Developing Single-Molecule Fluorescence Techniques and

Their Application in the DNA Nanotechnology Field

Thesis submitted in partial fulfillment

of the requirements for the degree of

“DOCTOR OF PHILOSOPHY”

by

Roman Tsukanov

Submitted to the Senate of Ben-Gurion

University of the Negev

January 2014

Beer-Sheva

Developing Single-Molecule Fluorescence Techniques and

Their Application in the DNA Nanotechnology Field

Thesis submitted in partial fulfillment

of the requirements for the degree of

“DOCTOR OF PHILOSOPHY”

by

Roman Tsukanov

Submitted to the Senate of Ben-Gurion

University of the Negev

Approved by the advisor______________

Approved by the Dean of the Kreitman School of Advanced Graduate

Studies____________

January 2014

Beer-Sheva

This work was carried out under the supervision of Dr. Eyal Nir

In the Chemistry Department

Faculty of Natural Science

i

Research-Student's Affidavit when Submitting the Doctoral Thesis

for Judgment

I Roman Tsukanov, whose signature appears below, hereby declare that:

√ I have written this Thesis by myself, except for the help and guidance offered by

my Thesis Advisors.

√ The scientific materials included in this Thesis are products of my own research,

culled from the period during which I was a research student.

___ This Thesis incorporates research materials produced in cooperation with others,

excluding the technical help commonly received during experimental work.

Therefore, I am attaching another affidavit stating the contributions made by myself

and the other participants in this research, which has been approved by them and

submitted with their approval.

Date: Student's name: Roman Tsukanov Signature: __________

ii

Acknowledgments

As a first member in Eyal’s research group, I have gained the unique experience of

being a part of establishing single molecule fluorescence lab out of scratch. Eyal’s

never-ending energy, ideas, dedication and aggressive problem solving made it

possible for me to build and optimize the performances of three optical setups, which

are currently serving all the lab members. I also thank Eyal for great project he gave

me and the high level equipment, which ensured the quality of the data. I am grateful

to Eyal for sending me to relevant scientific conferences, where I enjoyed very

much.

I indebted to many people who were interacting with me on this intensive period. I

would like to thank all former and present lab members, who provided me great

support and shared inter-disciplinary knowledge, which contributed largely to my

current expertise. So, Tommy Yaron, Miran, Michael, Noa, Rula, Hagai, Gai, Tapasi

thank you! Special thanks to Tommy Tomov for great ideas, discussions and fun.

I’m grateful to Yaron Berger and Michael Muzika, who were taking data for my

project and especially for night and early morning shifts in the lab. Thanks to Miran,

Rula and Noa for establishing the labeling procedure in our lab and rapid labeling,

once it was needed.

I thank our collaborators Dr. Doron Gerber and Yair Glick from Bar Ilan university

for providing us microfluidics equipment, sharing the knowledge and help in

problem solving. Yaron Berger, for taking an active part in early stages of

establishing the microfluidics technology in our lab and his great technical abilities

which were very helpful on the daily basis.

Prof. Victor Kagalovsky, for his precious advices.

Michael Lubker for help with electronics for the whole period of this work.

Hagai Drori for contribution in the development of Matlab software.

Dmitry and Yulia Matiuhin for the support.

I thank my lovely wife for huge patience, understanding and providing me support

during my research. I also thank her for giving a birth to our child, who became an

origin of inspiration and an energy source for finishing this work.

iii

Last but not least, I would like to express my appreciation and respect to my parents,

who have always encouraged me and were supporting me during all my academic

studies.

iv

Table of Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1. Measuring Biomolecular Dynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2. Single Molecule Fluorescence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.3. Probability Distribution Analysis (PDA) . . . . . . . . . . . . . . . . . . . . . . . . 2

1.4. DNA as a Model Molecule for Single-Molecule

Fluorescence Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . 3

1.5. DNA Hairpin. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.6. DNA Origami. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.7. DNA Hairpin Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.8. Dependency of the Dynamics on NaCl Concentration . . . . . . . . . . . . . . 9

1.9. DNA Nanotechnology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.10. DNA Motors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.11. Non-Autonomous Motors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2. Objectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3. Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1. Immobilization-Based sm-FRET-TIRF. . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1.1. Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1.2. Data Analysis Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1.3. Surface Immobilization of Biomolecules . . . . . . . . . . . . . . . . . . 20

3.2. Diffusion-Based Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.2.1. Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.2.2. Alternating Laser Excitation (ALEX) . . . . . . . . . . . . . . . . . . . . . 24

3.2.3. Data Analysis and Presentation . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.3. Probability Distribution Analysis (PDA) . . . . . . . . . . . . . . . . . . . . . . . . 27

3.3.1. PDA Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.3.2. Rational of the PDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.4. Calculating the Opening Rates using MFOLD and Transition State

Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.5. Sample Preparation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.5.1. Single-Stranded DNA Labeling . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.5.2. Origami Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

v

3.5.3. Annealing Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.5.4. Origami purification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.5.5. DNA Origami Structure Validation . . . . . . . . . . . . . . . . . . . . . . . 33

4. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.1. Comparison of the Fraction of Open State . . . . . . . . . . . . . . . . . . . . . . . 37

4.2. Comparison of Immobilized Hairpin-Only and Hairpin-Origami

Opening and Closing Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.3. Probability Distribution Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.3.1. Experimental E-Histograms and the PDA Fitting . . . . . . . . . . . . . 41

4.3.2. PDA Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.4. DNA Hairpin Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5. Combination of Microfluidics and Single-Molecule Fluorescence

Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

7. Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

8. Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

8.1. Microfluidics Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

8.2. Dwell-time analysis, semi-logarithmic plot . . . . . . . . . . . . . . . . . . . . . . . 54

8.3. PDA Calculation – E-histograms shape . . . . . . . . . . . . . . . . . . . . . . . . . . 55

8.4. Transition State Free Energies for Opening and Closing Reactions . . . . 56

9. Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

10. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

vi

List of Figures

Figure 1. Schematic of the influence of the shot-noise and the dynamics on the

shape of E-histogram. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Figure 2. Quasi-two-state behavior of DNA hairpin. . . . . . . . . . . . . . . . . . . . . . . 4

Figure 3. Unfolding reaction rates of DNA hairpins with different number of

base pairs in the stem. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . 5

Figure 4. DNA Origami. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Figure 5. Design of the DNA hairpins constructs studied in this work. . . . . . . . 9

Figure 6. Influence of ionic strength on DNA hairpin dynamics. . . . . . . . . . . . . 10

Figure 7. Non-autonomous motor operation and design. . . . . . . . . . . . . . . . . . . 11

Figure 8. A compression of the operational yield of motors operating using

fuels and hairpin-fuels measured after each step. . . . . . . . . . . . . . . . . . 12

Figure 9. Principle of the immobilized-based sm-FRET experiment. . . . . . . . . . 16

Figure 10. TIRF image of single DNA hairpin-only molecules A31-bp6-M

Immobilized to surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Figure 11. Pictures of immobilized-based TIRF setup, built as a part of this work. 17

Figure 12. Sample TIRF time trajectories for immobilized hairpin A31-bp6-S for

three different salt concentrations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Figure 13. Specific binding of Hairpin-Origami to coverslip by Biotin-

Avidin chemistry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Figure 14. Principle of diffusion-based setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Figure 15. Picture of diffusion-based optical setup, built as a part of this work. . . 24

Figure 16. Sorting capabilities of smALEX-FRET spectroscopy. . . . . . . . . . . . . . 25

Figure 17. DNA Origami structure and integrity . . . . . . . . . . . . . . . . . . . . . . . . . . 33

vii

Figure 18. Typical two-dimensional E/S-histogram and E- and S- one-dimensional

histograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Figure 19. Diffusion-based sm-FRET-ALEX measurements of dynamics of

hairpin-only A31-bp6-M and hairpin-origami A31-bp6-M. . . . . . . . . . 35

Figure 20. Immobilization-based sm-FRET-TIRF measurements of

hairpin dynamics. Data are of A31-bp6-M. . . . . . . . . . . . . . . . . . . . . . . 36

Figure 21. Very good agreement between the fraction of open state of

freely diffusing and immobilized hairpin-only and hairpin-origami . . 37

Figure 22. Very good agreement between the closing and opening dwell-

time histograms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Figure 23. Opening and closing dwell-time histograms. . . . . . . . . . . . . . . . . . . . . 39

Figure 24. Summary for immobilized-based transition rates. . . . . . . . . . . . . . . . . 40

Figure 25. PDA fit to hairpin-origami A31-bp6-FF E-histogram. . . . . . . . . . . . . . 41

