Single-molecule FRET and linear dichroism studies ofDNA breathing and helicase binding at replicationfork junctionsCarey Phelpsa,b,c, Wonbae Leea,b,c, Davis Joseb,c, Peter H. von Hippelb,c,1, and Andrew H. Marcusa,b,c,1

aOregon Center for Optics, bDepartment of Chemistry and Biochemistry, and cInstitute of Molecular Biology, University of Oregon, Eugene, OR 97403

Contributed by Peter H. von Hippel, August 19, 2013 (sent for review May 29, 2013)

DNA “breathing” is a thermally driven process in which base-paired DNA sequences transiently adopt local conformations thatdepart from their most stable structures. Polymerases and otherproteins of genome expression require access to single-strandedDNA coding templates located in the double-stranded DNA “inte-rior,” and it is likely that fluctuations of the sugar–phosphate back-bones of dsDNA that result in mechanistically useful local base pairopening reactions can be exploited by such DNA regulatory pro-teins. Such motions are difficult to observe in bulk measurements,both because they are infrequent and because they often occur onmicrosecond time scales that are not easy to access experimen-tally. We report single-molecule fluorescence experiments withpolarized light, in which tens-of-microseconds rotational motionsof internally labeled iCy3/iCy5 donor–acceptor Förster resonanceenergy transfer fluorophore pairs that have been rigidly insertedinto the backbones of replication fork constructs are simulta-neously detected using single-molecule Förster resonance energytransfer and single-molecule fluorescence-detected linear dichro-ism signals. Our results reveal significant local motions in the∼100-μs range, a reasonable time scale for DNA breathing fluctu-ations of potential relevance for DNA–protein interactions. More-over, we show that both the magnitudes and the relaxation timesof these backbone breathing fluctuations are significantly per-turbed by interactions of the fork construct with a nonprocessive,weakly binding bacteriophage T4-coded helicase hexamer initia-tion complex, suggesting that these motions may play a funda-mental role in the initial binding, assembly, and function of theprocessive helicase–primase (primosome) component of the bacte-riophage T4-coded DNA replication complex.

smFRET | single-molecule linear dichroism | thermal fluctuations |T4 primosome helicase | DNA helicase

DNA “breathing” is defined as the transient opening, due tothermal fluctuations, of nucleic acid base pairs at experi-

mental temperatures below the melting temperature of dsDNAduplex. Such fluctuations are thought to be important to thefunction of replication, transcription, recombination, and repairsystems, which depend on the ability of the relevant proteincomplexes to gain access to ssDNA templates that are locatedin the dsDNA “interior” (1–3). Furthermore, the ability of thegenome to spontaneously expose partial template sequences isa central feature of protein–DNA binding models in which openconformations serve as substrates that can be “trapped” by afunctional protein complex. McConnell and von Hippel (4) sug-gested that breathing of duplex DNA might be structurallydecomposed into a set of distinct breathing elements, each in-volving characteristic fluctuations of the sugar–phosphate back-bone and the nucleobases. For example, a possible motion mightinvolve compression along the double-helix axis (perhaps coupledwith twisting or bending), leading to the disruption of Watson–Crick hydrogen bonds without base unstacking. Such motionsmight serve as transition states for low-temperature hydrogenexchange. Another class of fluctuations might result in extensionof the DNA duplex along its double-helical axis, causing the faces

of adjacent bases to lift off of one another without breakinginterbase hydrogen bonds. Such fluctuations could be involvedin transition states leading to intercalation and base “flipping.”More-complex breathing fluctuations could be described ascombinations of the above (or other) breathing modes. Recently,equilibrium and steady-state spectroscopic approaches have beenused to monitor position-dependent DNA breathing that occursin the proximity of ss–dsDNA forks and primer–template (p/t)junctions of DNA constructs (5, 6). Such motions are believed tooccur in the microsecond range (7), a difficult regime to accessexperimentally. Methods to observe such breathing motions ofdsDNA backbones directly, or to determine how these motionsmight relate to the concerted and biologically relevant activities ofnucleic acid enzymes, have been largely unavailable.In this paper, we use single-molecule (sm) methods to ex-

amine DNA breathing and its role in the binding and assemblymechanisms of the bacteriophage T4 replication (primosome)helicase. Replication helicases are ATP-dependent molecularmotors that unwind dsDNA at replication forks to expose tem-plates for DNA synthesis (8, 9). The T4 replication system is anexcellent model for studies of replication in higher organisms, be-cause it is the simplest system to use a hexameric helicase–primase(primosome) subassembly to unwind the base pairs ahead of thepolymerases at the replication fork and to catalyze the synthesis ofthe RNA primers required for the reinitiation of lagging strand syn-thesis at the 3′ ends of the newly formedOkazaki fragments (10–12).In order for the gp41 helicase to recognize and interact with

the replication fork, it is necessary that the bases near the fork

Significance

Unique single-molecule fluorescence techniques were used tomonitor DNA “breathing” at and near the junctions of modelDNA replication forks on biologically relevant microsecond-to-millisecond time scales. Experiments performed in the absenceand presence of helicase complexes addressed the role of thesefluctuations in helicase function during DNA replication. Thesestudies simultaneously monitored single-molecule Förster res-onance energy transfer and single-molecule fluorescence lineardichroism of “internal” Cy3/Cy5 labels placed rigidly into theDNA backbones at positions near the fork junction. Our resultsshowed significant breathing at the fork junction that wasgreatly augmented by the presence of weakly bound helicase,followed by still larger fluctuations and strand separation afterfull duplex DNA unwinding by the complete tightly bound andprocessive helicase complex.

