dna-binding orientation and domain conformation of the e. coli

DNA-binding Orientation and Domain Conformation ofthe E. coli Rep Helicase Monomer Bound to a PartialDuplex Junction: Single-molecule Studies ofFluorescently Labeled Enzymes

Ivan Rasnik1†, Sua Myong1†, Wei Cheng2, Timothy M. Lohman2 andTaekjip Ha1*

1Department of PhysicsUniversity of Illinois atUrbana-Champaign, UrbanaIL 61801, USA

2Department of Biochemistryand Molecular BiophysicsWashington University Schoolof Medicine, St. Louis, MO63110, USA

The SF1 DNA helicases are multi-domain proteins that can unwindduplex DNA in reactions that are coupled to ATP binding and hydrolysis.Crystal structures of two such helicases, Escherichia coli Rep and Bacillusstearothermophilus PcrA, show that the 2B sub-domain of these proteinscan be found in dramatically different orientations (closed versus open)with respect to the remainder of the protein, suggesting that the 2Bdomain is highly flexible. By systematically using fluorescence resonanceenergy transfer at the single-molecule level, we have determined boththe orientation of an E. coli Rep monomer bound to a 30-single-stranded-double-stranded (ss/ds) DNA junction in solution, as well as the relativeorientation of its 2B sub-domain. To accomplish this, we developed ahighly efficient procedure for site-specific fluorescence labeling of Repand a bio-friendly immobilization scheme, which preserves its activities.Both ensemble and single-molecule experiments were carried out,although the single-molecule experiments proved to be essential here inproviding quantitative distance information that could not be obtainedby steady-state ensemble measurements. Using distance-constrained tri-angulation procedures we demonstrate that in solution the 2B sub-domainof a Rep monomer is primarily in the “closed” conformation when boundto a 30-ss/ds DNA, similar to the orientation observed in the complex ofPcrA bound to a 30-ss/ds DNA. Previous biochemical studies haveshown that a Rep monomer bound to such a 30-ss/ds DNA substrate isunable to unwind the DNA and that a Rep oligomer is required for heli-case activity. Therefore, the closed form of Rep bound to a partial duplexDNA appears to be an inhibited form of the enzyme.

q 2004 Elsevier Ltd. All rights reserved.

Keywords: helicase; protein conformation; single molecule spectroscopy;fluorescence resonance energy transfer*Corresponding author

Introduction

Helicases unwind double-stranded (ds) nucleicacids to generate the single-stranded (ss) inter-mediates that are essential for many cellularprocesses.3 These enzymes also translocate along

nucleic acids using the chemical energy from thebinding and/or hydrolysis of ATP or other nucleo-side triphosphates and hence are motor proteins.As such, a helicase must couple its conformationalchanges resulting from ATP binding and hydro-lysis to DNA unwinding and translocation. Ideally,in order to probe the mechanisms of theseenzymes, one would like to measure real-timestructural changes of a helicase–DNA complexduring each reaction step. Single-molecule fluor-escence techniques4,5 are promising for thisgoal because they can detect the conforma-tional changes of individual bio-molecules with

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

† I.R. & S.M. contributed equally to this work.

E-mail address of the corresponding author:[email protected]

Abbreviations used: ss, single-stranded; ds, double-stranded; FRET, fluorescence resonance energy transfer;PEG, poly-ethylene glycol.

doi:10.1016/j.jmb.2003.12.031 J. Mol. Biol. (2004) 336, 395–408

millisecond time resolution under physiologicalconditions.

Previously, we combined the single-molecule flu-orescence approach and a surface immobilizationscheme which does not interfere with biologicalactivity to probe the DNA unwinding mechanismof Escherichia coli Rep helicase.2 Fluorescence reson-ance energy transfer (FRET) between fluorophoresattached to the DNA reported the structuralchanges induced by the action of the enzyme,allowing us to measure DNA unwinding of only afew base-pairs and to detect unwinding stalls,DNA rewinding, and re-initiation of unwindingby a stalled enzyme–DNA complex, and to deducetheir underlying mechanisms.2

Rep, E. coli UvrD and Bacillus stearothermophilusPcrA are 30 to 50 DNA helicases belonging to theSF1 superfamily and share extensive sequencesimilarity6 as well as high structural homology.6 – 8

