesar. BCP solutions were filtered through 25 mm, 0.45 fxm pore Whatman PFPE

with GMF syringe filters. Silicon substrates were cleaved from silicon wafers (Silicon

Quest International) and cleaned by exposure to UV/O3. Stock solutions (4 mg/mL)

of each polymer were prepared by dissolving the solid polymer as received in anhy

drous toluene and stirring with a magnetic stir bar for 1 hour. After being loaded

with HAuCl4, the solutions were diluted to their final concentration for film forma

tion. In a glovebox, solid HAuCU was added in a 0.5 mol ratio (mole Au3+:mole

pyridine units) to each of the solutions. The solutions were stirred with magnetic

stirring until all of the solid was dissolved (usually 1 hour or more). Each solution

was filtered through a 0.45 (xm PTFE filter before use. Films were spin cast in an

N2 environment at 6000 RPM for 40 sec on clean silicon substrates. For nanoparticle

formation, these films were exposed to 0 2 plasma for 5 minutes (50W, 0.4 mbar O2).

Films were characterized by AFM (DI Nanoscope III) and nanoparticle arrays by

SEM (Sirion FEI XL30).

3.4.3 Experimental Results

Micelle films were characterized by measuring the nearest neighbor distances with

AFM (Figure 3.8). BCP solutions from 2 mg/mL down to 0.2 mg/mL gave semi-

ordered micelle films for all three BCPs tested. Above 2 mg/mL, the films tended to

have multilayers and order could not be discerned. Below 0.2 mg/mL, the observed

films were not continuous, and the micelles were no longer well-ordered. Attempts

were made to prevent film breakage by diluting the original BCP solution with 1

mg/mL polystyrene in toluene [77], but order was still degraded at lower BCP con

centrations. The largest interparticle distance shown in literature to date is 170 nm


Micelles Nanoparticles Nanoparticles

Center'of.sditie Center of sl ide After. UV'O.

Figure 3.8: AFM height images of a BCP (557k-75k PS-P4PV) film loaded with HA11CI4 (0.5 mole ratio) at sequential steps in surface modification. Height scale bars are shown in nanometers and each image is 5 x 5 //m. White areas correspond to micelle cores where HAuCU resides.

[75]; this approach consistently produced micellar films with nearest neighbor dis

tances > 200 nm, although there was some variation from the center of the slide

to the edge. Micelle density was observed to decrease linearly with decreasing BCP

concentration (Figure 3.9). Nanoparticle arrays formed from HAuCL4-loaded BCP

micelle films showed short range order comparable to that seen in the micelle films

by AFM. Nanoparticle diameter varied from 5-15 nm indicating either heterogeneous

loading of the micelles or material ablation during the plasma process. XPS analysis

confirmed that even after exposure to O2 plasma, the nanoparticles were composed

of elemental gold, as other groups have reported previously [78]. These results show

that BCPs with Mw > 200,000 are capable of patterning gold nanoparticle arrays

at distances > 200 nm. A typical particle nearest-neighbor distribution is shown in

Figure 3.10. Here, the mean NN distance is « 280 nm with a standard deviation of ~


100

— 80

£

& 60

c « 2 4° s is 20

• 172k-42k PS-PZVP • 464k-24k PS-P2VP AS57k-75kPS-P4VP

1

•

t

•

•

t • •

• •

0.2 0.4 0.6 0.8 1

BCP Concentration (mg/mL) 1.2

Figure 3.9: Micelle density, measured by AFM, was linearly related to the concentration of BCP solution.

15%. There is a performance hit to the fill factor due to the width of this distribution

when compared to an ordered array, as seen in Figure 3.11.

To demonstrate that these nanoparticle arrays can act as substrates for single

molecule deposition, attempts have been made to passivate the region between the

nanoparticles with a low molecular weight PEG-silane and then coupling 5'-thiolated

and 3'-Cy3 labeled DNA through a thiol-gold linkage. There has been some success

with this approach but the single molecule surface densities are still below the theo

retical limit. Further work to optimize the surface passivation and coupling chemistry

is required to achieve the desirable density and single DNA attachment. There have

been recent reports of fabricating regular C6o arrays by fullerene absorption onto gold

dots [79], which is one chemistry attachment scheme that might be explored.


150

v 100

I 50

iTT-y-q-0.2 0.25 0.3 0.35 0.4

NN distance / /xm

0.45 0.5

Figure 3.10: Nearest-neighbor distances for Au nanoparticles deposited via block co-polymer lithography. The mean NN distance is « 280 nm with a = 15%.

a 0.8

8 0.6

A \

1 S ®

0.6 0.8 1 1.2 < NN distanco / diffraction limit

0.8 1 1.2 1.4 1.6 1.8 <NNdistanco /diffractionlimit

Figure 3.11: The fill-factor for semi-ordered arrays depends on the width of the NN distribution. (A) A plot showing the fraction of resolvable particles (left axis) and the particle density per DLR as a function of particle NN distance (right axis). Curves for an ordered array (step function) and a semi-ordered array (sigmoidal function) are shown. (B) The fill factor of a semi-ordered array as a function of mean NN distance, assuming a = 15%.


3.5 Single Molecule Colloidal Epitaxy

The main drawback of pre-patterned arrays is that it is difficult to prevent two

molecules from coupling to the same feature. As seen in Figure 3.6, this results

in a large fraction of the features being unresolvable. If this limitation could be

overcome, the resolvable density of a prepatterned array would approach the circle

packing limit. This motivates a search for a deposition technique with a single rule:

once a molecule is bound to a random location on a surface, somehow prevent any

other molecules from binding within one diffraction limit of that molecule. Colloidal

epitaxy, described in the following section, uses steric hindrance as the means of

prevention.

The basic idea behind the approach is as follows. Colloids with a mean diameter

> the diffraction limit are functionalized with a fluorescently-labeled biomolecule of

interest. The decorated colloids are then self-assembled to form a monolayer on a

surface using the biomolecule as the tether to the surface. Through enzymatic or

chemical cleavage, the colloids are cleaved from the biomolecule to leave behind a

semi-ordered array of single biomolecules. If the length of the tether between the

colloid and the surface is small in comparison to the colloid diameter, this deposition

strategy will theoretically allow every deposited single molecule to be spaced by at

least a colloid diameter and be resolvable. A cartoon outlining the colloidal epitaxy

process is shown in Figure 3.12.


Monte Carlo simulations of single molecule colloidal epitaxy were implemented in

silico by only permitting a molecule to be immobilized if its NN was at least 1


300nm

(c)

biotin

P 0° 300nm

Figure 3.12: Illustration of single molecule colloidal epitaxy (not to scale). (1) Animated silica beads are coupled to thiolated double-stranded DNA decorated with Cy3 and biotin. (2) The DNA-bead conjugates are tethered to the surface through a biotin-neutravidin binding interaction. (3) Unoccupied neutravidin sites are filled with free biotin and the surface is incubated with a type II restriction enzyme to cleave the double-stranded DNA. (4) The beads are washed from the surface leaving behind an array of single molecules with a minimum pitch approximately equal to the colloid diameter.


diffraction limit away. Instead of using the surface density 6 as the metric that

determines fill factor, here the deposition attempt rate per unit area 6A is used. The

results of this simulation are shown in Figure 3.13. The data from the Monte Carlo

simulations was fit to a Langmuir isotherm giving the following relationship:

™ = i+nes ( 3 J 7 >

This equation gives a fairly good fit for small values of 6A', at larger values the fit

begins to underestimate the simulation predictions. Figure 3.13 shows that at high

6A, the "fill factor" 9epitaxy asymptotically levels off and reaches a maximum of «

0.70. This resolvable density exceeds all other methods by nearly a factor of two

and is approximately six times higher than can be achieved with random deposition.

Colloidal epitaxy is also nearing the hexagonal circle packing fill factor of « 0.907.

A simulated image of molecules deposited with this approach at a surface density of

59% is shown in Figure 3.14b.

3.5.2 Mass Transport Considerations

One can estimate the deposition time for colloidal epitaxy with the rate equation for

the association/dissociation of DNA-colloid constructs, which takes the form

dC ~K7~ = kon\Cs0 — CS)CW — K0ffCs (3.18)

where Cw is the concentration of the colloids in solution at time t, Co is the initial

concentration of colloids before reaction, Cs is the concentration of sites occupied,


Figure 3.13: Monte Carlo simulation of colloidal epitaxy. The x-axis, deposition attempts per DLR, is plotted on a linear scale (left) and a log scale (right). The blue circles show the simulation data; the red curves are fits using the Langmuir isotherm in equation 3.17.

CSQ is the total concentration of binding sites, and km and k0ff are the rate con

stants for the association/dissociation events with units M_ 1 sec -1 and sec -1. Using

the dimensionless parameters e = Dt/ti2 as the normalized diffusion time with h

as the height of the fluidic chamber and D as the diffusion coefficient of the col

loid, e = Co/Cgo as the relative adsorption capacity of the surface with respect to

the solution, Da = k^Cs^h?/D as the Damkohler number, and Kp = fc0///fconCo

as the dissociation constant with concentrations Cw = CW/CQ and Cs = Cs/Cso, in

dimensionless form is

dr = eDa{9w(l - 6S) - KD9S) (3.19)

Solving this differential equation subject to the initial condition that 6(T = 0) = 0,

and neglecting any time dependence of 6W, gives the solution

0S(T) = 9W + KD (1

-eDa(6w+KD)T ) (3.20)


0 10 20 30 40 50 0 10 20 30 40 SO

(a) Colloidal epitaxy at 38% surface density, (b) Colloidal epitaxy at 59% surface density.

Figure 3.14: Simulated images of colloidal epitaxy. Monte Carlo simulation of single molecules deposited on a surface via colloidal epitaxy at two different densities. Green spots are resolvable molecules, red spots are unresolvable molecules, and white area is unoccupied space.

Assuming that KD is small for a biotin-streptavidin binding interaction, Equation

3.20 simplifies to

es(r) = 1 - e-eDaewT (3.21)

The kon rate constant for streptavidin binding to biotinylated bovine serum albumin

has been reported to be 1.2 x 105 M~x sec -1 [80]. Since this constant is directly

proportional to the diffusion coefficients of the reacting species, biotin coupled to a

300 nm bead is expected to reduce the rate constant by two orders of magnitude

to « 103 M - 1 sec -1 . Using experimental values, one can then estimate that the

time constant for deposition in Equation 3.21 is 3 x 10~6 sec -1. This is in reasonable

agreement with our experimental observations of 38% coverage after 40 hrs of binding.


3.5.3 Methods

Double stranded DNA was coupled to 300 nm and 640 nm aminated silica colloids

(Corpuscular) as described elsewhere [81] with stoichiometrics experimentally deter

mined such that the average number of DNA molecules per colloid was near unity

(Figure 3.12a). The DNA sequence was designed to contain a Cy3 fluorophore, a

BsaH I restriction site, a 5'-biotin, and a 3'-thiol (Integrated DNA Technologies, se

quences 5'-Bio-CGC TCT ATC CTC CCT CCA TTC CAA CCA GAC GCC ACC

CTC AGT CAT TTG TA-SH- 3' and 5'-TAC AAA TGA CTG AGG GTG GCG

TCT GGT TGG AAT GGA GGG AGG ATA GAG CG -Cy3-3'). Imaging Cy3 with

a 1.45 N.A. objective gives a diffraction limited resolution of 240 nm; 300 nm colloids

were chosen to ensure slightly more than adequate spacing. Glass coverslips (Pre

cision Glass Optics, D-263T cut glass, 0.15 mm, 2"xl" 40/20 surface quality) were

RCA cleaned, coated with a polyelectrolye multilayer, and functionalized with Biotin-

PEO-Amine (Pierce) as previously described [82]. Surfaces were then washed with 1

mL of 10 mM Tris, 50 mM NaCl, pH 7.5 buffer and incubated for 45 minutes with 1

mg/mL neutravidin (Pierce) in 0.01 sodium azide, 20 mM Tris, 100 mM NaCl, pH 8

buffer. DNA-colloid constructs were resuspended in 100 /JL of 1% BSA, IX PBS, pH

7.4 buffer and allowed to bind to the neutravidin surface for 20 hours at 4°C (Figure

3.12b). The surfaces were then washed with IX PBS and the deposition process was

repeated with a fresh batch of DNA-colloid constructs. After the first deposition pe

riod of 20 hours, the fraction of beads containing DNA in solution decreased slightly

(~5%) due to deposition. Stopping and restarting the deposition the process serves

two purposes. First, it allows us to verify and quantify the coverage, and second, it

helps to improve the deposition rate and surface density for the second incubation.


After the final wash, unoccupied neutravidin sites on the surface were filled by incu

bation with 50 mM biotin for 30 minutes. The colloids were then removed by with

the addition of 10 units of BsaH I in IX buffer # 4 (New England Biolabs) and 1%

BSA for 2 hours at 37°C (Figure 3.12c) followed by washing the surface extensively

with dH20 (Figure 3.12d). The surface was imaged on a Nikon TE2000-S microscope

in total internal reflection fluorescence mode with a Hamamatsu ORCA-ER CCD.

A Cy3 filter set (HQ535/50, Q565LP, HQ610/75, Chroma) and a Nikon Plan Apo

TIRF 60X 1.45 NA objective with a low-fluorescence immersion oil was used (n =

1.515). Five minute movies were collected from « 50 fields of view over a 1 cm2 area.

Single molecules were identified based on the presence of single-step photobleaching

in their trajectories.

3.5.4 Experimental Results

Initial attempts at colloid cleavage used Hgal, a restriction endonuclease that cleaves

5/10 bp downstream of its 5 bp recognition site (Figure 3.15, top). To test the

enzyme's ability to cleave DNA on the beads, bulk experiment were done whereby

DNA-coated beads were treated with Hgal. The beads were then pelleted and the

supernatant was loaded on a polyacrylamide gel (Figure 3.16, left). Lane 3 shows

that Hgal was unable to cleave the DNA on the beads, even though it was able to

cleave free DNA in solution (Lane 5). Using the same DNA-coated beads, this bulk

experiment was repeated with BsaHI instead of Hgal. BsaHI has a similar recognition

sequence as Hgal but cleaves within the recognition sequence (Figure 3.15, bottom).

With this enzyme DNA cleavage was possible on the bead (Figure 3.16). The close

proximity of the Hgal cleavage site to the bead surface may result in unfavorable


Hgal 5' -2Bio- CGCTCTATCCICCCTCCATTeaVACCAGACGCCaCCqrCaGTCATTTGTA

Cy3- GCGAGATAGGAGGGAGGTAAGGTTGGTCTGCGGTGGGAGTCAeTAAACAT - 5'

-< 42 bp •'

cqrcaGTQ

GGAGTCAp

iGAgGCCJ

rcTGcbs

BsaH . .-5 ' - 2 B i o - CGCTCTATCCTCXCTX:aVTTCCAACCAGftEGCCACCCTC*GTCATTTGTA

C y 3 - GCGAGATAGGAGGGAGGTAAGGTTGGTCTG3KTGGGAGTCAGTAAACAT - 5

4 31 bp >

Figure 3.15: Restriction enzyme cleavage sites for colloidal epitaxy. Shown are the Hgal and BsaHI recognition sites (highlighted in green) and the restriction enzyme cut sites (vertical black lines).

steric interactions between the enzyme and the surface, thereby preventing activity.

Remarkably, a spatial difference of one helical turn (10 bp) was enough space for

cleavage with BsaHI.

Micron-sized colloids were capable of forming much more ordered arrays than

smaller colloids, as evidenced by the regular pattern in the power spectrum (Fig

ure 3.17b). Sub-micron sized colloids were capable of higher densities but less order

(Figure 3.17d,f). Figure 3.18 shows brightfield, SEM, and TIRF images of colloidal

epitaxy using 300 nm silica colloids pre- and post-enzymatic cleavage. Restriction en

zyme treatment and washing was > 90% successful at removing the tethered colloids.

Control experiments with non-biotinylated or non-thiolated DNA showed little non

specific binding and DNA without Cy3 showed no sign of fluorescence. Images were

processed using a custom script in MATLAB (see Appendix B) and single molecules

were identified based on their fluorescent intensities and observing step-wise photo-

bleaching in their trajectories. Figure 3.18a shows histograms of nearest-neighbor

distances for Cy3 labeled DNA patterned with two colloid sizes, 300 nm and 640 nm,

with peaks at 346 nm (a — 48 nm), and 641 nm (a = 101 nm), respectively. The


Figure 3.16: Restriction enzymes are sensitive to colloid proximity. Gel A, lane 1-dsDNA alone; lane 2- Cy3-ssDNA; lane 3- dsDNA coated beads treated with Hgal; lane 4- untreated dsDNA coated beads; lane 5- dsDNA treated with Hgal; lane 6-LIZ120 size standard. Gel B, lane 1- supernatant after coupling dsDNA to beads; lane 2- supernatant from first wash; lane 3- supernatant from second wash, lane 4-supernatant from third wash; lane 5- supernatant from fourth wash; lane 6- dsDNA coated beads treated with BsaHI; lane 7- untreated dsDNA beads; lane 8- LIZ120 size standard.


Figure 3.17: Brightfield images and power spectrum of colloidal epitaxy prepared surfaces. Colloid sizes are: (A,B) 2 //m, (C,D) 640 nm, and (E,F) 300 nm.


Figure 3.18: Colloidal epitaxy experimental results, (a) Bright field image of 300 nm beads immobilized on a neutravidin-coated glass coverslip through a biotinylated DNA tether. The image was taken in air after the colloid solution was removed with the surface still slightly wet. Drying was avoided primarily over concern that it would have undesirable effects on the DNA tether, (b) Scanning electron microscope image of 300 nm colloids tethered to the surface, (c) Bright field image of the surface following restriction enzymatic cleavage of the DNA tether and extensive washing, (d) Total internal reflection fluorescence image of Cy3-labeled DNA patterned on the surface after colloid removal. Scale bars are the same as in (a).


(a) 6

9 4

u c ai

<v

(b)

0.2 0.4 0.6 0.8 1 distance (urn)

§ 20

a*

o ft 10

.Q

£ 3

n rf n - i

density (features / urn2)

Figure 3.19: (a) Nearest-neighbor intensity histogram for fluorophores deposited via colloidal epitaxy with 300 nm colloids (white) and 640 nm colloids (grey). The cen-troid of each single molecule feature was identified with single pixel precision and the linear distance to the next nearest feature was calculated in pixels. Pixel distances were then converted to nanometers based on known pixel dimensions, (b) Histogram of single molecule densities produced by 300 nm colloidal epitaxy observed in different fields of view over a 1 mm2 area.

average feature density of a 1 mm2 field of view is shown in Figure 3.18b. The mean

experimental density was found to be 4.2 [xm~2 (a = 0.3 /xm~2), giving a fill factor

6S « 0.38. Colloidal epitaxy thus allows for over a three fold improvement over ran

dom deposition but the density is slightly lower than the 9.9 /zm-2 our simulations

predicted possible.

The colloid constructs were observed to form quasi-hexagonal islands on the sur

face in the SEM images (Figure 3.18b) and, to a lesser extent, in the fluorescence

images (Figure 3.18d). The presence of these islands in both wet and dry states sug

gests that the colloids are likely forming these structures during convective assembly.

This may be due to favorable colloid-colloid interactions during deposition or the non

uniform surface patterning of neutravidin, a phenomenon that has been previously

observed. In an attempt to improve deposition rates and densities, magnetic colloids


were used in conjunction with a weak magnetic field but there was limited success.

Colloids tended to either stack in multilayers or be held against the surface in random

orientations that did not favor binding, resulting in only a small fraction of colloids

remaining tethered to the surface.

3.6 Future Work for Colloidal Epitaxy

Colloidal epitaxy has demonstrated achieving « 38% surface coverage in practice,

but there is still much room for improvement in this approach. Longer incubation

times (up to one week) may further improve the surface density. Applications of

colloidal epitaxy may also be explored to pattern surfaces for many different single

molecule studies. For example, DNA fragments with appropriate sticky ends may be

ligated to the overhang created by the restriction digest for high throughput sequenc

ing [12], SNP genotyping, or enzymatic assays. Instead of using a restriction enzyme

to remove the beads, the DNA tether could also be broken through heat, chemical

treatment, or activation of a photocleavable or chemical-labile linker. This approach

is also not limited to depositing DNA on a surface: any molecule that can be hetero-

functionalized with a reversible linker may be deposited, including small molecules,

proteins, antibodies, viruses, or nanoparticles. Colloidal epitaxy relies on cheap and

easily accessible reagents and does not require any expensive instruments, and the

minimum spacing between molecules can easily be controlled by changing the colloid

diameter. This gain in throughput may be of benefit to any study where many single

molecules need to be interrogated in a massively parallel fashion.


3.7 Super-Resolution: Breaking the Diffraction Limit

There have been efforts to surpass the optical resolution limit using near-field scan

ning optical microscopy (NSOM) [83, 84], apertureless near-field scanning optical

microscopy [85], local enhancement using bowtie nano-antennas [86], structured illu

mination [65, 66], stimulated emission-depletion [67], and by fitting the point spread

function [68, 69, 70]. There are also two similar techniques called photo-activated

localization microscopy (PALM) [71, 72] and stochastic optical reconstruction mi

croscopy (STORM) [73] that improve spatial resolution by taking advantage of the

abundance of temporal resolution. Each of the aforementioned approaches is capable

of resolving more than one molecule within a single diffraction limited region (DLR),

although this is typically at the cost of increased image collection time and/or image

processing time.

Monte Carlo simulations were run to see if super resolution techniques could of

fer an advantage over standard microscopy techniques for imaging large-area single

molecule arrays. These simulations modeled a simple PALM/STORM imaging ses

sion where only a stochastic subset of the total fluorophores in a field of view are

excited and imaged during each imaging cycle. By cycling many times, virtually all

of the molecules can be switched on at least once to allow for the reconstruction of a

super-resolution image.

Figure 3.20 shows the fraction of molecules resolvable vs. cycle number for four

surfaces at different molecular densities. The simulations assume a microscope diffrac

tion limit of 250 nm, a super-resolution limit of 20 nm, and a probability p that any

given molecule be switched on during each cycle. There appears to be an important

trade-off between working at higher densities and the number of excitation/imaging


100 200 300 400 500 600 700 800

B ^ ^

r

— w

,

_

100 200 300 400 500 600 700 800

o c o V* u

r ^ ~ ^ ^

»-«,

•

•

100 200 300 400 500 600 700 800

. . . . .

fT V p = 5% p = 2% p=1% p = 0.5%

•

100 200 300 400 500 600 700 800

excitation/imaging cycle

Figure 3.20: Super-resolution Monte Carlo simulations. The fraction of molecules resolvable vs. cycle number for surfaces at four different densities (molecules per diffraction limited region) of (A) TT"1 = 0.32, (B) 0.20, (C) 0.10, (D) 0.01. Curves for different values of p are shown.


cycles required to image all (> 99%) of the molecules on a surface. Being able to

finely tune the fraction of molecules that switch on during each excitation cycle is

also important for optimization.

For the ideal density of 7r_1 = 0.32 molecules per super-resolution diffraction

limited region, over 800 cycles are required to image all of the molecules. For a

sub-optimal density (0.10) as few as 400 cycles are required. In this case, cutting

the number of cycles in half reduces the number of resolvable molecules by less than

half, which suggests that working at a sub-optimal density would likely give the

highest throughput. The requirement that > 99% of the molecules are imaged after

each dNTP incorporation necessitates more cycles than would typically be needed for

standard PALM/STORM image reconstruction. For example, if each cycle takes 0.1

sec, a surface with a density of 0.10 (318 molecules//um2) could have a throughput of

5 molecules/(/im2 sec). This assumes that the stage is capable of repeatedly returning

to the same position with a high degree of accuracy.

Table 3.1 illustrates how throughput scales with image acquisition and stage

move times. Here it becomes clear that a super-resolution approach will only offer a

throughput advantage if the imaging time is faster than stage moving time. Currently

image and stage moving times are around 100 msec, meaning that super-resolution

techniques would have to be able to excite, image, and switch off fluorophores in

10-50 msec to offer any real advantage. Signal/noise is probably the main limitation

on imaging speed right now. Imaging time improvements might be had in developing

brighter fluorophores, tweaking the fluor's local environment to maximize the absorp

tion cross section and quantum yield (solvent polarity, viscosity, metal enhancement,


Table 3.1: Theoretical throughput (molecules/(/Ltm2 sec)) of a super-resolution approach with an effective resolution of 20 nm. The image acquisition time (timg) and stage move time (£mow)are shown in seconds. These calculations assume randomly deposited single molecule arrays at five different surface densities (32% to 1%). For comparison, the throughput of a standard diffraction-limited microscope is shown in the right-most column.

Hmg

0.001 0.001 0.001 0.001 0.01 0.01 0.01 0.01 0.1 0.1 0.1 0.1 1 1 1 1

''move

0.001 0.01 0.1 1

0.001 0.01 0.1 1

0.001 0.01 0.1 1

0.001 0.01 0.1 1

32% 371.6 368.3 338.2 186.0 37.2 37.2 36.8 33.8 3.7 3.7 3.7 3.7 0.4 0.4 0.4 0.4

20% 439.5 434.6 391.1 195.6 44.0 43.9 43.5 39.1 4.4 4.4 4.4 4.3 0.4 0.4 0.4 0.4

10% 513.7 502.4 412.0 147.1 51.5 51.4 50.2 41.2 5.1 5.1 5.1 5.0 0.5 0.5 0.5 0.5

5% 661.7 633.3 443.3 110.8 66.5 66.2 63.3 44.3 6.6 6.6 6.6 6.3 0.7 0.7 0.7 0.7

1% 306.9 281.8 155.0 28.2 31.0 30.7 28.2 15.5 3.1 3.1 3.1 2.8 0.3 0.3 0.3 0.3

32%, no SR 1000.0 181.8 19.8 2.0

181.8 100.0 18.2 2.0 19.8 18.2 10.0 1.8 2.0 2.0 1.8 1.0


Table 3.2: Summary of the parameters that govern the total bandwidth of single molecule imaging. We assume a diffraction limit of 250 nm except for the super-resolution approaches, where we estimate a diffraction limit of 20 nm. 9 is the maximum obtainable "fill factor".

Method

Circle packing Colloidal epitaxy Ordered arrays

Semi-ordered arrays Random deposition

Super-resolution

PSFarea(f,m2)

TT(0.125)2

TT(0.125)2

TT(0.125)2

TT(0.125)2

TT(0.125)2

TT(O.OIO)2

9

|TT\/3 « 0.907 « 0.6-0.7

e-1 « 0.368 w0.21

(Tre)"1 « 0.117 (Tre)"1 » 0.117

Hmage \SGC)

0.1 0.1 0.1 0.1 0.1 120

''move V.S6CJ

0.1 0.1 0.1 0.1 0.1 0.1

temperature), further reducing the excitation volume to minimize scattering, and de

veloping faster, more sensitive detectors. In summary, super-resolution techniques

therefore offer slightly lower imaging bandwidth with the additional requirements of

accurate stage movement and post-acquisition image processing.

3.8 Throughput Comparison

Table 3.2 shows the key parameters that govern the bandwidth for each of these

methods. Here, 9 represents the maximum obtainable "fill factor" as determined

above by theory or Monte Carlo simulation. With this data, Equation 3.2 was used

to calculate the imaging bandwidth for each method. A summary of the imaging

bandwidths of these approaches is shown in Figure 3.21. The benefits of shrinking

the size of the PSF using super-resolution techniques are comparatively lost due to the

lengthy imaging times required. Ordered arrays that have higher fill factors and short

imaging times currently offer higher bandwidths, but if super-resolution techniques

can reduce the required imaging times, they may eventually surpass them. Overall,


super-resolution with random deposition

random deposition

semi-ordered arrays

ordered arrays

colloidal epitaxy

circle packing

i i 1 1 t i i i i

0 10 20 30 40 50 60 70 80 90 100 molecules yum2 sec1

Figure 3.21: Bandwidth comparison of single molecule deposition methods. The imaging bandwidth of each method was calculated using Equation 3.2 and the parameters shown in Table 3.2.

colloidal epitaxy is currently far and above the fastest and easiest approach with a

throughput nearing that of circle packing.