Author contributions: C.P., P.H.v.H., and A.H.M. designed research; C.P., W.L., and D.J.performed research; D.J. contributed new reagents/analytic tools; C.P., W.L., D.J., and A.H.M.analyzed data; and C.P., P.H.v.H., and A.H.M. wrote the paper.

The authors declare no conflict of interest.

See Commentary on page 17173.1To whom correspondence may be addressed. E-mail: petevh@molbio.uoregon.edu orahmarcus@uoregon.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1314862110/-/DCSupplemental.

17320–17325 | PNAS | October 22, 2013 | vol. 110 | no. 43 www.pnas.org/cgi/doi/10.1073/pnas.1314862110

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

junction be transiently exposed from behind the sugar–phosphatebackbones that shield them from the solvent environment. Asshown in Fig. 1 A–C, the T4 gp41 helicase hexamer binds weaklyto the replication fork junction, and is more likely to dissociatefrom the fork rather than to engage in processive dsDNA un-winding (13, 14). In contrast, the T4 primosome helicase, whichconsists of a gp41 hexamer and a gp61 monomer, binds tightlyat the replication fork junction and manifests very processiveunwinding activity in the presence of NTP. The breathing fluc-tuations of base pairs located in the immediate vicinity of areplication fork or a p/t junction are much more extensive thanthose of base pairs located deep within a dsDNA region (6, 13).A hypothetical free-energy surface (FES) describing base-pairdynamics near the fork junction is portrayed in Fig. 1D. In thismodel, base pairs can thermally fluctuate into the open state,which is separated from the native DNA structure by a free-energy barrier. The height of this barrier is likely dependent onseveral variables, such as salt concentration, pH, base composi-tion, and base pair sequence. By observing the fluctuations ofdsDNA constructs within regions that have been labeled withfluorescent dyes, it should be possible to obtain informationabout the FESs to which defined segments of DNA bases andsugar–phosphate backbones are subject, and also to deter-mine how these surfaces are affected by the presence of theT4 helicase.We here perform single-molecule experiments on DNA repli-

cation fork constructs that have been “internally” labeled with theFörster resonance energy transfer (FRET) donor–acceptor chro-mophores iCy3 and iCy5 (14). These cyanine dyes were rigidlyincorporated into the sugar–phosphate backbone using phos-phoramidite chemistry, with each fluorophore replacing an op-posed DNA base at each label position (Fig. 2A). In someexperiments, the fluorophores were positioned close to the rep-lication fork junction; in others, deep within theDNAduplex. Ourmeasurements use a polarized excitation scheme that simulta-neously monitors, on submillisecond time scales, the smFRETbetween the iCy3/iCy5 donor–acceptor chromophore pair and thefluorescence-detected linear dichroism (smFLD) of the iCy3 do-nor chromophore. The principle of these measurements is basedon the idea that a single-oriented DNA replication fork constructwill absorb light with different probabilities, depending on thepolarization of the laser field. By rapidly modulating the laserpolarization, and recording the phase of the modulation forevery detected signal photon, we were able to monitor the time-dependent projections of the absorption transition dipole momentonto arbitrarily defined laboratory axes. Whereas the smFRET

signal is sensitive to the relative separation and orientation of theiCy3/iCy5 chromophore labels, the smFLD signal is sensitive to theiCy3 orientation within the laboratory frame. DNA breathing andhelicase-facilitated duplex unwinding are expected to influence bothtypes of signals.Single-molecule FRET experiments have previously been used

to examine helicase binding at replication fork junctions (15).Moreover, polarization-resolved single-molecule experiments havepreviously been used to study the orientations of chromophoresembedded in complex systems, including glasses (16, 17), bio-logical macromolecules (18–20), conjugated polymers (21), andlight-harvesting complexes (22). The majority of these methodsare capable of sampling a single-molecule signal with time res-olution on the order of tens of milliseconds or longer. To observeDNA breathing, it was necessary to record single-molecule sig-nals with submillisecond time resolution.

smFLD and smFRETWe performed control experiments on a model DNA replicationfork construct that had been internally labeled using the donor–acceptor FRET chromophore pair iCy3/iCy5 as shown in Fig. 2A.In this way, both smFLD and smFRET signals could be simul-taneously monitored to provide information about DNA back-bone rigidity and the relative displacement of the conjugate strandsat the level of the Cy3/Cy5 probes. Ensemble linear dichroism (LD)measurements are often used to determine the orientations of ab-sorption transition dipole moments of chromophores embedded inuniaxially strained polymer films (23, 24). The LD of such samples

‘open’ forkconformation

’closed’ forkconformation

free

ener

gy

local conformational coordinate Q(e.g. inter-base separation, orientation)

+

lagging strand

leading strand

3’

5’ DNA·(gp41)6

gp61 primase

(gp41)6 helicase

DNA replication fork

DNA·(gp41)6·gp61

lagging strand

leading strand

A

C

B

D

Fig. 1. (A) The isolated hexameric gp41 helicase (blue spheres), the subunitsof which are bound together by intersubunit (and in this case non-hydrolyzable) NTP ligands (not shown for clarity), binds weakly to the ss–dsDNA junction of a replication fork, favoring dissociation over binding, asshown in B. Introduction of gp61 primase (orange ellipsoid), in a 6:1 gp41-to-gp61 subunit ratio, causes the resulting T4 primosome helicase complex tobind strongly to the replication fork, as shown in C. (D) A hypothetical FESdescribing local dsDNA fluctuations near the fork junction in the absenceof helicase.