Figure 26. E-histograms of the hairpin-only and hairpin-origami constructs

studied in this work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

Figure 27. Agreement between rates obtained using the diffusion-based PDA

method and the immobilization-based TIRF. . . . . . . . . . . . . . . . . . . . . . 43

Figure 28. Validation of PDA for hairpin A31-bp6-F free and attached to

origami in diffusion-based method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Figure 29. Opening rates and closing rates for series of four hairpins. . . . . . . . . . . 46

Figure 30. Picture of the microfluidics device positioned on the single-

molecule TIRF setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Figure 31. Efficiency and kinetics of DNA motors immobilized to surface

inside microfluidic device. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

viii

Figure 32. The range of rates that could be measured with single

molecule fluorescence techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Figure 33. Microfluidic chip scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Figure 34. Opening and closing dwell-time histograms of A31-bp6-M and

A31-bp6-S with fit to exponent function in semi-log plot. . . . . . . . . . . 54

Figure 35. The shape of E-histograms can be predicted by PDA calculation . . . . 55

Figure 36. Energy for opening and closing for the four hairpins . . . . . . . . . . . . . . 56

List of Tables

Table 1: Hairpins names and sequences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

ix

Abstract

The structural dynamics of large biomolecules such as DNA, RNA, and proteins can

be studied in great detail using single-molecule fluorescence techniques. Because of

the complexity of the experiments and the data analysis, however, various

experimental aspects are yet to be examined to validate the technique.

In my research I compared diffusion-based and immobilized-based single-molecule

fluorescence techniques by measuring a series of DNA hairpin molecules, different

only by their stem sequences. These DNA structures presumably have relatively

simple dynamics, making them a favored model system. The opening and closing

rates of hairpins immobilized to coverslip surfaces and hairpins attached to DNA

origami were measured using total internal reflection fluorescence (TIRF) and

compared to rates obtained for freely diffusing hairpins and DNA origami-bound

hairpins using the probability distribution analysis (PDA) method. The data from

diffusion- and immobilization-based techniques were consistent for all constructs

and for all NaCl concentrations examined, cross validating the TIRF and PDA

techniques. From the excellent agreement between rates of opening and closing of

free hairpins and of hairpins bound to origami, we concluded that the origami has no

influence on the hairpin dynamics and that the PDA method correctly separates the

diffusion from the dynamic component. The experimental opening rates were in

excellent agreements with rates predicted using MFOLD and transition state theory

for melting of duplexes with sequence identical to that of the stems, leading to the

conclusion that the hairpin unfolding mechanism resembles that of duplex melting.

The closing rates were identical for all hairpins suggesting that the folding reaction

depends on counterion concentration and not on stem sequence. Thus, the hairpin

loop influences the folding reaction and the stem influences the unfolding reaction.

Finally, I present a single-molecule fluorescence application in DNA

nanotechnology. The immobilization-based technique was combined with

microfluidics technology to operate a DNA bipedal walker that strides on a 100-

nanometer-sized DNA origami track with unprecedented efficiency and speed.

x

Keywords

Single-molecule fluorescence, FRET-ALEX Spectroscopy, DNA Dynamics, DNA

Hairpin, DNA Origami, DNA motors, TIRF, Microfluidics.

1

1. Introduction

1.1. Measuring Biomolecular Dynamics

The functions of biomolecules are strongly dependent on their molecular structure

and the dynamics with which these structures fold and unfold. Molecular structures

are often studied in high resolution using transmission electron microscopy (TEM),

X-ray crystallography, NMR, and atomic force microscopy (AFM). Acquiring

dynamic information, however, is complex. Because a solution ensemble of

biomolecules are often out of phase in respect to each other, to move beyond the

ensemble average and to obtain reliable dynamic information it is often necessary to

use the single-molecule fluorescence approach. As the physics Nobel laureate

Richard Feynman famously said, “It is very easy to answer many of these

fundamental biological questions; you just look at the thing!”, and single molecule

fluorescence techniques enable exactly that.

1.2. Single-Molecule Fluorescence

Single-molecule Förster resonance energy transfer (sm-FRET) is a powerful

technique for measuring real-time conformational dynamics of biomolecules. The

immobilization-based total internal reflection (TIRF) technique, in which each

molecule is continually observed, provides direct structural dynamics information on

the transitions between molecular states, and as a result, most sm-FRET

measurements of transition rates to the date have been conducted using this

technique. Immobilizing biomolecules, however, adds experimental complexity, and

the immobilization procedure and proximity to surface potentially influence the

dynamics. Previous pioneering studies of immobilized DNA hairpins yielded

somewhat ambiguous data due to the fact that only 2-5% of the hairpins were active

(exhibiting dynamics), suggesting that surface immobilization influenced the

dynamics.[1] Furthermore, the immobilized-based TIRF technique resolution is

limited by the EMCCD camera frame rate. Single-photon detectors can increase the

resolution; however, in this approach data accusation is tedious and slow as it

requires scanning the surface and acquiring data for one molecule at a time. As a

result, very few surfaced-immobilized systems with fast dynamics have been studied

thus far [2, 3].

2

1.3. Probability Distribution Analysis (PDA)

In diffusion-based approaches, the limitations due to immobilization are avoided

and, due to the use of single-photon detectors, the temporal resolution is potentially

higher. To obtain the dynamics from the shape of the resultant FRET efficiency

histogram (E-histogram) a semi-empirical probability distribution analysis (PDA)

method has been developed. The method includes statistical descriptions of the shot-

noise contribution [4-6], and the dynamics are incorporated by calculating the

expected distribution of mean E values of any assumed dynamics scenario [5, 7-9].

The influence of dynamics and shot-noise on the shape of the E-histogram is

illustrated in Figure 1. For a molecule with a single and fixed E value (Figure 1A),

the E-histogram has a single peak with a shape that can be described by a binomial

function that is dependent on the averaged E value and on the burst size distribution

[4, 5]. Bursts containing larger numbers of photons yield narrower histograms than

those with fewer photons. A molecule that interconverts between states during the

burst yields an E value that is the average of the E values of the states, weighted by

the fraction of time the molecule spends in each of the states during the burst [5, 7,

8]. The probability of detecting such events depends on the transition rates and on

the burst duration. A molecule that interconverts between E = 0.2 and E = 0.8 states

at a rate significantly slower than the time the molecule spends in the confocal spot

(burst duration) has low probability of interconverting during the burst, and, as a

result, the E-histogram shows two distinct peaks, each dominated by shot-noise

(Figure 1B).

Only minor intermediate E values, collectively called ‘bridge’ and corresponding to

molecules that undergo transitions, are observed. A molecule that interconverts

between the states at rates that are comparable to the diffusion time (Figure 1C)

yields the two peaks and a larger bridge. Even more frequent transitions result in the

formation of a single peak with an averaged E value (Figure 1D).

3

Figure 1. Schematic of the influence of the shot-noise and the dynamics on the shape of E-histogram.

In addition, PDA method and the relationship between the dynamics and the E-

histogram was reviewed elsewhere [10]. An alternative approach based on maximum

likelihood analysis has also been described [11].

The PDA method has been used experimentally to analyze the dynamics of DNA

hairpins [5, 7, 9], of LacY protein [12], and of DNA polymerase-I [13]. PDA has

been validated theoretically by demonstrating that the transition rates of numerically

simulated two-state system can be obtained with high accuracy [5-7]. Because the

shape of the E-histograms may be influenced by experimental artifacts, such as

population mixing, photobleaching, and misalignment of the donor and acceptor

detection volumes, experimental validation of the method is essential to demonstrate

reliability under experimental conditions. We will show here the first experimental

validation of the PDA method.

1.4. DNA as a Model Molecule for Single-Molecule

Fluorescence Studies

In principle, DNA, RNA, proteins, beads, and any large fluorescent complex can be

used to calibrate single-molecule fluorescence techniques. We chose to use DNA

hairpins because (1) these molecules can be synthesized in high yield and purity, (2)

DNA oligonucleotide synthesis is affordable, (3) DNA oligonucleotides are

chemically stable, ensuring the robustness and repeatability of the experiments, (4)

4

hairpin folding and unfolding is presumably a two-state system, (5) labeling

oligonucleotides with fluorescent dyes is a straightforward procedure, and, most

importantly, (6) modifications can be introduced in a design manner to control the

dynamics and stability (energy landscape) of the hairpin as we show in this work. A

wide range of interconversion rates were achieved by alterations of the stem

sequence and the NaCl concentration. The hairpin quasi-two-state dynamics proved

very valuable for the development and characterization of the single-molecule

fluorescence techniques.