Chapter 4

A Single Molecule Measurement of

the "Speed Limit" of DNA

Polymerase

4.1 Introduction

Fast and accurate DNA replication is required to ensure the faithful transfer of genetic

information to daughter cells. A number of endogenous and exogenous factors can

threaten genome integrity by obstructing the progression, stability, and restart of

replication forks during cell division. Although most paused forks are stable and

capable of resuming DNA synthesis, some pause sites are thought to be hotspots for

misinsertion, deletion, and recombination. Replication forks can be slowed or paused

by encountering DNA-binding ligands [87], genomic tRNA gene sites [88], stalled

ternary complexes of RNA polymerase [89], and DNA-bound protein [90]. Specific

52

CHAPTER 4. THE "SPEED LIMIT" OF DNA POLYMERASE 53

DNA sequences, including palindromic DNA capable of forming hairpin secondary

structure [91, 92, 93], slow zones [94], and trinucleotide repeats of (CGG)n/(CCG)„

or (CTG)n/(CAG)„ [95, 96] have also been shown to cause pausing with several types

of DNA polymerases. While these factors are thought to prevent fork movement along

the template by steric hindrance and may be regulated by the replisome itself, other

factors such as temperature, contaminants, nucleotide analogs [97], template tension

[98, 99, 100], and nonclassic pause sites such as Pyr-G-C [101] may interfere with

steps in the DNA polymerase reaction pathway.

Despite recent advances in single molecule techniques, there have only been a

handful of reports characterizing DNA polymerase behavior [98, 99, 102, 103, 104].

There is a need for single molecule experiments to help characterize the pausing

mechanisms described above and develop a more complete biophysical model for the

polymerase reaction pathway. This knowledge would also be extremely beneficial for

the optimization of DNA sequencing experiments. This chapter describes an attempt

to make some headway towards these goals by developing novel assays to study real

time DNA polymerase kinetics as a function of sequence and template secondary

structure.

4.2 Primer Extension with Fluorescently Labeled

Nucleotides

Much of the seminal work measuring Pol I(KF) incorporation kinetics was done in

bulk using radioactive nucleotides in single base incorporation experiments [105, 106].

There have only been a handful of precision measurements since then to measure DNA


5'TGCTGGGCTTTTGGTTTGTGGG 3'ACGACCCGAAAACCAAACACCCGACATACAAGAAGCCATCC-Biotin

+ Pol I and Cy3-dCTP

5'TGCTGGGCTTTTGGTTTGTGGGC 3'ACGACCCGAAAACCAAACACCCGACATACAAGAAGCCATCC-Biotin

Figure 4.1: Sequence of the primer and template used for single base incorporation experiments. The primer contained an internal Alexa647 fluorophore to act as a FRET acceptor for the single base incorporation of a Cy3-labeled FRET donor (dCTP).

polymerase kinetics, and none of them have been performed with single base resolu

tion. As a first step, real-time single nucleotide incorporation events were observed

on surface-immobilized templates.

4.2.1 Methods and Results

Synthetic oligonucleotides were obtained from Integrated DNA Techologies and the

primer-template sequences used in the first set of experiments are shown in Figure

4.1. The primer contained an internal Alexa647 fluorophore so that immediately

upon Cy3-dCTP incorporation, excitation of Cy3 results in FRET to Alexa647 and

a corresponding increase in emission at 647 nm.

Bulk experiments were first performed to verify that the system works as planned.

. Primer and template were annealed in IX PBS (1 fjM final DNA concentration) and

Cy3-dCTP was added to a final concentration of 100 nM. When excited at 532 nm,

the emission spectra of 50 ixL of this solution was monitored in a spectrophotometer

before and after the addition of 5 units of Pol I(KF) (New England Biolabs) (Figure

4.2). Before the addition of polymerase, there was a large emission peak at « 570


600

500

400

3

16 -̂ £• 300 'l/l c <v 4-»

.- 200

100

0

500 550 600 650 700 750

wavelength / nm

Figure 4.2: Single nucleotide incorporation experiment monitored in a bulk experiment. A solution containing annealed primer/template and Cy3-dCTP was excited at 532 nm before (blue) and after (red) addition of Pol I(KF).

nm corresponding to the direct excitation and emission from Cy3. Five minutes after

polymerase addition, however, the 570 nm peak dropped significantly in intensity and

a new peak at 650 nm appeared. As expected, incorporation of Cy3-dCTP against

the template guanosine placed the two fluorophores in close proximity and resulted

in efficient FRET.

The biotinylated DNA duplex was immobilized and imaged on neutravidin-coated

coverslips as described in Chapter 5. A split-field microimager (Photometries) was

placed in the detection path with a 610 nm dichroic mirror to split the emission

light into Cy3 / Alexa647 channels. Template locations were identified by excitation

with a 633 nm laser (Meshtel) while blocking the Cy3 emission channel. Pol I(KF),

Cy3-dCTP (GE Healthcare), and oxygen scavenging solution [70] were added to the

flowcell and images were acquired at 2 Hz on a Photometries Cascade II EM-CCD.


Figure 4.3: Sample split-field image of Cy3/Alexa647 single primer/templates. The left half shows Cy3 emission; the right half shows Alexa647 emission.

Cy3 excitation was achieved with a 532 nm laser (Meshtel) and the fluorescence emis

sion from both channels was monitored simultaneously (Figure 4.3). Single molecule

trajectories were analyzed using custom software written in MATLAB (for example

code, see Appendix).

A sample single molecule trajectory showing real-time single nucleotide incorpo

ration is shown in Figure 4.4. The red trace shows the intensity of a molecule in the

Alexa647 (FRET acceptor) channel; the green trace shows the intensity of that same

molecule in the Cy3 (FRET donor) channel. At time « 30 sec, there was a sudden

increase in FRET signal which indicated the incorporation of a Cy3-dCTP nucleotide.

A few seconds later, the Alexa647 fluorophore photobleached, resulting in a recovery

of Cy3 signal at the same position in the Cy3 channel. The anti-correlation of the

Alexa647 signal going off and the Cy3 signal going is strong evidence that these two


3000

2500

. 2000

| 1500

I 1000 c.

500

0

-500 0 20 40 60 80 100

time / sec

Figure 4.4: Sample two-color trajectory showing real-time single nucleotide incorporation by Pol I(KF). The red trace shows the integrated intensity of a feature in the Alexa647 channel; the green trace shows the corresponding intensity in the Cy3 channel.

fluorophores were coming from the same template.

While this shows that real-time single nucleotide incorporations can be detected,

the approach suffers from a few drawbacks. First, it requires a high concentration of

Cy3-dCTP nucleotides in solution. Even with TIRF excitation, this inevitably results

in a high background in both channels and makes detecting individual molecules diffi

cult. Second, and more importantly, it is impossible to actually measure a processive

synthesis rate with only a single nucleotide incorporation.

To address the latter of these two issues, another template sequence was designed

to first incorporate a Cy3-labeled dUTP, then six dark nucleotides, followed by an

Alexa647-labeled nucleotide (Figure 4.5). Under 532 nm laser excitation, one would

expect to see a signal come on in the Cy3 channel when the first nucleotide is in

corporated, and then some time later the Cy3 fluor would undergo FRET when the


5' CCTATCCCCTGTGTGCCTTG 3' GGATAGGGCACACACGGAACCTCTTCATTCTTCGTTTCTTATTCTTCGTTTCTTATTCTTCGTTT-Biotin

+ Poll(KF),dATRdGTP, dUTP-Cy3, dCTP-Alexa647

.^P incorporation at t ̂ 5'CCTATCCCCTGTGTGCCTTGGAGAAGO X-3' GGATAGGGGACACACGGAACCTCTTCATTCTTCGTTTCTTATTCTTCGTTTCTTATTCTTCGTTT-Biotin

^ A — ̂ B — incorporation at 12

5' CCTATCCCCTGTGTGCCTTGGAGAAGOAAGAAGC 3' GGATAGGGGACACACGGAACCTCTTCATTCTTCGTTTCTTATTCTTCGTTTCTTATTCTTCGTTT-Biotin

Figure 4.5: "Race track" template for polymerase kinetics with modified nucleotides. Cy3-dUTP is first incorporated, followed by six dark nucleotides, and Alexa647-dCTP. The time difference between these two events (t2 —1\) gives the time required by the polymerase to incorporate six dark and one labeled nucleotides.

Alexa647-labeled nucleotide is incorporated. The length of time between Cy3 signal

going on (ti) and Alexa647 signal going on (i2) would be a direct measurement of the

time required to incorporate the six dark nucleotides and Alexa647-dCTP. Similar

to the previous experiments, the primer/templates shown in Figure 4.5 were immo

bilized on a neutravidin-coated coverslip. Movies were taken as before following the

addition of Pol I(KF), # 2 buffer, oxygen scavenger, and 100 nM fluorescently-labeled

dNTPs.

Initial attempts to do this experiment with two labeled nucleotides in solution

were met with significant obstacles. The main problem stemmed from the use of the

split-field microimager as it was impossible to obtain a high enough FRET signal due

to the high background fluorescence in the donor channel. Removing the microimager

and imaging just the Alexa647 channel showed numerous incorporation events (Figure

4.6, but without the microimager the start times could not be calibrated. Running

this type of experiment at lower nucleotide concentrations (10-30 nM) has been done


=i IB

>* ns

c

1000

800

600

200

-200

-400

0 50 tOO 150 200 250 300 0 50 100 150 200 250 300

time /sec

Figure 4.6: Sample single-color single molecule FRET trajectory showing real-time nucleotide incorporation using the template shown in Figure 4.5. This trajectory was detected in the Alexa647 channel without a microimager present. Generation of FRET signal can be interpreted as Alexa647-dCTP being incorporated into a primer that already contained Cy3-dCTP.

with some success (Braslavsky et al, unpublished data), but the resulting kinetic

rates are not physiological. The Km of Pol I(KF) is on the order of 5 /xM [106]

so rates obtained at low dNTP concentrations are likely possible but are slow and

uninteresting.

Due to these issues, there was a need to design a new assay to measure real

time single molecule DNA polymerase kinetics. Ideally the assay would operate at

saturating concentrations of dNTPs, without the need for fluorescent nucleotides in

solution, and without the need to use force spectroscopy to apply tension to the

enzyme or template.

4.3 Strand Displacement Synthesis Through a DNA

Hairpin

In order to investigate what role sequence and secondary structure have in strand-

displacement synthesis rates, a FRET-based approach was developed to study single

polymerase molecules replicating through a DNA hairpin of known sequence. This


assay allowed for the measurement of polymerase activity in real-time with near sin

gle base resolution without the need to apply tension to the DNA molecule, which

has been shown to affect polymerase behavior [98, 99, 100]. Furthermore, with this

method the polymerase incorporates natural nucleotide triphosphates and does not

require fluorescently-labeled nucleotide analogs in solution, thereby enabling the mea

surement of the polymerase kinetics at saturating nucleotide concentrations.

4.3.1 Hairpin Design

A 259 nucleotide (nt) single-stranded DNA molecule was designed and synthesized

containing an internal double-stranded 33 base pair (bp) hairpin flanked by 94 nt

single-stranded tails, one of which had a 3'-biotin group (Figure 4.9A). The 3' base of

the hairpin contained an internal Cy3 FRET donor and the 5' base contained an in

ternal non-fluorescent FRET acceptor (Black Hole Quencher-2, BHQ-2) so that when

the hairpin was fully folded, quenching of Cy3 by BHQ-2 prevented any fluorescence

emission (Figure 4.9B). The length of the hairpin was chosen to make full use of the

dynamic range for this FRET pair: when the last base pair of the hairpin was broken

due to strand displacement replication, the distance between fluorophores was just

over twice the Forster radius (R0 = 5.02 nm, 33 bp of dsDNA =11.2 nm).

The effect of moving the position of the Cy3 fluorophore away from the base of the

hairpin stem was examined (Figure 4.7A). As the fluorophore was moved from position

-1 to -3 to -5, the quenched fluorescence of the folded hairpin increased (Figure 4.7B.

After primer extension, Cy3 emission increased for all positions (Figure 4.7B, but to

different levels. The reason for this discrepancy is unknown but it may be due to dye-

dye orientation differences. It is therefore fruitful to examine the relative improvement


5' S T ? *CTCTGCTGCGCCATCCGCGGTCTATACGCTAGGTT

T

jJfciGAGACGACGCGGTAGGCGCCAGATATGCGATCC T T

+15 +30

IK 500 540 580 620 660 700 500 550

wavelength / nm

600 650 700 750

Figure 4.7: The internal Cy3 position influenced the FRET efficiency. (A) Three different positions were examined for the Cy3 fluorophore at the -1 , -3, and -5 template positions. (B) Residual fluorescence of the fully quenched hairpin and the fully extended primer were measured in bulk on a spectrophotometer via excitation at 532 nm. (C) The relative fold improvement between pre- and post- extension was calculated for each fluorophore position.


in Cy3 recovery by normalizing to the original quenched fluorescence level (Figure

4.7C. Here the greatest enhancement factors are seen with the fmorophore in closest

initial proximity to BHQ-2. Therefore for the following experiments, templates with

Cy3 in the -1 position were used to minimize the residual fluorescence and maximize

the difference between the pre- and post- extension states.

Olig

o A

B

B'

Z

C

C

D

Y

Tab

le 4

.1:

DN

A o

ligon

ucle

otid

e se

quen

ces

for

hair

pin

ligat

ion.

Sequ

ence

5'-Phos-TCA TAG CCA GAT GCC CAG AGA TTA GAG CGC ATG ACA AGT AAA

GGA CGG TT-3'-Biotin

5'-P

hos-

CG

G A

TG

GC

G C

AG

CA

G A

G[iC

y3]C

AG

T T

CA

GT

C C

CA

CC

G A

CG

TT

T

GG

T C

AG

TT

-3'

5'-Phos-AGG ATC TTA CCA GAG AC/ iCy3/C AGT TCA GTC CCA CCG ACG TTT

GGT CAG TTC CAT CAA CA-3'

5'-AAC CGT CCT TTA CTT GTC ATG CGC TCT AAT CTC TGG GCA TCT GGC

TAT GAT GTT GAT GGA ACT GAC CAA ACG TCG GTG GG-3'

5'-Phos-AGT GAC GCC AAC GCA ATT AC[BHQ2-dT] CTG CTG CGC CAT CCG

CGG TCT ATA CGC TAG GTT TTT CCT AGC GTA TAG ACC G-3'

5'-AGT GAC GCC AAC GCA ATT AC[BHQ2-T] CTC TGG TAA GAT CCT AGG

TCT ATC CTG AAG GTT TTT CCT TCA GGA TAG ACC T-3'

5'-Anti-dig-GCC CTG AGA GAG TTG CAG CAA GCG GTC CAC GCT GGT TTG

CCC CAG CAG GCG AAA ATC CTG TTT GAT GGT GGT TCC GAA AT-3'

5'-TGC GTT GGC GTC ACT ATT TCG GAA CCA CCA TCA AAC AGG ATT TTC

GCC TGC TGG GGC AAA CCA GCG TGG ACC GCT TGC TGC AAC TCT CTC

AGG GC-3'

I Si

o to

§ O 2 3

Oi

CO


D N A hairpin construction

DNA hairpins were constructed through the stepwise enzymatic ligation of four sepa

rate synthetic oligonucleotides (Table 4.1, Integrated DNA Technologies and Operon).

In the first step, oligo A and oligo B were annealed to the complementary oligo Z in 20

mM Tris, 100 mM NaCl, pH 8.0 by heating to 95°C for 5 minutes and slowly cooling

at 0.1°C/s to 4°C. 20 units of E. coli DNA ligase and IX E. coli DNA ligase reaction

buffer (New England Biolabs) were then added and the reaction was held at 16°C for

3 hours, after which the ligase was inactivated by holding the reaction at 65°C for 20

minutes. Oligo C, oligo D, and oligo Y were added to the solution in equimolar con

centrations along with 100 units of Ampligase DNA ligase in IX Ampligase Reaction

Buffer (Epicentre Biotechnologies). The solution was then heated to 95°C for 5 min

utes to melt all the strands, rapidly cooled to 89°C, slowly cooled at 0.1°C/s to 65°C,

held at 65°C for 2 minutes, and then reheated to 95°C for 1 minute. This cooling and

reheating process was repeated 20 times to ensure high stringency ligation. The DNA

ligation mixture was then loaded on a 15% TBE-urea denaturing polyacrylamide gel

(Invitrogen) and run at 175V for 1 hour in a 65°C water bath. The gel was stained

with SYBR Green I (Invitrogen) for 20 minutes, the band corresponding to the 259

bp DNA product was excised from the gel, and the DNA was purified as described

in Invitrogen's GeneTrapper manual. The control sequence hairpin was synthesized

as above but with oligo B replaced with B' and oligo C replaced with C'.

4.3.2 Hairpin FRET Calibration

Glass surfaces (Precision Glass Optics, D-263T cut glass, 0.145 mm, 2"xl" 40/20

surface quality) were RCA cleaned and a plastic hybriwell (Grace Biolabs) was placed


Figure 4.8: Gel purification of the ligated hairpin. The ligation reactions were gel purified on a 15% TBE urea polyacrylamide gel run at 60°C for 1 hour at 174V. Red staining indicates SYBR-Green, blue indicates Cy3, and green indicates the LIZ-500 size standard. (A) Lanes 1-7 show 0.5 /JL, 1.0 yuL, 2 /xL, 4 //L, and 8 fjL of ligation reaction products. Lanes 8-11 are the products of a ligation reaction missing oligo C. Lane 12 contains the LIZ-500 size standard. (B) The top band on gel (A) corresponding to the 259 nt product was cut out and purified as described in the text. Lanes 1-2 contain 0.5 fiL of the purified product loaded on to a gel to assess purity and yield. Lane 3 contains the LIZ-500 size standard.


BHQ-2

Cy3

B 5' s T

_ACTCTGCTGCGCCATCCGCGGTCTATACGCTAGGT TT

^^GAGACGACGCGGTAGGCGCCAGATATGCGATCC T T

• 6 _ A C

G +1 +15 +30 3' T , GC 5 GC 5 §£ Tj. f BHQ-2

r51 , ^^iTCT^lGflHATCCHGGTCTATlCH^GG7 T 5' 3'

I^UU

. 1000 Z3

£ 800 >* OT 600 CD

c 400

200

n

c

'/^* '

*

•

'

5 10 15 20 25 30

position / bp

GACPAG.

35 #0x3

35

D 30

£-25

"c 20 .O % 15 o tt10

?CCHGGTCTAT|CJHAI LGGHCCAGATA|G^BT< ITCC T ,

a ^

200 400 600 800 1000 1200

intensity / a.u.

Figure 4.9: FRET can be used to identify the real-time position of single DNA polymerase molecules. (A) Cartoon of the DNA template used to measure real-time polymerase kinetics. Primed 259 nt DNA molecules containing internal 33-bp hairpins and flanking 94 bp tails were immobilized on a glass surface through a biotin-streptavidin linker. DNA replication through the hairpin resulted in a reduction in FRET efficiency between Cy3 and BHQ-2, giving rise to an increase in Cy3 fluorescence. (B) Sequence and structure of the two DNA templates used in these experiments. The pink boxes highlight the differences in sequence between the sample (top) and the control (bottom) hairpins. Sequence numbering refers to the template position awaiting nucleotide addition. (C) Cy3 fluorescence recovery data for the primer extension calibration experiments is shown in red. The position along the x-axis refers to the polymerase position along the template, the blue curve is the least-squares fit, and the dashed lines are 95% confidence intervals. The error bars on the data points are the standard error of the mean. (D) Intensity-to-position curve based on (C) to convert an arbitrary intensity to a template position. Each rectangle represents the 95% confidence interval for the expected intensity at each polymerase position along the template.


on the surface to create a fluidic chamber. The surface was incubated with 0.2 mg/mL

biotinylated-BSA in IX PBS for 20 minutes, washed extensively with IX PBS, and

then incubated with 0.5 mg/mL neutravidin in IX PBS for 20 minutes. Purified

DNA hairpin was annealed with a 5' biotinylated primer (oligo Z) as described above,

diluted to 10 pM in IX PBS, and incubated on the neutravidin surface for 20 minutes.

Surfaces were imaged on an inverted Nikon TE2000S microscope with a 60X 1.45

NA PlanApo TIR objective, a low-fluorescence immersion oil (n = 1.515), and an

automated XY stage (Mad City Labs). A Nikon TIRF attachment in conjunction

with a 532 nm diode-pumped sold-state laser (Meshtel) and a custom laser-launch

system (Thor Labs) was used for illumination. A Cy3 filter set (HQ535/50, Q565LP,

HQ610/75, Chroma) and a Photometries Cascade II EMCCD were used for filtering

and image acquisition, respectively.

Immediately prior to enzyme addition, surfaces were imaged to ensure the hairpin

was properly folded as evidenced by few or no Cy3 molecules detected in the field

of view. For the FRET calibration experiments, an enzyme mixture consisting of

IX polymerase # 2 reaction buffer (New England Biolabs), 100 /xM of dATP, dTTP,

dCTP, 10 units of DNA polymerase (Klenow exo~, New England Biolabs), 0.1 mg/ml

glucose oxidase, 0.2 mg/mL catalase, 10% w/w glucose, and 1 mM Trolox was flowed

into the chamber. After 5 minutes of incubation at room temperature, the surface was

washed extensively with IX PBS and 20 different fields of view were quickly imaged to

minimize photobleaching. Fluorescent intensities represented a polymerase position

at -6 bp with respect to the hairpin (ie. the polymerase active site is 6 bp away from

the first bp of the hairpin stem, see Figure 4.9). This process was repeated with the

dNTP mixes shown in Table 4.2 to partially extend the primer to a series of known


Table 4.2: Nucleotide combinations used for stepping through the hairpin.

Nucleotides used 1. dATP dCTP dTTP 2. dATP dCTP dGTP 3. dGTP dCTP dTTP 4. dATP dCTP dTTP 5. dATP dCTP dGTP 6. dATP dCTP dTTP 7. dATP dCTP dGTP 8. dATP dCTP dTTP dGTP

Stop position

-6 -2

+13 +17 +21 +28 +30 n/a

positions (Figure 4.10).

4.3.3 Single Molecule Kinetics Experiment

For the real-time kinetics experiments, surfaces were prepared as described above.

An enzyme mixture consisting of IX polymerase # 2 reaction buffer (New England

Biolabs), 100 /zM dNTPs, 10 units of DNA polymerase (Klenow exo- or 029, New

England Biolabs), 0.1 mg/ml glucose oxidase, 0.2 mg/mL catalase, 10% w/w glucose,

and 1 mM Trolox was flowed into the chamber. Images were immediately acquired

every 200 msec for approximately 20 minutes. Freshly prepared 1 M betaine was

added to the above mixture for the betaine experiments. For the temperature ex

periments, a commercial flow cell with internal heating elements (CFCS2, Bioptechs)

was used in conjunction with an external temperature controller to hold the sample

temperature at 37°C. Reaction mixtures were pre-heated to 37°C prior to addition.


5' S

position: -6

500 1000

position:+21

1500

*CTCTGCTGCGCCATCCGCGGTCTATACGCTAGGTTT "ZACIGAGACGACGCGGTAGGCGCCAGATATGCGATCC T ™

G 3' T GC GC GC TA

+1 +15 +30

position: -2 position:+13

0 500 1000 1500

position: +28

0 500 1000 1500

position: +30

500 1000 1500 0 500 1000 1500 0

intensity/a.u.

500 1000 1500

position:+17

0 500 1000 1500

position: full extension

0.2

0.1

J11IL 0 400 800 1200 1600

Figure 4.10: Partially extended primers permit FRET distance calibration. (Top) Primed DNA hairpins were partially extended to known positions on the template by adding polymerase and only three of the four dNTPs. Template stop positions (highlighted in red) refer to the adjacent position awaiting nucleotide addition. The dNTP sets and the corresponding stop positions are shown in Table 4.2. (Bottom) Each set of intensities was fit to a Gaussian distribution to determine the mean intensities and the s.e.m. for the FRET calibration curve.


Image analysis

Custom software was written in MATLAB to automatically identify and track the

XY coordinates of single molecules using a previously published particle tracking

algorithm [107]. A 3x3 pixel grid surrounding each molecule was integrated and the

5x5 pixel perimeter was used to calculate the background. Raw and smoothed net

intensity counts for each trajectory were then plotted and manually analyzed. Steps

were identified and distinguished from gradual non-stepped growth using a custom

step-fitting algorithm [108]. This algorithm generates 'S-values' which are a measure

for the quality of a step-fit for a given number of steps over an entire trajectory, and

the maximum value of S should correspond to the best fit. For noisy trajectories,

step trains containing a slight excess of steps were chosen to ensure that all true steps

were included. Some trajectories like were scored as having multiple distinct pauses

using this algorithm. A few others showing slow extension were not scored because

there was no clear maximum S value; any S chosen yielded steps that were within

the RMS noise. Dwell times were calculated as the time between steps. Synthesis

rates were calculated by measuring the trajectory slope between the mean intensities

of two consecutive pauses or a pause and full extension.

4.3.4 Discussion and Results

In order to correlate the FRET signal with template position, a series of primer ex

tension experiments were done by adding DNA polymerase and only three of the four

deoxynucleotide triphosphates (dNTPs). After the first extension step, the flow cell

was thoroughly washed to remove any unincorporated nucleotides, and the process

was repeated multiple times with a different set of three dNTPs (see Methods section


above). This enabled primer extension to known positions on the template and al

lowed for the observation of the corresponding reduction in FRET signal as the donor

and acceptor were forced apart. Before extension began, the distance R between the

two dyes was small and the FRET efficiency was calculated as

EFRET = TTogF (41)

As the primer was extended and the hairpin unwound, R increased and EFRET de

creased, leading to recovery of Cy3 signal. The observed Cy3 intensity was thus given

by

Icvs = IMAX ( 1 - ( , , / J L V 3 ) 1 (4.2) 1

}+T£Y where IMAX was the maximum Cy3 intensity achievable when the dyes were as far

apart as possible (assuming 0.34 nm per base pair, the maximum dsDNA separation

was 2 x 33 bp + 5 bp loop = 71 bp « 24 nm). A weighted least-squares fit of the data

to Equation 4.2 gave IMAX — 987 and R0 = 15.3 bp « 5.2 nm, which is consistent with

the vendor's reported Forster radius for Cy3 and BHQ-2. There was some width to

the distributions of Cy3 intensities at each position which may be due to dephasing,

signal to noise variation, or Cy3-BHQ-2 orientation effects (Figure 4.10). However,

by performing this alignment procedure on an ensemble of records at each position,

a calibration curve was generated to relate the observed fluorescent intensity and the

polymerase position along the template (Figure 4.9C). Using this calibration curve

and its 95% confidence limits, an algorithm was developed to convert any arbitrary

fluorescent intensity into a corresponding polymerase location along the template

(Figure 4.9D).


Using an unconventional FRET configuration comprised of a non-fluorescent quencher

instead of a traditional fluorescent acceptor offers a number of key advantages. First,

there is no need to monitor two emission wavelengths simultaneously. This reduces

both the number of optical elements in the detection path and the background fluo

rescence, thereby improving signal-to-noise and spatial precision. In a single detector

setup, this also allows one to monitor twice as many molecules and generate twice

as much data per experiment compared to a donor/acceptor split-field setup. Using

a quencher also greatly simplifies the data analysis: there is no need to track and

correlate features between two movies because the net intensity of each feature can

be directly converted into a FRET efficiency (Figure 4.9D).

The real-time kinetics of DNA replication was examined in the presence of all

four dNTPs with two different DNA polymerases: the Klenow fragment of DNA

Polymerase I (exo-) from Escherichia coli (Pol I(KF)) and the replicative polymerase

from the Bacillus subtilis bacteriophage 029. The trajectories for both polymerases

exhibited heterogeneous behaviors that could be classified into four categories: fast

replication without pausing (Figure 4.11A), fast replication with a single pause (Fig

ure 4.1 IB), fast replication with multiple pauses (Figure 4.11C), and slow replication

(Figure 4.1 ID). Slow replication occurred in only a small portion of the traces and

while its origin is unclear, it could stem from the template sticking to the surface or

unfavorable polymerase-surface interactions. Approximately two thirds of the DNA

polymerases extended the entire template without a pause, while about a third of

the polymerases paused at least once. The pauses occurred at highly stereotyped

positions and we took advantage of the single molecule approach to measure both the

distribution between pausing and non-pausing enzymes and the relationship between


the pauses and the burst synthesis speeds of the polymerase. This data that would

be impossible to obtain with conventional bulk techniques.