B

prism

objective

pinhole

1 MHz s / pphase-modulation

sample x-y scanning stage

long-passfilter

phase synchronousphoton-countingdetectors

dichroicmirror

C

Cy5

Lead

ing

stra

nd

Lagg

ing

stra

nd

Bio

tin

3’ 5’

3’5’

Neutravidin

PEG

iCy5

10-m

er (d

s)16

-mer

(ds)

GTP

iCy3

T4 h

elic

ase

dT29

34-m

er (s

s)

GTP

A

xy

z

p

s

x

z

y

^^

^

Fig. 2. The smFRET and smFLD experimental layout. (A) The T4 gp41 heli-case binds to the d(T)29 loading sequence on the lagging strand of a modelDNA replication fork construct. An assembled (gp41)6∙gp61∙DNA primosomecomplex can unwind the duplex region of the DNA in the presence of GTP.The strands within the dsDNA region are internally labeled with the FRETdonor–acceptor chromophores iCy3 and iCy5, respectively. (B) TIRF excitationscheme and detection method. The polarization of the excitation beam ismodulated at 1 MHz. The p-polarization component points in the directionof the y axis, and the s-polarization component is contained within the x–zplane. (C) Orthogonally polarized directions used to measure the FLD signal,from the perspective of the incident beam.

Phelps et al. PNAS | October 22, 2013 | vol. 110 | no. 43 | 17321

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

is defined as the difference between the absorption of light polar-ized parallel and perpendicular to the strain axis.

LD=Ak −A⊥ [1]

In FLD, the relative population of molecules excited in thetwo polarization directions is determined by measuring the dif-ference between the corresponding fluorescence intensities. Inour smFLD experiments, the DNA replication fork substrateswere chemically attached to the surface of a glass microscopeslide using biotin/NeutrAvidin linkages (Fig. 2A and SamplePreparation Methods). The DNA fork construct, with its rigidlyattached iCy3 donor chromophore, was thus oriented with re-spect to the laboratory frame. A total internal reflection fluo-rescence (TIRF) microscope was used to illuminate the samplewith an evanescent field, as shown in Fig. 2B. The polarization ofthe incident laser beam was rapidly modulated (at 1 MHz) sothat the sample was alternately excited using plane polarizationsin the directions –45° (designated as k) and +45° (designatedas ⊥) relative to the surface normal (Fig. 2C; see instrumentdetails provided in SI Text, Instrumentation for Single-MoleculeFluorescence-Detected Linear Dichroism and Single-Molecule För-ster Resonance Energy Transfer). Because the absorption proba-bility depends on the square projection of the laser field onto theexcitation transition dipole moment, the emission probabilitydepends on the phase ϕ of the polarization modulation cycle. Themodulation amplitude of the fluorescence intensity is a measure ofthe orientation of the absorption transition dipole moment.The 1-MHz polarization modulation of the excitation beam

was implemented using acousto-optic Bragg cells placed in thebeam path of the laser (SI Text, Instrumentation for Single-MoleculeFluorescence-Detected Linear Dichroism and Single-MoleculeFörster Resonance Energy Transfer). The sample image was raster-scanned using a computer-controlled x-y piezo-scanning micro-scope stage (NPS-XY-100A; Queensgate), and the stage positionwas held fixed when the fluorescence from a single molecule wasdetected. Donor and acceptor fluorescence from a single mole-cule was masked using a pinhole, directed along separate pathsusing a dichroic beam-splitter, and separately detected using twofast photon-counting detectors (SPCM-AQR-16, 175-μm activearea; Perkin-Elmer Optoelectronics). The detection times of in-dividual photoelectron counts were stored on a computer andadditionally assigned values of the laser polarization phase ϕ,which was subdivided into 64-bin increments. We thus recordedtime- and ϕ-dependent trajectories of the donor (acceptor) sig-nals, IDðAÞðt;ϕÞ=

PNDðAÞn= 1 δðt− tDðAÞn Þ expðiϕDðAÞ

n Þ, where tDðAÞn andϕDðAÞn are the time and phase of the nth donor (acceptor) signal

count, respectively, and NT =ND +NA is the total number ofsignal counts. In principle, single-molecule kinetic informationcould be extracted from these data with a time resolution of ∼1 μsand a polarization phase resolution of ∼2π/64 ≈ 0.1 rads. Fromthe signal trajectories ID=Aðt;ϕÞ, we constructed the cumulativedonor–acceptor signal IT = ID + IA and the FRET efficiency pa-rameter, defined as IA=IT . We note that the presence of the FRETacceptor chromophore does not alter the meaning of the FLDmeasurement because the acceptor signal depends on the excita-tion of the donor chromophore, which was excited with polariza-tion specified by ϕ. Two separate approaches were taken toprocess these data. The first approach was used to visualize thesmFLD and smFRET trajectories on millisecond time-scales,and the second served to resolve the tens-of-microseconds dy-namics of the DNA sugar–phosphate backbone.Because the fluorescence intensity from single iCy3/iCy5-