1.5. The DNA Hairpin

The hairpin is a nucleic acid secondary structure motif, formed by the hybridization

of complementary segments; these structures are frequently observed in DNA and

RNA. DNA hairpin structures are involved in many biological processes, such as the

regulation of gene expression and DNA recombination and may facilitate mutagenic

events [14-16]. Hairpin structures are not static. In a simplified description, the

conformational ensemble can be divided into two dominant states, the open state and

the closed state (quasi-two-state behavior). The closed state is characterized by a low

enthalpy, due to pairing of complementary bases to form the stem. The open state is

stabilized by high entropy, due to the large number of configurations available by the

single-stranded chain [17-20].

Figure 2. Quasi-two-state behavior of DNA hairpin. The hairpin fluctuates between open and

closed states. The open state is characterized by high entropy at a local minimum of free energy; the

closed state is characterized by low enthalpy, also yielding a local minimum of free energy.

Conversion between the states requires passing through an energy barrier involving high free energy

transition states.

5

Hairpin dynamics is of relevance to several scientific fields. In DNA

nanotechnology, hairpins are utilized for powering DNA-based artificial motors and

for controlling reaction rates and hierarchies [17-23] and were used as biosensors

[24, 25]. Because of the presumably uncomplicated structure and dynamics, hairpins

are often used as model systems in the development of experimental [1, 5, 9, 26-28]

and computational techniques [29-31]. Hairpin dynamics have been studied by

fluorescence correlation spectroscopy (FCS) [26, 32-36], temperature-jump (T-jump)

and stopped-flow (ion-jump) [37, 38], single-molecule optical trapping [28, 39], and

sm-FRET techniques [1, 5, 9, 27, 40, 41].

Despite the extensive study, important aspects of hairpin dynamics remain

unresolved [33, 38, 42]. Experimental uncertainties could result from different time

scales available for the experimental techniques applied, leading to discrepancies in

the assignment and interpretation of the obtained rates [33, 36, 38] (Figure 3).

Figure 3. Unfolding reaction rates of DNA hairpins with different number of base pairs in the stem.

Various techniques used for the measurements are categorized by FCS (filled circle), T-jump (open

circle), single molecule force measurement (filled square), and smFRET (open square). Large

uncertainties arise, especially for stem with six base pairs. Reproduced from Van Orden et al. [33]

In early FCS studies the results were explained using a two-state model, and the

obtained rates were assumed to reflect transitions between open (unfolded) and

6

closed (the presumably fully zipped, folded) states [26, 32]. Recent dual-beam FCCS

measurements [33, 34], FCS measurements of diffusion-decelerated hairpins [36],

and stopped-flow (ion-jump) measurements [42] led to questioning of this

interpretation. These data suggest that the fast relaxation rates (< 1 ms) observed in

FCS measurements are transitions into and out of one or more metastable

intermediates rather than between fully open and fully closed states. In addition, the

different relaxation times observed after T-jump and ion-jump perturbations may be

the result of populating distinctly different conformational ensembles, a scenario that

rules out a simple two-state model and indicates that the folding energy landscape is

rather rugged. The FCS technique is sensitive to many time scales, but the obtained

rates cannot be straight forwardly associated with a particular transition. Ion-jump

and T-jump techniques are excellent for measuring kinetics from a defined state;

however, the change in conditions introduces additional complications, and the data

may not directly reflect dynamics in equilibrium. Here we will use direct single-

molecule measurements to unambiguously determine the hairpin opening and

closing rates.

1.6. DNA Origami

A beautiful new technology called DNA-origami, recently developed by Rothemund

[43], offers exciting new possibilities for assembling complex molecular structures.

Here, a long bacterial ss-DNA called a scaffold is annealed with up to several

hundred short synthetic ss-DNA called staples. The staple sequences are designed to

bind to nonconsecutive complimentary scaffolds’ sequences, forcing the scaffold to

fold into 2D or 3D designed structures. Using available computer software,

researchers can easily program desired shapes by annealing staples and scaffold to

form structures of up to several hundred nanometers (Figure 4).

7

Figure 4. DNA Origami. Top row diagrams showing the bend of helices at crossovers. Color

indicates the base-pair index along the folding path. Lower rows, AFM images. All images and panels

without scale bars are the same size, 165 nm times 165 nm. Scale bars for lower AFM images: b, 1

µm; c–f, 100 nm. Reproduced from [43].

The highly specific and designable origami structure can serve as a template for

assembling molecules besides DNA. Elongated staples that branch out from the

DNA-origami can bind to a complementary DNA sequences that are pre-attached to

various molecules. The chemistry of attaching DNA to nanotubes, gold

nanoparticles, proteins, quantum dots, and many other molecules is well known,

therefore, it is possible to use DNA origami to organize these molecules in space in a

very precise manner [44].

In this work we use a rectangular shape DNA origami for the following purposes: (1)

as a platform for biocompatible investigation, (2) to slow the diffusing through the

confocal spot and by that making slow rates available for PDA, (3) to examine the

influence of the coverslip on the hairpin dynamics by comparing results of hairpin

directly immobilized to the surface to that attached to origami which is immobilized

on the coverslip and, (5) to operate a DNA walker on a DNA.

1.7. DNA Hairpin Design

The DNA hairpins were designed with poly(A) loops of 31 nucleotides (A31) and

six base-pair stems; the stem sequences differed in the numbers of G-C and A-T

pairs. The hairpin names and sequences are given in Table 1.

8

Name Hairpin Sequence # of G·C

pair

A31-bp6-S 5’-TCGCCG-(A)31-CGGCGA-3’ 5

A31-bp6-M 5’-TCGCCT-(A)31-AGGCGA-3’ 4

A31-bp6-F 5’-TGGGTT-(A)31-AACCCA-3’ 3

A31-bp6-FF 5’-TGGATT-(A)31-AATCCA-3’ 2

Table 1. Hairpin names and sequences.

The hairpins were examined when anchored to origami (hairpin-origami) or not

anchored to origami (hairpin-only). Previous diffusion-based sm-FRET and FCS

studies found that hairpin dynamics was influenced by the presence of a

complementary strand located adjacent to the stem [45] and by the fluorophores due

to, among other reasons, dye-dye interactions.[41] Our hairpins were designed to

minimize such possible interfering interactions. Two T bases (Figure 5B; spacer)

separated the stem section from the dsDNA linker [9], and the donor-acceptor

interactions were minimized by placing the donor and the acceptor such that when

the hairpin was closed, the donor-acceptor distance remained significant (larger than

10 base pairs, Figure 5B) [5]. A 35-bp long, double-stranded DNA (dsDNA-linker,

Figure 5B) was used to distance the hairpin section from the coverslip or the origami

surfaces. To enable binding of the hairpin-only construct to the avidin-coated

coverslip, the 5′ end of the bottom strand of the dsDNA linker was extended with

TTTT-biotin. The hairpin was anchored to the origami by extending the 5′ end of the

bottom strand of the linker with TT attached to origami staple. To anchor the origami

to the avidin-coated coverslip, four biotinylated origami staples were introduced in

the annealing process. The hairpins were labeled with donor and acceptor

fluorophores (ATTO-550 and ATTO-647N, respectively) positioned such that the

hairpin open state yielded low FRET values (E) and the closed state yielded high E

values. To enable differentiation between the hairpin-origami and residual free

hairpins in the diffusion-based experiments by means of the ALEX technique, an

additional strand labeled with ATTO-647N was anchored to the origami sufficiently

far from the hairpin to prevent energy transfer (Figure 5C).

9

Figure 5. Design of the DNA hairpins constructs studied in this work. The hairpin sequences and

labeling positions were designed to minimize possible interfering interactions between the

fluorophores when the hairpins are closed, and the TT spacer was introduced to minimize interactions

between the stem and the duplex. The origami was labeled with an additional acceptor to enabled

separation of hairpin-only events from hairpin-origami using the ALEX technique [46].

1.8. Dependency of the Dynamics on NaCl Concentration

The phosphate groups in DNA and RNA are negatively charged resulting in

intramolecular repulsion that destabilizes compact folded conformers and stabilizes

the unfolded structurally extended state (Figure 6). In the presence of positively

charged counterions this repulsion is reduced due to screening [47-49]. Thus, the

fold state dominates in high NaCl concentrations whereas the unfolded state is

dominant at low salt concentrations [26, 38, 45, 46].

(C)

10

Figure 6. Influence of ionic strength on DNA hairpin dynamics.