One potential drawback of this approach is that blinking fiuorophores may in

correctly be interpreted as real extension events. If a Cy3-Cy5 FRET pair were

used this would certainly be a concern as Cy5 is known to exhibit frequent blinking

[109, 110]. In this FRET system a false positive extension could be caused by one

of two events: either BHQ-2's quenching ability switches off before extension begins

or a dark Cy3 fluorophore switches on after extension has finished. BHQ-2 blinking

was not significantly observed while imaging folded hairpins in the absence of poly

merase or dNTPs, is not thought to occur by the vendor, and has not been reported

in the literature. Dark Cy3 fiuorophores can make up 12% of a population and have

a dark-state lifetime of «2 sec; the on-blink and off-blink rates are 0.36 and 0.26

times per molecule over 120 sec, respectively [111]. Given the infrequent blinking and

short off times of Cy3, the probability of it being off during any real replication event

is very low. It is therefore not surprising that only 1% of the trajectories exhibited

identifiable blinking after full extension (signal going from on to off and then back

to on). Importantly, for the kinetics experiments intermediate intensity pauses were

observed in approximately 50% of the trajectories, a phenomenon that cannot be

produced by fluorophore blinking. The calibration experiments (Figure 4.9C) with

polymerases stalled on the template did not show any significant deviation from the

expected FRET behavior, so protein-induced quenching is not thought to be a sig

nificant source of error. Taken together, the majority of the signals are believed to

be due to reductions in the Cy3-BHQ-2 FRET efficiency and not blinking or protein

quenching.


1200

800

400

Pol I: 44% ^29 :51%

CO

to c CD

in 50 100 150

Pol I: 2% <p29: 4%

1200

800

400

Pol I: 6% ^29: 7%

150

Pol I: 22% ^29:15%

150

Pol I: 2% <p29: 1%

50 100 150 50 100 150 time / s

50 100 150

Figure 4.11: DNA replication exhibited heterogenous pausing. Single molecule trajectories of extension through the hairpin exhibited four distinct patterns. Molecules exhibited rapid extension without pausing (A), rapid extension with a single pause (B), rapid extension with multiple pauses (C), and slow extension (D) prior to photo-bleaching. A significant fraction of the trajectories showed full extension followed by a single step (E) or multiple step (F) decreases in Cy3 intensity before photobleaching. The grey curves are the raw data, the blue curves are the smoothed data over 5 raw data time points, and the red steps are generated by a custom step-fitting algorithm (see Methods section). The percentages show the fraction of all trajectories for each polymerase that demonstrated that particular pattern at 23°C.


The positional accuracy of individual trajectories depended on both template

location and signal-to-noise, with a maximum attainable resolution of «2 bp. For

example, a well-behaved trajectory that paused with a mean intensity of 562 ± 47

a.u. (s.d.) was called at +15 or +16 bp; a noisier trajectory that paused with a mean

intensity of 447 ± 168 a.u. (s.d.) was called at +14 ± 2 bp. By aligning an ensemble

of trajectories for Pol I(KF) and 029, polymerase pause positions were localized with

near single base accuracy over the template from +8 bp to +18 bp (Figure 4.12A

and B). Outside of this window the overlapping confidence limits of the calibration

curve prevented pause localization accuracy better than ± 3 bp. For Pol I(KF), over

85% of the pauses were located between +13 and +17 bp, a region that is GC-rich

and includes a Pyr-G-C motif at +16 bp. Similarly, over 82% of the 029 pauses were

located between +14 and +18 bp. Pyr-G-C motifs at +8 bp and +26 bp showed

weak pausing for Pol I(KF) and no pausing for 029, but the identification accuracy

of pause events is lower at those locations.

A significant reduction in the frequency of pausing was observed for both enzymes

with the addition of 1 M betaine to the reaction mixture (Figure 4.12A and B).

Betaine is a zwitteronic osmoprotectant that is thought to alter DNA stability such

that GC-rich regions melt at AT-rich temperatures [112], a finding that has led to

its inclusion in PCR formulations for improved amplification of difficult templates

[113, 114]. Betaine has also been used in bulk studies to suppress replication pausing

at certain Pyr-G-C sequences [101], suggesting that these motifs within the hairpin

might cause the pauses.

Increasing the temperature of the Pol I(KF) reaction from 23°C to 37°C also

reduced the frequency of pausing significantly, which is in agreement with previous


8%

7%

6%

5%

4%

3%

2%

1%

0%

|23C MMBetaine I37C

JL _s_ GAGACGACGCGGTAGGCGCCAGATATGCGATCC +1 5 10 15 20 25 30

CAGAGACCATTCTTGGATCCAGATAGGACTTCC +1 5 10 15 20 25 30

8%

7%

6%

5%

4%

3%

2%

1%

0%

8%

7%

6%

5%

4%

3%

2%

1%

0%

B

' IJBLII ' GAGACGACGCGGTAGGCGCCAGATATGCGATCC

+1 5 10 15 20 25 30

D

: rfflTT, - ' CAGAGACCATTCTTGGATCCAGATAGGACTTCC

+1 5 10 15 20 25 30

template position

Figure 4.12: Pause frequency as a function of polymerase position. Pause frequency was sequence dependent and was suppressed with betaine or elevated temperature. The mean intensity of each pause was mapped to the template sequence and normalized to the frequency of pausing for the given conditions. (A) and (C) show Pol I(KF) data for sample and control sequences; (B) and (D) show the data for 029.


bulk studies by Mytelka et al. [101]. These authors suggested that pauses at Pyr-G-C

sequences might be caused by difficulties in the polymerase fingers-closing conforma

tional change, as at the time this transition was thought to be rate-limiting and the

most sensitive to changes in temperature. However, a recent report [115] showed that

the slow prechemistry step is likely not the fingers-closing transition, raising the pos

sibility that these pauses are associated with an earlier DNA template rearrangement

step that might be sequence dependent.

In order to verify that the observed pausing was sequence dependent and not

due to the hairpin secondary structure, a control DNA molecule was constructed

with the same overall structure and length as the original (Figure 4.9B, top) but

with a different stem sequence (Figure 4.9B, bottom). The control sequence removed

all three occurrences of the Py-G-C motif while maintaining as much similarity to

the original sequence as possible. Importantly, only two bases were changed in the

region where most pausing was found to occur: 5'-TAGGCGCCA-3' was changed

to 5'-TAGGATCCA-3'. Both polymerases showed over a 5-fold reduction in pause

frequency with the control sequence compared to the original sequence (Figure 4.12C

and D). This result confirmed that the secondary structure of the hairpin was not

playing a role in the observed pausing. It also suggested that the central 5'-CG-3'

motif, either by itself or in the context of the surrounding sequence, was responsible

for the pause efficiency. The distribution of pause lifetimes was consistent with single-

step Poisson statistics. At 23°C Pol I(KF) exhibited a mean pause lifetime of 14.5

sec for the sample template (Figure 4.14A). In the reduced population of trajectories

that showed pausing even with the addition of betaine or heat, a 40% decrease in

the mean pause lifetime was observed (Figure 4.14B-C). A similar reduction in pause


0.25 C <u 3 O" 0.2 <U

<4—

T3 °-15

<D N

^ 0-1

0.05

t_

0.25

0.15

0.05

p. B

— i

iln 8 10 12 0 2 4 6

S/N

8 10 12

Figure 4.13: Histograms of S/N ratios for (A) trajectories during a pause and (B) after full extension. Over 100 individual trajectories were analyzed and the mean S/N ratio for each one was calculated by taking the net signal and dividing it by the RMS noise (standard deviation of the signal). The mean values were calculated to be 2.6 for pause intensities and 5.6 for full intensities.

lifetime was observed for Pol I(KF) acting on the control template (Figure 4.14D).

The mean pause lifetimes for 029 with the sample template were less than for Pol

I(KF) (Figure 4.14E), and a reduction in lifetimes was not observed with betaine

addition (Figure 4F) or on the control sequence (Figure 4.14G). Short time period

pauses are likely undersampled due to limitations on our sampling bandwidth and

signal to noise, which results in some of the pause lifetime distributions to have a

distinct rise and decay. The pause intensities had a mean signal to noise ratio of 2.6

while the full extension intensities had a mean signal to noise ratio of 5.6 (Figure 4.13).

Pause frequencies and lifetimes are impossible to obtain through ensemble studies and

this knowledge may help constrain the biophysical models of the polymerase kinetic

pathway.

The pauses observed in this study are clearly distinct from those observed in pre

vious single molecule DNA polymerase experiments. T7 and Pol I(KF) polymerases


N O

50%

40%

30%

20%

10%

0%

, / \

7=14.5 s

0 25 50 0 25 50 0 25 50

0 25 50 0 25 50

pause time / s

Figure 4.14: Pause lifetimes for Pol I(KF) and 029 were measured under different conditions for the sample and control sequences. The red curves are normalized single exponential fits given by / = exp(—t/r), where r is the mean pause lifetime. (A) Pol I(KF) at 23°C with the sample template; (B) Pol I(KF) at 23°C with 1 M betaine with the sample template; (C) Pol I(KF) at 37°C with the sample template, (D) Pol I(KF) at 23°C with the control template, (E) 029 at 23°C with the sample template, (F) 029 at 23°C with 1 M betaine with the sample template, and (G) 029 at 23°C with the control template.


have been shown to exhibit long pauses of heterogeneous length under template ten

sion [98, 99]. These pauses were thought to be due to fluctuating hairpins, exogenous

DNA hybridization, or template damage. The entire T7 replisome was also shown

to halt leading-strand synthesis due to primase activity on the lagging strand [103].

Similar experiments with the E. coli replisome identified a relationship between pause

times and core polymerase concentration, suggesting that pauses can be caused by

enzyme dissociation events [104]. A recent study on real-time DNA sequencing with

single 029 polymerase molecules identified pause sites that corresponded to regions

with predicted template secondary structure [16]. However, these experiments were

performed with a rate-limiting dNTP concentration which complicates the discovery

of novel pause sites. Here 029 and Pol I(KF) are shown to be susceptible to sequence-

dependent pausing at saturating dNTP concentrations and the pauses are known not

due to template secondary structure.

DNA synthesis is therefore a combination of high activity rapid synthesis that

reflects the intrinsic "speed limit" of DNA polymerase with low activity pauses; these

single molecule measurements enable us to separate these contributions and measure

their kinetics. DNA synthesis burst rates were calculated by measuring the slopes of

the trajectories between pauses (Figure 4.15A inset). The mean rates for Pol I(KF)

at 23°C without (Figure 4.15A) and with betaine (Figure 4.15B) were similar at 16

± 14 nt/s (s.d.) and 24 ± 20 bp/s (s.d.) respectively. Pol I(KF) had a higher mean

replication rate of 25 ± 20 nt/s (s.d.) at 37°C (Figure 4.15C). Although synthesis

rates are dependent on template sequence, temperature, and buffer conditions, the

rates measuredg are 2-3 fold faster than the only other Pol I(KF) single molecule

measurement in the literature [98], a low tension force spectroscopy study that did


not have sufficient time resolution to separate pauses from bursts. Assuming that

Pyr-G-C pause sites occur once every 32 base pairs in a random template and given

that the pause motif is 25% efficient, one would expect the net synthesis rate over long

templates to be RS6 nt/s, a rate that is comparable to the previous single molecule

measurement. A recent bulk study of Pol I(KF) reported a strand displacement

synthesis rate over an 18 bp template of 1.2 nt/s [116], but this template contained

three Pyr-G-C pause motifs. Based on the results of our single molecule experiments,

Pol I(KF) would likely have 1.1 sec of synthesis time and (14.5 sec) (26%) (3) = 11.3

sec of pause time over this template, giving an estimated ensemble rate of 1.5 nt/s,

in good agreement with the bulk study. For 029 at 23°C, a wide distribution of

synthesis burst rates were observed ranging from a few nt/s up to 150 nt/s with a

mean rate of 48 ± 32 nt/s (s.d.) (Figure 4.15D). These results are consistent with

reported 029 bulk rates of 38 nt/s at a higher temperature (30°C) that likely included

pausing events [117]. This assay is only limited by its integration time in being able

to measure very fast rates. For example, a 100 nt/sec synthesis event detected at 10

Hz should provide at least three frames of variable intensity.

For Pol I(KF) molecules that paused, the mean replication rate over the template

sequence up until the pause site was 11 ± 1 nt/s (s.e.m.); after the pause site, the

mean rate increased to 18 ± 2 nt/s. Similar to Pol I(KF), 029 molecules that paused

underwent faster replication after arrest (41 ± 3 nt/s (s.e.m.)) than before (24 ±

4 nt/s (s.e.m.)). Accelerating synthesis rates were also observed in trajectories that

did not pause (Figure 4.16). The small difference in rates before and after pausing

may be due to the transition near the end of the hairpin from strand displacement

replication to displacement-free replication. It could also be an artifact related to the


T3 O

"(5 £

50 100 150

30%

25%

20% j

15%

10%

5%

0% ..

c

-

Tltoi 50 100 150 0

rate / bp/s

Figure 4.15: "Speed limit" replication rates for Pol I(KF) and 029 were measured for single DNA polymerase molecules under different conditions. For trajectories that paused, only data in the intervals between pauses was analyzed; for trajectories that did not pause, data over the entire interval was analyzed. Rates were determined by calculating the average intensities at the start and finish of active replication through the hairpin, converting the intensities to polymerase positions, and then calculating the slope (A, inset). The normalized distribution of rates are shown for (A) Pol I(KF) at 23°C, (B) Pol I(KF) with 1 M betaine at 23°C, (C) Pol I(KF) at 37°C, and (D) 029 at 23°C.


presence of the fluorophore in the template.

4.4 The Energy Landscape and Kinetics of

Hairpin Refolding

Surprisingly, approximately 25% of all complete extensions did not exhibit single

step photobleaching from the peak intensity; instead, either a single step decrease

to a lower intensity (Figure 4. HE) or multiple step decreases (Figure 4.1 IF) were

observed prior to loss of signal. The newly-synthesized double-stranded DNA con

tained an inverted repeat so this may have been due to the stepwise refolding of two

complementary hairpins to form a cruciform structure. The average refolding pause

time was measured to be 7.4 sec with an average intensity of 380 ± 26 a.u. (s.d.)

(Figure 4.17). A previous study measuring Holliday junction branch migration re

ported an average dwell time of 9.7 sec between steps for four different sequences [119].

This is in reasonably good agreement with these measurements, especially given that

photobleaching may systematically reduce the measured dwell times.

The presence of betaine drastically reduced the frequency and duration of pauses

during refolding, which suggests that they arose due to difficulties in melting GC

base pairs. To explore this idea, the theoretical energy landscape of the FRET-

observable refolding transition was calculated using mfold (Figure 4.18) [118]. The

average refolding pause intensity corresponded with a DNA structure containing two

28 bp hairpins (Figure 4.19A); further hairpin growth required overcoming an energy

barrier and breaking additional GC base pairs in the non-hairpin strands. This cal

culation is based on work by Karymov et al. [120] who reported that when Cy3/Cy5


•<p29

Pol l(KF) 23C

Pol l(KF) 37C

Poll(KF)23C + betaine

(^29 + betaine

Pol l(KF) 23C control sec

0.2 0.4 0.6 0.8 1

time /s

1.2 1.4 1.6

Figure 4.16: Average trajectories for polymerase molecules that did not pause. Trajectory data was interpolated with a cubic spline on a 20X finer mesh to further improve the time resolution. For each experimental condition, over 100 of these interpolated trajectories were sampled to produce the "average" trajectories above. Sampling was done by first identifying the start and finish times for each trajectory during which replication was observable. Each time interval was then divided equally into eight subintervals and the intensity at each time point was recorded. The time and intensity values for each of these subintervals were then averaged to produce the corresponding "average" trajectory.


10 20 30 pause time / s

200 400 600 800 1000 intensity /a.u.

Figure 4.17: Histograms of pause durations and intensities locations during hairpin refolding. The bins were normalized based on the frequency of refolding for all trajectories that reached a fully extended state. (A) Refolding pause times in the absence or presence of 1 M betaine. The red curve is a normalized single exponential fit given by / = e - t / r , where r is the mean pause lifetime. (B) Average intensities of pauses during refolding with and without betaine. The average intensity was measured to be 380 ± 26 a.u.

are separated across a Holliday junction by 10 bp or 12 bp, the FRET efficiencies

are 73% or 54%, respectively. Given that the Forster radius of Cy3/Cy5 is 5.3 nm,

these distances across the Holliday junction work out to be 4.5 nm and 5.2 nm. The

mean refolding intensity for Cy3/BHQ2 was 380 a.u. which corresponds to a FRET

efficiency of 60%. Integrated DNA Technologies reported that the Forster radius of

this pair is 5.02 nm. Using these numbers, the distance between the Cy3/BHQ2 was

calculated to be «4.7 nm, which is more consistent with 10 bp of separation than 12

bp assuming 2 bp steps. The presence of GC-rich sequences at the base of the stem

are consistent with this conclusion, suggesting that most of the refolding pauses are

due to cruciforms trapped in this energy well. Note that the geometry of the system

is significantly different in the absence of DNA polymerase and in the presence of an

extruded cruciform and the construct does not have the same 2 bp spatial resolution


-429

-430

-431 I

o J-432 U

^ 3 3 <a

-434

•435

-436

24 25 26 27 28 29 30 31 32 33 34 35 hairpin length / bp

Figure 4.18: The observable energy landscape of cruciform structural transitions. The Gibbs free energies for every hairpin length and the corresponding intermediates were calculated using mfold [118]. The hairpins of integer length have fully paired nucleotides; the hairpins with non-integer lengths contain 4 unpaired nucleotides in the cruciform center (e.g. Figure 4.17B). The structures of the labeled intermediates (A-G) are given in Figure 4.19.

J i i i i i i i i


in this configuration.

Cruciform extrusion has been well studied and there are considerable kinetic bar

riers that need to be surpassed for it to occur (31), so it is perhaps surprising how

often it was observed immediately following replication. Additional study is required

to identify what factors are facilitating extrusion; one possible explanation is that

DNA polymerase exerts sufficient torsional stress on the template to reduce the en

ergy barrier for extrusion to occur.

4.5 Future work

The real-time single molecule observation of Pol I(KF) revealed that the polymerase's

intrinsic "speed limit" is 2-3 fold faster than reported by previous single molecule

measurements. Replication can be interrupted by heterogeneous sequence-dependent

pauses that are independent of template secondary structure, and knowledge of these

pause frequencies and lifetimes would otherwise be impossible to obtain from ensem

ble studies. The pauses were highly localized to a short GC-rich sequence motif and

were relieved with the addition of heat or betaine, suggesting that they may be asso

ciated with a structural template rearrangement step during the polymerase reaction

pathway. Future single molecule experiments should help fully elucidate the mecha

nism of DNA polymerase sequence-dependent pausing and provide insight into how

other endogenous and exogenous factors act to slow or stall replication forks.


Figure 4.19: Structure of the DNA cruciform intermediates as shown with mfold. The Gibbs free energy of each structure is shown in Figure 4.18. Structure (A) is the suspected intermediate responsible for the majority of pausing during hairpin refolding.

Chapter 5

High Tempera ture Single Molecule

Imaging with Colloidal Lenses

5.1 Introduction

Although single molecule fluorescence spectroscopy was first demonstrated at liquid

helium temperatures (1.8 K) [2, 3], the field has since grown to include wide-field

room temperature observations [121] largely due to advances in brighter fluorophores,

better objectives, and more sensitive detectors. This has opened the door for many

chemical and biological systems to be studied at native temperatures at the single

molecule level both in vitro [7, 4, 5] and in vivo [10, 8, 9]. However, systems and

phenomena that operate at temperatures above 37°C remain difficult to study at the

single molecule level due to the index matching fluid requirements of most commercial

high NA objective lenses. These fluids act as a thermal conductor between the sample

and the objective and sustained exposure to high temperature can cause the objective

to fail. This has prevented the single molecule study of thermophilic organisms, the

89

CHAPTER 5. SINGLE MOLECULE COLLOIDAL LENSING 90

interactions of their protein repertoire, and the temperature dependent unfolding

kinetics of nucleic acids and proteins. This is one reason why there are no published

reports of single molecule fluorescence detection at high temperature. The challenge is

to develop an optical technique that is capable of detecting single fluorescent molecules

with high collection efficiency while operating at a long working distance without an

index matching fluid.

Wenger et al. [122] recently demonstrated that a latex microsphere in conjunction

with a low numerical aperture objective lens can be used for fluorescence-correlation

spectroscopy with near-single molecule sensitivity. Previous work in the field demon

strated the ability of hemispherical solid immersion lenses [123] and spherical colloidal

lenses to construct a rotational sensor [124], to create an optomagnetic dimmer [125],

to generate two-dimensional micropatterns [126, 127] and nanopatterns [128], and to

optically couple one dimensional arrays of colloidal particles [129, 130]. An efficient

nanolens system based on gold nanospheres was also shown to exhibit a strong electro

magnetic surface-enhanced Raman scattering (SERS) enhancement [131]. However,

none of the aforementioned techniques achieved true single molecule sensitivity. In

this chapter, high index of refraction micron-sized colloidal lenses are shown to be

capable of achieving single molecule sensitivity by incorporating a focusing element

in immediate proximity to an emitting molecule or nanoparticle; the optical system is

completed by a low numerical aperture optic which can have a long working distance

and an air interface. The colloid acts as a lens and dramatically improves the photon

collection efficiency of the optical system. To demonstrate the potential applications

of colloidal lenses in single molecule spectroscopy, 2.0 //m Ti0 2 colloids were used with

with a 20X 0.5 NA air objective to image single quantum dot nanoparticles at 23°C,


single DNA and protein molecules at 23°C and 70°C, and the real-time dynamics of

mesophilic and thermophilic DNA polymerases at 23°C and 70°C, respectively.

5.2 Theory of Colloidal Lensing

5.2.1 Geometric Optics

Geometric optics arguments can be used to estimate how much a micron-sized col

loidal lens increases the amount of light collected from a point source. The configu

ration is illustrated in Figure 5.1 A and has been previously described for the lensing

of a smaller fluorescent microshere [124]. For configurations where the radius of the

colloidal lens is much greater than the distance S from the lens to the photon source,

as is the case for a micron-sized lens and a nanometer-sized chemical linker, the light

ray exit angle from the colloidal lens is given by [124]:

<f/-e" = 2wr1 (— one) -e (5.1)

where n2 is the index of refraction of the colloid, rix is the index of refraction of the

environment, 9 is the angle of incidence, and </>' — 6" is the exit angle. Equation

5.1 is shown with ni = 1.33 for a variety of different values of n2 in Figure 5.IB.

By selecting spherical colloids composed of a very high index of refraction material,

such as amorphous Ti0 2 (n2 ~ 2.0), the absolute value of the exit angle given by

Equation 5.1 in water (ri\ = 1.33) is always less than 25° even for very large angles

of incidence. If 0o is the semi-aperture of the external microscope objective, any exit

ray with 0' — 6" < 0o should therefore be collected. Based on this argument, TiC>2


colloidal lenses can therefore improve the effective numerical aperture of a low NA

air objective (0.5 NA, 60 = 30°) to NA e / / = (2.0) sin(84°) = 1.99. Higher values of 9

likely cause the light to undergo total internal reflection within the colloid resulting

in some losses.

5.2.2 Maxwell's Equations

Maxwell's Equations describe the interactions of electric (B) and magnetic (H) fields

with one another and with matter and sources. They can be solved analytically for

simple systems but generally requite a numerical solution. The equations for the

evolution of the fields are:

~ = - V x E - J B - ( 7 B B (5.2)

^ = V x H - J - ^ D (5.3)

B = txK (5.4)

D = eE (5.5)

where D is the displacement field, e is the dielectric constant, J is the current density,

JB is the magnetic charge current density, B is the magnetic flux density, pi is the

magnetic permeability, and H is the magnetic field.

Solving Maxwell's Equations for the colloidal lensing system should give a more

accurate prediction for how the system actually behaves. In order to approximate

the solutions to these equations numerically, a finite difference time domain (FDTD)

method was used. FDTD methods divide space and time into a finite rectangular

grid and evolve the equations over time. In the colloidal lensing system, the aim is to


Figure 5.1: Colloids can be used as lenses for quantum dot nanoparticles and single fluorophores. (A) Light ray tracing of photons emitted from a nearby single fluo-rophore undergoing lensing through a colloid. The diagram defines the angles 9, 9', 9", 4>, $ and the distances r and 5 as previously described [124]. (B) A plot showing the ray exit angle ((f>'-9") vs. the emission angle from the fluorophore (9). The dashed horizontal lines define the ray collection limits for a 0.5 NA objective. Increasing the index of the colloid (n2) improves the light collection efficiency. (C) Scanning electron microscopy image of a single 2.0 (im T i0 2 colloid confirms they are spherical in shape. Scale bar = 1 [im. (D) 3D simulation of a point source (free space A0 = 570 nm) in close proximity to a 2.0 lira spherical dielectric (n2 = 2.0) in an aqueous environment (n! = 1.33). The E2 fields are shown in blue/red in the z = 0 plane.


calculate the power transmitted by a point source at frequency u through the colloid.

This can be written as the integral of the Poynting vector over a plane on the far side

of the colloid:

P(u) = Rn- [E^X)* x H„(x)d2x (5.6)

In order to make sense of a transmission spectrum it should be be normalized and

compared to something else. Simulations were therefore run typically run once with

the colloid present and once with the colloid absent. Each dataset was then normalized

to the total power emitted by the dipole during the same time interval. This was

accomplished by setting up 6 flux planes that completely surrounded the dipole and

summing the total power passing through them all.

5.2.3 FDTD Simulation Methods

Finite difference time domain simulations were performed using a freely available

software package (Meep, http://ab-initio.mit.edu/wiki/index.php/Meep) on an 8-core

server with 16GB of RAM running Debian Linux 2.6. A three dimensional region of

space 13 /im x 13 /im x 13 /zm was discretized so that the mesh size was A = Ao/20 or

smaller to ensure accuracy for near-field effects. The medium was assigned an index

n\ = 1.33 with perfectly-matched layers as boundary conditions to absorb all waves

incident on them, with no reflections. A dielectric sphere of index ri2 and radius

r was placed at the origin and a dipole emitter was positioned a distance S away

from the sphere along the positive y-axis. The dipole was oriented perpendicular to

the x-y plane and began emitting as a continuous Gaussian source centered at the

wavelength A0 with width 0.05Ao. Square flux planes in the x-z plane were positioned

at y = -2.75 fj,m and varied in size according to the collection angle of the objective

http://ab-initio.mit.edu/wiki/index.php/Meep


being modeled. For example, an objective with a half-collection angle of 60° focused

at the dipole location would have a light collection cone radius of (tan(60°)2.75 fim)

= 4.76 fim at this y-position, so its flux plane was modeled with dimensions of 9.5

//m x 9.5 pLm. Half-collection angles ranging from 1° to 65° were included for all of

the calculations. Simulations were run for at least 50 periods or to the point where

running for additional periods did not change the calculated power through the flux

plane, whichever was longer. Each of the key parameters of the system (AOJ r, ni,

8) was systematically varied to determine the how it affects the calculated power.

Sample Meep code is provided in Appendix C.

5.2.4 FDTD Simulation Results

Simulations were performed with the FDTD method to confirm the geometric optics

predictions and to explore how key parameters of the system affect the collection

efficiency [132]. A point source was positioned next to a spherical colloid and the

total power P{u>) transmitted through a plane on the opposite side of the lens was

calculated (Figure 5.ID). Different plane dimensions were chosen to represent the

semi-aperture light collection of 14 different microscope objectives (1° < 0o < 65°).