Both Rep and PcrA consist of four major domainscalled 1A, 2A, 1B and 2B (Figure 1). Crystal struc-tures of Rep bound to ssDNA show two dramati-cally different Rep monomer conformations(“open” and “closed”) that differ from each otherby a large reorientation (1308 swiveling) of the 2Bdomain while the other domains remain essentiallyunchanged (Figure 1a). The crystal structure of aPcrA monomer in complex with a partial duplexDNA (dsDNA with a short 30 oligodeoxynucleotidetail) is found in the closed form while the structureof a PcrA monomer alone is in the open form. Sinceboth Rep and PcrA crystals were formed at highsalt concentrations, conditions under which DNAunwinding is not favored in vitro, and becausecrystal packing could affect the orientation of thehighly flexible 2B domain, it is not clear a prioriwhether a single conformation of the 2B domainoccurs under solution conditions that supportDNA unwinding, or whether both conformationsare accessible.

Both Rep and UvrD require more than a mono-mer to initiate DNA unwinding in vitro.1,2,9 Yet,monomers of both Rep and UvrD are able to trans-locate along ssDNA with a directional bias (30 to 50)(C. Fischer and T. M. L. et al., unpublished results),similarly to what has been shown for PcrAmonomers.10,11 In both open and closed forms inthe Rep crystal structures,6 the 30 end of ssDNA isoriented toward the 1A and 1B domains while the50 end is oriented towards the 2A domain (Figure1b). The same orientation for the 30 ssDNA tail isobserved in the complex of a PcrA monomerbound to a ss/ds DNA.7 Thus far, the preciseDNA-binding orientation in solution has not beendetermined for any non-hexameric helicase.

In order to detect structural changes of theenzyme and enzyme–DNA complex directly, flu-orescent probes can be attached to the protein site-specifically. Measurements using fluorescentlylabeled helicases have been reported for the E. coliDnaB, a hexameric helicase.12 The Rep helicase is agood model system for this type of fluorescencestudy since (i) its crystal structure is known, allow-

ing the rational choice of labeling sites, (ii) exten-sive biochemical information is available,3 (iii) itcan bind to DNA as a monomer, thus simplifyingthe interpretation of FRET data, and (iv) none ofthe five native cysteine residues are essential.Our goal is to combine site-directed mutagenesisand ensemble and single-molecule fluorescenceanalysis, leveraged by the crystallographic struc-tural information, to lay a solid foundation forexploring the structural transitions of protein–DNA complexes.

Figure 1. Rep–ssDNA structure and fluorescent label-ing sites. a, Rep structures in the open and closedforms. The four sub-domains of the Rep protein arecolored as follows. 1A (residues 1–84, 198–275) in yel-low, 1B (residues 85–197) in green, 2A (residues 276–373, 544–) in red, and 2B (residues 374–543) in blue. b,The residues on the Rep protein used for site-specificlabeling are indicated on the Rep monomer structure inthe closed form bound to a ssDNA (each Cys-lightmutant has only one cysteine at one of the eight sitesshown). The sites are marked with filled symbols if vis-ible in the view and with open symbols if invisible. Cir-cles indicate sites on the 2B domain (473 and 486),squares are sites predicted to be closer to the 30 end ofthe ssDNA (43, 97 and 233), and diamonds are sites pre-dicted to be closer to the 50 end of the ssDNA, hence tothe partial duplex junction (310, 316 and 333). Oligo-dT(only nine out of 16 bases in the crystal structure) isshown in a space-filling model and its polarity (30 versus50 ends) is marked.

396 DNA-binding Orientation of Rep Helicase Monomer

Results

Choice of protein labeling sites

The sites on the Rep protein chosen for site-directed mutagenesis were alanine or serine resi-dues that are not conserved among Rep, UvrDand PcrA, that are not contained within theknown helicase motifs, and that are well-exposedon the surface of the protein. Eight labeling siteswere chosen (Figure 1), distributed strategicallyamong all four sub-domains, two on domain 1A,one on domain 1B, three on domain 2A, and twoon domain 2B. Based on the Rep crystal structurein complex with (dT)16,

6 three of these sites are pre-dicted to be closer to the 30 end of the ssDNA (47,97 and 233), while three others are predicted to becloser to the 50 end of the ssDNA (310, 316, and333), and thus also predicted to be closer to thepartial duplex junction of the DNA used in thepresent studies. The two sites on the flexible 2Bdomain (473 and 486) were also chosen to deducethe orientation of the 2B sub-domain relative toboth the DNA as well as the remainder of the Repprotein (open or closed).