labeled molecules was of the order of ∼25,000 s−1, we integratedthe cumulative signal IT over ∼800 modulation cycles to obtaina ϕ-dependent probability distribution, from which the smFLDsignal could be determined. We extracted the smFLD signalthrough a post data-acquisition procedure that was operationallysimilar to the algorithm used by a lock-in amplifier (25). Fora single molecule of fixed orientation, the integrated histogram

of photoelectron counts vs. polarization phase ϕ approximatesa smoothly varying sinusoidal function f ðϕÞ=A+C cosðϕ+ϕ0Þ,where 2C is the modulation amplitude (equal to the FLD signal),A is a constant offset, and ϕ0 is a known phase determined by themodulation electronics. We numerically calculated the smFLDsignal by multiplying the function f ðϕÞ by cosðϕ+ϕref Þ, where weset ϕref =ϕ0, and subsequently performing an average over thephase angle ϕ:

Dcos

�ϕ+ϕref

�f ðϕÞ

Eϕ=C2=FLD4

: [2]

We define the reduced FLD (FLDr) = FLD/Nhist, where Nhistis the number of photoelectron counts contributing to the in-tegrated histogram. The quantity FLDr is independent of signalintensity, and thus depends only on the orientation of the mo-lecular transition dipole moment. We found that accurate resultscould be obtained with as few counts as Nhist = 20. Thus, asmFLDr signal trajectory could be obtained from a moleculewith an emission count rate of ∼25,000 s−1 and a time resolutionof ∼800 μs.The second method we used to process our trajectories is

based on an analysis of time correlation functions (TCFs).We focus on the fluctuating smFRET and smFLDr signals,FRETðtÞ= δFRETðtÞ+FRET and ITðt;ϕÞ= δITðt;ϕÞ+ IT , re-spectively, which were recorded using a DNA replication forkconstruct at equilibrium. We define the TCFs GFRETðτÞ=hFRETðτÞFRETð0Þi and GFLDðτÞ= hIpTðτÞITð0Þi, which describethe decay of correlations of the fluctuating smFRET and smFLDrsignals, respectively, with increasing time interval τ. The anglebrackets indicate an average over time according to

GFRET=FLDðτÞ=Z∞

−∞

SðtÞS p ðt+ τÞdt: [3]

In Eq. 3, SðtÞ represents the fluctuating smFRET or smFLDrsignal. The TCFs, so defined, are expected to decay on a timescale associated with the same underlying microscopic processesthat give rise to spontaneous fluctuations of the system awayfrom its stable equilibrium configuration (26). We note thatthe short time limit of Gð0Þ= hS2i represents the mean squaremagnitude of the fluctuating signal, whereas the long time limitGð∞Þ= hSi2 represents the square mean signal. The TCF anal-ysis provides access to time scales much faster than the histogramintegration method described previously, because the latter relieson multiple detection events to reconstruct a sinusoidal wave-form. Because the TCF method correlates the properties of in-dividual photoelectron signal counts, we were able to extractkinetic information from single-molecule trajectories with ∼20-μstime resolution.We performed control experiments to calibrate the smFLD

and smFRET signals from known test samples, and to ensurethat our measurements were not influenced by the dynamics oflaser-induced excited triplet states (SI Text, Sample Preparationand Model DNA Replication Fork Constructs).

ResultsWe investigated the dynamics of the two iCy3/iCy5-labeled DNAreplication fork constructs (SI Text, Sample Preparation andModel DNA Replication Fork Constructs). For both types of DNAconstructs, the iCy3/iCy5 chromophores replaced nucleotidebases on opposite strands directly across from one another.The “duplex-labeled” construct had the iCy3/iCy5 pair placeddeep in the dsDNA region of the fork construct, whereas the“fork-labeled” construct contained the iCy3/iCy5-labeled probepair at the ss–dsDNA junction. Using these DNA constructs, wewere able to simultaneously measure the smFRET and smFLDrsignals. We initially examined the duplex-labeled DNA construct

17322 | www.pnas.org/cgi/doi/10.1073/pnas.1314862110 Phelps et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

and its unwinding by the T4 primosome helicase complex. In theabsence of any helicase, we observed that smFRET signals wereconstant over time, with an average FRET efficiency IA=IT ≅ 0.7.Photobleaching from these samples occurred only occasionally,presumably because the fluorophores were protected from thesolvent environment by the sugar–phosphate backbones of thedsDNA (14). In the absence of unwinding proteins, the smFLDrsignals also remained at a constant value, which could lie anywherewithin the range −1 to 1. Example control data sets for duplex-labeled constructs are presented in SI Text, Control Experiments.Upon introducing the T4 hexameric helicase and 6 μM ATP