While keeping all other parameters constant, simulations were run to determine how

changing 8 affects P (Figure 5.2A). For 8 < 0.1 mm the simulations predicted that

a colloidal lens-aided 0.5 NA air objective (0O = 30°) has at least the same effective

NA as a 1.4 NA oil immersion objective (0O = 65°). This is in good agreement with

the geometric predictions. When the point source was placed very near to the surface

(8 < 10 nm), the colloid was actually able to focus more than 50% of the total power

emitted. This may be due to the colloid being able to support a higher mode density.


The relationship between the index n-i of the colloid and P is shown in Figure

5.2B. High index colloids provide the greatest relative enhancement for objectives with

small semi-apertures, also in agreement with geometric predictions. For example, the

collection efficiency of a n2 = 2.4 colloidal lens-aided 0.3 NA air objective is improved

nearly 8-fold compared to using the objective alone. Even high NA objectives see a

modest 2-fold improvement with the use of n^ — 1.6-1.8 colloids. The radius of the

colloid r plays a critical role on the detected power: for a A0 = 570 nm point source,

only colloids with r > 750 nm showed any appreciable enhancement for 0o > 20° (Fig.

5.2C). The calculated P for different A0 is shown in Figure 5.2D. The presence of the

2.0 [xm colloid gave similar enhancement profiles for A0 = 400 nm and 500 nm, but

the power slightly decreased for shorter or longer wavelengths.

5.3 Single Quantum Dots as Rotational Probes

To test the photon collecting limits of high index colloidal lenses, they were first

coupled to single quantum dot nanoparticles. This coupling process is governed by

Poisson statistics and the concentration of each species can be manipulated so that the

majority of colloidal lenses contain no more than one quantum dot. Carboxylated

titania colloids (2.0 /mi, Corpuscular Inc.) were washed three times in IX MES

buffer (Pierce) and activated with 1 mg/mL EDC and Sulfo-NHS in IX MES for 45

minutes. The colloids were then washed with 100 mM carbonate buffer, pH 8, and

covalently coupled to aminated quantum dots (Qdot 525 ITK amino, Invitrogen) with

stoichiometry such that the majority of colloids had zero quantum dots and only a

fraction had a single dot.


10 20 30 40 50 60 0 10 20 30 40 50 60

D ^ ^ =

/

300 nm

10 20 30 40 50 60 0 10 20 30 40 50 60

collection angle/0

Figure 5.2: 3D finite difference time domain simulations were performed containing a single dipole emitter in an aqueous environment (ni = 1.33) while varying key parameters of the system. Unless otherwise specified, the point source was positioned at y = 0.010 /im (0,0.010,0) and had a free space A0 = 570 nm; the lens had an index ni = 2.0, was centered at (0,-1,0), and had a radius of r = 1.0 /im. The integrated power at the plane y = -2.75 was calculated and used to estimate the amount of light collected by an objective focused at the origin. The size of the plane was varied to model objectives with different semi-apertures. Each dataset was normalized to the total power emitted from the dipole. (A) The point source was moved a distanced 8 away from the surface of the colloid along the positive y-axis. (B) The refractive index of the lens was varied from n2 = 1.33 to 2.40. (C) The radius of the lens was varied from r = 0.010 jttm to r = 1.25 /xrn while keeping the center of the lens at (0,—r, 0). (D) The free space wavelength of the point source was varied from A0 = 450 nm to 700 nm.


Imaging was performed by placing a quantum dot colloid solution on the surface

a 3-aminopropyltriethoxysilane (APTES, Sigma) coated glass coverslip, covering it

with a second coverslip, and sealing the edges with nail polish to prevent evaporation

or fluid flow. The APTES surface is positively charged at neutral pH so it electro

statically and non-specifically immobilizes the colloids. The integrated intensity of

the quantum dot colloid conjugates was measured with both 20X 0.5 NA and 60X

1.45 NA objectives. The total intensity minus the average background for single con

jugates imaged with the 0.5 NA objective was on average within 10% obtained with a

1.45 NA objective, in good agreement with the theoretical predictions. In the absence

of the colloidal lenses it was not possible to observe single quantum dots with a 20X

0.5 NA objective.

Next, the quantum dot colloidal lenses were sandwiched in aqueous solution be

tween two clean RCA coverslips. The colloids undergo constant rotational and trans-

lational diffusion under these conditions (Figure 5.3A). When the quantum dot is

close to being in alignment with the microscope's optical collection axis, the mea

sured fluorescent intensity is significantly greater than when it is out of alignment.

Brightfield and epi-fluorescent images containing thousands of colloidal lenses under

going Brownian motion were taken with a 20X 0.5 NA air objective and a Hamamatsu

ORCA-ER CCD with a 500 msec integration time. While the majority of the lenses

were observed to be free of quantum dots, the fluorescent intensity fluctuations of

approximately 30 colloidal lens quantum dot conjugates were selected based on their

peak intensities being consistent with a single quantum dot emitter. Each conjugate

was traced over a period of 5 minutes to give over 500 different blinking events from

which to generate rotational diffusion statistics.


The probability P(t) that the intensity of a given quantum dot is above a set

threshold at time t given that the intensity was above the threshold at t = 0 can be

estimated from these fluctuations [124]. If a is the azimuthal angle over which the

signal is enhanced, the solution to the one dimensional diffusion equation with initial

condition \6\ < a is:

where Dr is the rotational diffusion constant of a sphere:

Here r\ is defined as the viscosity and a is the sphere's radius. By summing over all

multiples of 2n, Equation 5.7 can be made periodic in 9. The probability density

function for the distribution in angles is given by

n=oo

P ( M ) = Y, Pinf(d + 2rnr,t) (5.9)

n=—oo

with a total probability distribution of

/•a/2

P(t)= / p(9,t)dd (5.10) J-a/2

This integral was approximated numerically as previously described [124]. Figure

5.3 shows P(t) for quantum dot conjugates in three different glycerol-water solutions

of varying viscosities (3.4 cP, 6 cP, 12 cP). For each experiment, measured data was

least-squares fit to the theoretical solution to the one-dimensional diffusion equation


QD 7\v

enhanced dim signal signals

0.8

2 ' Q.

0.2

2 3 4 time / s

Figure 5.3: Colloidal lenses enable imaging of single quantum dot nanoparticles. (A) Quantum dot-colloid conjugates are free to rotate and translate in aqueous solution. When the pair aligns with the optical collection axis, the observed fluorescent intensity is enhanced. (B) Probability correlation functions for the fluorescent intensity of colloidal lens-quantum dot conjugates in three different glycerol-water solutions. The circles represent the observed rotational statistics of approximately 30 conjugates and 500 blinking events over 5 minutes. An fixed intensity cutoff was chosen for all three experiments to define when lensing was taking place. Error bars represent the standard error of all conjugates for each solution. The solid lines represent least-square fits to the experimental data with a = 0.3 rad to represent the set intensity cutoff. The rotational diffusion coefficient was the single free parameter. Rotational diffusion coefficients were found to be D^cP = 0.0146 sec -1, D6cp = 0.0247 sec -1, and DZACP = 0.0501 sec -1. The inset shows typical intensity data from a single lens-quantum dot conjugate (77 = 3.4 cP) observed over 1 minute.


1500

D ^ 1000

+^ 'w C $i 500 c

"0 10 20 30 40 50 0 5 0 1 00 150

time / sec

Figure 5.4: Static imaging of single quantum dots without and with colloidal lenses. Grey trajectories show the raw data; red show the raw data smoothed over 5 time points. (A) Quantum dot blinking was observed with a 60X 1.45 NA objective, but was not significantly observed when coupled to TiC>2 colloids and imaged with a 20X 0.5 NA objective (B).

[124]. Experimentally determined rotational diffusion constants were found to be

within 10% of the predicted values based on glycerol-water viscosities [133]. Signal

to noise ratios (S/N) were calculated by integrating the intensity of a 3 x 3 box

of pixels around fluorescent features, subtracting the average background per pixel,

and dividing by the root-mean-square variation in the signal. Peak S/N for single

quantum dots in the absence of colloidal lensing with a 60X 1.45 NA objective was

«10; with a 20X 0.5 NA objective with colloidal lensing it

One potential source of error in this experiment is quantum dot blinking. When

the quantum dots were immobilized on the APTES surfaces at low concentrations,

intermittent blinking behavior was observed on timescales ranging from 1-10 sec (Fig

ure 5.4A). However, this behavior was not observed in the presence of the colloidal


lenses with a low NA objective (Figure 5.4B). One possible explanation is that quan

tum dot blinking is suppressed by electromagnetic interactions with the nearby metal

or metal oxide, a phenomena also reported by others [134, 135]. There should also be

zero correlation between quantum dot blinking "on-time" and the solution viscosity.

5.4 Enhanced Fluorescence Effects

There has been recent interest in exploring the fluorescence spectral properties of or

ganic fluorophores near metal films [136, 137] and nanoparticles [138, 139, 140]. Much

of this work has been focused on Au and Ag nanoparticles and island films, where an

increase in photostability, brightness, and radiative decay rates of fluorophores has

been reported. There have also been reports of enhanced fluorescence near metal-

oxide surfaces [141] and nanoparticles [142]. To see if TiC>2 colloids give similar ef

fects, the following experiment was performed. A dilute solution of Cy3-streptavidin

(1 ng/mL, Invitrogen) was incubated on a biotin-BSA coated glass coverslip (see

Chapter 3) for 20 minutes. Surfaces were then imaged on a microscope with epi-

fluorescence excitation with a 60X 1.45 NA objective (Figure 5.5A) and fluorescence

intensities (Table 5.1 and lifetimes were recorded (Figure 5.5C).

The surface was then incubated with a 1.3% solution of biotinylated 2.0 fxm. TiC*2

colloids overnight, unbound colloids were washed away, and the surface was imaged

again under the same illumination conditions (Figure 5.5B). In this configuration

the colloids were not in the light collection path of the microscope. However, Cy3

fluorophores showed a «45% increase in fluorescence lifetime (Figure 5.5D) and a

wl5% increase in photons/sec emitted (Table 5.1) when they were near TiC*2 colloids

compared to when they were absent. An increase in the Cy3 radiative decay rate


>-u c

a-

N

£

0.40

0.30

0.20

0.10

20 40 60 80 0 20 40 60

Cy3 photobleaching lifetime / sec

Figure 5.5: Enhanced fluorescence effects of Ti0 2 colloids. Cy3-streptavidin was imaged using a 60X 1.45 NA objective in an epi-fluorescence configuration without (A) and with (B) biotinylated-Ti02 colloids coupled to the protein. Note that in this configuration the lensing effects of the colloid are not used. The distribution of measured Cy3 fluorescence lifetimes for each configuration are shown in (C) and (D), respectively. Red curves are single exponential fits to the distributions.


Table 5.1: Photon statistics for colloidal Ti02-enhanced fluorescence.

Objective

60X 0.5 NA epi 60X 0.5 NA epi

colloids

N Y

dry/wet

dry dry

photons/sec

3.1 x 104

3.6 x 104

< r > /sec

11.2 16.3

total photons

3.5 x 105

5.9 x 105

from the excited state is one possible explanation for this effect. Decreasing the time

molecules spend in the excited state reduces the time they are able to react with singlet

molecular oxygen, thereby reducing the window of opportunity for photobleaching.

This may also increase the number of excitation-emission cycles each fluorophore is

capable of going through prior to photobleaching, possibly leading to the observed

increase in fluorescent intensity.

5.5 Single Molecule Imaging with Colloidal Lenses

The ray optics arguments and FDTD simulations predict that high index colloids

should enable the same light collection ability as the highest NA objectives currently

available. To test this hypothesis, colloids were coupled to single molecules of Cy3-

labeled streptavidin as previously described. RCA clean glass coverslips were coated

with biotinylated bovine serum albumin (Pierce) and functionalized with a sparse

population Cy3-streptavidin (Invitrogen). The optical setup for the epi-fluorescence

configuration was based on a Nikon TE-2000S inverted microscope equipped with

a Nikon Plan Fluor 20X 0.5 NA air objective. Sample illumination was provided

by a mercury arc lamp and filtered with a Cy3 filter cube (HQ535/50x, Q565LP,

HQ610/75m, Chroma Technology Corp). Fluorescence emission was collected by the

60X 1.45 NA oil immersion objective and detected by a high sensitivity 512x512


pixel Photometries Cascade II EM-CCD. After verifying that the molecule density-

was « 1 molecule / 100 /xm2, biotin-decorated 2.0 yum Ti02 colloids were coupled to

the surface-bound protein. Following a 4 hour incubation, unbound colloids were

washed away, the surface was dried, and the colloids were imaged using brightfield

and epifluorescence excitation with a 20X 0.5 NA air objective (Figure 5.6A and

B). Signals were only detected where colloids were immobilized. Each signal showed

a stepwise decrease to background which was indicative of the presence of a single

fluorophore (Figure 5.6B). The surface was heated from 23°C to 70°C using resistor

heaters and single step photobleaching was still observed even at high temperature

(Figure 5.6C). In the absence of colloids no signal was detected with a 20X 0.5 NA

objective.

The light collection efficiency was similar in each of these experiments («7 x 103

photons /sec). This was slightly less than the collection efficiency of a 60X 1.45 NA

objective in the absence of colloidal lenses («3 x 104 photons/sec), possibly because

the fluorophores were never perfectly aligned with the optical axis of the lensing

system. Interestingly, the photostability of the fluorophores when excited through

the colloid with a 20X 0.5 NA objective at 23°C increased by over 4-fold (r =16.3

sec to 70.9 sec). This may be due to the colloid increasing the local incident incident

field on the fluorophore through a "Lightning Rod" effect [143, 144].

Raman scattering by water molecules prevented single fluorophore detection in

aqueous solution with colloidal lenses and epifluorescence excitation. However, re

ducing the excitation volume with prism TIR to minimize the background did allow

for single fluorophore detection. Cy3-labeled dsDNA was coupled to TiC*2 colloids to

empirically give « 1 molecule per colloid; the constructs were then allowed to bind


Cy3-streptavidin 23t 2000 epifluorescence in air

=3

CO 15001

.-^lOOO CO

c <D 500

4-J

c

Cy3-streptavidin 23t Cy3-streptavidin 70t Cy3-DNA 23°C Cy3-DNA 70t epifluorescence in air epifluorescence in air TIR in water TIR in water

50 100

23C, dry sample

20 40 60 80 0 20 40 60 80 0 20 40 60

t ime / sec 70C, dry sample 23C, wet sample 70C, wet sample

0 50 100 150 200 250 0 50 100 150 200 250 10 30 50 70 90 10 30 50 70 90

Cy3 photobleaching time / sec

Figure 5.6: Static imaging of single fiuorophores with colloidal lenses. Raw data is shown in grey and 5-point smoothed data is shown in red. (A) Brightfield fluorescent images of a 2.0 ^m Ti02 colloid coupled to a surface through a fluorescently-labeled tether. Using a tether of Cy3-streptavidin and epifluorescence excitation of a dry sample, single step photobleaching was observed at (B) 23°C and at (C) 70°C. With a tether of Cy3-labeled dsDNA in aqueous solution excited via prism TIR, stepwise photobleaching was also observed (D) 23°C and (E) 70°C. Photobleaching lifetimes of Cy3 varied between dry (F-G) and wet (H-I) samples and decreased with temperature for both configurations.


Table 5.2: Photon statistics for Ti02 colloidal lensing.

Objective

20X 0.5 NA epi 20X 0.5 NA epi

20X 0.5 NA TIR 20X 0.5 NA TIR

T / ° C

23 70 23 70

dry/wet dry dry wet wet

photons/sec

8.2 x 103

7.8 x 103

6.1 x 103

5.4 x 103

< r > /sec

70.9 39.1 31.0 15.0

total photons

5.8 x 105

3.0 x 105

1.9 x 105

0.8 x 105

S/N

3.2 2.5 2.0 1.9

non-specifically to an aminated glass surface. Unbound colloids were washed away

and the Cy3-labeled DNA was imaged in a custom flow cell. At 23°C, signals were

only detected where colloids were immobilized and each signal showed stepwise pho-

tobleaching (Figure 3D). Stepwise photobleaching was also observed at 70°C after

taking careful precautions to minimize evaporation in custom made high tempera

ture flow cell. Cy3 fluorescence lifetimes were measured for hundreds of trajectories

from each of these experiments (Figure 5.6E-H). A «2-fold reduction in lifetimes was

observed at higher temperatures for both the dry and wet samples, which might be

explained by an increase in the reaction rates of molecules in excited or triplet states

with singlet molecular oxygen.

Single molecule imaging at high temperatures requires careful attention to the

details of the flowcell and the optical setup. For prism-TIR excitation, the viscosity

and index of the immersion oil between the prism and the top coverslip can change

with temperature. This can result in the prism actually moving during the course of

an experiment. Another challenging issue to overcome is sample loss due to evapo

ration. The flowcell has to be sealed extremely well to withstand the vapor pressure

that builds up within the chamber. Sealing the flowcell with epoxy prevents reagent

exchanges, which makes more complicated assays difficult to execute.


5.6 Measuring Polymerase Kinetics with Colloidal

Lenses

To demonstrate that colloidal lenses can enable a biological experiment that was

previously impossible due to the temperature limitations of high NA objectives, the

real-time single molecule kinetics of a thermophilic DNA polymerase were measured

at 70°C (Therminator from Thermococcus 9°N). First, however, the kinetic assay was

verified with a mesophilic polymerase in the presence of a TiC>2 colloid.

5.6.1 Escherichia coli Pol I(KF) Activity

Hairpin Synthesis

Virtually the same 259 nt DNA molecule containing an internal double-stranded 33

base pair (bp) hairpin was used as described in Chapter 5. The 3' base of the hairpin

contained an internal Cy3 FRET donor and the 5' base contained an internal non-

fluorescent FRET acceptor (Black Hole Quencher-2, BHQ-2) so that when the hairpin

was fully folded, quenching of Cy3 by BHQ-2 prevented any fluorescence emission.

The 5' tail of the construct was modified with an anti-digoxigenin group for coupling

to the surface and a 3' biotin group for coupling to the colloid.

Colloid Functionalization

Ti02-COOH colloids (2.0 fj,m, Corpuscular) were washed three times in water and

twice with a crosslinking buffer (0.025 M MES, pH 5.0). The carboxylated beads

were activated with 40 mg EDC (l-ethyl-3-[3-dimethylaminopropyl]carbodiimide hy

drochloride, Thermo Scientific) and 20 mg sulfo-NHS (N-hydroxysulfosuccinimide,


Thermo Scientific) in 400 /zL of MES buffer (0.025 M MES, pH 5.0). The colloids

were placed slowly on a rotating mixer at room temperature for 60 min to allow the

EDC and sulfo-NHS to form NHS esters. Following this reaction, the colloids were

washed twice with MES buffer and with 200 fiL of coupling buffer (0.1 M sodium

bicarbonate, 0.25 M sodium chloride, pH 8.0). The colloids were then functionalized

with a biotin-PEG layer through the addition of 20 mM amine-PEG2-Biotin (EZ-

Link Amine-PEG2-biotin, Thermo Scientific) in coupling buffer. This reaction was

left overnight to allow the primary amine groups of the biotin-PEG linker to react

with the NHS esters on the colloid surface. Excess amine-PEG2-biotin was removed

by washing twice with PBS. Neutravidin (0.5 mg/ml, Thermo Scientific) in IX PBS

buffer was added to the biotinylated colloids and allowed to react for at least 5 hrs.

The colloids were then washed twice with IX PBS and resuspended in 400 /J,L of

storage buffer (0.05 M sodium phosphate, 0.25 M NaCl, 0.1% Tween-20, pH 7.4).

Surface Functionalization

To immobilize the hairpin duplex on the surface, an anti-digoxigenin coating was

created by incubation of a solution of 0.2 mg/mL anti-digoxigenin (Fab fragments

from sheep, Roche Applied Science, Mannheim, Germany) for 5 hrs. Unbound anti-

digoxigenin was washed away with 3 mL IX PBS. Hairpin functionalized colloids were

introduced to the flowcell and incubated overnight. The flow cell was then washed

with 2 mL of IX PBS buffer.


Prism TIR Microscope

The single-molecule fluorescence experiments were performed by using a total internal

reflection (TIR) wide-field microscope built around an inverted Nikon TE-2000 with

a 20X 0.5 NA air objective (Nikon Plan Fluor). A diode-pumped 532 nm green laser

(CrystaLaser) was used as the excitation light source. Using a beam expander, the

beam diameter was increased and guided to a 20 cm focusing lens. By adjusting this

lens, the excitation beam was directed to a fused silica prism optically coupled to

the top coverslip surface through a refractive index matching oil (Nikon Type NF).

With the incident angle of the laser beam adjusted to (=369°, total internal reflection

was achieved at the interface between the coverslip and the sample solution. The

emission signal was collected with the air objective lens, passed through a 550 nm long-

pass filter (E550LP, Chroma Technology), and a bandpass emission filter (HQ580/60,

Chroma Technology) before entering the EM-CCD camera (Cascade II 512B, Roper

Scientific, Trenton, NJ). The exposure time was 500 ms for the room temperature

and 350 ms for the high temperature experiments. The laser power at the prism was

adjusted to 300 W/cm2.

Mesophilic Polymerase Assay

The extension mixture was comprised of 5 units of DNA Pol I(KF) from Escherichia

coli, 100 fiM of each dNTP, 0.1 mg/ml glucose oxidase, 0.2 mg/mL catalase, 10%

w/w glucose, 1 mM Trolox, in buffer # 2 from New England Biolabs. As described

in Chapter 4, control experiments taken with a 60X 1.45 NA oil objective without

colloidal lenses showed fast replication (Figure 5.7B), fast replication with a single

pause (Figure 5.7C), fast replication with multiple pauses (not shown), and stepwise


7 < <BHQ-2

SCy3

Pol l(KF) 23C 60X 1.45 NA

B

50 100 150

Pol l(KF) 23C 20X 0.50 NA

Therminator 70C 20X 0.50 NA

0 50 100 150

time/s

H

fi I •• .h,..u.1 liit

0 50 100

Figure 5.7: Single molecule DNA polymerase activity observed with colloidal lenses and a 20X 0.5 NA air objective. (A) Cartoon of the DNA template used to measure real-time polymerase kinetics. Primed 259 nt DNA molecules containing internal 33-bp hairpins and flanking 94 nt tails were immobilized on a glass surface. DNA replication through the hairpin resulted in a reduction in FRET efficiency between Cy3 and BHQ-2, giving rise to an increase in Cy3 fluorescence. Trajectories of this process imaged with a 60X 1.45 NA oil-immersion objective without colloidal lenses show different behaviors, including (B) fast extension, (C) pausing during fast extension, and (D) refolding of the hairpin into a cruciform structure. Trajectories exhibiting similar behaviors were observed with a 20X 0.5 NA air objective with colloidal lenses at 23°C with Pol I(KF) (E-G) and at 70°C with Therminator (H-J).


refolding of the hairpin after replication (Figure 5.7D).

It was not possible to see any single molecules on the surface after switching to

a 20X 0.5 NA air objective under these conditions. With colloidal lenses, however,

addition of the extension mixture to the flowcell generated signal and replication

through the hairpin was observed. Single molecule trajectories taken with the colloidal

lenses and a 20X 0.5 NA air objective showed fast replication (Figure 5.7E), fast

replication with a single pause (Figure 5.7F), fast replication with multiple pauses

(not shown), and stepwise refolding of the hairpin after replication (Figure 5.7G).

The fact that it is possible to detect strand displacement synthesis events similar to

those observed with a high NA objective illustrates the sensitivity and capability of

colloidal lensing for biological assays. The pause efficiency was slightly reduced (25%

to 10%) with the colloid present; this may be due to increased template tension. If Pol

I(KF) or the dNTPs were omitted from the extension mixture there was no recovery

of Cy3 signal.

5.6.2 Thermococcus 9PN-7 Therminator Activity

The assay was slightly modified to accommodate operating at high temperature.

Surface functionalization

Amine-reactive glass surfaces were prepared as follows. Circular coverslips (D-263T,

Precision Glass and Optics) were RCA cleaned and then rinsed in water and ace

tone. The acetone was replaced with a 2% solution of 3-aminopropyl triethoxysilane

(APTES, Sigma-Aldrich) in acetone and the surfaces were incubated for 60 min at


room temperature. The aminosilylated surfaces were washed thoroughly with ace

tone and dried under a steam of air. The slides were then baked for 30 min at 120°C.

The silanization procedure leaves the surface covered with positively charged amine

groups. Next, the surfaces were reacted with a solution of 20 mg/mL epoxy-PEG-

epoxy (MW 5,000, Laysan) in 100 mM sodium bicarbonate buffer (pH 8.3) with 0.3

M K2SO4. This reaction was carried out by dropping 100 \iL of the PEG solution

on to the surface and immediately covering it with a second surface. Surfaces were

incubated for 4 hrs, followed by carefully rinsing with deionized water and drying

with nitrogen gas. These surfaces were stable for several months when stored under

argon atmosphere.

Colloid Functionalization

TiCVCOOH colloids were activated with EDC/Sulfo-NHS as described above. The

washed colloids were treated with 2% poly(ethyleneimine) (50 wt. % in H20, Sigma

Aldrich) in 100 mM carbonate buffer, pH 8, overnight. The colloids were then

washed thoroughly with carbonate buffer and water. The aminated colloids were

then washed three times with DMF followed by incubation in 1,4-phenylene diisoth-

iocyanate (Acros) in DMF with 10% pyridine (CHROMASOL Plus, Sigma-Aldrich)

for 2 hrs. The colloids were washed twice with DMF, twice with methanol, and twice

with acetone. The amine reactive colloids were then coated with the aminated primer

at 20 nM (Oligo Z with a 5'-NH2 group) in 100 mM carbonate buffer, pH 9, overnight.

Next, succinic anhydride (Sigma-Aldrich) in methylpyrrolidinone (Fluka) and borate

buffer was used to passivate the charged surface for 4 hours. After washing with wa

ter, the colloids were passivated again with 1% NH4OH for 10 minutes. The colloids


were washed three times with IX PBS and were then ready for hybridization.

Initial experiments using non-passivated beads resulted in high background flu

orescence. This was attributed to the unreacted amine groups, present on the PEI

surface, non-specifically binding the hairpin through an electrostatic interaction with

the backbone phosphate groups. It was therefore desirable to block these unreacted

amine groups to remove the positive charge. Reacting the amine groups of the PEI-

treated colloids with succinic anhydride achieved this aim. This was probably due to

the negative charge of the carboxylate group repelling the negatively charged DNA.

Covalent D N A Surface Attachment

A flow cell was created by placing a 2 mm thick polydimethylsilicone (PDMS) chip

on the PEG-epoxy coated surfaces. The PDMS chip contained one 30 /JL circular

chamber, 100 //m thick, that was accessible through inlet and outlet holes. A 3'

aminated oligo (500 pM, sequence 5'-CTG GGG CAA ACC AGC GTG GAC CGC

TTG CTG CAA CTC TCT CAG GGC-NH2-3') was injected into the flow cell and

incubated overnight in a humid chamber. The oligonucleotide surfaces were washed

twice with 100 mM carbonate buffer and IX PBS buffer to remove unbound DNA.

The surface was passivated with a solution containing 1M phosphate buffer for 3 hrs.

Before hybridization, the surface was rinsed with PBS buffer and 3x SSC. The hairpin

oligo, diluted in 3x SSC, was added to the flow cell and annealed at 55°C for 1 hr.

After incubation, the flow cell was washed with in 3x SSC followed by lx SCC/0.1%

SDS at 30°C. The colloids (decorated with the primer, described above) were diluted

in 2x SCC and added to flow cell. Annealing was achieved by incubation at 55°C for 8

hrs in a specially design hybridization cassette. The flow cell was then rinsed with lx


SCC/0.1% SDS, followed by two rinses in IX PBS buffer, to remove unbound colloids.

The PDMS flow cell was peeled off and the slide was ready for the experiment.