1.9. DNA Nanotechnology

The DNA nanotechnology field picked up momentum in the early 1980s with the

recognition that DNA is not just a one-dimensional molecule. By careful design of

synthetic sequences, branched DNA motifs can be formed. Using their single

stranded ends, these motifs can be further organized to create 3D structures that are

“limited only by imagination and a few physical properties”, to quote Seeman [50].

Indeed, in the following years, researchers constructed a variety of DNA structures

[51-53], and, most relevant to this, DNA-origami [43]. Recent advances in DNA

nanotechnology has led researchers to suggest that this technology could be

developed, and harnessed, for the benefits of other fields, including the study of

biomolecules [54, 55].

1.10. DNA Motors

Motivated by the success in constructing DNA dynamical structures and inspired by

biological motors in nature, researchers began exploring the possibility of creating

artificial DNA-based motors. Several autonomous motors, which achieve

unidirectionality [17, 18] by damaging or changing the track, were published.

Turberfield and his coworkers [56] published a bipedal walker that, at least in

principle, is autonomous, processive, and bidirectional (direction can be switched by

changing the fuel sequence). In our work, we chose to sacrifice autonomy in favor of

controllability, processivity, and directionality.

11

1.11. Non-Autonomous Motors

Most relevant to our study, Shin and Pierce [57] (Figure 7, our design) developed a

processive bipedal DNA walker that traveled hand-over-hand by moving its rear feet

to the front for each step. Two legs, each with a different sequence, walk on a track

containing four different footholds. Each step requires the sequential addition of two

instruction strands: the first lifts the back feet from the track and the second

reconnects the feet just lifted from the rear to the front of the walker. Such a design

requires alternating additions of fuel and anti-fuel for each step, and a total of eight

additions are needed to complete a cycle, altogether, enabled increase external

controllability. By giving up autonomy and externally dictating the sequence of fuel

and anti-fuel and their incubation periods at each step, the authors obtain control

over travel direction and gain increased processivity while keeping the walker and

the track chemically unchanged. Our group further developed this motor by

incorporating the track into origami (Figure 7).

Figure 7. Non-autonomous motor operation and design. (A1-6) Fuels attaches foothold to walker leg,

and anti-fuel detaches leg from foothold. Sequential addition of fuels and anti-fuels results in walker

striding along the track. (B) Top view of the origami track.

Non-autonomous motor operational yield (defined as the fraction of devices that

operate as intended) can be further improved by rational design of asymmetrical

hairpin-fuels that by regulating the reaction hierarchy avoid consecutive binding.

The best yield which have been achieved in our lab in a diffusion-based experiment

is 4% per reaction, see Error! Reference source not found. (reproduced from

Tomov and Tsukanov, 2013 [58]).

12

Figure 8. A compression of the operational yield of motors operating using fuels and hairpin-fuels

measured after each step [58].

The basic limitation in a diffusion-based experiment follows from the fact that fuels

and anti-fuels accumulate in the solution and interact with the motors to inhibit

further reaction. Immobilizing the motors to the coverslip surface inside a

microfluidics device should enable removal of the excess fuels and anti-fuels.

Combining the single-molecule technique with the microfluidics technology is a

significant challenge but, once enabled, will contribute to the development of new

generation of highly efficient and fast DNA motors that will be able to perform many

tasks.

13

2. Objectives

General Aim

Our first aim was to develop a single-molecule fluorescence toolkit capable of

accurate acquisition of structural dynamic information of biomolecules and, more

specifically, to examine and validate the diffusion-based and the immobilized-based

methods. This was conducted by using a series of four DNA hairpins as a model

system and a rectangular DNA origami as a platform to which the hairpin were

attached. After validating the spectroscopic methods, the dynamic data obtained for

the four hairpins was analyzed in terms of the hairpin states and transition

mechanisms. Finally, DNA-made molecular motors, developed by me in

collaboration with other members of our group, were created as an example of usage

of single-molecule fluorescence methods in the field nanotechnology.

Aim 1: Development and validation of single-molecule fluorescence techniques

The dynamics of a series of DNA hairpin-only and hairpin-origami, differing only in

their stem sequences, were measured using the diffusion-based and immobilized-

based techniques. The opening and closing rates and the fraction of open state of

hairpin-only and hairpin-origami were measured using the two techniques. The

following experimental aspects were examined:

By comparing the dynamics of hairpin immobilized to coverslip to the dynamics

of hairpin anchored to origami, which is in turn immobilized to coverslip, we

were able to determine the influence of the coverslip on the hairpin dynamics.

By comparing the dynamics of free hairpin to that of hairpin anchored to

origami we examined the influence of origami on the dynamics of the nearby

hairpins.

The diffusion-based PDA method was verified by comparison of the opening

and closing rates of freely diffusing hairpins to those of immobilized hairpins.

To further verify the PDA method and to demonstrate that the method can

correctly separate the dynamics component from the diffusion component, we

compared the opening and closing rates of hairpin-only and hairpin-origami; the

14

latter diffuses more slowly. Using the origami we extended the time window

available to PDA to slower dynamics.

The validity of the immobilized-based technique was examined by analysis of

the single exponential of the dwell-time histograms.

Aim 2: Detailed study of DNA hairpin dynamics

The second aim was to study the dynamics of the two-state hairpin model system.

The hairpins structures, the presence/absence of intermediate states, the opening and

closing rates, the opening and closing energies, and the folding and unfolding

mechanisms were examined. More specifically:

The presence/absence of intermediate states was examined using the fast

snapshot ability of the diffusion-based method.

The hairpin opening and closing reaction orders were examined by analysis of

the shape (single-, double-exponential decay) of the dwell-time histograms

obtained from intensity time trajectories in immobilization-based method.

The influence of different stem sequences (i.e., different thermodynamic

stabilities) on the dynamics were examined both in diffusion- and

immobilization-based techniques.

The influence of the buffer ionic strength on the opening and closing rates was

determined by measuring the dynamics under different NaCl concentrations.

The unfolding mechanism was investigated by comparing the opening rates to

rates predicted from MFOLD and transition state theory.

Compression of the data obtained by the two single-molecule fluorescence

techniques, with and without the origami, and for four hairpins different only

in their stem sequence, cross validated the techniques, the use of origami, and

our conclusions regarding hairpin structural dynamics.

Aim 3: Combination of immobilization-based technique with microfluidics

technology for the development of DNA-made motors possessing high

operational yield and speed

15

We demonstrate the application of single-molecule fluorescence in DNA

nanotechnology. Immobilized-based experiments were combined with microfluidics

technology to operate a DNA motor on a DNA origami track. The motor was

immobilized inside the microfluidic channel. This novel approach enabled efficient

fuel and anti-fuel exchange, significantly improving motor operational yield and

speed relative to previously described designs.

3. Methods

3.1. Immobilization-Based sm-FRET-TIRF

3.1.1. Experimental Setup

The sm-FRET-TIRF experiments were carried out on an in-house built optical setup.

In brief, a green CW laser beam (532 nm, MLL-FN-532, Changchun New Industries

Optoelectronics Tech. Co., Ltd.) was aligned into a single-mode fiber. After the

fiber, the beam was collimated, expanded by factor of 4.16X and then focused

(achromatic lens 180 mm, Thorlabs AC508-180-A) on the back focal-plane of a high

numerical aperture oil objective (NA 1.45, 60×, Olympus America, Melville, NY)

mounted on a commercial inverted microscope (IX71, Olympus America, Melville,

NY). The excitation laser intensity was tuned to meet two requirements, according to

the experimental conditions. First, it has to be strong enough to ensure sufficient

signal-to-noise ratio and weak enough to allow molecules of study to contain several

transitions in time trajectory. We found that power of 30-100 mW (depends on the

hairpin rates) allows identification of several transitions before photo-bleaching

(time trajectories of 4-12 seconds, camera frame rate of 5-15 msec, respectively).

The emitted fluorescence was separated from the excitation light by a dichroic mirror

(ZT532/638RPC, Chroma), split based on their wavelength (donor and acceptor) by

a second dichroic mirror (FF650-Di01, Semrock), passed through a filter (band-pass

filter, FF01-580/60, Semrock, for the donor channel and a long-pass filter BLP01-

635R, Semrock for the acceptor channel), and focused into a fast CCD camera

(IXON DU-897E, Andor), donor channel on the left and the acceptor channel on the

right.