into the sample chamber, DNA-unwinding events could be ob-served that were characterized by a sudden drop in the iCy5fluorescence intensity and a simultaneous increase in the iCy3fluorescence intensity, defined as a drop in the FRET efficiency.Similar helicase-unwinding FRET-conversion events have beenobserved previously with this system (14). smFLDr traces werealso significantly affected by these events. In Fig. 3 A–C, wepresent simultaneously measured single-molecule iCy3/iCy5 in-tensities, representing smFRET and smFLDr trajectories, re-spectively, of the duplex-labeled DNA replication fork constructin the presence of the processively unwinding T4 helicase–pri-mase system (300 nM gp41 and 50 nM gp61) and 6 μM ATP.Before the unwinding event at the ∼17.5-s tick mark, the iCy5intensity remained high and constant, resulting in a FRET effi-ciency of over 0.7. During this time, the smFLDr signal fluctuatednear a nonzero constant value. Immediately after the unwindingevent at ∼17.5 s, characterized by a drop in the smFRET signal,the smFLDr signal fluctuated broadly around zero because theiCy3-labeled strand had been freed from the tethered dsDNAconstruct by the helicase. In Fig. 3 Right, we also show selectedpolarization ϕ-dependent signal distributions, which were usedto determine the smFLDr trajectory shown in Fig. 3B.Having established that our experiment was able to simulta-

neously monitor the smFRET and smFLDr trajectories, we nextsought to observe the breathing dynamics of the system usingmodel replication fork constructs with the iCy3/iCy5 probes

placed at either the ss–dsDNA fork junction or deep within theduplex region. Our initial studies focused on observing dynamicsrevealed by calculating two-point TCFs for the smFRET andsmFLDr trajectories, with the goal of observing DNA breathingfluctuations at equilibrium. The TCFs should reveal thermallyactivated backbone fluctuations as the iCy3/iCy5 probes samplethe FES shown hypothetically in the simplified diagram of Fig.1C. The decays of the TCF report on the timescale of theseequilibrium fluctuations, and the functional form of the decay,can provide information about the nature of the fluctuations.In Fig. 4, we show examples of TCFs calculated from simul-

taneously recorded smFRET and smFLDr trajectories. Thoughthe paired TCFs derived from smFRET and smFLDr signal tracesdecayed on time scales of the order of ∼100 μs, they generallyappeared to be nonidentical, reflecting the distinct origins of thetwo types of signals. The decays of the TCFs should be sensitive tothe motions of the rigidly integrated probes, and thus of the DNAsugar–phosphate backbones, and hence likely reflect importantaspects of DNA breathing. To ensure that potential “flickering”dynamics of long-lived excited triplet states did not significantlyaffect our results (27), we performed control studies to check thatthe observed TCFs were independent of the excitation intensity ofthe laser (SI Text, Control Experiments). In Fig. 4, we show resultsfrom the fork-labeled DNA construct in the absence of helicaseproteins. Additional examples of data sets are presented in SI Text,Control Experiments, and show those of the same DNA constructin the presence of gp41 helicase hexamers that had been assem-bled using nonhydrolyzable GTPγS as the NTP ligand. Thesehelicase hexamers, assembled with nonhydrolyzable NTPs in theabsence of primase, bind weakly to the ss–dsDNA junction of thereplication fork constructs, but cannot catalyze duplex unwinding(5, 13, 14).Our experimentally derived TCFs are consistent with earlier

reports from ensemble fluorescence correlation spectroscopy(FCS) experiments performed by Altan-Bonnet et al. (7), inwhich DNA breathing dynamics were observed in the 30- to 100-μsrange. These authors proposed a model to explain the functionalform of the TCFs measured in their solution phase experimentsthat was based on a rate equation in which a DNA breathing“bubble” can open or close by 1 bp, with the closing rate muchfaster than the rate of opening. The TCFs produced using thismodel have a functional form

GðτÞ∝�1+

τ

2τc

�erfc

� ffiffiffiffiffiffiffiτ

4τc

r �−

ffiffiffiffiffiffiffiτ

πτc

re−τ=4τc : [4]

We have used Eq. 4 to fit the TCFs calculated from our single-molecule experiments, including those shown in Fig. 4, and inmany cases we found that this model fit our data very well. Ap-proximately 50% of the molecules we observed in the absence ofunwinding proteins fit this model precisely, whereas ∼25% of themolecules we observed in the presence of unwinding proteins alsofit the model. For those molecules in which there is a discrepancy,the model generally does not match the fastest time componentsof the decays, and therefore a refinement to the model may benecessary to capture the shortest time dynamics to which ourexperiments are sensitive. Because the closing rate of the DNAbubble should be faster than the opening rate, the relaxationtimes reported in ref. 7 and in our data can be considered torepresent the lifetime of the open state of the bubble. Themoderate agreement we observe between this model and ourexperimental data suggests the possibility that the chromophore-labeled region of the DNA construct may experience multiplebubbles of various sizes over the course of the observation timewindow. We note that the average lifetime we observed in oursystem is ∼4–5 times larger than that observed in ref. 7, which islikely due to the different labeling schemes used in the twostudies. The internally labeled iCy3 and iCy5 chromophores re-place native bases in our DNA strands, and this relatively rigidconfiguration might extend the lifetime of spontaneously formedbubbles in the local dsDNA sequence surrounding the probes. In

0 5 1 0 1 5 2 0 2 5 3 0

0.2

0.4

0.6

0.8

-0.5

0.0

0.5

1.0

100

200

300

400

500

600

00.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

time (sec)10 20 30 40 50 60

0 10 20 30 40 50 60

0 10 20 30 40 50 60

inte

grat

ed s

igna

lin

tegr

ated

sig

nal

inte

grat

ed s

igna

l

No.