Thermophilic Polymerase Assay

The extension mixture was comprised of 5 units of Therminator (from Thermococcus

gPN-7, New England Biolabs), 100 (M of each dNTP, 0.1 mg/ml glucose oxidase, 0.2

mg/mL catalase, 10% w/w glucose, 1 mM Trolox, in buffer # 2 from New England

Biolabs. The flowcell and extension mixture were pre-heated in an oven held at 75°

for 30 minutes. The reaction was started by adding 20 piL of mixture to the surface

and immediately sandwiching another coverslip on top. The chamber was sealed

twice with 5-minute epoxy, allowed to dry, and placed in a custom flowcell (CFCS2,

Bioptechs) with a 50 /im thick Teflon gasket for imaging. This flow cell was made

of stainless steel and was capable of holding the sample completely sealed to prevent

evaporation. For reactions performed at 70°C, a thin circular resistor heater was

placed on the bottom of the flow cell using a thermal paste. Heating was achieved by

applying a voltage to the heater with a variable DC power supply. The temperature

of the cell was monitored by placing a thermocouple in close proximity to the imaging

region.

Data collection proceeded as for the mesophilic enzyme experiments. Figure 5.7H-

J shows three sample trajectories measured with the colloidal lenses of Therminator

DNA polymerase activity at 70°C. Heterogeneous behavior was observed with some

trajectories showing fast replication (Figure 5.7H), fast replication with a single pause

(Figure 5.71), and stepwise refolding (Figure 5.7J). However, the frequency of pausing

was drastically reduced with Therminator at 70° to «5% of all trajectories.


5.6.3 Replication Rates Measured with Colloidal Lenses

Incorporation rates for Pol I(KF) and Therminator were measured by first generating

a calibration curve between Cy3 intensity and DNA polymerase position based on

the location of the known Pyr-G-C pause motif. The intensities along each trajectory

were then converted into positions and the rates were calculated by measuring the

slope of the trajectory between two steady states (see Chapter 4 for more detail).

The measured rate of 12 nt/sec for Pol I(KF) (Figure 5.8A) was slightly slower than

measured in the absence of the colloid with a 60X 1.45 NA objective (17 nt/sec).

This may be because the replicating polymerase follows the helical backbone of the

template and has to rotate the colloid. Therminator's measured rate of 18 nt/sec

(Figure 5.8B) is almost double the reported bulk rate of 10 nt/sec, suggesting that

the pausing behavior we observe with our single molecule experiments are included

in the ensemble studies. This separation of burst synthesis from sequence-specific

pausing can only be realized in a single molecule experiment.

5.7 Future work

Colloidal lenses enable not only single molecule sensitivity with low numerical optics,

but also the whole new field of high temperature single molecule spectroscopy. The

ability to image very large fields of view at high temperatures with single molecule

sensitivity, free of the restraints of index matching fluids and short working distances,

could be of great interest in biology, chemistry, and nanoscience. For example, single

molecule sequencing-by-synthesis technologies currently employ a mesophilic DNA


c o

• &

o CO i _

M—

"D

ize

ro £ o c

0.35

0.30

0.25 0.20

0.15

0.10

0.05

0

Pol l(KF) Therminator

0 20 40 60 80 0 20 40 60 80

synthesis rate / nt/sec

Figure 5.8: Replication rates measured with colloidal lenses and a 20X 0.5 NA objective for (A) Pol I(KF) at 23°C and (B) Therminator at 70°C. Rates were measured as described in Chapter 4.

polymerase to sequentially incorporate fluorescently-labeled nucleotides into a grow

ing complementary strand. However, the ability to use a thermophilic polymerase

would offer a number of key advantages: improved enzyme heat stability, better abil

ity to incorporate nucleotide analogs, and the capacity to melt templates that are

GC-rich or have a high degree of secondary structure. The total throughput of single

molecule DNA sequencing methods is also limited by the number of templates visible

in a single field of view, a bottleneck that could also be improved by an order of mag

nitude with the aid of colloidal lenses. Finally, this technology can be used to improve

the sensitivity of miniaturized, hand held microscopes and biological detectors.

Chapter 6

Surface Plasmon Resonance

Enhanced TIRF Microscopy

6.1 Introduction

As introduced in the previous chapter, a fluorescent dye near a metal or metal-oxide

objects can experience changes in both its excitation and emission properties. Planar

metallic systems, however, have been studied for decades largely because they are

easier to prepare with high precision and reproducibility. Single molecule metal-

enhanced fluorescence studies on planar surfaces have used prism-based total internal

reflection microscopy (TIR) [145] or epi-illumination scanning confocal microscopy

[146] to perform excitation and image collection. Theoretical investigations of single

molecule fluorescence detection near a metal layer indicated that the excitation and

emission processes are mediated by surface plasmons [147, 148].

Surface plasmons, or surface plasmon polaritons, are electromagnetic waves that

propagate parallel to a metal or dielectric interface. An electron or light beam can be

118

CHAPTER 6. SPR ENHANCED TIRF MICROSCOPY 119

used to excite surface plasmons in a resonant manner by matching its impulse to the

plasmon. The surface plasmon resonance (SPR) is very sensitive to changes at the

interface boundary, which has led to its use for surface binding detection. Fluores

cence microscopy, however, takes advantage of the fact that the electromagnetic field

generated by the surface plasmons is stronger than that generated by total internal

reflection, giving rise to enhanced fluorescence signals.

While both prism-TIR and scanning confocal approaches demonstrated enhance

ment effects, there are some drawbacks. Prism-TIR limits the amount of fluidic

integration and automation possible on the flowcell, as accessibility is limited on both

sides of the sample. Epi-fluorescence scanning confocal setups can remove this limi

tation but are slow to acquire images. For these reasons, developing a through-the-

objective TIR fluorescence microscopy platform would be ideal. Although it would

necessitate performing both excitation and detection through the thin metal film,

it would allow for integrated fluidics and fast imaging for high throughput single

molecule applications.

The focus of this chapter is to explore the feasibility of using through-the-objective

TIR for single molecule imaging on a metal film. Specifically, the aims are to assess

the surface chemistry, the quenching of non-specifically bound fluorophores, and the

enzymatic activity of a DNA polymerase. The fabrication process begins by coating

a low-autofluorescence coverslip with a thin metal film and functionalizing it with ap

propriate chemistry for the specific attachment of fluorescently labeled biomolecules.

The evanescent field generated by total internal reflection is enhanced by the produc

tion of surface plasmons in the metal film [145, 147, 148]. Surface plasmons tend to

stay longer along the surface and produce a stronger electromagnetic field than that


generated by total internal reflection. As a result, fluorophores within « 20 nm of the

surface exhibit intensity enhancement. This could be of particular benefit for single

molecule DNA sequencing platforms where signal-to-noise ratios are critical for rapid

image acquisition.

There is an additional important benefit of having a thin metal film near the flu

orophores of interest. As mentioned in Chapter 2, single molecule DNA sequencing

approaches can require fluorescently-labeled dNTPs to be washed across a surface

many times throughout the course of a run. This process inevitably results in the

nonspecific binding of fluorescently-labeled dNTPs to the surface, resulting in in

creased background fluorescence and false-positive features. However, the presence of

the thin metal film can quench any excited fluorophores near the surface (within a few

nm) by a mechanism of fluorescent energy transfer into the surface plasmon modes

of the metal [149, 150]. Many alternative surface attachment chemistries have intrin

sic properties designed to enhance specific molecule binding but do little to directly

inhibit or suppress the effects of the nonspecific binding. The corresponding increase

in background fluorescence as labeled molecules are washed across the surface, com

bined with the limited fluorescent intensity and lifetime of any single fluorophore,

imposes restrictions on the overall imaging capabilities of any single molecule surface

chemistry.

6.2 Methods

Glass coverslips (Precision Glass Optics, D-263T cut glass, 0.15 mm, 2"xl" 40/20 sur

face quality) were RCA cleaned and coated with 5-50 nm Au layers using a sputter


coater (Cressington 108). Control slides were coated with a polyelectrolyte mul

tilayer and functionalized with Biotin-PEO-Amine (Pierce) as previously described

[82]. To prepare a self-assembled monolayer (SAM) on the Au film, 10 mg of 11-

amino-undecanethiol (Dojindo Inc.) was dissolved in in 1 mL absolute ethanol to

give a 42 mM solution. This stock solution was then diluted in ethanol to give a 0.5

mM working solution. Au-coated surfaces were completely immersed in the solution

in a vertical chamber for 18-24 hrs. The surfaces were then rinsed with ethanol,

ultrapure water, and dried with N2.

NH

2 N

H2

NH

2 N

H2

> N

H2

t«

NH

2 .>

* N

H2

NH

2

Fig

ure

6.1:

Au-

coat

ed s

urfa

ces

are

subs

trat

es f

or s

elf-

asse

mbl

ed

mon

olay

ers

(SA

Ms)

. (1

) R

CA

cle

an c

over

slip

s w

ere

drie

d an

d a

thin

lay

er o

f A

u w

as s

putt

er-c

oate

d on

one

sid

e.

(2)

A s

elf-

asse

mbl

ed

mon

olay

er o

f 11

-am

ino-

unde

cane

thio

l w

as f

orm

ed o

n th

e su

rfac

e th

roug

h a

thio

l-A

u bo

nd,

(3)

follo

wed

by

trea

tmen

t w

ith

SMC

C t

o cr

eate

a

thio

l-re

acti

ve s

urfa

ce.

(4)

Thi

ol-f

unct

iona

lize

d D

NA

was

the

n be

rea

cted

wit

h th

e su

rfac

e to

for

m a

cov

alen

t bo

nd (

5).

S3

pi I I 3 o

Co O

h->

to

to


To prepare the surfaces for DNA attachment, 2.18 mg of sulfo-SMCC (436.37 g

mol -1 , Pierce) was dissolved in 5 mL of 0.1 M triethanolamine (TEA), pH 7.0, to

create a 1 mM sulfo-SMCC solution. Hybriwells (Grace Biolabs) were placed on the

surfaces and the SMCC solution was incubated at room temperature for 45 minutes.

Surfaces were then washed in with IX PBS buffer, pH 7.4.

Synthetic DNA was obtained from Integrated DNA Technologies (template 5'-SH-

(CTG CTA)5 CTG CTA CTA c cca caa ace aaa age cca gac-3' and primer 5'-Cy3-

GAC TGG GCT TTT GGT TTG TGG G-3'). Oligos were resuspended to 100 /zM

final concentration in ultrapure water. To prepare DNA duplexes for single molecule

incorporation experiments, 1.0 fih primer and 1.0 fjL thiolated template were diluted

to a 10 iih final volume in 10 mM Tris, 100 mM NaCl, pH 7.5. The solution was

tip mixed and annealed on an MJ Thermal Cycler by heating to 96°C and slowly

cooling to 4°C at 0.1°C per second. Thiolated DNA was then reduced using bond-

breaker TCEP as instructed by the manufacturer (Pierce) and diluted to 10-100 pM

for incubation on the SMCC-treated surfaces. The DNA was allowed to react for

«30 min, after which the surfaces were washed extensively with Tris buffer. Surfaces

were then imaged on a custom-built free space total internal reflection fluorescence

microscope with a 60X 1.45 NA Olympus objective and a Hamamatsu ORCA-ER

CCD. Oxygen-scavenging solution was prepared as previously described [70].

6.3 Results

Au coated and 11-amino-undecanethiol/Au surfaces were characterized using x-ray

photoelectron spectroscopy (XPS). The spectra for the Au surfaces showed Auger

peaks primarily due to Au, as expected (Figure 6.2). There was a small amount of


Binding Energy (eV)

Figure 6.2: XPS spectra of Au-coated surfaces.

carbon contamination likely coming from reaction with the atmosphere. The SAM-

treated Au surface showed a substantial increase in carbon content and a small amount

of nitrogen (Figure 6.3). Angle-resolved XPS was also used to measure the depth of

each layer (data not shown), and it was found that the nitrogen layer was « 1.4 Aand

the carbon layer 19.9 A.

To characterize the ability of the Au film to quench fluorophores in close prox

imity, 10 nM Cy3-dCTP (Amersham Biosciences, now GE) in IX TE was incubated

for 5 minutes on three different surfaces (Au, 11-amino-undecanethiol SAM on Au,

and polyelectrolyte multilayer (PEM)). The nucleotides were then washed away and

the surface was imaged. This process Cy3-dCTP incubation, surface washing, and

imaging was repeated two more times. The Au coated surfaces showed no sign of

any Cy3 fluorophores, either due to zero non-specific binding or absolute quenching,

similar to that of a RCA clean surface. PEM surfaces showed a significant amount

of clearly-resolvable non-specifically bound nucleotides over the course of the experi

ment (Figure 6.4A-C). The alkanethiol-coated Au surface did not contain any clearly


Binding Energy (eV!

Figure 6.3: XPS spectra of an 11-amino-undecanethiol SAM on an Au-coated surface.

resolvable fluorophores, but the background fluorescence increased by « 20% (Figure

6.4D-F).

Next, a single base incorporation experiment was attempted on the Au-SAM sur

faces. Primed DNA templates were covalently attached to the SAM through a cova-

lent thiol bond (see Methods section above). Primer-template locations were imaged

and the Cy3 fluorophores on the primer were photobleached. Pol I(KF) and Cy3-

dCTP were then added to the flowcell and allowed to react for 15 minutes at room

temperature. The surface was then washed with Tris buffer and an oxygen-scavenging

solution was added for post-incorporation imaging (Figure 6.5). Although polymerase

was active on the SAM/Au surfaces and it was possible to observe single nucleotide

incorporation events, it only occured on »10% of the initial templates. This may have

been due to the template orientation with respect to the surface: if they were mostly

lying flat on the SAM, DNA polymerase binding would have been largely hindered.

The through-the-objective TIRF configuration required that the emitted light

had to pass through the Au film for detection. This inevitably resulted in some


Figure 6.4: Cy3-dUTP was quenched on Au-coated surfaces. 10 nM Cy3-dCTP in T50 buffer was incubated in a hybriwell on a surface for 5 minutes, the chamber was rinsed with buffer, and the process was repeated three times. The surfaces were then imaged and sample fields of view are shown after each washing cycle on a PEM surface (A-C) and 50 nm Au-coated surface with a SAM of 11-amino-undecanethiol (D-F).

Figure 6.5: Pol I(KF) incorporates Cy3-dCTP on a SAM-Au surface. Green features show the location of each primer-template complex and red features show the location of Cy3-dCTP incorporation. In this field of view, three different templates showed incorporation events (circled).


signal loss and reduced sensitivity. It remains unclear whether the metal-enhancement

fluorescence is great enough to offset the losses and make the effort worthwhile; further

experiments are required to quantify each of these contributions. Even with the

reduced amount of light available to detect, it was still possible to observe the stepwise

photobleaching of single molecules. There was also a significant background present

on the SAM-Au surfaces that appeared similar to Newton's rings. A confocal setup

might be able to filter out these reflections [146], but additional work could be done

to reduce this background for wide-field SPR-TIRF.

6.4 Future work

Using metal surfaces for single molecule imaging is a relatively simple and inexpensive

way to simultaneously quench non-specifically bound species and enhance fluorescent

signals further from the surface. There is also an important trade-off between having

a thick enough Au film for enhancement to be significant versus the corresponding

reduction in light collection efficiency of having to image through a thicker metal

film. Additional work needs to be done to quantify the observed metal-enhanced

fluorescence as a function of the film thickness.

Chapter 7

DNA Polymerases and Nucleotide

Analogs

7.1 Introduction

DNA polymerases are classified into seven general families based on evolutionary ori

gin and structural similarity: families A (pol I), B (pol a, 8, e, £), C (pol III), D (from

Archaea), X (pol /?,/z, cr, A), Y (pol IV, V, 77,6, K), and RT (reverse transcriptases).

All of these enzymes share some degree of sequence homology and general struc

tural similarity, including a right-hand shaped DNA binding cleft that features palm,

thumb, and finger subdomains [151]. Although each family has a distinct secondary

structure, all polymerases appear to make use of a common catalytic mechanism for

template directed DNA synthesis.

Polymerases are widely exploited in modern molecular biology as common tools

for polymerase chain reaction (PCR), single nucleotide polymorphism (SNP) geno-

typing, gene expression studies, and DNA sequencing. A particularly useful feature

128

CHAPTER 7. DNA POLYMERASES AND NUCLEOTIDE ANALOGS 129

of certain polymerases is their ability to incorporate modified nucleotides even in the

presence of their natural counterparts. The ability of being able to maintain incor

poration fidelity while tolerating fluorescent labels has become a critically important

process for a wide variety of applications. Most notably, the rapid completion of the

draft sequence of the human genome [28, 27] was largely due to advances both in

capillary electrophoresis technology and Sanger dideoxy sequencing chemistry [25].

The data generated from this project has increased the interested in comparative

genomics research, with a desire to learn about the relationships between disease,

genetic variability, and pharmaceutical response. However, the amount of sequence

data needed for this requires cheaper, faster, and higher throughput technologies to

be developed.

As described in Chapter 2, single molecule sequencing approaches potentially offer

the highest levels of throughput for the lowest cost [12, 16, 15]. These approaches use

DNA polymerase to incorporate a fluorescently-labeled nucleotide at every position

during replication. The identity of the incorporated base is read out either in real-time

or by stopping the reaction after each step. If a biophysical model for how polymerases

behave while incorporating modified nucleotides can be developed, it may allow for

the rational design of better enzymes for single molecule DNA sequencing and other

applications.

7.2 Polymerase Structures

Here the structural basis for polymerase promiscuity is examined by looking at the

crystal structures of six common polymerases: Taq from Thermus aquaticus (1QSS,

1QSM, 1QTM), Pol I (Klenow Fragment) from Escherichia coli (1D8Y), Tgo from


Thermococcus gorgonarius (1TGO), Vent from Thermococcus sp. 9°N-7 (1QHT), T7

from T7 bacteriophage (1T7P), and reverse transcriptase from HIV-1 (1J50). All

of the protein crystal structures and primary sequences were obtained in PDB or

FASTA file format, respectively, from the RCSB Protein Data Bank1. Structure

selection was based on a three main criteria: the quality of the crystal structure,

the family of the polymerase crystallized, and the biological relevance of the enzyme

in modern molecular biology. Taq and T7, two commonly used B family enzymes,

were selected because they share a common geometry of bound DNA, nucleotide,

and metal ions in their tertiary structures [151], and they both have high resolution

DNA-complexed structures available. Tgo and Vent, two recently discovered B family

enzymes, were selected as they are structurally quite different from the a family

enzymes, yet also have the ability to repeatedly incorporate modified nucleotides at

high rates [152]. HIV-1 RT was chosen to represent a structurally and evolutionary

different family of enzymes for comparison. A DNA-unbound form of Pol I (KF)

was also selected to act as a scaffold for the modeling of Cy5-dCTP at the active

site to explore how the fluorophore group may interact with the enzyme. Although

some of these polymerases have multiple crystal structures available, the most recent,

best data-containing, highest resolution structures complexed with a dsDNA and/or

dNTP in the active site were selected for this analysis.

Pair wise sequence alignment was done using BioEdit Sequence Alignment Editor

v5.0.9 [153] while allowing the ends of the sequences to slide. Figure 7.1 was generated

from Perkin Elmer's technical datasheet for R6G-dGTP and GE's datasheet for Cy5-

dCTP. The protein structure figures in this chapter were created in PyMol v0.932

1http://www.rcsb.org 2http://pymol.sourceforge.iiet

http://www.rcsb.org

http://pymol.sourceforge.iiet


or SwissPDV Viewer[154] 3. For Pol I (KF) modeling, hydrogens were added with

MolProbity [155] to DNA-unbound enzyme prior to energy minimization with ORBIT

[156]. The dCTP-Cy5 substrate was constructed with Biograf (Molecular Simulations,

San Diego, CA) based on the position of the ribose and nitrogenous base rings of the

complexed dCTP substrate in 1KFD [157].

7.2.1 Crystal Structure Quality

Table 7.1 presents a summary of the crystallographic and refinement data for the

structures used in the following analysis. The quality of each of the crystal structures

is discussed briefly below.

Taq (1QSS, 1QSM, 1QTM): Each of these structures was determined to 2.3

A with dsDNA and a different ddNTP complexed in the active site, giving a fairly

precise definition for the positions of the substrate relative to the protein. The data

collection completion of the measured reflections was good, with 94.9% total and

87.1% in the highest resolution shell, and the symmetry total Rsym was 88.0%. Phases

were determined by molecular replacement with the previously solved ternary closed

structure [158]. For refinement, completion was also acceptable (90.1% / 82.6%), and

Kwark and R/ree were fairly close and reasonable for this resolution (23.5% / 27.6%).

Deviation from r.m.s. bond lengths and bond angles was also acceptable (0.007 A

and 1.501°, respectively).

T7 (1T7P): As with Taq, the DNA-T7 complex crystal structure was quite good.

Data was collected from 20-2.2 A with high completion (94.1% / 89.2%) and accurate

symmetry (3.3% / 9.1%). Phases were obtained via MAD with Se-derivatives [160]

3http://us.expasy.org/spdbv/

http://us.expasy.org/spdbv/


Table 7.1: Summary of crystallographic and refinement data, Part I [159, 160, 161]. These structures contain bound duplex DNA and a dNTP.

P D B entry code Data collection

Resolution limit (A) Reflections, observed/unique

Completeness (%), overall/outer Rsym(%)> overall/outer shell

Refinement Resolution range (A)

Unique reflections Completeness (%), overall/outer

Rworfc(%)

R/ree(%) r.m.s. bond lenghts (A) r.m.s. bond angles (°)

1QSS

30-2.3 290,633 / 26,823

94.9 / 87.1 8 . 0 / -

30-2.3 25,948

90.1/82.6 23.5 27.6

0.007 1.501

1T7P

20-2.2 153,202 / 58,916

94.1 / 89.2 3.3 / 9.1

20-2.2 -

7-24.0 27.9

0.006 1.19

1J50

40-3.5

7-91.5 / 83 11.8 / -

10-3.50 39,033

V-26.2 33.8

-

-

and refinement yielded a 2.2 A structure with Rworfc/R/ree values that were good for

this resolution (24.0% / 27.9%). Deviation from r.m.s. bond lengths and bond angles

was also reasonable (0.006 A and 1.19°, respectively).

HIV-1 RT (1J50): The only available structure of HIV-1 RT bound to dsDNA

was determined at a relatively mediocre 3.5 A final resolution, with acceptable com

pleteness (91.5% / 83%), and good symmetry (Rsym = 11.8). Phases were derived

from a previous protein model [162] and refinement was satisfactory for this resolu

tion with Rcrys and R/ree at 26.2% and 33.8%, respectively. At this resolution it is

inappropriate to make predictions about hydrogen bonding patterns or salt bridges,

but it is still possible to evaluate the overall fold of the protein and the orientation

of DNA binding can still be evaluated.

Vent (1QHT): Vent crystal data was collected to 3.0 Afinal resolution with high

completeness (99.0% / 91.1%) and good symmetry. The authors determined the


Table 7.2: Summary of crystallographic and refinement data, Part II [163, 164, 165]. These structures do not contain bound DNA.

P D B entry code Data collection

Resolution limit (A) Reflections, observed/unique

Completeness (%), overall/outer RSym(%), overall/outer shell

Refinement Resolution range (A)

Unique reflections Completeness (%), overall/outer

Rtoorfc(%)

R/ree(%) r.m.s. bond lenghts (A) r.m.s. bond angles (°)

1QHT

25-3.0 164,954 / 22,290

99.0 / 91.1 8 . 2 / -

25-2.25 55-813

93.5/61.9 23.9 30.8

--

1D8Y

20-2.08

- / -94.6 /-

- / -

20-2.08 44,062

V-21.7 23.3

0.008 1.33

1TGO

15-3.0 136,953 / 21,529

91.1 / -7 . 0 / -

25-2.5 30,451

91.1/86.0 7.1 (26.2)

20.9 (27.1) 0.008

1.5

phases via multiple isomorphous replacement with a number of native and derivative

crystals due to non-isomorphism [163]. Refinement with these multiple structures

gave a 2.25 A resolution structure with questionable completeness (93.5% / 61.9%)

and moderately different R-factors (23.9% / 30.8%). This structure was solved with

DNA-unbound enzyme.

Tgo (1TGO): Data collection for Tgo gave data from 15-3.0 A resolution with

good completeness (91.1%) and reasonable symmetry (Rsym = 7.0%). Phases were

determined by multiple isomorphous replacement and anomalous scattering with data

from low-salt containing crystals [165]. Refinement yielded a 2.5 A structure with

good completeness (91.1% / 86.0%) and similar R^ys and R/r-ee factors in the outer

shell (26.2% / 27.1%). This structure was solved with DNA-unbound enzyme.

Pol I(KF) (1D8Y, 1KFD): Although extensively studied, there are no crystal

structures available for Pol I (KF) that have dsDNA bound within the catalytic


groove, presumably because it exists in a highly disordered conformation within this

region [157]. Furthermore, the only crystal structure available for Pol I(KF) with

a nucleotide bound in its active site is, unfortunately, at a relatively low resolution

(3.9 A, 1KFD, [157]). To examine the mechanism of dCTP-Cy5 incorporation by Pol

I(KF), the relative positions of the ribose ring and nitrogenous base from the 1KFD

structure were superimposed on to the 2.08 A resolution DNA-unbound structure

1D8Y. Although any substrate-binding induced conformation changes were lost by

doing in this operation, the RMSD between the two structures was relatively low (0.28

A, which is within experimental error). This suggests that the binding of a nucleotide

results in minimal conformational changes, if any. Larger differences would likely be

seen in dsDNA-bound/unbound structures.

By moving to the higher resolution structure, there is the advantage of being

able to reliably look at individual interactions between the fluorophore and nearby

side chains. The 1D8Y structure was solved to 2.08 A resolution with good total data

completeness (94.6%) and very similar R^orfc/R/ree values (21.7% / 23.3%). Deviation

from r.m.s. bond lengths and bond angles were also acceptable (0.008 A and 1.33°,

respectively). Phases were obtained via molecular replacement with the previously

solved, lower resolution structure (1KLN, [157]).

7.2.2 Sequence Alignment

A total sequence alignment was done pair wise between each of the six polymerases,

and the percent identity and similarity for each of these alignments is illustrated

in Table 7.3. As expected, the primary sequence of the B family polymerase from

Thermococcus gorgonarius (1TGO) shares 91% sequence identity with the B family


Table 7.3: Pair wise sequence alignment identity and similarity between six polymerases. Values shown are (total identity %, total similarity %). For comparison, each polymerase was also aligned to the DNA-binding protein human topoisomerase I (1A31) and carboxypeptidase A (1ARL).

PDB

1D8Y 1TGO 1QHT 1T7P 1HQU 1A31 1ARL

1QSS

27,43 14,34 15,34 18,40 10,24 9,27 5,16

1D8Y

-

12,33 12,32 20,41 11,32 10,29 6,14

1TGO -

-

91,96 15,33 13,29 6,13 7,13

1QHT

-

--

16,34 9,24 11,28 7,17

1T7P -

---

9,18 12,31 7,19

1J50 -

----

11,23 8,27

polymerase from Thermococcus sp. 9°N-7 (1QHT). Taq also shares some degree of

total similarity with Pol I(KF) (27%, 43%), as do Pol I(KF) and T7 (20%, 41%).

Similarities between the other polymerases are on the same low level as those seen

with another DNA binding protein, human topoisomerase I, with approximately 10%

identity and 25% similarity. All of the polymerases showed very low identity when

compared to the protein-binding enzyme carboxypeptidase with « 7% identity and

15% similarity.

These findings are not unexpected as each of these enzyme families is overall

structurally different, and each is derived from a fairly evolutionary distinct source.

As we shall see, however, the polymerase domains of all of these families show some

striking similarities that may play a role in nucleotide analog incorporation.


Figure 7.1: 2D structures for R6G-dGTP and Cy5-dCTP. Both of these structures are commonly used as nucleotide analogs for fluorescently labeling DNA for a wide variety of molecular biology applications.

7.3 Accomodating Nucleotide Analogs

Nucleotide analogs with fluorophores attached to the nucleobase typically have linkers

on the 5' position in purine rings or at the 7' position in pyrimidine rings (Figure 7.1).

As a result, the flurophores extend into the volume surrounding the major groove of

DNA and, depending on the length of the linker between the nitrogenous base and the

flurophore, may extend further. The horizontal distance across the planar conjugated

ring system is approximately 10 A and 16 A for R6G and Cy5, respectively. The alkyl

linker allows Cy5 to adopt multiple conformations, which may increase its ability to

adapt its orientation according to the structure of the enzyme. R6G, however, is

limited to fewer conformations due to the presence of a rigid benzyl-ring in the linker

region (Figure 7.1).