16

Figure 9. Principle of the immobilized-based sm-FRET experiment. (A) Total Internal Reflection

Fluorescence (TIRF) setup: The excitation lasers are focused on the side of the back focal-plane of a

high numerical aperture objective, creating an evanescent field of ~100 nm, thereby reducing the

background florescence. A low concentration of fluorescently labeled molecules is immobilized on a

coverslip surface via biotin-avidin chemistry, and a flow cell (or a microfluidic chip) allows

exchanging buffer during the experiment. (B) Total internal reflection optical path for objective-type

TIRF: the beam enters on the side of the objective. (C) After having been split to donor and acceptor

channels, the emitted photons are imaged on a fast CCD camera and recorded as a function of time.

(D) Time-traces of each individual molecule are analyzed by means of FRET.

17

Figure 10.TIRF image of single DNA hairpin-only molecules A31-bp6-M immobilized to surface.

On the left - green channel, on the right - red channel.

Figure 11. Pictures of immobilized-based TIRF setup, built as a part of this work. (A) Excitation

path, microscope and microfluidic chip. (B1) Emission path, green filter and slit, (B2) imaging lens

and dichroic mirror, separating based on wavelength to green and red channels.

(A)

(B1) (B2)

((A) ((B)

18

3.1.2. Data Analysis Procedures

After acquiring the movies, data processing has been performed with in-house built

Matlab software (MathWorks, Natick, Massachusetts).

3.1.2.1. Overlapping the Donor and Acceptor Images

For each movie, several donors and the corresponding acceptors spots were manually

selected. To each spot, a two-dimensional Gaussian was automatically fitted,

yielding the X and Y positions of the donor and the acceptor spots. Based on these

sets of positions, a non-linear polynomial transformation was applied to overlap the

donor and acceptor images, compensating for optical aberrations and imperfect

alignment of the optical setup.

3.1.2.2. Generating Donor and Acceptor Time Intensity Trajectories

For each movie, 15-50 of the brightest pixels in the acceptor or donor channels were

manually selected followed by automatic selection of the corresponding spots in the

donor or acceptor channel. The intensity of each of the selected pixels was summed

with the intensity of the 8 surrounding pixels (altogether, a 3×3 box, centered on the

brightest pixel). The per-pixel averaged background was estimated by averaging the

intensity of the 16 surrounding pixels (around the 3×3 box).The background was

subtracted from the intensity (after multiplication by 9 to reflect the 3×3 pixels

contribution to the signal). This operation was performed for the donor and the

acceptor channels, for all the selected spots and for all the movie frames, generating

15-50 donor and acceptor background-corrected intensity trajectories for each movie.

E time trajectories were calculated by dividing the intensity in the acceptor trajectory

into the sum of intensities in the acceptor and donor time trajectories (as in a

conventional calculation of E).

19

Figure 12. Sample TIRF time trajectories for immobilized hairpin A31-bp6-S for three different salt

concentrations: (A) low salt, 10 mM NaCl; (B) medium salt, 25 mM NaCl; (C) high salt

concentration, 70 mM NaCl. Upper plots show the intensities in green and red channels, lower plot

(violet) is FRET trajectory. Time bin is 30 msec.

3.1.2.3. Selecting the Best Trajectories

First, the end-time of each trajectory was determined as the time at which the sum of

the donor and acceptor intensities (for any given time bin) falls under a certain

threshold (well above the background noise). If all the time bins were above the

threshold, the trajectory’s duration was the same as that of the movie. Second, the

averaged intensity per bin (sum of donor and acceptor channels divided into the

number of bins) was calculated. Finally, to ensure data quality, only time trajectories

with average intensity and duration above a certain thresholds were further

considered. The data were then analyzed in two ways.

20

3.1.2.4. Generating E Histograms and Calculating the Fraction of Open State

E time trajectories from more than 100 individual molecules (several movies) were

projected and accumulated to generate each E histogram (bin = 0.01). Two

prominent histogram peaks are observed; the open state peak with E < 0.35 and the

closed state peak with E > 0.7. The fraction of open state was calculated by dividing

the size of the open state peak into the sum of sizes of the closed and open state

peaks.

Generating Open and Closed States Dwell-Time Histograms and Calculating

Opening and Closing Times. For each E, time trajectory periods at which E < 0.5

were considered as open state and periods at which E > 0.5 were considered as

closed state. Open and closed states dwell-time histograms were generated from

more than 100 E time trajectories (accumulated from several movies). These

histograms were fitted to single- or double-exponential functions, from which,

opening and closing time constants (and, respectively) were calculated. To prevent

bias caused by photo-bleaching, for each time trajectory, the last time segment in

each trajectory (open or closed states) was ignored.

3.1.3. Surface Immobilization of Biomolecules

To regulate the immobilization process, good control over the solutions volumes,

incubation periods, and solutions flow rates has to be achieved. It was done using a

flow channel (Ibidi sticky Slide VI, Martinsried, Germany). The lower coverslip (the

one which faces the objective) was pre-treated with HF (sonication 1 min) and then

washed thoroughly with distilled water. The immobilization process included

following steps: (i) introduction of 60 µL (1 mg/mL) of biotin-coated BSA (BSA-

biotin A8549 Sigma-Aldrich) into the channel, incubation for 5 min, and thorough

washing with 0.5 mL T50 (Tris 10 mM, 50 mM NaCl); (ii) introduction of 60 µL

(0.2 mg/mL) of NeutrAvidin (ImmunoPure NeutrAvidin Protein, Pierce, Rockford,

USA) followed by thorough washing with 0.5 mL T50; (iii) gentle and slow

injection of 60 µL of biotinylated hairpin-only or hairpin-origami in concentration ~

5-10 pM in imaging buffer (Tris 10 mM, EDTA 1 mM, 2-3 mM Trolox and different

concentrations of NaCl) and then incubation for 5-15 minutes; and finally (vi)

thorough and gentle washing with the measurement solution (0.5 mL) to remove the

21

unbound molecules from the channel. The immobilization was performed in room

temperature of 22-230 C.

3.1.3.1. Specificity Validation of Biomolecules Immobilization

To test the specificity of the biotin-avidin immobilization procedure, several

validation experiments were conducted. The surface was treated with biotin avidin in

all the experiments. First, high concentration (~50pM) of hairpin-only molecules

lacking biotin were injected into the flow channel. After incubation of 15 minutes the

molecules were washed with the buffer solution, and then no immobilized molecules

were detected (in both channels). In next experiment (conducted in the same flow

channel), hairpin-only molecules with biotin modification at 5' of the bottom

sequence were injected at the concentration of ~10 pM. Tens of molecules were

attached to the surface within tens of seconds (data not shown). Extensive wash did

not influence the number of molecules attached to surface. Similar check was

conducted with biotinylated origami-hairpin. First, origami-hairpin lacking

biotinylated staples was very slowly injected into flow channel. No immobilized

origami molecules were detected (Figure 13A1-2). Then, after slow injection of 10

pM hairpin-origami solution and incubation for 5 min tens of immobilized molecules

seen in both channels (Figure 13B1-2).

Figure 13. Specific binding of Hairpin-Origami to coverslip by Biotin-Avidin chemistry. Less than

1% of the hairpin-origami without biotin attached to surface (non-specifically, A1-2), and even less

than 1% for hairpin-only (data not shown). Hairpin-Origami with biotin attached to surface

specifically after incubation of 5-10 minutes (B1-2).

(A2) (B1) (B2) (A1) (A1)

22

Additional validation procedure proved then if the BSA-biotin or avidin layer

missing in the surface then both biotinylated and non-biotinylated molecules did not

attach to the surface. Then, only proper biotin-avidin immobilization procedure can

be used to attach molecules to surface in a specific manner.

It is important to mention, that while immobilizing the biotinylated origami-hairpin

complex specifically, slow injection (few microliters in 10 min) is crucial. Once the

non-biotinylated hairpin-origami molecules were injected in fast fashion, numerous

molecules were attached to surface, due to non-laminar flow in the flow channel

(Data not shown).