Cys-light Rep mutants retain helicase activityin vivo and in vitro

The site-directed mutants of Rep in which all ofthe native Cys residues have been replaced and asingle Cys residue has been introduced to a specificposition, referred to as Cys-light mutants, could beover-expressed in E. coli and were found in thesoluble fraction. Western blot analysis was used toconfirm that the expressed protein is the Rep pro-tein. Since E. coli Rep helicase is the only E. colihelicase that can support fX174 phagereplication,13 we used plaque assays with fX174phage in an E. coli strain that has a deletion for thewild-type rep gene and showed that all of theseRep mutants can support fX174 phage replication,and thus are active in vivo. Our labeling and purifi-cation scheme resulted in highly pure preparationsof all Rep mutants as checked by SDS-polyacryl-amide gel electrophoresis of the labeled mutants.The ssDNA ((dT)70)-stimulated steady-stateATPase activity was reduced by only 25% for thelabeled mutants (70 ATP hydrolyzed per Rep persecond for the wild-type and 50 ATP hydrolyzedper Rep per second for mutants). In addition, themultiple turnover DNA unwinding activities ofthe unlabeled Rep mutants, measured via fluor-escence using an 18 bp duplex containing a 30

(dT)20 tail (labeled by Cy3 and Cy5 at thejunction),2 were within 15% of the value obtainedwith wild-type Rep. Although under the steady-state unwinding conditions used the initiationrate rather than the actual unwinding rate is rate-limiting, we conclude that the steady-state ATPaseand DNA unwinding activities of the Cys-lightRep proteins are very similar to those of wild-typeRep.

Fluorescent labeling and purification arestreamlined and efficient

Labeling of the Cys-light proteins with fluoro-phores was performed while they were still boundto a Ni-NTA column. After the labeling reaction,any free fluorophores were removed via extensivewashing of the column with buffer A (50 mM Tris(pH 8.0 at 4 8C), 150 mM NaCl, 25% (v/v) glycerol).Additional removal of any remaining free fluoro-phores was achieved by a subsequent purificationstep using a ssDNA–cellulose column. The label-ing efficiencies were determined by comparing theprotein absorbance at 280 nm (extinction coefficientof Rep e ¼ 76; 800 M21 cm21) with the absorbanceof the fluorophores at their respective absorptionmaxima (extinction coefficients of Cy3 at 544 nmand Cy5 at 647 nm are 150,000 M21 cm21 and250,000 M21 cm21, respectively), after correctingfor the UV absorbance contributed by the fluoro-phores at 280 nm. The labeling efficiencies rangedfrom 85% to 95% with the average being 90%. Thesite-specificity of labeling was at least 90% becausecontrol reactions with Cys-free protein show label-ing efficiency below 10%. The single-moleculemeasurements described below further demon-strate that the labeling is highly specific for thesingle cysteine residues.

Ensemble solution experiments to determinethe orientational bias of labeled Rep proteinwhen bound to a partial duplex DNA

To test the binding of labeled Rep proteins to ourstandard DNA substrate used for helicase studies(an 18 bp duplex with a 30 ssDNA tail composedof (dT)19 or (dT)20; Figure 2), we performed bulksolution titrations in which acceptor (Cy5)-labeledRep protein was added in 1 nM (or smaller) incre-ments to a 5 nM solution of the donor (Cy3)-labeled DNA. As shown in Figure 2a, uponaddition of protein, the acceptor signal on the pro-tein increased while the donor signal on the DNAdecreased. The apparent FRET efficiency, calcu-lated as the ratio of the acceptor peak intensityand the sum of the peak intensities of the donorand acceptor, increased with increasing proteinconcentration and reached saturation at similarprotein concentrations for Rep protein labeled ateach of the three protein labeling sites (43, 333,and 473). Fitting the titration curves using a 1:1single-site binding model (Materials and Methods)yielded apparent equilibrium-binding constants of3.6(^0.2) nM for Rep protein binding to DNA inwhich the donor was at the end of the ssDNA tail(Figure 2b) and 5.0(^0.3) nM for DNA in whichthe donor was at the ss/duplex DNA junction(Figure 2c). Therefore the observed fluorescencesignal changes arise from Rep binding to theDNA. The binding affinity in the low nanomolarrange is consistent with our previous single-molecule measurements of Rep monomer bindingto a partial duplex DNA.2 We conclude that the

DNA-binding Orientation of Rep Helicase Monomer 397

DNA-binding properties of Rep are not signifi-cantly affected by the mutagenesis or fluorescentlabeling.