Taq Polymerase (Family A): Figure 7.2 illustrates the closed state of the Taq


Figure 7.2: Crystal structure of Taq polymerase. (A) dsDNA in the active site of Taq polymerase. The template strand is in pink, the nacent strand is in green, and the incoming dNTP is in yellow. (B) There is a large valley that follows along the major groove of the dsDNA that likely accommodates any bound flurophores. (PDB filename: 1QSS; created with PyMol).

polymerase complex after dNTP binding. The nascent base pair is present in a narrow

pocket surrounded by both protein and DNA with one side formed by the O helix in

the fingers domain and the other side formed by the n+1 incorporated dNMP and

template base (Figure 7.2). The "exquisite" tightness of this pocket is thought to

limit only correct Watson-Crick base pairs from forming within it [159]. This closed

complex is formed after the fingers domain of Taq rotate inwards by 46° towards the

active site when the enzyme binds DNA [151]. There do not appear to be any direct

interactions between the O helix and the D loop in this structure, so the distance

between the fingers and thumb domains could be variable, depending on the size of

the substrate. The opening of the pocket is parallel to the plane of the incoming

nucleotide in the region at the start of the major groove (Figure 7.3). Although R587

and R660 appear to interact with incoming nucleotides, their side chains are flexible

and could likely accommodate a fluorophore during synthesis. These two residues are


Figure 7.3: Close up of the active site of Taq (left) and T7 (right). The DNA template is shown in pink, the primer in green, the incoming nucleotide in yellow, and the enzyme in white. Hydrogen bonds are shown as dashed green lines. In Taq, R587 and R660 interact with the recently incorporated and incoming nucleotides in the major groove space. T7 protein-DNA interactions at the active site are predominantly in the minor groove space.

well positioned to form hydrogen bonds with the carbonyl group on the fluorophore

linker. The flexibility of the linker itself could allow the fluorophore's charged ring

system to associate with nearby oppositely charged amino acids.

After the first modified nucleotide incorporation, there needs to be sufficient space

along the enzyme catalytic groove for subsequent movement of the dye along the

growing DNA strand. In Taq, the minor groove contains multiple hydrogen bond

ing interactions between the protein and the DNA phosphate backbone (Figure 7.3;

[159]), in contrast to the major groove, which forms the floor of a large valley void

of protein (Figure 7.2b). The valley varies in depth from 15-25 A; at the vertical

entrance to the valley, the width varies from 5-15 A across, while near the bottom of


Figure 7.4: The active site of T7 polymerase. (A) dsDNA in the active site of T7 polymerase. The template strand is in pink, the nascent strand is in green, and the incoming dNTP is in yellow. The overall similarity between T7 and Taq (Figure 2) is clearly evident from the crystal structures. (B) Similar to Taq, there is a large valley that follows along the major groove of the dsDNA that likely accommodates any bound fluorophores. In T7, however, the valley is partly obstructed by a short loop from the palm domain (red) and an inter-domain hydrogen bonding interaction over the valley between Q539 from the fingers domain and E367 from the thumb domain (A - grey residues) (PDB filename: 1T7P; created with PyMol).

the valley, this distance increases up to 25 A. In their full extended form, the fluo

rophores in Figure 7.1 can only extend vertically out of the major groove by 10-12 A

suggesting that this valley is likely critical for the sequential incorporation of fluores-

cently labeled nucleotides. As nascent DNA migrates away from the active site, any

attached fluorophore groups can travel through the valley to minimize unfavorable

electrostatic and steric interactions (Figure 7.3). Given the largely planar structure

of the conjugated ring system in the fluorophore groups, they may create favorable

aromatic stacking interactions by lining up parallel to the major groove.


T7 (Family A): Protein-DNA interaction at T7's active site primarily occurs

through the minor groove (Figure 7.4), leaving the major groove relatively open and

able to accommodate a fluorescent dye. As with the other family A polymerase Taq,

T7 bound to dsDNA has the major groove largely devoid of protein intrusion. There

are, however, a few notable exceptions: the palm domain contains a short loop region

comprised of amino acids 113-117 that protrudes into the major groove at a point

10-12 bp downstream of the active site, and there is an inter-domain interaction

containing two hydrogen bonds between Q539 of the fingers domain and E367 of

the thumb domain (Figure 7.4). The effect of the palm domain intrusion on initial

incorporation is likely small, although it may influence the fidelity of subsequent

incorporations. The interaction between the fingers and thumb domain could serve

to stabilize the closed form protein-DNA complex, but it also creates a narrower

active site pocket that may limit the space available for the incorporation of modified

nucleotides.

HIV-1 RT (RT Family): While the structure of HIV-1 RT appears to be signif

icantly different from the A and B family of polymerases, it still contains finger, palm,

and thumb domains that surround the active site (Figure 7.5). Like T7, the major

ity of the interactions at the active site are through the minor groove, leaving the

major groove pocket relatively open and available for fluorophore occupation. One

location within the active site minor groove, position 184, appears to be important in

forming Van der Waals interactions ( 3.7 A; this is somewhat questionable given the

3.5 A resolution of this structure) with the incoming nucleotide. It is thought that a

M184I mutation is responsible for RT resistance to inhibitors with unfavourable steric

interactions, such as 2',3'-dideoxy-3'-thiacytidine (Figure 7.5, [161]). GE's CyScript


Figure 7.5: Crystal structure of HIV-1 reverse transcriptase. (A) Surface representation of HIV-1 RT bound to duplex DNA. As with the A family polymerases, the major groove of DNA bound to HIV-1 RT is largely void of protein, creating space for fluorophore labels. Domains are coloured as previously described. (B) Cartoon illustration of HIV-1 RT, with 1184 labeled in the active site and R356 identified as protruding into the major groove. (PDB filename: 1J50; both figures were created in PyMol).

Reverse Transcriptase contains a number of point mutations at this and other sites

to improve its ability to integrate cyanine-labeled nucleotides into cDNAs, suggest

ing that minor groove steric hindrance may actually be important in fluorophore

incorporation.

Away from the active site, the majority of the protein-DNA interactions continue

to occur through the minor groove [151], with the major groove relatively free from

protein (Figure 7.5). The fact that an open major groove is conserved in both A-family

polymerases and reverse transcriptases suggests there may have been evolutionary

pressures that selected for this property. What biological function, if any, an open

major groove may have during synthesis is unclear. Like T7, there is a short loop

region in the palm domain approximately one alpha helical turn away from the active

site that protrudes slightly into the major groove, most notably by residue R356


Figure 7.6: Tgo and Vent polymerases. (A) Cartoon illustration of Tgo DNA polymerase with finger, palm, and thumb domains coloured as previously described. Active site residues are identified and shown in yellow. (B) Cartoon illustration of Vent DNA polymerase. (PDB Filenames: 1TGO, 1QHT. Prepared in PyMol).

(Figure 7.5). Due to the low resolution of this crystal structure, however, it is difficult

to make any predictions about how R356 may interact with specific nucleotides.

Tgo/Vent (Family B): Tgo and Vent are both derived from the Thermococcus

genus and therefore have a high degree of structural similarity (Figure 7.6). The total

improved RMSD (calculated with SwissPDV) of these two structures is only 1.04 A,

which is lower than the resolution of the structures themselves, suggesting that their

structural motifs should be largely analogous. The fingers domain and the thumb

domain are separated by a 15 A cleft that contains the residues critical for polymerase

activity (Figure 7.6). Unlike the enzymes previously discussed, where the fingers and

thumb went from an open to closed conformation, the Tgo/Vent finger and thumb

domains appear to go from a closed to open state in order to accommodate duplex

DNA [163]. The closed state seems to be important for thermal stability, as it is

maintained by two disulfide bridges, shortened loops, and an increase in electrostatic

interactions at subdomain interfaces. Reports by Augustin et al. [152] that Tgo


and Vent have an improved ability over Taq to synthesize DNA exclusively from Cy5-

labeled nucleotides suggest that this conformational change can create sufficient space

for many fluorescently-labeled nucleotides to be incorporated in series. However,

without DNA-bound structures for Tgo or Vent, it is difficult to offer a structural

explanation for where and how much space is created during the transition.

Pol I(KF) (Family A): The overall structure of the polymerase domains of

Taq, T7, and Pol I(KF) are extremely similar [166]. In this structure (Figure 7.7) Pol

I(KF) only has dCTP bound in the active site. The catalytic groove is approximately

10 A wide, 25 A deep, and 20 A tall, with the critical residues along the surface of the

fingers domain. Pol I(KF)'s thumb and finger domains appear much closer together

than in Taq, forming more of a closed groove through the enzyme instead of a valley

(Figure 7.7a). There do not appear to be any finger-thumb domain interactions

like those seen in T7. In this closed conformation, dsDNA cannot reach the active

site without multiple steric conflicts, suggesting that the fingers and thumb domain

undergo a significant "swelling" conformational change like Tgo and Vent in order

to accommodate the double helix upon binding [167]. This is in contrast with fellow

family members Taq and T7, whose finger domains close inwards when the enzyme is

complexed with DNA [160]. This also suggests that Pol I(KF)'s ability to open wide

enough in order to accommodate a DNA strand, an incoming nucleotide, and a bulky

fluorophore group may be a key limiting factor in its ability to incorporate labeled

nucleotides.

Even though the overall conformation of the catalytic groove is likely different

in the dsDNA-bound form of Pol I(KF), this structure can still be used to make


Figure 7.7: Pol I(KF) fragment with Cy5-dCTP bound. (A) Surface representation of Pol I(KF) with Cy5-dCTP bound at the active site. The fingers domain (gold), thumb domain (cyan), and palm domain (red) are coloured accordingly. Note that this represents the open, DNA-unbound conformation of Pol I(KF). (B) An energy-minimized Cy5-dCTP (blue) in the active site of Pol I(KF) forms a number of favourable interactions with R682, K635, P680, and the N-terminus of a nearby alpha helix. (PDB filename: 1D8Y used in conjunction with 1KFD.

predictions about how a fluorescently labeled nucleotide may interact with the pro

tein at the active site. As illustrated in Figure 7.7, the energy minimized model of

Cy5-dCTP in the active site shows a number of favorable hydrogen bonding interac

tions between the fluorophore SO 3 - group and K635, and between the linker carbonyl

oxygen and R682. There are also Van der Waals packing interactions between the

conjugated fluorophore ring system and the protein backbone around P680 and be

tween the linker and R682 side chain. The alpha helix from E684 to F693 has its

positively charged dipole moment oriented in a position that lines up directly with

the negatively charged it orbitals of the conjugated ring system.

This model is useful in illustrating the relative sizes of the Cy5-dCTP and the

closed conformation of the active site and in identifying potentially favorable interac

tions between the fluorophore and enzyme. However, this model could be, and likely


is, entirely incorrect for what happens in reality. In the presence of dsDNA, the con

formation of the active site cleft is different and the entire cleft is more crowded. A

lack of space limits the number of conformations available to the Cy5-group and likely

excludes the minimum energy conformation found here. Without a structure of Pol

I(KF) complexed to dsDNA in the active site, it is difficult to predict how or whether

multiple Cy5 moieties could be packed within the cleft. In the present conformation,

taking into account the size and shape of dsDNA from the Taq crystal structure, it

would appear such packing would be limited to three or four groups. This is consis

tent with previously found experimental data that Pol I(KF) is unable to synthesize

more than f«5 bp of DNA exclusively from Cy5-labeled nucleotides [12, 168, 152] but

is capable of synthesizing DNA exclusively from nucleotides tagged with a smaller

fluorophore [167].

This analysis demonstrates that a number of different polymerase families may

accommodate the incorporation of fluorescently modified nucleotides for three main

reasons. First, the major groove of duplex DNA forms the base of a large valley

void of protein intrusion, thereby providing space to any incorporated fluorescently

labeled nucleotides. Second, the flexibility of the fingers domain, combined with the

localized fidelity of the Watson-Crick base pair active site pocket, may further allow

for nucleotides to be processed. Third, the fluorophore group itself may form favorable

interactions with protein side chains both at the active site, along the catalytic groove

as DNA is synthesized, or with other fluorophores. These findings suggest that in

order to engineer a new polymerase with an improved ability to incorporate modified

nucleotides, non-critical residues that occupy the major groove at the active site or

along the valley should be considered for mutation to smaller amino acids. Selecting


residues to increase finger or thumb domain flexibility or to improve fluorophore

binding is likely beyond the current capabilities of rational protein design.

It would be extremely useful to have crystal structures for each of these poly

merases complexed with DNA and/or dNTPs that have been fluorescently labeled to

varying degrees at different positions. This would allow for validation or improvement

of these proposed models for how polymerases are able to tolerate the incorporation of

modified nucleotides. It may be difficult to solve such structures, however, as the fluo-

rophores can exist in multiple states due to the flexibility of their linkers. Introducing

structural rigidity into the linkers to improve crystallization may have the undesired

effect of reducing incorporation efficiency. Until these issues can be resolved, we are

limited to developing explanations based on existing crystal structures. However,

even these may not be a good model for predicting how fluorescently labeled DNA

interacts with polymerases. In the structures examined here, B-form DNA becomes

A-form within 2-3 bp from the active site [160], which results in a widened DNA major

groove and a narrowed minor groove. Experimental evidence suggests dsDNA with

one strand completely labeled with a dye will undergo a transition from right-handed

DNA to left-handed Z-DNA [167]. While the major groove in A- and B-DNA appears

to have the ability to accommodate bulky fluorophore groups during synthesis, the

major groove in Z-DNA is very shallow and wide. The transition from right-handed

DNA to Z-DNA while in the catalytic groove could open the possibility for numerous

protein-DNA interactions not seen in these crystal structures.


7.4 Custom Nucleotide Variants

As discussed in the previous sections, commercially available fluorescently-labeled

nucleotides are capable of being incorporated by wild-type DNA polymerases. How

ever, most sequencing-by-synthesis applications require complete replacement of ev

ery dNTP by a labeled dNTP to avoid the creation of sequence gaps. Incorporating

sequentially-labeled dNTPs limits the total read length due to steric hindrance be

tween the fluorophores and the enzyme, fluorophore-fluorophore interactions, or both.

To overcome this limitation, various groups have proposed moving the fluorophore

from the nucleobase to the 7-phosphate [169, 170, 171]. Since DNA polymerases

naturally induce the cleavage of the a — /?-phosphoryl bond upon nucleotide incorpo

ration, these nucleotides release the pyrophosphate leaving group and the attached

fluorescent label simultaneously. The resulting product is natural, "unscarred" DNA.

An alternative approach is to synthesize fluorescently-labeled nucleotides with ei

ther longer [168] or cleavable nucleobase linkers [172, 173, 174, 33], both of which have

demonstrated success in sequencing applications. Reversible terminating nucleotides

have the property of terminating DNA synthesis after incorporation, and have been

reported using either N6-alkyl [175], 3'-0-allyl [176], 3'-0-azidomethyl [33], or 3'-0-

(2-nitrobenzyl) [177] modifications. Removal of the terminating group either optically

or chemically allows for synthesis to proceed. Being able to incorporate a single base

and terminate synthesis for detection is especially important for sequencing through

homopolymer regions.


hydroxylamine or esterase

•

Figure 7.8: dUTP-17-E-Cy5 with an internal ester group. The \moa{abs) = 650 nm. Addition of hydroxylamine or an esterase results in cleavage of the linker and the release of the fluorophore.

7.5 Longer Linkers with Internal Esters

A novel nucleotide analog, dUTP-17-E-Cy5 (shown in Figure 7.8), contains an internal

ester in the linker between the nucleobase and the fluorophore. This feature facilitates

the cleavage of the tether by either hydroxylamine or esterase enzyme treatment.

dUTP-17-E-Cy5 was chemically synthesized and characterized by Carolyn Woodroofe

and Brian Stoltz at Caltech using a similar approach as previously described [178].

To test the potential biological applications of nucleotide variant, a simple assay

was designed to examine both the efficiency of incorporation and the ability to cleave

the fluorophore with an esterase. Oligos were commercially synthesized (Integrated

DNA Technologies) and the sequences of the primer and template are shown in Fig

ure 7.9. For the assay, 10 fjM primer and template were annealed in 20 mM Tris

100 mM NaCl pH 7.5 buffer. Extension reactions were prepared containing 1 /xM

annealed primer/template, 2.5 units of Pol I(KF) (New England Biolabs), 100 /JM of


Cy3-5'-TGCTGGGCTTTTGGTTTGTGGG 3'-ACGACCCGAAAACCAAACACCCGACATACAAGAAGCCATCC-5'

Figure 7.9: The sequences of the primer and template used for the nucleotide assay are shown. dUTP-17-E-Cy5 should incorporate opposite the template adenine bases shown in red.

each dNTP, in # 2 buffer from NEB. Reactions proceeded at 37°C for 1 hour followed

by purification using a QiaQuick Nucleotide Removal Kit (Qiagen) as per the manu

facturer's instructions. Samples were then loaded on a 15% TBE urea polyacrylamide

gel (Invitrogen) and run for 85 minutes at 174V. Gels were imaged on an GE Typhoon

9410 scanner.

The first set of experiments were designed to validate the assay with unlabeled

nucleotides. Figure 7.10 shows that when only one nucleotide was added, Pol I(KF)

only extended the primer if it was dCTP (Lane 5). Further extension was only pos

sible with the addition of the other unlabeled nucleotides (Figure 7.11 Lanes 2,3,5).

Figure 7.10 Lane 6 shows that when dUTP-17-E-Cy5 was added alone, some misin-

corporation occurred against the template guanosine. If dUTP-17-E-Cy5 and dCTP

or dCTP/dGTP were added (Lanes 7 and 8, respectively), the primer was expected

to be extended +2 and +4 nucleotides, respectively. However, the addition of the

dUTP-17-E-Cy5 resulted in a much larger shift between the primer and these prod

ucts. This made interpreting the band sizes of lanes 6-8 difficult, but the general trend

of longer products with more nucleotides was consistent, including with all three un

labeled nucleotides present (Figure 7.11 Lane 1). It was interesting to see that when

the four unlabeled nucleotides were included alongside dUTP-17-E-Cy5 (Figure 7.11,

Lanes 6 and 8), the full unlabeled extension product was detected (41 bp) along with

slower moving, Cy5-labeled products.


Figure 7.10: Pol I(KF) fidelity with single base extension reactions containing unlabeled and labeled nucleotides, Part I. Green bands show the location of the Cy3-labeled primer; red bands show the location of primers labeled with Cy5. DNA polymerase and the following nucleotide combinations were added: Lane 1- dATP; Lane 2- dGTP; Lane 3- dTTP; Lane 4- LIZ 120 size standard; dCTP; Lane 6- dUTP-17-E-Cy5; Lane 7- dCTP, dUTP-17-E-Cy5; Lane 8- dCTP, dGTP, dUTP-17-E-Cy5.

Figure 7.11: Pol I(KF) fidelity with multiple base extension reactions containing unlabeled and labeled nucleotides, Part II. Lane 1- dCTP, dGTP, dATP, dUTP-17-E-Cy5; Lane 2-dCTP, dTTP; Lane 3- dCTP, dTTP, dGTP; Lane 4- LIZ 120 size standard; Lane 5-dCTP,dGTP,dATP,dTTP; Lane 6- dCTP, dTTP, dGTP, dATP, dUTP-17-E-Cy5; Lane 7- no dNTPs; Lane 8- dNTP mix,dUTP-17-E-Cy5.


Figure 7.12: Cy5 nucleotides can be incorporated by Pol I(KF). Top: Primer and template used in the nucleotide incorporation assay. The primer is 22 bp long before extension begins. Lane 1- dUTP-10-Cy5; Lane 2- dUTP-17-E-Cy5; Lane 3- dCTP and dUTP-10-Cy5; Lane 4- LIZ 120 size standard; dCTP and dUTP-17-E-Cy5; Lane 6- dCTP, dGTP, dUTP-17-E-Cy5; Lane 7- dCTP, dGTP, dUTP-17-E-Cy5; Lane 8-dCTP; Lane 9- dTTP.

To determine if this effect was specific for dUTP-17-E-Cy5, commercially available

dUTP-10-Cy5 (Amersham Biosciences, now GE Healthcare) was used as a substrate

against the same template (Figure 7.12). In Lane 1 the polymerase was presented

only with the commercially available Cy5-10-dUTP, a nucleotide that should not

be incorporated opposite the template C, and it indeed showed little incorporation.

This was in contrast to Lane 2 where dUTP-17-E-Cy5 was added alone, where a

higher band suggesting incorrect incorporation was clearly evident. Lanes 3 and 5

show the incorporation of dCTP and dUTP-10-Cy5 or dUTP-17-E-Cy5, respectively.

The polymerase now correctly incorporated the commercial dUTP-10-Cy5 to give

the primer +2 product (Lane 3). Fidelity was lost with dCTP and dUTP-17-E-Cy5

(Lane 5 shows a broad smear of extension products) but seemed to improve with the

addition of more nucleotides (Lanes 6-7).

Each of the extension reactions in Figure 7.12 were subsequently treated with


Figure 7.13: Nucleotides with ester-containing linkers can be cleaved with an esterase. Lane assignments are as in Figure 7.12.

100 units of esterase from porcine liver (Sigma Aldrich) in 75 ^L 10 mM boric acid

buffer, pH 8.0. At 15 minute intervals, 0.77 /xL of 100 fjM NaOH was added to each

reaction and mixed well. After 1 hour of incubation, the reactions were purified using

the QiaQuick kit and run on a gel as described above. The scanned gel showing the

cleavage products is shown in Figure 7.13. Esterase treatment resulted in decrease of

fluorophore from lanes 2, 4, and 7 as expected, but there was also a general reduction

in signal in all lanes. This may be due to nonspecific nuclease activity resulting in

widespread DNA hydrolysis (Sigma claims that the esterase is at least 95% pure).

In any case, the promiscuity of Pol I(KF) to incorporate this nucleotide opposite the

wrong base under certain nucleotide conditions is cause for concern.

7.6 Prospects

While linkers containing esters and other cleavable groups show promise for sequenc

ing applications, the dUTP-17-E-Cy5 analog examined here requires additional work

to improve its fidelity. Using enzymes to cleave fluorophore linkers is also clearly


troublesome due to issues associated with purity and specificity. Linkers that can be

cleaved chemically or with light are promising alternatives and are worth exploring

further.

Appendix A

C + + Code for Monte Carlo

Simulations

Listing A.l: Example C + + code for Monte Carlo simulations

i / / a simple program that generates N molecules at random X—Y coordinates

and determines

// the number of resolvable features based on a set diffraction limit.

3 ̂ include <iostream>

#include < c s t d l i b >

5 #include <ctime>

#include <cmath>

7 #include <va la r ray>

#include <fstream>

9

#def ine M 75 / / define the grid length and width

ii #def ine DIFFDIST 1 / / set the diffraction limit to a distance of 1 unit

using namespace std ;

13

154

APPENDIX A. C++ CODE FOR MONTE CARLO SIMULATIONS

i n t ma in ( )

15 {

c o n s t c h a r *FILENAME = " F I L E 0 2 0 4 0 9 - 7 5 . t x t " ; / / filename to

output data

17 u n s i g n e d i n t S1=0;

l ong Nmax=20000;

19 v a l a r r a y < f l o a t > X l (20000) ;

v a l a r r a y < f l o a t > Y l ( 2 0 0 0 0 ) ;

21 v a l a r r a y < i n t > u l ( 2 0 0 0 0 ) ;

//float dist [500][500j;

23 f l o a t d = 0 ;

/ / set evil seed

25 s r a n d ( ( u n s i g n e d ) t i m e ( NULL));

/ / open file to write data to

27 o f s t r e a m fout (FILENAME) ;

for ( i n t N = l ; N<8000; N+=200) / / run the simulation for N=l

N=8000

29 {

s t d :: cout « N « " J ' ;

3i fout « N « " J' ;

for ( i n t c t r = 0 ; c t r < 5; c t r + + ) / / run 5 simulations

each N

33 {

for ( i n t a =0 ; a < N; a++)

35 {

X l [ a ] = 0 ;

37 Y l [ a ] = 0 ;

u l [ a ] = 0 ;

39 }

A. C++ CODE FOR MONTE CARLO SIMULATIONS 156

for ( i n t c = 0; c < N; C++) / / generate random

X—Y positions

{

XI [c]=M*( f l o a t ) r and ()/RANDJVtAX; / /

for random positions

Yl [ c]=M* ( f l o a t ) rand () /RMCLMAX;

/ / XI[c]= int (M*(float)rand()/RANDMAX);

//for binning

// Yl[c]= int (M*(float)rand0/RANDMAX);

// font « Xl[c] « " " « Yl[c] « " ";

// write coordinates to file for debugging

}

for ( i n t i = 0; i < N; i++) / / compare array

against itself

{

for ( i n t j = i + 1 ; j < N; j + + )

{

d = s q r t ( ( X l [ i ] - X l [ j ] ) * ( X l [ i ] - X l [ j

] ) + ( Y l [ i ] - Y l [ j ] ) * ( Y l [ i ] - Y l [ j

] ) ) ;

i f (d < DIFFDIST && i != j )

{

u l [ i ] = l ; / / set the

index for each

molecule to 1

u l [ j ] = l ;

}

APPENDIX A. C++ CODE FOR MONTE CARLO SIMULATIONS 157

59 //std::cout« dist[i][j]« "

J

//font « dist[i][j] « " ";

}

}

63 S1=0;

for (int b = 0; b < N; b++) / / count up

resolvable molecules in array 1

65 {

if (Xl[b] > 2 &fe Xl[b] < (M-2) &fe Yl[b]

> 2 && Yl[b] < (M-2)) / / only

examine a subset of molecules to

eliminate edge effects

{

if (ul[b]==0)

69 {

S1=S1+1;

}

}

}

std :: cout « S l « " J" ;

75 fout « S1«V" ;

}

77 s td: : cout « std : : endl;

fout « s td: : endl;

}

fout .c lose() ;

si return 0;

}

Appendix B

MATLAB Code for Image

Processing and Analysis

Listing B.l: Test

clear a l l ;

2 run=double( zeros (1 ,2)) ;

frames =700;

4 for m=0:6

digl=m+48;

e for k=0:9

dig2=k+48;

s for p=0:9

dig3=p+48;

o filename = [ 'G: \ !2008\08-07\072808\3\ img-0 ' ,digl , dig2 , dig3 , ' .

TIF ' ] ;

a=double (imread ( f i lename)) ;

.2 b=imcrop(a,[2 81 167 100 100]);

current=pkfnd(b,1600,3) ;

158

APPENDIX B. MATLAB CODE FOR IMAGE PROCESSING 159

14 run=cat (1 ,run , current) ;

i6 e n d ;

end;

is end;

20 % f i l ter out duplicates in run

uni=unique (run , ' rows ') ;

22 uni( l ,:) =[];

cy3=uni;

24 numberpeaks=size (cy3) ;

frame = l;

26 for m=0:6

digl=m+48;

28 for k=0:9

dig2=k+48;

30 for p=0:9

dig3=p+48;

32 filename = ['G:\!2008\08-07\072808\3\img-0' ,digl , dig2 , dig3 , ' .

TIF' ];

a=double (imread (filename)) ;

34 b=imcrop(a,[2 81 167 100 100]);

for o = l:numberpeaks(l)

36

cy3_n=imcrop(b,[cy3(o,l)-2 cy3(o,2)-2 4 4]);

38 cy3_im=imcrop(b,[cy3(o,l)-l cy3(o,2)-l 2 2]);

40 cy3_intensity(o, frame )=double (sum (sum (cy3-im))) ;


cy3_back(o, frame )=double( sum (sum (cy3_n))— sum(sum(cy3_im)

) ) ;

42

cy3_net (o, frame)=cy3-intensity (o , frame) — 9/16*cy3_back(o ,

frame) ;

44 end;

46 frame=frame + l;

end;

48 end;

end ;

50

%code to screen out short lived species.