Cou

nts

Cy3Cy5

FLDr

FRE

T

‘duplex-labeled’fork construct

A

B

C

Fig. 3. (A) Single-molecule iCy3/iCy5 signals were recorded during a T4 pri-mosome–helicase unwinding experiment. An unwinding event occurred at the17.5-s tick mark. (B) The simultaneously recorded smFLDr signal exhibitsa change in behavior coincident with the unwinding event defined by thesmFRET conversion efficiency shown in C. Horizontal dashed lines (red/green)indicate the magnitude of the fluctuations immediately before and after theunwinding event. (Right) Laser polarization ϕ-dependent signal distributionsthat were used to determine the FLDr signal shown in B. The donor–acceptorchromophore labels are placed deep within the double-stranded region of theDNA replication fork construct (SI Text, Sample Preparation and Model DNAReplication Fork Constructs).

Phelps et al. PNAS | October 22, 2013 | vol. 110 | no. 43 | 17323

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

contrast, the chromophores used in ref. 7 were attached usingflexible linkers outside of the DNA, leaving native bases intact, sothat the dynamics observed are likely perturbed by the enhancedmobility of the probe labels.To examine the effects of the presence and initial binding of

the helicase protein on DNA breathing, we defined the re-laxation time τc as the time required for the TCF to decay toa factor 1/e of its initial value. In Fig. 5A, we show histograms ofthe relaxation times (τc) obtained from the TCFs of the smFRETdata using DNA fork constructs that had been labeled eitherdeep in the duplex region (blue) or at the replication forkjunction (red). We characterized these data using the gammadistribution function f ðτcÞ= ½βαΓðαÞ�−1ðτcÞα−1e−τc=β, which is anappropriate distribution for the time domain. Here, the param-eters α and β describe the degree of skewedness and the widthof the distribution, respectively, and ΓðzÞ= R∞

0 tz−1e−tdt is thegamma function. We also present the corresponding rate domain(kc = 1=τc) histograms, which are fit to Gaussian distributions inSI Text, Control Experiments. In the absence of unwinding pro-teins (Fig. 5A), the duplex-labeled construct exhibits a slightlyfaster average decay time (blue, 201 μs) in comparison with the

fork construct (red, 289 μs). The faster decay time for the duplexsample suggests that the closing rate of a spontaneously formedbubble in the duplex region of dsDNA is faster than that ofa similarly formed unstable conformation near the ss–dsDNAfork junction. DNA breathing fluctuations have been shown tooccur much more frequently at replication fork junctions than atsequences positioned well within duplex DNA regions (5, 6). Therelative magnitudes of the lifetimes we have observed for thefork- and duplex-labeled DNA constructs are consistent withprevious observations (7).The relaxation times of equilibrium fluctuations observed in

our data can be dramatically lengthened by the introduction ofthe GTPγS-stabilized hexameric gp41 helicase, which binds weaklyto replication fork junctions. In this experiment we assembled thegp41 helicase (300 nM gp41, no gp61) using 6 μM nonhydrolyzableGTPγS rather than NTP. Under these conditions, the hexamerichelicase forms and binds weakly at the fork junction, as depictedin Fig. 1 A–C, but unwinds the dsDNA sequence by only one basepair (13, 14). As shown in Fig. 5B, we observed that the distri-bution of relaxation times for the DNA construct with chromo-phore labels placed near the replication fork junction weredramatically broadened (blue), with an average decay time of650 μs. Clearly, the presence of the gp41 helicase at the iCy3/iCy5-labeled replication fork junction extended the lifetimes ofopen (unpaired) DNA bubbles formed in that region, andbroadened their distribution. Furthermore, both the width andthe average magnitude of fluctuations at the replication forkincreased in the presence of the T4 helicase. In Fig. 5C, weplot the distribution of the magnitude of the relative fluctuationsof the smFLDr signal hS2i− hSi2, which can be used to characterizethe width of the distribution of open DNA fork conformations.Though both the duplex- and fork-labeled DNA constructs exhibita relatively narrow distribution of open conformations (Fig. 5C,Left and Center), the average magnitude and width of the fluc-tuations at the replication fork junction increases in the pres-ence of the T4 helicase (Fig. 5C, Right). The combination ofthese results for the helicase-dependent distributions of relaxationtimes and fluctuation magnitudes is consistent with the notion thatmicrosecond time-scale breathing fluctuations at and near the forkjunction likely play a significant role in helicase-binding mecha-nisms at this locus.

1

5

2.0

2.8

SM FRET TCF SM FLDr TCF

10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2

(sec) (sec)

c = 516 sec

c = 143 secGFR

ET(

) (×1

02 )

GFL

D(

) (×1

02 )

= 8.9 × 10-3

= 2.1 × 10-2

= 6.1 × 10-2

= 9.8 × 10-3

‘fork-labeled’construct

A B

Fig. 4. Examples of single-molecule FRET and FLDr TCFs. TCFs determinedfrom (A) FRET and (B) FLDr trajectories obtained from a single DNA modelfork construct in the absence of helicase proteins. Red lines represent opti-mized fits to Eq. 4, where the fit parameter hSi2 is the mean square mag-nitude of the signal fluctuations, hS2i is square average signal, and τc is thecorrelation time. The donor–acceptor chromophore labels are placed at thess–ds fork junction of the DNA construct (SI Text, Sample Preparation andModel DNA Replication Fork Constructs).