3.2. Diffusion-Based Technique

3.2.1. Experimental Setup

The sm-FRET-ALEX experiments were carried out on an in-house built optical

setup. In brief, a green CW laser beam (532 nm, CL532-025-L, Crystal Laser) was

aligned/misaligned into a single-mode fiber using an acousto-optic modulator

(AOM; R23080-2-LTD, Neos Technologies, Melbourne, FL), alternating with a red

diode laser (640 nm, Cube 640-40C, Coherent Europe, Utrecht, NL) that was

electronically switched on/off. The AOM and the red laser were computer controlled

with a 12.4-µs on-time, a 12.6-µs off-time, and a phase shift of 12.5 µs. The laser

intensity rise and fall times were less than 50 ns, and there was no overlap time

between the lasers. The laser beams were combined by a dichroic mirror (Z532RDC,

Chroma) and coupled into a single-mode fiber (P1-460A-FC-2, Thorlabs). The laser

intensities were tuned such that the doubly labeled species would yield S = ~ 0.5 (90

µW for the green laser and 70 µW for the red laser, measured after the fiber while

alternating). After collimation (objective PLCN10×/0.25, Olympus), the combined

green and red beams were introduced into a commercial inverted microscope (IX71,

Olympus America, Melville, NY) and focused about 70 µm inside the sample

solution with a water-immersion objective (NA 1.2, 60X, Olympus America,

Melville, NY). The emitted fluorescence was separated from the excitation light by a

dichroic mirror (ZT532/638RPC, Chroma), focused into a 100-µm pinhole (P100S,

Thorlabs), re-collimated, split by a second dichroic mirror (FF650-Di01, Semrock),

passed through filters (band-pass filter, FF01-580/60, Semrock, for the donor

23

channel and a long-pass filter BLP01-635R, Semrock for the acceptor channel), and

focused into two single-photon avalanche photodiodes (SPAD; SPCM-AQRH-13,

Perkin-Elmer Optoelectronics, Fremont, CA). The TTL signals of the two SPADs

were recorded as a function of time by a 12.5-ns resolution counting board (PCI-

6602, National Instruments, Austin, TX) and in-house prepared LabView acquisition

software.

Figure 14. Principle of diffusion-based setup. (A) Set up: An alternating donor-excitation laser (Dex)

and an acceptor-excitation laser (Aex) are focused via the objective to create a diffraction limited spot.

Picomolar concentrations of fluorescently labeled samples are freely diffusing into and out of the

confocal spot. The donor dye (D) absorbs a photon that originated from the Dex laser and either emits

a photon or transfers the energy to the acceptor dye (A), which, in turn, emits a photon. Alternatively,

the acceptor dye directly absorbs a photon that originated from the Aex laser and then emits a photon.

The emitted photons are collected by the objective, split based on their wavelengths, and detected by a

single photon detector (APD). (B) Binned photon time trajectories: Separate bursts of photons, each

corresponding to a single molecule, are detected. (C1-2) Schematic representation of E and S

histograms of molecules having different fluorescent dyes stoichiometry and E values.

24

Figure 15. Picture of diffusion-based optical setup, built as a part of this work.

3.2.2. Alternating Laser Excitation (ALEX)

The sm-ALEX reports on the stoichiometry of fluorophores presence in a given

molecular system. By detecting the presence/absence of several labeled components,

ALEX method enables excellent monitoring of the structural integrity of a complex.

By labeling parts of a complex with different fluorophores, the method enables

monitoring complex assembly and disassembly reactions [59]. In addition, ALEX

enables sorting population of interests from a mixture containing other populations.

We show ALEX sorting capabilities on experimental data for hairpin-origami

measurement. Selecting only the correct ALEX stoichiometry ratio (2 acceptors and

1 donor in the case of hairpin-origami) and rejecting events with different

stoichiometry ratios help filtering the data from contribution of various unwanted

species that were formed in the annealing and filtration procedures, Figure 16.

25

Figure 16. Sorting capabilities of smALEX-FRET spectroscopy. Real experimental data for hairpin-

origami A31-bp6-M. (A1-6) Different species present in origami-hairpin sample. (B1) 2D ALEX-

FRET plot. (B2) E-histogram of the populations with correct S-ratio (pink rectangle. (B3) S-

histogram, S = ~ 0.33 corresponds to the complete hairpin-origami complex (2 acceptors and 1

donor).

For A31-bp6-M hairpin-origami the species that were formed are as follows (Figure

16): (A1) S = ~ 0.99 – donor only – origami without hairpin loop and additional

acceptor. (A2) S = ~ 0.5 – one donor one acceptor - origami without hairpin loop but

with additional acceptor (could be also large burst of hairpin-only in an open state).

(A3) S = ~ 0.33 – one donor two acceptors – the correct complex: hairpin-origami

with an additional acceptor (hairpin in an open state). (A4) S = ~ 0.5 - one donor one

acceptor - hairpin-origami complex in a closed state, without additional acceptor

(could be also large burst of hairpin-only in a closed state); (A5) S = ~ 0.33 – one

donor two acceptors – the correct complex: hairpin-origami with an additional

acceptor (hairpin in a closed state); (A6) S = ~ 0.1 – acceptor only – origami missing

hairpin.

26

3.2.3. Data Analysis and Presentation

3.2.3.1. Burst Search

Data analysis was performed with the in-house written LabView software as

described before [23, 27]. The beginnings and endings of bursts were determined by

the all-photons-burst-search (APBS, parameters: L = 200, M = 10, and T = 500 s for

the hairpin-only and L = 2000, M = 100, and T = 2500 s for the hairpin-origami).

For each burst, E and S were calculated according to Eq. 1 and Eq. 2, respectively,

binned (0.01 bin size), and plotted on one- dimensional E and S histograms and on a

two-dimensional E/S histogram. The E-histograms were smoothed with a running

average for visualization purposes.

3.2.3.2. Calculation of E and S Values

Because in ALEX experiments two lasers alternatively excite the donor and the

acceptor dyes, the calculation of E is somewhat different from that in a conventional

single laser experiment (Eq. 1).

𝐸 =

(1)

where D is the number of photons recorded in the donor channel, and A is

the number of photons recorded in the acceptor channel during times in which the

donor laser is on (donor laser “on time”), as commonly defined in ALEX

experiments [5, 23, 27]. No correction was made for donor photons leaking into the

acceptor channel (donor leakage). Stoichiometry, S, is calculated by dividing the sum

of the photons recorded in the donor and the acceptor channels during donor laser

on-time by the sum of the photons recorded in both channels during donor laser and

acceptor laser on-times (Eq. 2).

=

(2)

where D and A are the sums of photons recorded in the donor and the acceptor

channels during donor laser and acceptor laser on-times, respectively.

27

3.2.3.3. Correction of γ-Factor Bias Problem

In our setup and using ATTO-550 and ATTO-647N the product of the detection

efficiency and the quantum yield of the donor are larger than that of the acceptor (the

ratio is known as the γ-factor).[60] This is evident from the fact that a low E

population (open hairpins) exhibits higher S values than those of a high E population

(closed hairpins) when no correction was applied (data not shown). This

phenomenon leads to a higher probability to detect open hairpin events than closed

hairpin events and numerical simulations show that this problem cannot simply be

solved by multiplying by a γ-dependent factor (data not shown). To correct for this

bias, therefore, 23% of the donor photons were stochastically deleted from the data

files (before performing burst search), such that the S values of the low and the high

E populations were identical (see Figure 18 for example). Numerical simulations

show that such photon deletion corrects for the γ-factor bias (data not shown).

3.2.3.4 Generation of E-Histograms and Calculation of the Fraction of Open State

E-histograms were generated from events having the correct S values (0.37 < S <

0.63 for the hairpin-only and 0.25 < S < 0.40 for the hairpin-origami). For the

calculation of the fraction of open state events with 0.1 < E < 0.35 were considered

as open state, and events with 0.7 < E < 0.95 were considered as closed state. The

fraction of open state was calculated by dividing the number of open state events into

the sum of open and closed state events (area under the corresponding E-histograms

peaks). Intermediate E values, corresponding to transitions between the states, were

ignored.

3.3. Probability Distribution Analysis (PDA)

The opening and closing rates (kop and kcl) and the averaged E values of the open and

the closed states (Eop and Ecl) were calculated by fitting the E-histograms using

previously described methods [5] with some minor modifications. A slightly

different PDA method was developed and explained in detail elsewhere [7-9], and

alternative approach, based on maximum likelihood analysis, has also been

developed [3, 11, 61].

28

3.3.1. PDA Algorithm

The PDA algorithm calculates a semi-empirical E-histogram and fits it to the

experimental E-histogram by optimizing the dynamic parameters. The following

procedure is used:

(i) Choose an oversampling factor N, a physical model that describes the dynamics

of the molecule and realistic initial parameters (in our case: Eop, Ecl, kop and kcl).

(ii) For each burst, calculate the overall time the molecule spent in each of the two

states (τop and τcl) using a Monte Carlo simulation (running for a time duration

equal to the burst duration, BD, of the corresponding burst).

(iii) Calculate the shot-noise-free E value (Esnf) using Eq. 3.

(iv) Calculate the shot-noise-dependent E value (Esn) using Eq. 5.

(v) Repeat steps (ii) through (iv) N times for each burst and add the results to an E-

histogram.

(vi) Divide the resultant E-histogram by N.