52 net_filtered=zeros (1, frames)

count = l;

54 for f = 1: s i ze ( cy3-.net (: , 1) ,1)

j = s i z e ( f ind ( c y 3 . n e t (f , : ) >300)) ;

se k=size(find(cy3_net(f ,:) >1500)) ;

if (j(2)>40)

58 if (k(2)<10)

net_filtered (count , 1: end)=cy3_net (f ,:) ;

6o cy3_filtered (count ,l)=cy3(f , 1) ;

cy3-filtered (count ,2)=cy3(f ,2) ;

62 count=count + l;

end;

64 end;

end;

66

%calculate the running average of every trace

http://cy3-.net

http://cy3.net


68 n e t - f i l t e r e d - a v e r a g e = [];

count = l;

70 for c = l : s i z e ( n e t . f i l t e r e d (: , 1) ,1) ;

n e t - f i l t e r e d - a v e r a g e (count , : )=smooth( n e t - f i l t e r e d (count , : ) , 'moving ' ) ;

72 c o u n t = c o u n t + l ;

end;

74

f i g u r e (997)

76 imagesc (b)

78 x = l : l : f r a m e s ;

y = l : l : 2 0 4 8 ;

so count = l ;

for k = l :35 % # f i g u r e s

82 f i g u r e (k)

e l f ;

84 for h = l:40 % 12 p l o t s per f ig

s u b p l o t ( 5 , 8 , h )

86 p l o t ( x , n e t . f i l t e r e d (count , : ) , ' b l u e ' ) ;

hold on;

88 p l o t ( x , n e t - f i l t e r e d _ a v e r a g e (coun t , : ) , ' r e d ' , 'L ineWidth ' ,1)

t i t l e ( [ c y 3 _ f i l t e r e d (count ,1) , ( c y 3 - f i l t e r e d (count , 2 ) ) ])

90 h o l d o f f ;

ylim([0 1500])

92 x l i m ( [ 0 140])

c o u n t = c o u n t + l ;

94 end;

end;

Appendix C

Meep Code for FDTD Simulations

Listing C.l: Example bash script to call meep

# / b i n / b a s h

2

mpirun —np 6 / usr/bin/meep—mpi lens?=t rue radi=1.25 xlens=1.25 r u n l l l l 0 8

. c t l | tee 1111- s l . 25 .ou t

4 mpirun —np 6 /usr/bin/meep—mpi lens?=t rue radi=1.00 xlens=1.00 r u n l l l l 0 8

. c t l | tee 1111-s l .00 .ou t

mpirun —np 6 /usr/bin/meep—mpi lens?=t rue radi=0.75 xlens=0.75 r u n l l l l 0 8

. c t l | tee l l l l - s 0 . 7 5 . o u t

6 mpirun —np 6 /usr/bin/meep—mpi lens?=t rue radi=0.50 xlens=0.50 r u n l l l l 0 8

. c t l | tee 1111-sO .50.out


. c t l | tee l l l l - s 0 . 2 5 . o u t

s mpirun —np 6 /usr/bin/meep—mpi lens?=t rue radi=0.10 xlens=0.10 r u n l l l l 0 8

. c t l | tee l l l l - s 0 . 1 0 . o u t


. c t l I tee l l l l - s 0 . 0 5 . o u t

162

APPENDIX C. MEEP CODE FOR FDTD SIMULATIONS 163

10 mpirun — up 6 /usr/bin/meep—mpi lens?=t rue radi=0.01 xlens=0.01 r u n l l l l 0 8

. c t l | tee 1111-sO.Ol.out

Listing C.2: Example meep code for FDTD simulations

define—param sx 13) ; size of cell in X d i rec t ion

define— param sy 13) ; size of cell in Y d i rec t ion

define—param sz 13) ; size of cell in Z d i rec t ion

define n.env 1.33)

s e t ! geometry —lat t ice (make l a t t i c e ( s ize sx sy s z ) ) )

s e t ! default—material (make d i e l e c t r i c (index n .env) ) )

define xlens 1) ; x center of lens

define y—flor 0) ; y pos i t ion of fluor

define z—flor 0) ; z pos i t ion of fluor

define yfluxposl 0) ; y center of flux box

define zfluxposl 0) ; z center of flux box

define xfluxposl 2.75) ; x center of flux box

define xfluxwid 0) ; width of detector for flux in x

define s ix ty5 11.8) ; these angles are only valid for xfluxposl =2.75

and xfluor@0

define s ixty 9.53)

define f i f ty5 7.85)

define f i f ty 6.55)

define fourty5 5.5)

define fourty 4.62)

define t h i r t y 5 3.85)

define t h i r t y 3.18)

define twenty5 2.56)

define twenty 2.00)


( d e f i n e f i f t e e n 1.47)

26 ( d e f i n e t e n 0 .97)

( d e f i n e f ive 0 .48)

28 ( d e f i n e one 0 .096)

(define—param l e n s ? t r u e ) ; i f t r u e , i n s e r t l ens

30

( s e t ! geometry

32 ( i f l e n s ?

( l i s t

34 (make s p h e r e

( c e n t e r x l e n s 0 0)

36 ( r a d i u s r a d i )

( m a t e r i a l (make d i e l e c t r i c ( i ndex t i t a n ) ) ) ) )

38 ( l i s t

(make s p h e r e

40 ( c e n t e r —5 —5 0)

( r a d i u s 0 .01)

42 ( m a t e r i a l (make d i e l e c t r i c ( i ndex 1 . 3 3 ) ) ) ) ) ) )

44 (define—param fcen 1.75) ; p o i n t s o u r c e f requency

(define—param df 0 .05) ; p u l s e wid th ( i n f r e q u e n c y )

46

( s e t ! s o u r c e s ( l i s t

48 (make s o u r c e

( s r c (make con t inuous—src ( f r equency f c e n ) ) )

50 (component Ez)

( c e n t e r x—flor y—flor z — f l o r ) ) ) ) ; se t f l u o r o p h o r e

l o c a t i o n

52


(set! pml—layers ( l is t (make pml (thickness 0.5))))

54

(set— param! resolution 30) .; resolution

56

(define—param nfreq 100) ; number of frequencies at which to compute

flux

58 (define fluxPlanel.60

(add—flux fcen df nfreq

60 (make flux—region

(center xfluxposl yfluxposl zfluxposl)

62 (size xfluxwid sixty sixty))))

(define fluxPlanel-50

64 (add—flux fcen df nfreq

(make flux—region

66 (center xfluxposl yfluxposl zfluxposl)

(size xfluxwid fifty fifty))))

68 (define fluxPlanel_40




72 (size xfluxwid fourty fourty))))

(define fluxPlanel.30


(make flux—region


(size xfluxwid thirty thir ty))))



so (make flux—region

APPENDIX C. MEEP CODE FOR FDTD SIMULATIONS


82 (size xfluxwid twenty twenty))))

(define fluxPlanel_10


(make flux—region


(size xfluxwid ten ten))))





92 (size xfluxwid sixty5 sixty5))))



(make flux—region


(size xfluxwid fifty5 fifty5))))



IOO (make flux—region


102 (size xfluxwid fourty5 fourty5))))



(make flux—region


(size xfluxwid thirty5 thirty5))))




no (make f lux—region

( c e n t e r x f l u x p o s l y f l u x p o s l z f l u x p o s l )

112 ( s i z e x f luxwid twen ty5 t w e n t y 5 ) ) ) )

( d e f i n e f l u x P l a n e l _ 1 5

ii4 (add—flux fcen df n f req

(make f lux—region

ii6 ( c e n t e r x f l u x p o s l y f l u x p o s l z f l u x p o s l )

( s i z e x f luxwid f i f t e e n f i f t e e n ) ) ) )

us ( d e f i n e f l u x P l a n e l _ 5

(add—flux fcen df n f req

120 (make f lux—region

( c e n t e r x f l u x p o s l y f l u x p o s l z f l u x p o s l )

122 ( s i z e x f luxwid f ive f i v e ) ) ) )

( d e f i n e f l u x P l a n e l - 1

124 (add—flux fcen df n f req

(make f lux—region

126 ( c e n t e r x f l u x p o s l y f l u x p o s l z f l u x p o s l )

( s i z e x f luxwid one o n e ) ) ) )

128 ( run—unt i l 50

(a t—beginn ing ou tpu t—eps i lon )

130 (at—end ou tpu t—ef ie ld —z))

132 ( d i s p l a y — f l u x e s f l u x P l a n e l - 6 5 )

( d i s p l a y — f l u x e s f l u x P l a n e l _ 6 0 )

134 ( d i sp l ay—f luxes f l u x P l a n e l _ 5 5 )

( d i s p l a y — f l u x e s f l u x P l a n e l - 5 0 )


( d i sp l ay—f luxes f l u x P l a n e l _ 4 0 )


APPENDIX C. MEEP CODE FOR FDTD SIMULATIONS

(display—fluxes

140 (display—fluxes

(display—fluxes


(display—fluxes


(display—fluxes

f luxPlanel_30)

f luxPlanel_25)

f luxPlane l -20)

f luxPlane l -15)

f luxPlane l -10)

f l uxP lane l . 5 )

f l u x P l a n e l . l )

Bibliography

[1] W. E. Moerner and L. Kador. Optical detection and spectroscopy of single

molecules in a solid. Phys. Rev. Lett, 62(21):2535-2538, May 1989.

[2] M. Orrit and J. Bernard. Single pentacene molecules detected by fluorescence

excitation in a p-terphenyl crystal. Phys. Rev. Lett, 65(21):2716-2719, Nov

1990.

[3] W. P. Ambrose, Th. Basche, and W. E. Moerner. Detection and spectroscopy

of single pentacene molecules in a p-terphenyl crystal by means of fluorescence

excitation. The Journal of Chemical Physics, 95(10):7150-7163, 1991.

[4] W. Patrick Ambrose, Peter M. Goodwin, James H. Jett, Alan Van Orden,

James H. Werner, and Richard A. Keller. Single molecule fluorescence spec

troscopy at ambient temperature. Chemical Reviews, 99(10):2929-2956, 1999.

[5] S. Weiss. Fluorescence spectroscopy of single biomolecules. Science,

283(5408): 1676-1683, 1999.

[6] W. E. Moerner and Michel Orrit. Illuminating Single Molecules in Condensed

Matter. Science, 283(5408): 1670-1676, 1999.

169

BIBLIOGRAPHY 170

[7] W. E. Moerner. A dozen years of single-molecule spectroscopy in physics, chem

istry, and biophysics. The Journal of Physical Chemistry B, 106(5) :910-927,

2002.

[8] Yasushi Sako, Shigeru Minoghchi, and Toshio Yanagida. Single-molecule imag

ing of egfr signalling on the surface of living cells. Nat Cell Biol, 2(3):168-172,

2000. 1465-7392 10.1038/35004044 10.1038/35004044.

[9] GJ Schutz, G Kada, VP Pastushenko, and H Schindler. Properties of lipid mi-

crodomains in a muscle cell membrane visualized by single molecule microscopy.

The EMBO Journal, 19:892-901, 2000.

[10] Ji Yu, Jie Xiao, Xiaojia Ren, Kaiqin Lao, and X. Sunney Xie. Probing

Gene Expression in Live Cells, One Protein Molecule at a Time. Science,

311(5767): 1600-1603, 2006.

[11] M. Sauer, B. Angererb, W. Ankenbauerb, Z. Foldes-Pappd, F. Gobela, K. T.

Hana, R. Riglerd, A. Schulza, J. Wolfruma, and C. Zande. Single molecule dna

sequencing in submicrometer channels: state of the art and future prospects.

Journal of Biotechnology, 86(3):181-201, 2001.

[12] I. Braslavsky, B. Hebert, E. Kartalov, and S. R. Quake. Sequence information

can be obtained from single dna molecules. Proceedings of the National Academy

of Sciences of the United States of America, 100(7) :3960-3964, 2003.

[13] M. J. Levene, J. Korlach, S. W. Turner, M. Foquet, H. G. Craighead, and

W. W. Webb. Zero-mode waveguides for single-molecule analysis at high con

centrations. Science, 299(5607)-.682-686, 2003.

BIBLIOGRAPHY 171

[14] Yann Astier, Orit Braha, and Hagan Bayley. Toward single molecule dna se

quencing: Direct identification of ribonucleoside and deoxyribonucleoside 5-

monophosphates by using an engineered protein nanopore equipped with a

molecular adapter. Journal of the American Chemical Society, 128(5): 1705-

1710, 2006.

[15] Timothy D. Harris, Phillip R. Buzby, Hazen Babcock, Eric Beer, Jayson Bowers,

Ido Braslavsky, Marie Causey, Jennifer Colonell, James DiMeo, J. William Ef-

cavitch, Eldar Giladi, Jaime Gill, John Healy, Mima Jarosz, Dan Lapen, Keith

Moulton, Stephen R. Quake, Kathleen Steinmann, Edward Thayer, Anastasia

Tyurina, Rebecca Ward, Howard Weiss, and Zheng Xie. Single-Molecule DNA

Sequencing of a Viral Genome. Science, 320(5872): 106-109, 2008.

[16] John Eid, Adrian Fehr, Jeremy Gray, Khai Luong, John Lyle, Geoff Otto, Paul

Peluso, David Rank, Primo Baybayan, Brad Bettman, Arkadiusz Bibillo, Keith

Bjornson, Bidhan Chaudhuri, Frederick Christians, Ronald Cicero, Sonya Clark,

Ravindra Dalai, Alex deWinter, John Dixon, Mathieu Foquet, Alfred Gaertner,

Paul Hardenbol, Cheryl Heiner, Kevin Hester, David Holden, Gregory Kearns,

Xiangxu Kong, Ronald Kuse, Yves Lacroix, Steven Lin, Paul Lundquist, Con-

gcong Ma, Patrick Marks, Mark Maxham, Devon Murphy, Insil Park, Thang

Pham, Michael Phillips, Joy Roy, Robert Sebra, Gene Shen, Jon Sorenson,

Austin Tomaney, Kevin Travers, Mark Trulson, John Vieceli, Jeffrey Wegener,

Dawn Wu, Alicia Yang, Denis Zaccarin, Peter Zhao, Frank Zhong, Jonas Ko-

rlach, and Stephen Turner. Real-time dna sequencing from single polymerase

molecules. Science, page 1162986, 2008.

BIBLIOGRAPHY 172

[17] James Clarke, Hai-Chen Wu, Lakmal Jayasinghe, Alpesh Patel, Stuart Reid,

and Hagan Bayley. Continuous base identification for single-molecule nanopore

dna sequencing. Nature Nanotechnology, 4:265-270, 2009.

[18] Tobias Sjoblom, Sian Jones, Laura D. Wood, D. Williams Parsons, Jimmy

Lin, Thomas D. Barber, Diana Mandelker, Rebecca J. Leary, Janine Ptak, Na

talie Silliman, Steve Szabo, Phillip Buckhaults, Christopher Farrell, Paul Meeh,

Sanford D. Markowitz, Joseph Willis, Dawn Dawson, James K. V. Willson,

Adi F. Gazdar, James Hartigan, Leo Wu, Changsheng Liu, Giovanni Parmi-

giani, Ben Ho Park, Kurtis E. Bachman, Nickolas Papadopoulos, Bert Vogel-

stein, Kenneth W. Kinzler, and Victor E. Velculescu. The Consensus Coding

Sequences of Human Breast and Colorectal Cancers. Science, 314(5797) :268-

274, 2006.

[19] Jerrod J. Schwartz and Stephen R. Quake. High density single molecule surface

patterning with colloidal epitaxy. Applied Physics Letters, 91(8):083902, 2007.

[20] Jerrod J. Schwartz, Stephen R. Quake, and Milan Mrksich. Surface chemistry

and deposition techniques. US Patent Application No. 20090053690, 2009.

[21] Randy Stoltenberg, Jerrod J. Schwartz, Stephen R. Quake, and Zhenan Bao.

Patterned nanoparticle arrays as platforms for single-molecule fluorescence

imaging. In preparation, 2009.

[22] Randy Stoltenberg, Jerrod J. Schwartz, Stephen R. Quake, and Zhenan Bao.

Patterned nanoparticle arrays as platforms for single-molecule fluorescence

imaging. US provisional patent filed., 2009.

BIBLIOGRAPHY 173

[23] Jerrod J. Schwartz and Stephen R. Quake. A single molecule measurement of

the "speed limit" of dna polymerase. In review, 2009.

[24] Jerrod J. Schwartz, Stavros Stavrakis, and Stephen R. Quake. High temperature

single molecule observation of a thermophilic enzyme with colloidal lenses. In

preparation, 2009.

[25] F. Sanger, S. Nicklen, and A. R. Coulson. DNA sequencing with chain-

terminating inhibitors. Proceedings of the National Academy of Sciences of

the United States of America, 74(12) :5463-5467, 1977.

[26] Lloyd M. Smith, Jane Z. Sanders, Robert J. Kaiser, Peter Hughes, Chris Dodd,

Charles R. Connell, Cheryl Heiner, Stephen B. H. Kent, and Leroy E. Hood.

Fluorescence detection in automated dna sequence analysis. Nature, 321:674-

679, 1986.

[27] J. Craig Venter et al. The sequence of the human genome. Science, 291:1304-

1351, 2001.

[28] Eric S. Lander et al. Initial sequencing and analysis of the human genome.

Nature, 409:860-921, 2001.

[29] Levy S, Sutton G, Ng PC, Feuk L, Halpern AL, and et al. The diploid genome

sequence of an individual human. PLoS Biology, 5(10), 2007.

[30] David A. Wheeler, Maithreyan Srinivasan, Michael Egholm, Yufeng Shen, Lei

Chen, Amy McGuire, Wen He, Yi-Ju Chen, Vinod Makhijani, G. Thomas Roth,

Xavier Gomes, Karrie Tartaro, Faheem Niazi, Cynthia L. Turcotte, Gerard P.

Irzyk, James R. Lupski, Craig Chinault, Xing zhi Song, Yue Liu, Ye Yuan,

BIBLIOGRAPHY 174

Lynne Nazareth, Xiang Qin, Donna M. Muzny, Marcel Margulies, George M.

Weinstock, Richard A. Gibbs, and Jonathan M. Rothberg. The complete

genome of an individual by massively parallel dna sequencing. Nature, 452:872-

876, 2008.

[31] Dna sequencing of a cytogenetically normal acute myeloid leukaemia genome.

Nature, 456(7218) :66-72, 2008. 0028-0836 10.1038/nature07485 10.1038/na-

ture07485.

[32] The diploid genome sequence of an asian individual. Nature, 456(7218):60-65,

2008.

[33] David R. Bentley et al. Accurate whole human genome sequencing using re

versible terminator chemistry. Nature, 456:53-59, 2008.

[34] Daniel Branton, David W Deamer, Andre Marziali, Hagan Bayley, Steven A

Benner, Thomas Butler, Massimiliano Di Ventra, Slaven Garaj, Andrew Hibbs,

Xiaohua Huang, Stevan B Jovanovich, Predrag S Krstic, Stuart Lindsay, Xin-

sheng Sean Ling, Carlos H Mastrangelo, Amit Meller, John S Oliver, Yuriy V

Pershin, J Michael Ramsey, Robert Riehn, Gautam V Soni, Vincent Tabard-

Cossa, Meni Wanunu, Matthew Wiggin, and Jeffery A Schloss. The potential

and challenges of nanopore sequencing. Nature Biotechnology, 26:1146-1153,

2008.

[35] John J. Kasianowicz, Eric Brandin, Daniel Branton, and David W. Deamer.

Characterization of individual polynucleotide molecules using a membrane

channel. Proceedings of the National Academy of Sciences of the United States

of America, 93(24): 13770-13773, 1996.

BIBLIOGRAPHY 175

[36] David W. Deamer and Daniel Branton. Characterization of nucleic acids by

nanopore analysis. Accounts of Chemical Research, 35(10):817-825, 2002.

[37] Hagan Bayley. Sequencing single molecules of dna. Current Opinion in Chemical

Biology, 10(6):628 - 637, 2006. Model systems / Biopolymers.

[38] Nurit Ashkenasy, Jorge Snchez-Quesada, Hagan Bayley, and M. Reza Ghadiri.

Recognizing a single base in an individual dna strand: A step toward dna se

quencing in nanopores. Angewandte Chemie International Edition, 44(9): 1401-

1404, 2005.

[39] Nick Mitchell and Stefan Howorka. Chemical tags facilitate the sensing of indi

vidual dna strands with nanopores. Angewandte Chemie International Edition,

47(30):5565-5568, 2008.

[40] Breton Hornblower, Amy Coombs, Richard D Whitaker, Anatoly Kolomeisky,

Stephen J Picone, Amit Meller, and Mark Akeson. Single-molecule analysis of

dna-protein complexes using nanopores. Nature Methods, 4:315-317, 2007.

[41] Scott L. Cockroft, John Chu, Manuel Amorin, and M. Reza Ghadiri. A single-

molecule nanopore device detects dna polymerase activity with single-nucleotide

resolution. Journal of the American Chemical Society, 130(3) :818-820, 2008.

[42] William K. Thomas and William Glover. Direct Sequencing by TEM of Z-

Substituted DNA Molecules. Next Generation Genome Sequencing, pages 103-

116, 2008.

[43] Jonas Korlach, Patrick J. Marks, Ronald L. Cicero, Jeremy J. Gray, Devon L.

Murphy, Daniel B. Roitman, Thang T. Pham, Geoff A. Otto, Mathieu Foquet,

BIBLIOGRAPHY 176

and Stephen W. Turner. Selective aluminum passivation for targeted immo

bilization of single dna polymerase molecules in zero-mode waveguide nanos-

tructures. Proceedings of the National Academy of Sciences, 105(4):1176-1181,

2008.

[44] William J. Greenleaf and Steven M. Block. Single-Molecule, Motion-Based

DNA Sequencing Using RNA Polymerase. Science, 313(5788) :801-, 2006.

[45] Kristina M. Herbert, Arthur La Porta, Becky J. Wong, Rachel A. Mooney,

Keir C. Neuman, Robert Landick, and Steven M. Block. Sequence-resolved

detection of pausing by single rna polymerase molecules. Cell, 125(6):1083-

1094, 2006.

[46] Dmitry Pushkarev, Norma Neff, and Stephen R. Quake. Submitted, 2009.

[47] Z. Gueroui, C. Place, E. Freyssingeas, and B. Berge. Observation by fluorescence

microscopy of transcription on single combed dna. Proceedings of the National

Academy of Sciences of the United States of America, 99(9):6005-6010, 2002.

[48] Y. Y. Liu, P. Y. Wang, S. X. Dou, W. C. Wang, P. Xie, H. W. Yin, and X. D.

Zhang. Ionic effect on combing of single dna molecules, and observation of their

force-induced melting by fluorescence microscopy. Journal of Chemical Physics,

121(9):4302-4309, 2004.

[49] P. Bjork, A. Herland, I. G. Scheblykin, and O. Inganas. Single molecular

imaging and spectroscopy of conjugated polyelectrolytes decorated on stretched

aligned dna. Nano Letters, 5(10): 1948-1953, 2005.

BIBLIOGRAPHY 177

[50] D. Mijatovic, J. C. T. Eijkel, and A. van den Berg. Technologies for nanofluidic

systems: top-down vs. bottom-up - a review. Lab on a Chip, 5(5):492-500,

2005.

[51] H. Yokota, J. Sunwoo, M. Sarikaya, G. van den Engh, and R. Aebersold. Spin-

stretching of dna and protein molecules for detection by fluorescence and atomic

force microscopy. Analytical Chemistry, 71(19):4418-4422, 1999.

[52] H. Nakao, M. Gad, S. Sugiyama, K. Otobe, and T. Ohtani. Transfer-printing

of highly aligned dna nanowires. Journal of the American Chemical Society,

125(24) :7162-7163, 2003.

[53] V. Dugas, J. Broutin, and E. Souteyrand. Droplet evaporation study applied

to dna chip manufacturing. Langmuir, 21(20):9130-9136, 2005.

[54] M. Fujita, W. Mizutani, M. Gad, H. Shigekawa, and H. Tokumoto. Patterning

dna on mu m scale on mica. Ultramicroscopy, 91(l-4):281-285, 2002.

[55] S. A. Lange, V. Benes, D. P. Kern, J. K. H. Horber, and A. Bernard. Mi-

crocontact printing of dna molecules. Analytical Chemistry, 76(6)-.1641-1647,

2004.

[56] G. J. Zhang, T. Tanii, T. Funatsu, and I. Ohdomari. Patterning of dna nanos-

tructures on silicon surface by electron beam lithography of self-assembled

monolayer. Chemical Communications, (7):786-787, 2004.

[57] L. M. Demers, D. S. Ginger, S. J. Park, Z. Li, S. W. Chung, and C. A. Mirkin.

Direct patterning of modified oligonucleotides on metals and insulators by dip-

pen nanolithography. Science, 296(5574): 1836-1838, 2002.

BIBLIOGRAPHY 178

[58] N. L. Rosi and C. A. Mirkin. Nanostructures in biodiagnostics. Chemical

Reviews, 105(4):1547-1562, 2005.

[59] F. Matsumoto, M. Kamiyama, K. Nishio, and H. Masuda. Highly ordered

nanopatterning of dna with 40 nm diameter using anodic porous alumina sub

strate. Japanese Journal of Applied Physics Part 2-Letters Express Letters,

44(8-ll):L355-L358, 2005.

[60] P. Pammer, R. Schlapak, M. Sonnleitner, A. Ebner, R. Zhu, P. Hinterdorfer,

O. Hoglinger, H. Schindler, and S. Howorka. Nanopatterning of biomolecules

with microscale beads. Chemphyschem, 6(5):900-903, 2005.

[61] J. P. Hoogenboom, D. L. J. Vossen, C. Faivre-Moskalenko, M. Dogterom, and

A. van Blaaderen. Patterning surfaces with colloidal particles using optical

tweezers. Applied Physics Letters, 80(25) : 4828-4830, 2002.

[62] John William Strutt (III Lord Rayleigh). Investigations in optics, with spe

cial reference to the spectroscope. Philosophical Magazine, VIII:261-274,403-

411,477-486, 1879.

[63] H Nyquist. Certain topics in telegraph transmission theory. Trans. AIEE,

47:617-644, 1928.

[64] W. Weibull. A statistical distribution function of wide applicability. Journal of

Applied Mechanics, 18(3):293-297, 1951.

[65] M. G. L. Gustafsson. Nonlinear structured-illumination microscopy: Wide-field

fluorescence imaging with theoretically unlimited resolution. Proceedings of the

BIBLIOGRAPHY 179

National Academy of Sciences of the United States of America, 102(37): 13081,

2005. 0027-8424.

[66] M.G.L. Gustafsson. Surpassing the lateral resolution limit by a factor of two

using structured illumination microscopy. Journal of Microscopy, 198(2) :82-87,

2000.

[67] Stefan W. Hell and Jan Wichmann. Breaking the diffraction resolution limit

by stimulated emission: stimulated-emission-depletion fluorescence microscopy.

Opt. Lett, 19(ll):780-782, 1994.

[68] M. P. Gordon, T. Ha, and P. R. Selvin. Single-molecule high-resolution imaging

with photobleaching. Proceedings of the National Academy of Sciences of the

United States of America, 101(17):6462-6465, 2004.

[69] X. H. Qu, D. Wu, L. Mets, and N. F. Scherer. Nanometer-localized multiple

single-molecule fluorescence microscopy. Proceedings of the National Academy

of Sciences of the United States of America, 101(31):11298-11303, 2004.

[70] A. Yildiz, J. N. Forkey, S. A. McKinney, T. Ha, Y. E. Goldman, and P. R.

Selvin. Myosin v walks hand-over-hand: Single fluorophore imaging with 1.5-

nm localization. Science, 300(5628):2061-2065, 2003.

[71] Samuel T. Hess, Thanu P.K. Girirajan, and Michael D. Mason. Ultra-high res

olution imaging by fluorescence photoactivation localization microscopy. Bio

physical Journal, 91(11):4258 - 4272, 2006.

BIBLIOGRAPHY 180

[72] Eric Betzig, George H. Patterson, Rachid Sougrat, O. Wolf Lindwasser, Scott

Olenych, Juan S. Bonifacino, Michael W. Davidson, Jennifer Lippincott-

Schwartz, and Harald F. Hess. Imaging Intracellular Fluorescent Proteins at

Nanometer Resolution. Science, 313(5793): 1642-1645, 2006.

[73] M. J. Rust. Sub-diffraction-limit imaging by stochastic optical reconstruction

microscopy (storm). Nature methods, 3(10):793, 2006. 1548-7091.