A B

C

Fig. 5. Dynamics of breathing fluctuations at thereplication fork in the absence and presence ofhelicase. (A) Histograms of relaxation times (τc)obtained from the analysis of smFRET/smFLD trajec-tories for DNA fork constructs, which were labeledwith iCy3/iCy5 placed deep in a duplex region (blue)or at the replication fork junction (red) in the ab-sence of helicase protein. (B) A comparison is shownfor a fork-labeled construct in the absence (red) andthe presence (blue) of the “frozen” hexameric heli-case (gp41 ∙ GTPγS)6 composed of 300 nM gp41 and6 μM GTPγS. (C) A comparison is shown of histo-grams of the relative magnitudes hS2i− hSi2 of thefluctuating smFLDr signal for duplex-labeled DNA(Left), fork-labeled DNA (Center), and fork-labeledDNA + (gp41 ∙ GTPγS)6 (Right). Histograms of therelaxation times were characterized using the gammadistribution function, given in the text, with skewed-ness and width parameters α (dimensionless) and β(in units of milliseconds), respectively.

17324 | www.pnas.org/cgi/doi/10.1073/pnas.1314862110 Phelps et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

DiscussionObservations of the local conformations of macromoleculesthrough smFRET and smFLD should provide insights into thefunctional mechanisms of the “protein machines” that manipu-late DNA. As a helicase processively unwinds a DNA duplex, thelocal conformation and flexibility of the opposing DNA strandsmust change considerably. Such motions will be reflected byfluctuations in the orientation and separation of fluorophores thatsubstitute for nucleotides within the sugar–phosphate backbone,and these motions can be simultaneously monitored using smFLDand smFRET.The smFLD method is a useful complement to smFRET be-

cause it can help to avoid misinterpretation of false smFRETsignals. For example, if the acceptor fluorophore of a coupleddonor–acceptor FRET pair was to undergo a photobleach event(permanent or temporary), there would be a drop in acceptorintensity coincident with a rise in donor intensity. An acceptorphotobleach event could thus be misinterpreted as a suddenseparation between the donor–acceptor pair. This ambiguity canbe addressed by simultaneously monitoring the smFLD signal,because a FRET conversion event due to donor–acceptor pairseparation will be accompanied by a significant change in thesmFLD signal, whereas an acceptor photobleach event will not.Our finding that the presence of the gp41 hexameric helicase

greatly enhances the widths and average values of both the life-time and the magnitude distributions of “open” conformations atthe fork junction implies that these thermally populated minoritystates of the DNA substrate play a role in the helicase-bindingmechanism. The significance of our results lies in the context ofunderstanding the energetic factors that contribute to the func-tional assembly of biological macromolecular machines. An openquestion is whether the initial steps of helicase–primase (primo-some) assembly occur at the replication fork junction througha cooperative mechanism in which the binding of the gp41 hex-americ helicase fundamentally alters the nature of DNA fluctu-ations at the fork junction and thereby facilitates the subsequentbinding of the gp61 primase during initial primosome helicaseassembly, as well as during the subsequent unwinding reaction.Alternatively, DNA breathing and helicase binding might not

influence one another, in which case the formation of the primo-some–DNA complex must reflect a random coincidence of events.Though the latter scenario can describe the chemical kinetics ofsmall molecular systems, which may be characterized by a station-ary potential energy landscape, our current results appear to sup-port the more complex former situation, in which a highly mutabletime-evolving energy landscape can respond to locally changingenvironments, including here the presence of the weakly boundhelicase complex at the DNA replication fork junction.

MethodsInstrumentation for smFLD and smFRET. A smFRET instrument was designed touse rapid modulation of the excitation beam. The detection system involvedcustom electronics and analysis to perform single photon-counting phase-sensitive techniques. Instrument details and theoretical considerations toperform these experiments are provided in SI Text, Instrumentation forSingle-Molecule Fluorescence-Detected Linear Dichroism and Single-Mole-cule Förster Resonance Energy Transfer.

Sample Preparation and Model DNA Replication Fork Constructs. Model DNAreplication fork constructs, which were labeled with the iCy3/iCy5 FRETchromophore pairs, were purchased from Integrated DNA Technologies.Microfluidic sample chambers were constructed from microscope slides andcoverslips. Details of the sample chamber construction, cleaning procedures,and reagents used are given in SI Text, Sample Preparation and Model DNAReplication Fork Constructs.

ACKNOWLEDGMENTS. The authors thank Shamir A. Kansakar, who partic-ipated in the Summer Program for Undergraduate Research at the Universityof Oregon during the summer of 2012, for his assistance in the preparationof samples and sample cells used in our smFLD and smFRET experiments;Clifford Dax for his technical assistance in developing the custom digitalelectronics used in our smFRET/smFLD detection system; Neil Johnson forhelpful conversations; and Steve Weitzel for providing the T4 gp41 and gp61preparations used in these studies. This work was supported by NationalInstitutes of Health/National Institute of General Medical Sciences GrantGM-15792 (to P.H.v.H.); Office of Naval Research Grant N00014-11-0193 (toA.H.M.), and National Science Foundation, Chemistry of Life Processes ProgramGrant CHE-1105272 (to A.H.M.). P.H.v.H. is an American Cancer Society ResearchProfessor of Chemistry.