(vii) Improve on Eop, Ecl, kop, and kcl to achieve best agreement between the

calculated and the experimental E-histograms using a chi-squared minimization.

The histograms may be slightly smoothed to assist in the optimization.

3.3.2. Rational of the PDA

Oversampling factor: The purpose of the oversampling factor (N = 200, step (i)) is to

reduce statistical noise caused by the binomial random number generator and the

Monte Carlo simulation and by the finite number of bursts collected. For a data set

containing 1000 bursts and burst size similar to that in Figure 18, N = 200 reduced

the noise almost entirely.

Calculating shot-noise-free E value (Esnf): The fitting procedure requires assumption

of a physical model that describes the hairpin dynamics (step (i)). We found that a

model consisting of two states that interconvert stochastically at fixed rates (a two-

state model with first-order transitions) described all data collected for this work.

Therefore, the Monte Carlo simulation (step (ii)) was performed according to a two-

state model. The simulation stochastically draws open and closed states dwell-times

from exponentially distributed dwell-times (with typical kcl and kop rates,

29

respectively) for overall time duration equal to the burst duration. The overall times

the molecule spent in each state (τop and τcl) were then calculated by summing over

all the open and the closed state dwell-times. Notice that τop and τcl are the sum over

all the durations a simulated molecule spends in the open and closed states, not the

typical opening and closing times (i.e., not 1/kop or 1/kcl). For each burst, the shot-

noise-free E value (Esnf, step (iii)) was calculated by summing the E values of the

states (Eop, Ecl) weighted by the fraction of the time the molecule spent in each of the

states (τop /BD and τcl /BD) according to Eq. 3:

𝐸 =

(3)

Calculating the final E value (Esn): The final shot-noise-dependent E value (Esn, step

(iv)) was calculated for each burst by first drawing the number of acceptor photons

(A) from a binomial random number generator ( , according to Eq. 3, LabView

v. 7.1, National Instruments) and then calculating Esn by dividing A by BS (BS, the

sum of the donor plus acceptor photons in a burst), according to Eq. 4:

| = ( )𝐸

( 𝐸 )

(4)

𝐸 =

(5)

Fitting procedure: Fitting of the calculated E-histogram to the experimental E-

histogram (step (vii)) was achieved by minimizing chi-squared. This can be carried

out manually or by using any automatic algorithm. We used a self-programed

algorithm (LabView) that searches for the set of free variables (Eop, Ecl, kop and kcl)

that result in a minimum chi-squared (calculated from the differences between the

experimental and the calculated histograms).

Smoothing the histograms: The experimental and the calculated E-histograms may

not be smooth due to a low number of bursts or due to photon statistics (deviation of

rational numbers of acceptor photons to rational number BS), [4, 5] and this can

cause difficulties in the minimization of chi-squared. In such cases, we

recommended that chi-squared be calculated on smoothed experimental and

calculated E-histograms. Smoothing was carried using a running averaged algorithm.

Over smoothing is not recommended because it can erase histogram features.

30

Additional width of the E-histograms: Although the theory that describes the shot-

noise contribution is well understood, none of the experimental results to date,

including data on molecules that presumably have a fixed donor-acceptor distance,

show a shot-noise width [4, 5, 7]. For reasons which are not understood [62], the

widths of the experimental histograms are always broader than expected from the

shot-noise calculation. It is customary, therefore, to broaden the calculated

histograms by a Gaussian distribution of E values or donor-acceptor distances. Here

we broadened the E-histograms by the equivalent of 1.4 Angstroms, which best fit

the results. This was achieved by converting the Eop and Ecl to distances (Rop and Rcl,

and using R0 = 50 Angstroms), recalculating the distance by drawing new distances

from a random Gaussian number generator (with a mean value Rop and Rcl, and

standard deviation 1.4-Angstroms width), and then recalculating Eop and Ecl. This

calculation was conducted between steps (ii) and (iii).

Background photons: We ignored the negligible contribution of background photons

(typically less than 2% of the signal), as it was previously shown that background of

that scale has only marginal effects on the E-histogram [5].

3.4. Calculating the Opening Rates using MFOLD and

Transition State Theory

Transition states energies cannot be directly calculated using the nearest-neighbor

model and the MFOLD program [63]. Instead, we used MFOLD to calculate the free

energies of formation (melting) of 6-bp duplexes with sequences identical to that of

the hairpin stems at 22 °C and at the corresponding Na+ concentrations. Assuming

that in melting of two strands the transition state is the unzipping of the last base-

pair, the difference between the transition state and the melted state energy is

expected to come primarily from the translational entropy (the many micro-states the

unbounded strands can occupy in the volume). Because the translational entropies of

the four pairs of single-stranded DNA (stems) are expected to be very close, we

assume that the difference in the heights of the barrier for opening of the four

hairpins is similar to the difference in stabilities calculated using the nearest neighbor

approach.

31

To calculate the opening rates from these energies we used standard transition state

theory:

= (6)

where is the observed opening rate, is the free energy height of the barrier

with respect to the free energy of the closed state, and the pre-exponential factor

reflects the transition rate in the absence of an energy barrier. The exact value of

factor is unknown. Woodside et al. [28] found that data from pulling experiments of

DNA hairpins using optical trap are best fitted by = . In these two

papers, Eq. 6 is called ‘transition state theory’; however, the formulations used are

identical. In second part of this work, which summarizes data for opening and

closing rates of four hairpins, we adjusted the pre-exponential factor such that the

MFOLD/TST opening rates fit best the experimental opening rates, and the best fit

was achieved for = . With this value all energies decreased by

around 0.5-0.7 kJ/mol with respect to energies calculated using = .

3.5. Sample Preparation

3.5.1. Single-Stranded DNA Labeling

HPLC-purified bottom and top strands of DNA were purchased from IDT

(Coralville, LA, USA) with a C6 dT internal amino modifier (iAmMC6T) in position

10 from the 3' end and in position 1 from the 5' end. These positions were labeled

with ATTO-550 and ATTO-647N (ATTO-TECH GmbH, Siegen, Germany),

respectively, and HPLC purified (reverse-phase C18, Amersham Bioscience,

Uppsala, Sweden). Typical labeling yields were ~ 70% and purities after HPLC were

> 99% as determined by reintroduction into the HPLC. To prevent loss of the

adenine bases during storing, the molecules were maintained in basic conditions (pH

> 8).

32

3.5.2. Origami Design

A DNA origami rectangle was prepared following Rothemund’s design.[43]

M13mp18 single-stranded DNA was used as the scaffold (New England BioLabs,

Ipswich, MA, USA), and the staples were unpurified. The hairpin bottom strand

contained a sequence identical to that of one of the origami staples (r-1t16f) such that

it was incorporated into the origami in the annealing process and branched out of the

origami plane; the original r-1t16f staple was not introduced. An additional acceptor

labeled strand (elongated with another staple’s sequence) was added to annealing.

This helped separate residual hairpin-only from hairpin-origami based on S

values[46] .

3.5.3. Annealing Procedures

Hairpin-only: The top and bottom strands were annealed at 94 °C (1.5 µM in 10 µL

TE-NaCl buffer [10 mM Tris, pH 8.0, 1 mM EDTA], and 100 mM NaCl) and then

gradually cooled (30 min) to room temperature (using PCR).

Hairpin-origami: The annealing mixture consist of 2 nM scaffolds and 10 fold excess

of staples and 20 fold excess of top and bottom strands in 50 µL of 50X TAE buffer

(2 M Tris, 50 mM EDTA, 2 M acetic acid) and 12 mM MgAc; we found that with

this high concentration of buffer, the yield was 50% higher than 1X TAE was used.

To anneal, strands were incubated at 95 °C for 5 min, cooled to 60 °C at 1 °C/2 min,

and cooled again to room temperature at 1 °C/5 min (using PCR).

3.5.4. Origami purification

After the annealing, the origami was filtered through a size-exclusion column[64]

(Sephacryl S-300 HR, dsDNA cut-off of 118 bp; GE Healthcare, Little Chalfont,

UK) that was hand-packed with 750 µL of liquid resin and then spin at 1000 g for 2

minutes, yielding 500 µL of dry volume. The column was equilibrated by washing 3

times with 500 µL of 1X TAE buffer (40 mM Tris, 1 mM EDTA, 40 mM acetic

acid) and centrifuging for 1 minute at 1000 g to remove excess buffer. Origami

sample of 50 µL was added and centrifuged at 1000 g for 4 minutes; this was

repeated twice to achieve complete removal of excess staples. The origami structural

integrity was validated using AFM (data not shown).