The apparent FRET efficiency observed at satur-ating protein concentrations varies among the Repproteins with Cy5 at the different positions. In par-ticular, Rep with Cy5 at residue 43 shows higherFRET than Rep with Cy5 at residue 333, when theCy3 donor is at the end of the 30 ssDNA tail,whereas the reverse is true if the Cy3 donor isplaced at the ss/duplex DNA junction. The relativedifferences in the FRET values among the experi-ments performed with fluorophores located atdifferent positions on either the protein and/orthe DNA are quite clear and were reproducible fortwo or more independent preparations of labeledproteins. These data suggest that residue 43 iscloser to the 30 end of the ssDNA tail while residue333 is closer to the ss/duplex DNA junction. Wetherefore conclude that Rep binds to a ss/duplex

DNA in solution with a bias toward the orientationobserved in the X-ray crystallographic structures.6,7

However, it is difficult to obtain absolute FRETefficiencies from bulk solution measurements dueto the likelihood of incomplete labeling and thepresence of inactive acceptor species as well asthe possibility of multiple-binding stoichiometries.Furthermore, the steady-state bulk solutionmeasurements here do not provide informationon the magnitude of the orientational bias. Forexample, if the protein binds 30% of the time withthe reverse orientation, we would not be ableto detect the minority species in a steady-stateensemble FRET measurement unless FRET valuesof these species are known a priori. We also notethat the gel-FRET technique,14 in which FRET ismeasured after electrophoretic separation of thedesired complexes, is not applicable here becausethe Rep–DNA complexes dissociate with a half-life of ,100 seconds (data not shown), which is

Figure 2. Ensemble FRET studies of fluorescent Rep monomer binding to a fluorescently labeled partial duplexDNA. a, Fluorescence emission spectra are shown for a partial duplex DNA with Cy3 donor labeled at the 30 end ofthe (dT)20 ssDNA tail (Cy3-DNA) at 5 nM concentration in buffer A and in the presence of 3 nM and 7 nM acceptor-labeled Rep protein (Cy5-Rep43: Rep with single cysteine at 43 position and labeled with Cy5). As Cy5-Rep43 isadded to the Cy3-DNA, the fluorescence emission peak of the donor near 565 nm decreases while the acceptor peaknear 668 nm increases. Accompanying cartoons illustrate the donor (D), either at the partial duplex junction or at theend of the 30 ssDNA tail, and the acceptor (A) on the Rep protein (shaded oval). b, Results of equilibrium titrations ofa partial duplex DNA with a Cy3 donor label (D) at the ss/duplex junction with Rep labeled with Cy5 (A) at a loneCys at residue 43, 33 or 473. Apparent FRET efficiency is plotted versus [Cy5-Rep] for the three different protein label-ing sites. The continuous curves show the non-linear least squares fits of these data to a one-site binding model. c,Apparent FRET efficiency versus [Cy5-Rep] for equilibrium titrations of a partial duplex DNA with a Cy3 (D) labelattached to the end of 30 ssDNA tail. The same three Cy5-labeled Rep proteins described in (b) were used for thesetitrations. The continuous curves show the non-linear least squares fits of these data to a one-site binding model.


far shorter than the time required for gel-electro-phoresis. We have therefore used single-moleculeFRET measurements to overcome these difficulties.Such measurements can probe fully labeled mono-mer–DNA complexes in less than one second.

Fluorescently labeled Rep protein bindsspecifically to DNA, not to the PEG surface

Surface immobilization of the DNA allows asingle-molecule observation to be made forextended times, allowing access to dynamic pro-

cesses that might not be observed otherwise. How-ever, a fundamental problem to be addressed ishow to avoid perturbations to the system resultingfrom surface immobilization. In particular, DNA-binding proteins that are basic may interactstrongly with the negatively charged quartz sur-face. To avoid this, we have used a poly-ethyleneglycol (PEG)-coated surface, which has beenshown to reject the non-specific surface adsorptionof proteins if it forms a dense coating.2,15,16 OurPEG surface includes a small fraction of PEG mol-ecules possessing end-modified biotin, thereby