1. Delagoutte E, von Hippel PH (2002) Helicase mechanisms and the coupling of heli-

cases within macromolecular machines. Part I: Structures and properties of isolated

helicases. Q Rev Biophys 35(4):431–478.2. Choi CH, et al. (2004) DNA dynamically directs its own transcription initiation. Nucleic

Acids Res 32(4):1584–1590.3. Peyrard M, Cuesta-López S, James G (2009) Nonlinear analysis of the dynamics of DNA

breathing. J Biol Phys 35(1):73–89.4. McConnell B, von Hippel PH (1970) Hydrogen exchange as a probe of the dynamic

structure of DNA. I. General acid-base catalysis. J Mol Biol 50(2):297–316.5. Jose D, Weitzel SE, von Hippel PH (2012) Breathing fluctuations in position-specific

DNA base pairs are involved in regulating helicase movement into the replication

fork. Proc Natl Acad Sci USA 109(36):14428–14433.6. Jose D, Datta K, Johnson NP, von Hippel PH (2009) Spectroscopic studies of position-

specific DNA “breathing” fluctuations at replication forks and primer-template junctions.

Proc Natl Acad Sci USA 106(11):4231–4236.7. Altan-Bonnet G, Libchaber A, Krichevsky O (2003) Bubble dynamics in double-

stranded DNA. Phys Rev Lett 90(13):138101.8. Nossal N (1994) Molecular Biology of Bacteriophage T4 (Am Soc Microbiol, Wash-

ington, DC), pp 43–53.9. Alberts BM (1987) Prokaryotic DNA replication mechanisms. Philos Trans R Soc Lond B

Biol Sci 317(1187):395–420.10. Richardson RW, Nossal NG (1989) Characterization of the bacteriophage T4 gene 41

DNA helicase. J Biol Chem 264(8):4725–4731.11. Dong F, Gogol EP, von Hippel PH (1995) The phage T4-coded DNA replication helicase

(gp41) forms a hexamer upon activation by nucleoside triphosphate. J Biol Chem

270(13):7462–7473.12. Dong F, von Hippel PH (1996) The ATP-activated hexameric helicase of bacteriophage

T4 (gp41) forms a stable primosome with a single subunit of T4-coded primase (gp61).

J Biol Chem 271(32):19625–19631.13. Jose D, Weitzel SE, Jing D, von Hippel PH (2012) Assembly and subunit stoichiometry

of the functional helicase-primase (primosome) complex of bacteriophage T4. Proc

Natl Acad Sci USA 109(34):13596–13601.

14. Lee W, Jose D, Phelps C, Marcus AH, von Hippel PH (2013) A single-molecule view ofthe assembly pathway, subunit stoichiometry, and unwinding activity of the bacte-riophage T4 primosome (helicase-primase) complex. Biochemistry 52(18):3157–3170.

15. Ha T, et al. (2002) Initiation and re-initiation of DNA unwinding by the Escherichia coliRep helicase. Nature 419(6907):638–641.

16. Weston KD, Goldner LS (2001) Orientation imaging and reorientation dynamics ofsingle dye molecules. J Phys Chem B 105:3453–3462.

17. Wei C-YJ, Lu CY, Kim YH, Vanden Bout DA (2007) Determining if a system is het-erogeneous: The analysis of single molecule rotational correlation functions and theirlimitations. J Fluoresc 17(6):797–804.

18. Ha T, Laurence TA, Chemla DS, Weiss S (1999) Polarization spectroscopy of singlefluorescent molecules. J Phys Chem B 103:6839–6850.

19. Beausang JF, Sun Y, Quinlan ME, Forkey JN, Goldman YE (2012) Orientation androtational motions of single molecules by polarized total internal reflection fluores-cence microscopy (polTIRFM). Cold Spring Harb Protoc, 10.1101/pdb.top069344.

20. Beausang JF, Shroder DY, Nelson PC, Goldman YE (2013) Tilting and wobble of myosinV by high-speed single-molecule polarized fluorescence microscopy. Biophys J 104(6):1263–1273.

21. Thomsson D, Lin H, Scheblykin IG (2010) Correlation analysis of fluorescence intensityand fluorescence anisotropy fluctuations in single-molecule spectroscopy of conju-gated polymers. ChemPhysChem 11(4):897–904.

22. Tubasum S, Cogdell RJ, Scheblykin IG, Pullerits T (2011) Excitation-emission polarizationspectroscopy of single light harvesting complexes. J Phys Chem B 115(17):4963–4970.

23. Widom JR, et al. (2013) Electronic transition moments of 6-methyl isoxanthopterin—a fluorescent analogue of the nucleic acid base guanine. Nucleic Acids Res 41(2):995–1004.

24. Nordén B, Rodger A, Dafforn T (2010) Linear Dichroism and Circular Dichroism: ATextbook on Polarized-Light Spectroscopy (RSC, Cambridge, UK).

25. Horowitz P, Hill W (1986) The Art of Electronics (Cambridge Univ Press, New York).26. Chandler D (1987) Introduction to Modern Statistical Mechanics (Oxford Univ Press,

New York).27. Fink MC, Adair KV, Guenza MG, Marcus AH (2006) Translational diffusion of fluo-

rescent proteins by molecular Fourier imaging correlation spectroscopy. Biophys J91(9):3482–3498.

Phelps et al. PNAS | October 22, 2013 | vol. 110 | no. 43 | 17325

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021