[74] Mathieu Foquet, Kevan T. Samiee, Xiangxu Kong, Bidhan P. Chauduri,

Paul M. Lundquist, Stephen W. Turner, Jake Freudenthal, and Daniel B. Roit-

man. Improved fabrication of zero-mode waveguides for single-molecule detec

tion. Journal of Applied Physics, 103(3):034301, 2008.

[75] Masato Aizawa and Jillian M. Buriak. Block copolymer templated chemistry

for the formation of metallic nanoparticle arrays on semiconductor surfaces.

Chemistry of Materials, 19(21):5090-5101, 2007.

[76] Peidong Yang, Dongyuan Zhao, David I. Margolese, Bradley F. Chmelka, and

Galen D. Stucky. Block copolymer templating syntheses of mesoporous metal

oxides with large ordering lengths and semicrystalline framework. Chemistry of

Materials, ll(10):2813-2826, 1999.

[77] Ryan D. Bennett, Andrew C. Miller, Naomi T. Kohen, Paula T. Hammond,

Darrell J. Irvine, and Robert E. Cohen. Strategies for controlling the planar

arrangement of block copolymer micelles and inorganic nanoclusters. Macro-

molecules, 38(26): 10728-10735,- 2005.

BIBLIOGRAPHY 181

[78] G. Kastle, H.-G. Boyen, F. Weigl, G. Lengl, T. Herzog, P. Ziemann, S. Rieth-

muller, O. Mayer, C. Hartmann, J.P. Spatz, M. Moller, M. Ozawa, F. Ban-

hart, M.G. Gamier, and P. Oelhafen. Micellar nanoreactors-preparation and

characterization of hexagonally ordered arrays of metallic nanodots. Advanced

Functional Materials, 13(11):853-861, 2003.

[79] C Papadopoulos, R S van den Boogaard, S Clutterbuck, G M Lindsay, and

S R Winnitoy. Large-area ordered c60 nanoparticle arrays. Nanotechnology,

18(24):245306 (4pp), 2007.

[80] T. Gervais. Mass transport and surface reactions in microfluidic systems. Chem

ical engineering science, 61(4):1102, 2006. 0009-2509.

[81] M. Hegner. Dna handles for single molecule experiments. Single molecules,

1(2):139, 2000. 1438-5163.

[82] E. P. Kartalov, M. A. Unger, and S. R. Quake. Polyelectrolyte surface inter

face for single-molecule fluorescence studies of dna polymerase. Biotechniques,

34(3):505-510, 2003. 0736-6205.

[83] E. Betzig, J. K. Trautman, T. D. Harris, J. S. Weiner, and R. L. Kostelak.

Breaking the Diffraction Barrier: Optical Microscopy on a Nanometric Scale.

Science, 251 (5000): 1468-1470, 1991.

[84] E. Betzig, A. Lewis, A. Harootunian, M. Isaacson, and E. Kratschmer. Near field

scanning optical microscopy (nsom): Development and biophysical applications.

Biophysical Journal, 49(1):269 - 279, 1986.

BIBLIOGRAPHY 182

[85] T. J. Yang, Guillaume A. Lessard, and Stephen R. Quake. An apertureless near-

field microscope for fluorescence imaging. Applied Physics Letters, 76(3) :378-

380, 2000.

[86] David P. Fromm, Arvind Sundaramurthy, P. James Schuck, Gordon Kino, and

W. E. Moerner. Gap-dependent optical coupling of single bowtie nanoantennas

resonant in the visible. Nano Letters, 4(5):957-961, 2004.

[87] Irina V. Smolina, Vadim V. Demidov, and Maxim D. Frank-Kamenetskii. Paus

ing of dna polymerases on duplex dna templates due to ligand binding in vitro.

Journal of Molecular Biology, 326(4):1113-1125, 2003.

[88] Atul M. Deshpande and Carol S. Newlon. Dna replication fork pause sites

dependent on transcription. Science, 272(5264): 1030-1033, 1996.

[89] Maria M. Krasilnikova, George M. Samadashwily, Andrey S. Krasilnikov, and

Sergei M. Mirkin. Transcription through a simple dna repeat blocks replication

elongation. The EMBO Journal, 17:50955102, 1998.

[90] Yasushi Takeuchi, Takashi Horiuchi, and Takehiko Kobayashi. Transcription-

dependent recombination and the role of fork collision in yeast rdna. Genes

Dev., 17(12):1497-1506, 2003.

[91] P. Bedinger, M. Munn, and B. M. Alberts. Sequence-specific pausing during

in vitro dna replication on double- stranded dna templates. J. Biol. Chew,.,

264(28): 16880-16886, 1989.

[92] Richard J. LaDuca, Philip J. Fay, Christy Chuang, Charles S. McHenry, and

Robert A. Bambara. Site-specific pausing of deoxyribonucleic acid synthesis

BIBLIOGRAPHY 183

catalyzed by four forms of escherichia coli dna polymerase iii. Biochemistry,

22(22):5177-5188, 1983.

[93] Prancene J. Lemoine, Natasha P. Degtyareva, Kirill Lobachev, and Thomas D.

Petes. Chromosomal translocations in yeast induced by low levels of dna poly

merase: A model for chromosome fragile sites. Cell, 120(5):587-598, 2005.

[94] Rita S. Cha and Nancy Kleckner. Atr homolog mecl promotes fork progression,

thus averting breaks in replication slow zones. Science, 297(5581) :602-606, 2002.

[95] Seongman Kang, Keiichi Ohshima, Miho Shimizu, Sorour Amirhaeri, and

Robert D. Wells. Pausing of dna synthesis in vitro at specific loci in ctg

and egg triplet repeats from human hereditary disease genes. J. Biol. Chem.,

270(45):27014-27021, 1995.

[96] George M. Samadashwily, Gordana Raca, and Sergei M. Mirkin. Trinu

cleotide repeats affect dna replication in vivo. Nat Genet, 17(3):298-304, 1997.

10.1038/ngll97-298 10.1038/ngll97-298.

[97] Taurai Tasara, Bernhard Angerer, Martine Damond, Holger Winter, Sabine

Dorhofer, Ulrich Hubscher, and Mario Amacker. Incorporation of reporter

molecule-labeled nucleotides by dna polymerases, ii. high-density labeling of

natural dna. Nucl. Acids Res., 31(10):2636-2646, 2003.

[98] Berenike Maier, David Bensimon, and Vincent Croquette. Replication by a

single dna polymerase of a stretched single-stranded dna. Proceedings of the

National Academy of Sciences of the United States of America, 97(22): 12002-

12007, 2000.

BIBLIOGRAPHY 184

[99] Gijs J. L. Wuite, Steven B. Smith, Mark Young, David Keller, and Carlos Busta-

rnante. Single-molecule studies of the effect of template tension on t7 dna poly

merase activity. Nature, 404(6773): 103-106, 2000. 0028-0836 10.1038/35003614

10.1038/35003614.

[100] Michelle D. Wang, Mark J. Schnitzer, Hong Yin, Robert Landick, Jeff Gelles,

and Steven M. Block. Force and velocity measured for single molecules of rna

polymerase. Science, 282(5390):902-907, 1998.

[101] D. S. Mytelka and M. J. Chamberlin. Analysis and suppression of dna poly

merase pauses associated with a trinucleotide consensus. Nucl. Acids Res.,

24(14):2774-2781, 1996.

[102] Sangjin Kim, Paul C. Blainey, Charles M. Schroeder, and X. Sunney Xie. Mul

tiplexed single-molecule assay for enzymatic activity on flow-stretched dna. Nat

Meth, 4(5):397-399, 2007. 1548-7091 10.1038/nmethl037 10.1038/nmethl037.

[103] Jong-Bong Lee, Richard K. Hite, Samir M. Hamdan, X. Sunney Xie, Charles C.

Richardson, and Antoine M. van Oijen. Dna primase acts as a molecular brake

in dna replication. Nature, 439(7076):621-624, 2006. 0028-0836 10.1038/na-

ture04317 10.1038/nature04317.

[104] Nathan A. Tanner, Samir M. Hamdan, Slobodan Jergic, Karin V. Loscha,

Patrick M. Schaeffer, Nicholas E. Dixon, and Antoine M. van Oijen. Single-

molecule studies of fork dynamics in escherichia coli dna replication. Nat Struct

Mol Biol, 15(2):170-176, 2008. 1545-9993.

BIBLIOGRAPHY 185

[105] R. D. Kuchta, V. Mizrahi, P. A. Benkovic, K. A. Johnson, and S. J. Benkovic.

Kinetic mechanism of dna polymerase i (klenow). Biochemistry, 26(25):8410-

8417, 1987.

[106] Michael E. Dahlberg and Stephen J. Benkovic. Kinetic mechanism of dna poly

merase i(klenow fragment): identification of a second conformational change

and evaluation of the internal equilibrium constant. Biochemistry, 30(20) :4835-

4843, 1991.

[107] John C. Crocker and David G. Grier. Methods of digital video microscopy for

colloidal studies. Journal of Colloid and Interface Science, 179(1):298 - 310,

1996.

[108] Jacob W. J. Kerssemakers, E. Laura Munteanu, Liedewij Laan, Tim L. Noet-

zel, Marcel E. Janson, and Marileen Dogterom. Assembly dynamics of micro

tubules at molecular resolution. Nature, 442(7103): 709-712, 2006. 0028-0836

10.1038/nature04928 10.1038/nature04928.

[109] R. Sabanayagam Chandran, S. Eid John, and Meller Amit. Long time scale

blinking kinetics of cyanine fluorophores conjugated to dna and its effect on

f[o-umlaut]rster resonance energy transfer. The Journal of Chemical Physics,

123(22):224708, 2005.

[110] Ivan Rasnik, Sean A. McKinney, and Taekjip Ha. Nonblinking and long-lasting

single-molecule fluorescence imaging. Nat Meth, 3(ll):891-893, 2006. 1548-7091

10.1038/nmeth934 10.1038/nmeth934.

BIBLIOGRAPHY 186

[111] Colin Echeverria Aitken, R. Andrew Marshall, and Joseph D. Puglisi. An oxy

gen scavenging system for improvement of dye stability in single-molecule fluo

rescence experiments. Biophysical Journal, 94(5):1826-1835, 2008. 0006-3495.

[112] William A. Rees, Thomas D. Yager, John Korte, and Peter H. Von Hippel.

Betaine can eliminate the base pair composition dependence of dna melting.

Biochemistry, 32(1):137-144, 1993.

[113] W. Henke, K. Herdel, K. Jung, D. Schnorr, and S. A. Loening. Betaine improves

the per amplification of gc-rich dna sequences. Nucl. Acids Res., 25(19):3957-

3958, 1997.

[114] N. Baskaran, R. P. Kandpal, A. K. Bhargava, M. W. Glynn, A. Bale, and S. M.

Weissman. Uniform amplification of a mixture of deoxyribonucleic acids with

varying gc content. Genome Res., 6(7):633-638, 1996.

[115] Catherine M. Joyce, Olga Potapova, Angela M. DeLucia, Xuanwei Huang, Van-

dana Purohit Basu, and Nigel D. F. Grindley. Fingers-closing and other rapid

conformational changes in dna polymerase i (klenow fragment) and their role

in nucleotide selectivity. Biochemistry, 47(23) :6103-6116, 2008.

[116] Kamalendra Singh, Aashish Srivastava, Smita S. Patel, and Mukund J. Modak.

Participation of the Fingers Subdomain of Escherichia coli DNA Polymerase I

in the Strand Displacement Synthesis of DNA. J. Biol. Chem., 282(14): 10594-

10604, 2007.

[117] Mara S. Soengas, Crisanto Gutirrez, and Margarita Salas. Helix-destabilizing

activity of [phi] 29 single-stranded dna binding protein: Effect on the elongation

BIBLIOGRAPHY 187

rate during strand displacement dna replication. Journal of Molecular Biology,

253(4):517-529, 1995.

[118] Michael Zuker. Mfold web server for nucleic acid folding and hybridization

prediction. Nucl. Acids Res., 31(13):3406-3415, 2003.

[119] Sean A. McKinney, Anne-Cecile Declais, David M. J. Lilley, and Taekjip Ha.

Structural dynamics of individual holliday junctions. Nat Struct Mol Biol,

10(2):93-97, 2003. 1072-8368 10.1038/nsb883 10.1038/nsb883.

[120] Mikhail Karymov, Douglas Daniel, Otto F. Sankey, and Yuri L. Lyubchenko.

Holliday junction dynamics and branch migration: Single-molecule analysis.

Proceedings of the National Academy of Sciences of the United States of Amer

ica, 102(23):8186-8191, 2005.

[121] Eric Betzig and Robert J. Chichester. Single Molecules Observed by Near-Field

Scanning Optical Microscopy. Science, 262(5138): 1422-1425, 1993.

[122] Jerome Wenger, Davy Gerard, Heykel Aouani, and Herve Rigneault. Disposable

microscope objective lenses for fluorescence correlation spectroscopy using latex

microspheres. Analytical Chemistry, 80(17):6800-6804, 2008.

[123] Kazuko Koyama, Masahiro Yoshita, Motoyoshi Baba, Tohru Suemoto, and

Hidefumi Akiyama. High collection efficiency in fluorescence microscopy with

a solid immersion lens. Applied Physics Letters, 75(12):1667-1669, 1999.

[124] J. P. Brody and S. R. Quake. A self-assembled microlensing rotational probe.

Applied Physics Letters, 74(1): 144-146, 1999.

BIBLIOGRAPHY 188

[125] L. E. Helseth, H. Z. Wen, and T. M. Fischer. Colloidal optomagnetic dimmer.

Langmuir, 22(8):3941-3944, 2006.

[126] M. H. Wu and G. M. Whitesides. Fabrication of arrays of two-dimensional mi-

cropatterns using microspheres as lenses for projection photolithography. Ap

plied Physics Letters, 78(16):2273-2275, 2001.

[127] M. H. Wu, K. E. Paul, and G. M. Whitesides. Patterning flood illumination

with microlens arrays. Applied Optics, 41(13):2575-2585, 2002.

[128] Euan Mcleod and Craig B. Arnold. Subwavelength direct-write nanopatterning

using optically trapped microspheres. Nature Nanotechnology, 3:413-417, 2008.

[129] D. McGloin, A. E. Carruthers, K. Dholakia, and E. M. Wright. Optically bound

microscopic particles in one dimension. Physical Review E, 69(2), 2004. Part 1

021403.

[130] S. A. Tatarkova, A. E. Carruthers, and K. Dholakia. One-dimensional optically

bound arrays of microscopic particles. Physical Review Letters, 89(28), 2002.

283901.

[131] Janina Kneipp, Xiangting Li, Margaret Sherwood, Ulrich Panne, Harald

Kneipp, Mark I. Stockman, and Katrin Kneipp. Gold nanolenses generated

by laser ablation-efficient enhancing structure for surface enhanced raman scat

tering analytics and sensing. Analytical Chemistry, 80(11):4247-4251, 2008.

BIBLIOGRAPHY 189

[132] A. Farjadpour, David Roundy, Alejandro Rodriguez, M. Ibanescu, Peter

Bermel, J. D. Joannopoulos, Steven G. Johnson, and G. W. Burr. Improv

ing accuracy by subpixel smoothing in the finite-difference time domain. Opt.

Lett, 31(20):2972-2974, 2006.

[133] JB Segur and HE Oberstar. Viscosity of glycerol and its aqueous solutions.

Industrial and Engineering Chemistry, 43(9):2117-2120, 1951.

[134] K. T. Shimizu, W. K. Woo, B. R. Fisher, H. J. Eisler, and M. G. Bawendi.

Surface-enhanced emission from single semiconductor nanocrystals. Phys. Rev.

Lett, 89(11):117401, Aug 2002.

[135] Jian Zhanga Yi Fua and Joseph R. Lakowicz. Suppressed blinking in single

quantum dots (qds) immobilized near silver island films (sifs). Chemical Physics

Letters, 447:96-100, 2007.

[136] Yi Fu and Joseph R. Lakowicz. Enhanced fluorescence of cy5-labeled oligonu

cleotides near silver island films: A distance effect study using single molecule

spectroscopy. The Journal of Physical Chemistry B, 110(45):22557-22562, 2006.

[137] Jian Zhang, Yi Fu, Krishanu Ray, Mustafa H. Chowdhury, Henryk Szmacin-

ski, Kazimierz Nowaczyk, and Joseph R. Lakowicz. Single molecule photo-

physics near metallic nanostructures. Single Molecule Spectroscopy and Imag

ing, 6862(1):68620S, 2008.

[138] Mustafa H. Chowdhury, James Pond, Stephen K. Gray, and Joseph R. Lakowicz.

Systematic computational study of the effect of silver nanoparticle dimers on the

BIBLIOGRAPHY 190

coupled emission from nearby fluorophores. The Journal of Physical Chemistry

C, 112(30):11236-11249, 2008.

[139] Jian Zhang, Yi Fu, Mustafa H. Chowdhury, and Joseph R. Lakowicz. Single-

molecule studies on fluorescently labeled silver particles: Effects of particle size.

The Journal of Physical Chemistry C, 112(1): 18-26, 2008.

[140] Joseph R. Lakowicz. Radiative decay engineering: Biophysical and biomedical

applications. Analytical Biochemistry, 298(1):1 - 24, 2001.

[141] Adam Dorfman, Nitin Kumar, and Jong-in Hahm. Highly sensitive biomolecular

fluorescence detection using nanoscale zno platforms. Langmuir, 22(11):4890-

4895, 2006.

[142] Gu Cuiping, Huang Jiarui, Ni Ning, Li Minqiang, and Liu Jinhuai. Detection of

dna hybridization based on sno2 nanomaterial enhanced fluorescence. Journal

of Physics D: Applied Physics, (17):175103, 2008.

[143] Tomokatsu Hayakawa, S. Tamil Selvan, and Masayuki Nogami. Field enhance

ment effect of small ag particles on the fluorescence from eu[sup 3 + ]-doped

siofsub 2] glass. Applied Physics Letters, 74(11):1513-1515, 1999.

[144] Konstantin Sokolov, George Chumanov, and Therese M. Cotton. Enhancement

of molecular fluorescence near the surface of colloidal metal films. Analytical

Chemistry, 70(18):3898-3905, 1998.

[145] Hiroaki Yokota, Kiwamu Saito, and Toshio Yanagida. Single molecule imag

ing of fluorescently labeled proteins on metal by surface plasmons in aqueous

solution. Phys. Rev. Lett, 80(20):4606~4609, May 1998.

BIBLIOGRAPHY 191

[146] F. D. Stefani, K. Vasilev, N. Bocchio, N. Stoyanova, and M. Kreiter. Surface-

plasmon-mediated single-molecule fluorescence through a thin metallic film.

Phys. Rev. Lett, 94(2):023005, Jan 2005.

[147] J. Enderlein. Single-molecule fluorescence near a metal layer. Chemical Physics,

247:1-9, August 1999.

[148] Jrg Enderlein. A theoretical investigation of single-molecule fluorescence detec

tion on thin metallic layers. Biophysical Journal, 78(4):2151 - 2158, 2000.

[149] Edward H. Hellen and Daniel Axelrod. Fluorescence emission at dielectric and

metal-film interfaces. J. Opt. Soc. Am. B, 4(3):337-350, 1987.

[150] I. Pockrand, A. Brillante, and D. Mobius. Nonradiative decay of molecular

excitation at a metal interface. Nuovo Cimento B Serie, 63:350-357, May 1981.

[151] Sylvie Doubli, Michael R Sawaya, and Tom Ellenberger. An open and closed

case for all polymerases. Structure, 7(2):R31 - R35, 1999.

[152] M. A. Augustin, W. Ankenbauer, and B. Angerer. Progress towards single-

molecule sequencing: enzymatic synthesis of nucleotide-specifically labeled dna.

Journal of Biotechnology, 86(3):289 - 301, 2001. DNA-Sequencing at the Single

Molecule Level.

[153] Thomas Hall. BioEdit: a user-friendly biological sequence alignment editor and

analysis program for Windows 95/98/NT. Nucleic Acids Symposium Series,

41:95-98, 1999.

BIBLIOGRAPHY 192

[154] Nicolas Guex and Manuel C. Peitsch. Swiss-model and the swiss-pdb viewer: An

environment for comparative protein modeling. Electrophoresis, 18(15):2714-

2723, 1997.

[155] Simon C. Lovell, Ian W. Davis, W. Bryan Arendall III, Paul I. W. de Bakker,

J. Michael Word, Michael G. Prisant, Jane S. Richardson, and David C.

Richardson. Structure validation by c-alpha geometry: phi, psi and c-beta

deviation. Proteins: Structure, Function, and Genetics, 50(3):437-450, 2003.

[156] D. Benjamin Gordon and Stephen L. Mayo. Radical performance enhancements

for combinatorial optimization algorithms based on the dead-end elimination

theorem. Journal of Computational Chemistry, 19(13):1505-1514, 1998.

[157] LS Beese, V Derbyshire, and TA Steitz. Structure of DNA polymerase I Klenow

fragment bound to duplex DNA. Science, 260(5106):352-355, 1993.

[158] Y Li, S Korolev, and G Waksman. Crystal structures of open and closed forms

of binary and ternary complexes of the large fragment of Thermus aquaticus

DNA polymerase I: structural basis for nucleotide incorporation. The EMBO

Journal, 17(24) :7514-7525, 1998.

[159] Gabriel Waksman Ying Li. Crystal structures of a ddATP-, ddTTP-, ddCTP,

and ddGTP- trapped ternary complex of Klentaql: Insights into nucleotide

incorporation and selectivity. Protein Science, 10(6):1225-1233, 2001.

[160] Sylvie Doublie, Stanley Tabor, Alexander M. Long, Charles Richarson, and

Tom Ellenberger. Crystal structure of a bacteriophage T7 DNA replication

complex at 2.2 A resolution. Nature, 391:251-258, 1998.

BIBLIOGRAPHY 193

[161] Stefan G. Sarafianos, Kalyan Das, Arthur D. Clark, Jianping Ding, Paul L.

Boyer, Stephen H. Hughes, and Edward Arnold. Lamivudine (3TC) resistance

in HIV-1 reverse transcriptase involves steric hindrance with B-branched amino

acids. Proceedings of the National Academy of Sciences of the United States of

America, 96(18): 10027-10032, 1999.

[162] Y Hsiou, J Ding, K Das, AD Clark, Jr, SH Hughes, and E Arnold. Structure of

unliganded hiv-1 reverse transcriptase at 2.7 resolution: implications of con

formational changes for polymerization and inhibition mechanisms. Structure,

4(7):853 - 860, 1996.

[163] A.Chapin Rodriguez, Hee-Won Park, Chen Mao, and Lorena S Beese. Crystal

structure of a pol [alpha] family dna polymerase from the hyperthermophilic

archaeon thermococcus sp. 9n-7. Journal of Molecular Biology, 299(2):447 -

462, 2000.

[164] Marianna Teplova, Scot T. Wallace, Valentina Tereshko, George Minasov, Al

ice M. Symons, P. Dan Cook, Muthiah Manoharan, and Martin Egli. Structural

origins of the exonuclease resistance of a zwitterionic RNA. Proceedings of the

National Academy of Sciences of the United States of America, 96(25): 14240-

14245, 1999.

[165] Karl-Peter Hopfner, Andreas Eichinger, Richard A. Engh, Frank Laue, Wal-

traud Ankenbauer, Robert Huber, and Bernhard Angerer. Crystal structure of

a thermostable type B DNA polymerase from Thermococcus gorgonarius. Pro

ceedings of the National Academy of Sciences of the United States of America,

96(7):3600-3605, 1999.

BIBLIOGRAPHY 194

[166] Youngsoo Kim, Soo Hyun Eom, Jimin Wang, Dae-Sil Lee, Se Won Suh, and

Thomas A. Steitz. Crystal structure of Thermus aquaticus DNA polymerase.

Nature, 376:612-626, 1995.

[167] Susanne Brakmann and Petra Nieckchen. The Large Fragment of Escherichia

coli DNA Polymerase I Can Synthesize DNA Exclusively from Fluorescently

Labeled Nucleotides. ChemBioChem, 2(10):773-777, 2001.

[168] Christopher J. Lacenere, Neil K. Garg, Brian M. Stoltz, and Stephen R. Quake.

Effects of a modified dye-labeled nucleotide spacer arm on incorporation by

thermophilic dna polymerases. Nucleosides, Nucleotides, and Nucleic Acids,

25(1):9-15, 2006.

[169] Brent A. Mulder, Steve Anaya, Peilin Yu, Keun Woo Lee, Anvy Nguyen, Ja

son Murphy, Richard Willson, James M. Briggs, Xiaolian Gao, and Susan H.

Hardin. Nucleotide modification at the gamma-phosphate leads to the improved

fidelity of HIV-1 reverse transcriptase. Nucl. Acids Res., 33(15):4865-4873,

2005.

[170] Greg Bashford, Don Lamb, Dan Grone, Bob Eckles, Kevin Kornelsen, Lyle

Middendorf, and John Williams. Automated bead-trapping apparatus and con

trol system for single-molecule dna sequencing. Opt. Express, 16(5):3445-3455,

2008.

[171] Jonas Korlach; Arek Bibillo; Jeffrey Wegener; Paul Peluso; Thang T. Pham;

Insil Park; Sonya Clark; Geoff A. Otto; Stephen W. Turner. Long, processive

enzymatic dna synthesis using 100 dye-labeled terminal phosphate-linked nu

cleotides. Nucleosides, Nucleotides, and Nucleic Acids, 27(9):1072-1082, 2008.

BIBLIOGRAPHY 195

[172] Robi D. Mitra, Jay Shendure, Jerzy Olejnik, Edyta-Krzymanska-Olejnik, and

George M. Church. Fluorescent in situ sequencing on polymerase colonies.

Analytical Biochemistry, 320(1):55 - 65, 2003.

[173] Tae Seok Seo, Xiaopeng Bai, Hameer Ruparel, Zengmin Li, Nicholas J. Turro,

and Jingyue Ju. Photocleavable fluorescent nucleotides for dna sequencing on a

chip constructed by site-specific coupling chemistry. Proceedings of the National

Academy of Sciences of the United States of America, 101(15):5488-5493, 2004.

[174] Tae Seok Seo, Xiaopeng Bai, Dae Hyun Kim, Qinglin Meng, Shundi Shi, Hameer

Ruparel, Zengmin Li, Nicholas J. Turro, and Jingyue Ju. Four-color dna se

quencing by synthesis on a chip using photocleavable fluorescent nucleotides.

Proceedings of the National Academy of Sciences of the United States of Amer

ica, 102(17):5926-5931, 2005.

[175] Weidong Wu, Brian P. Stupi, Vladislav A. Litosh, Dena Mansouri, Demetra

Farley, Sidney Morris, Sherry Metzker, and Michael L. Metzker. Termina

tion of DNA synthesis by N6-alkylated, not 3'-0-alkylated, photocleavable 2'-

deoxyadenosine triphosphates. Nucl. Acids Res., 35(19):6339-6349, 2007.

[176] Hameer Ruparel, Lanrong Bi, Zengmin Li, Xiaopeng Bai, Dae Hyun Kim,

Nicholas J. Turro, and Jingyue Ju. Design and synthesis of a 3-O-allyl photo

cleavable fluorescent nucleotide as a reversible terminator for DNA sequencing

by synthesis. Proceedings of the National Academy of Sciences of the United

States of America, 102(17):5932-5937, 2005.

[177] Michael L. Metzker, Ramesh Raghavachari, Stephen Richards, Swanee E. Ja-

cutin, Andrew Civitello, Kevin Burgess, and Richard A. Gibbs. Termination of

BIBLIOGRAPHY 196

DNA synthesis by novel 3'-modifieddeoxyribonucleoside 5'-triphosphates. Nucl.

Acids Res., 22(20):4259-4267, 1994.

[178] Neil K. Garg, Carolyn C. Woodroofe, Christopher J. Lacenere, Stephen R.

Quake, and Brian M. Stoltz. A ligand-free solid-supported system for sono-

gashira couplings: applications in nucleoside chemistry. Chemical Communica

tions, (36):4551-4553, 2005.

technologies for high throughput single molecule...

Documents