Figure 3. Fluorescently labeled Rep protein binds specifically to DNA substrates attached to a PEG surface. a, Imagesof donor and acceptor channels show that non-specific binding of Rep protein (labeled with Cy3 donor at residue 43)to the PEG surface is negligible. Only weakly fluorescent impurities, with less than a third of the intensity of the fluoro-phores conjugated to the protein, are visible. b, Donor-labeled proteins bind efficiently to Cy5-acceptor-labeled DNAmolecules immobilized to the surface as shown schematically in (e), giving rise to hundreds of fluorescent spots inboth imaging channels. c, A donor channel image showing the result of non-specific binding of Cy3-donor-labeledRep to a bare quartz surface. d, A donor channel image showing the non-specific binding of donor-labeled proteinsto a BSA-coated quartz surface. e, Schematic illustration of single-molecule binding of a donor-labeled Rep protein toan acceptor-labeled DNA immobilized to a PEG surface. f, Examples of single-molecule binding events measuredwith Cy3-labeled Rep (1 nM) (labeled at a single Cys at position 43). Red and green curves represent the signals ofacceptor and donor, respectively. g, Examples of single-molecule binding events measured with Cy3-labeled Rep(1 nM) (labeled at a single Cys at position 333). No signal is detected before Rep binding because direct excitation ofthe Cy5 acceptor is insignificant at the excitation wavelength used (532 nm) and Cy3 donor-labeled Rep protein dif-fusing in solution contributes only to the background counts. Upon protein binding to the DNA, fluorescence signalsappear abruptly (marked by arrows). After the binding, either (i) the acceptor photobleaches, leading to donor signalincrease (top graph in (f)), or (ii) total signal disappears abruptly (bottom graph in (f) and top and bottom graphsin (g)).


allowing specific immobilization of a biotinylatedDNA molecule while minimizing non-specificadsorption of protein to the underlying surface.

The single-molecule FRET measurements wereperformed as follows. (1) The streptavidin-coatedPEG surface was prepared. (2) 1 nM donor-labeledRep protein in imaging buffer was added to thesample to test for non-specific binding. Typicallyless than ten fluorescent spots appeared in the50 mm £ 100 mm imaging area in the absence ofDNA (Figure 3a). (3) Biotinylated, acceptor-labeledDNA (10–100 pM in 10 mM Tris–HCl (pH 8.0),50 mM NaCl, 0.1 mg/ml BSA) was added. NoDNA binding was observed if the DNA lackedbiotin or if streptavidin was omitted. (4) 0.25–1 nM donor-labeled Rep in the imaging buffer wasadded, resulting in hundreds of fluorescent spotsin both donor and acceptor channels (Figure 3b).Movies of 5–60 seconds duration were obtained atten frames per second and analyzed to obtain thetime records of donor and acceptor fluorescencesignals of individual molecules.

On the PEG surfaces, at least 95% of the Rep pro-tein binding events were due to specific binding tothe DNA (see below). In contrast, the addition of a1 nM solution of the labeled Rep to a bare quartzsurface (Figure 3c) or a BSA-coated surface (Figure3d) resulted in very high degrees of non-specificsurface binding so that single molecules could notbe discerned as isolated spots. It was thereforeessential to use the PEG surface for our studies.

Direct observation of a single Rep monomerbinding to DNA

For single-molecule measurements, the proteinwas labeled with Cy3 (donor) and the DNA waslabeled with Cy5 (acceptor). This allows us toeasily distinguish a single monomer-binding eventfrom multiple monomer binding events. When aprotein binds to a DNA molecule in the middle ofa time record (illustrated in Figure 3e), a suddenappearance of fluorescence signals is observed(Figure 3f and g). On the average, about half thebinding events showed the donor-only signalswhile the other half showed both the donor andacceptor signals. It is highly unlikely that the bind-ing events showing FRET are due to non-specificbinding to the PEG surface, since the probabilitythat random binding of protein to the surfacewould occur within a 10 nm radius of an acceptor-labeled DNA is ,0.001% for the typical surfacedensities of DNA used. Control experimentswithout DNA showed a much reduced-bindingfrequency (Figure 3a). In addition, a large fractionof the donor-only binding events are likely due tofluorescently inactive acceptor species as has beencommonly observed in single-molecule FRETexperiments.17 Therefore, we conclude that thevast majority of the protein-binding events arespecific to DNA.