33

3.5.5. DNA Origami Structure Validation

To ensure that rectangular DNA origami structures formed as designed, AFM images

were taken. The images shown in Figure 17 reveal the rectangular DNA origami

structures.

Figure 17. DNA origami structure and integrity. (A-C) Atomic force microscope images at three

magnifications.

(B)

(C)

(A)

34

4. Results

We first examined the hairpin dynamics using the diffusion-based sm-FRET-ALEX

technique [5, 27]. The hairpin-only constructs have a single donor and a single

acceptor chemically linked to each oligonucleotide. The hairpin-origami constructs

consisted of a single donor on the hairpin and two acceptors, one placed on the

hairpin and the other on the origami. To reduce possible influence of fluorophore

bleaching, incomplete labeling, or missing components on the E-histograms, only

bursts with the expected fluorophore stoichiometry were selected. Therefore, the E-

histograms were constructed from events with S values around 0.5 for the hairpin-

only and S values around 0.33 for the hairpin-origami.

Figure 18. Typical two-dimensional E/S-histogram and E- and S- one-dimensional histograms. (A)

Hairpin-only A31-bp6-S (B) Hairpin-origami A31-bp6-S measured at 25 mM NaCl concentration.

At 25 mM NaCl concentration A31-bp6-S opened and closed very slowly (5 s-1

).

Thus, only a very minor bridge between the doubly labeled open and closed states

was observed. The very minor bridges observed between doubly labeled open and

closed states and the donor-only and the acceptor-only population indicate very

minor fluorophore bleaching and blinking. A31-bp6-M and A31-bp6-S, hairpin-only

and hairpin-origami (3 pM concentration) were measured at buffers with a range of

NaCl concentrations.

35

The E-histograms of A31-bp6-M hairpin only and hairpin-origami are presented in

Figure 19.

Figure 19. Diffusion-based sm-FRET-ALEX measurements of dynamics of (left) hairpin-only A31-

bp6-M and (right) hairpin-origami A31-bp6-M. E-histograms obtained over a range of NaCl

concentrations (1-100 mM).

Two prominent peaks were observed in all E-histograms, corresponding to open and

closed states, respectively, demonstrating that the hairpin is predominantly a (quasi)

two-state molecule. In all cases, the fraction of hairpins in the closed state increased

with increasing NaCl concentration.

Unlike the diffusion-based method, which provides only a millisecond snapshot of

the state of the hairpin, the immobilization-based sm-FRET-TIRF provides both a

snapshot (with duration depending on the camera’s frame rate, 5-15 ms, in this work)

and a time evolution of the state of individual hairpins (time trajectories, Figure 12).

This enables determination of the fraction of open state, as in the diffusion-based

experiments, and of the time periods the hairpin spent in each state (dwell-times).

From the dwell-times, the transition rates can be directly obtained. The opening rates

are calculated from the close state dwell-time histograms and the closing rates from

the open state dwell-time histograms.

A31-bp6-M and A31-bp6-S, hairpin-only and hairpin-origami were immobilized on

coverslips, and experiments were performed at different NaCl concentrations. For

the hairpin-only experiments, a 35-bp dsDNA linker modified with biotin connected

the hairpin to the coverslip coated with avidin (Figure 20B1). The hairpin was

connected to the origami through a 35-bp dsDNA linker, and the origami was

immobilized on the coverslip through a biotin-avidin interaction (Figure 20B2). The

data were analyzed in two ways. First, E values calculated for each individual

36

molecule and for each camera frame were incorporated into an E-histogram (Figure

20C1-2). Second, open and closed state dwell-times were determined from the time

trajectories and were incorporated into open and closed dwell-time histograms

(Figure 20D1-2). These histograms were fitted using a single- or a double-

exponential function from which the closing and opening rates were determined.

No inactive hairpins were detected in any of the trajectories that passed a certain

intensity threshold.

Figure 20. Immobilization-based sm-FRET-TIRF measurements of hairpin dynamics. Data are of

A31-bp6-M. (A) Typical donor and acceptor emissions recorded by the EMCCD camera and typical

donor and acceptor intensity time trajectories originating from an individual molecule and the

corresponding E time trajectory. (B1-2) Schematic of immobilization of hairpin-only and hairpin-

origami on a coverslip. (C1) E-histograms of hairpin-only and (C2) E-histograms of hairpin-origami

measured in NaCl concentrations ranging from 1 to 100 mM. (D1) Open state and (D2) closed state

dwell-time histograms of hairpin-only and fit to single-exponential functions from which closing and

opening rates were calculated [46].

37

4.1. Comparison of the Fraction of Open State

To examine the influence of the origami and the coverslip on the hairpin dynamics,

we calculated the fraction of hairpin in the open state by dividing the size of the low

E peak by that of the sizes of the low plus high E peaks in each of the E-histograms.

The results for the freely diffusing and immobilized hairpin-only and hairpin-origami

are presented in Figure 21.

Figure 21. Very good agreement between the fraction of open state of freely diffusing and

immobilized hairpin-only and hairpin-origami measured at a range of NaCl concentrations. (A) A31-

bp6-M, (B) A31-bp6-S.

For both hairpins, the fractions of open state in each of the NaCl concentrations

were, within the experimental noise (~4%), essentially identical. The fractions were

different for the two hairpins, however.

4.2. Comparison of Immobilized Hairpin-Only and

Hairpin-Origami Opening and Closing Rates

To further investigate possible influence of the coverslip and the origami on the

hairpin dynamics, we compared the shape of the open and closed state dwell-time

histograms of hairpin-only to that of hairpin-origami for A31-bp6-M and A31-bp6-S.

The shapes of the dwell-time histograms are essentially the same for hairpin-only

and for hairpin-origami (Figure 22A-B, blue and red symbols, respectively). Except

for the closed state dwell-time histograms of A31-bp6-S (Figure 22B2), all histograms

were fit reasonably well by a single-exponential function (Figure 22A1, 5A2, and

5B1). Within the experimental noise, the calculated opening and closing rates are

similar for hairpin-only and hairpin-origami at all NaCl concentrations (Figure

22C1-2, blue and red symbols, respectively). The closed state dwell-time histograms

of A31-bp6-S were not described well by single-exponential functions and were,

38

therefore, fit to a double-exponential function. The results indicate that kinetic of

opening is dominant (>80%) by a slow component and an additional but minor fast

component also exists. We suggest that the dominant slow component is the correct

closing of stem because it is the most stable state and we the slow rate in this work

for the comparison between hairpin-only and hairpin-origami, and for activation

energy calculation.

Figure 22. Very good agreement between the closing and opening dwell-time histograms measured

for immobilized hairpin-only and hairpin-origami and rates measured at different NaCl concentrations

for (left) A31-bp6-M and (right) A31-bp6-S; hairpin-only is indicated by blue symbols and hairpin-

origami by red symbols. (A1-2) Closing dwell-time histograms. (B1-2) Opening dwell-time

histograms. The solid lines in A1, A2, and B1 are fits to single-exponential functions, and solid lines

in B2 are fits to double-exponential functions. (C1-2) Calculated closing (closed symbols) and

opening (open symbols) rates. For the opening of A31-bp6-S, only the dominant slow (exponential)

component is presented. Solid lines are to guide the eye.

39

Based on the agreement between the fractions of open state and between the dwell-

time histograms of hairpin-only and hairpin-origami, we conclude that (i) the

origami does not influence the hairpin dynamics, (ii) immobilization on the coverslip

glass does not influence the hairpin dynamics, and (iii) the diffusion-based and the

immobilization-based techniques are in very good agreement.

Summaries of dwell time analyses for A31-bp6-M and A31-bp6-S are presented in

Figure 23. Only the A31-bp6-S closed-state dwell time showed double-exponential

behavior (Figure 23B1); the rest of the curves for open- and closed-state dwell times

showed clear single-exponential behaviors (Figure 23 A1-2, B2). For semi-

logarithmic plots of the dwell-time histograms see Figure 34 in the Appendix 9.2.

Figure 23. Opening (left) and closing (right) dwell-time histograms of A31-bp6-M (top) and A31-bp6-S

(bottom) all fitted with single-exponential functions, besides B1, which is fitted with double-

exponential function.

40

Summary of the opening and closing rates of hairpin-only A31-bp6-M and A31-bp6-S

measured using the immobilized-based method is presented in Figure 24.

Figure 24. Summary for immobilized-based transition rates. Opening rates (open symbol) and closing

rates (closed symbol) of A31-bp6-M (red) and A31-bp6-S (black) hairpin-only.

The obtained closing and opening rates showed that the ionic strength influenced the

opening and closing rea