After the initial-binding event, one of two resultsfollowed. Either (i), the acceptor underwent photo-

bleaching, leading to an increase in donor signal(top panel in Figure 3f), or (ii), the total signal dis-appeared abruptly (bottom panel in Figure 3f andtop and bottom panels in Figure 3g). The averagetime until the disappearance of the total signalwas reduced at higher laser intensities and anindependent measurement showed that dis-sociation of Rep–DNA complexes is substantiallyslower (,0.01 s21) than the rate of loss of the totalsignal. Therefore, it is likely that the observationsof a loss of total signal are due to photobleachingof the donor fluorophore. At least 95% of thedonor photobleaching events were single-stepped,consistent with the signal originating from a singlefluorophore event. Since the labeling efficiency ofthe Rep protein is high (,90%), this suggests thatRep binds to the DNA predominantly as a mono-mer under these conditions. This also indicatesthat the probability of having more than one fluor-ophore attached to a single protein is less than10%. A similar analysis also showed that over 90%of the Rep protein bound to the DNA at any timeis monomeric. These results indicate that reliablestudies of Rep monomer binding to a single DNAmolecule are possible via single-molecule exper-iments of this type.

Single-molecule measurements suggest thatthe orientation of Rep binding to the DNAis definitive

The donor and acceptor intensities (ID and IA) ofindividual DNA molecules were obtained by aver-aging the first ten frames (one second) of timetraces after correction for the cross-talk betweenthe two signal channels. Single-molecule FRET effi-ciency E was calculated as E ¼ 1=ð1 þ hIA=IDÞ;where h is a parameter reflecting the relative quan-tum yields and relative detection efficiencies of thedonor and acceptor signals.18 The value of h wasdetermined experimentally as the ratio betweenthe absolute changes in IA and ID upon acceptorphotobleaching18 and is close to 1 in our exper-imental configurations. Therefore, the single-mol-ecule FRET efficiency E was determined fromE ¼ 1=ð1 þ IA=IDÞ; and its histograms are shown inFigure 4 for three protein labeling sites (the donorat 43, 333, and 473) and two DNA labeling sites(the acceptor at the end of the ssDNA tail or at thepartial duplex junction).

Approximately, half of the population of DNAmolecules was observed to cluster in a peak cen-tered at E , 0; representing the donor-only specieslikely due to the population of molecules whoseCy5 acceptor on the DNA is inactive. Negativevalues of E arise from signal fluctuations at theacceptor channel near the background level. Theacceptor-labeled DNA molecules are not detectableby themselves since the absorbance of Cy5 at theexcitation wavelength (532 nm) is approximately25 times lower than its peak absorption. Eachsingle-molecule FRET histogram in Figure 4 alsoshows an additional peak centered at a non-zero


value of E that arises from a Rep monomer bindingto a DNA. The positions of the peaks are clearlydifferent for the different labeling sites and theserelative differences in peak position were reprodu-cible over independent surface preparations andindependent preparations of labeled proteins.Furthermore, these differences in peak positionsamong the different labeling sites are fully consist-ent with the bulk solution measurements shownin Figure 2. We also note that single FRET peaksare observed from each histogram withoutadditional peaks that might occur if the reverse-binding orientation is significantly populated. Thissuggests that the fluorescent labeling is highlysite-specific and DNA binding occurs with a defi-nite polarity.

The width of the peaks in the histograms canarise from several sources. At the instrumentallevel, uneven sensitivity of different areas of thecamera, imperfect mathematical mapping betweentwo images (donor and acceptor channels), andintensifier noise are among the possible sources.

Statistical noise due to the limited number of pho-tons can also be significant in single-moleculemeasurements. Heterogeneities in the spectralproperties, quantum yields, and dipole orien-tations of the fluorophores can also contribute tothe width. At the molecular level, the presence ofseveral alternative protein-binding sites on theDNA is another possible source. Since Rep canbind non-specifically to both ssDNA and duplexDNA, but with higher affinity to ssDNA, there aremultiple potential non-specific-binding sites for aRep monomer on the 30-(dT)20 ssDNA region ofthe DNA substrate. It is also possible that theRep–DNA complex can exhibit multiple confor-mations that can be manifested in the broadeningof the FRET distribution or in the asymmetricpeaks seen in some cases. Finally, some of theaforementioned possible sources of noise may beaffected by the exact configuration in which aRep–DNA complex interacts with the localenvironment of the surface. Here, we do notattempt to de-convolute the relative contributionsof these potential sources. Instead, we focus on theaverage values of E determined by Gaussian fittingof the peaks in the histograms.

Binding orientation information is independentof the type of fluorophores used

In principle, the spectral properties or orien-tations of the fluorophores in the labeled Rep

Figure 4. Single-molecule FRET studies of fluores-cently labeled Rep monomer binding to fluorescentlylabeled DNA to determine the orientation of Rep boundto the DNA. FRET efficiency between a single Cy3-labeled Rep and a Cy5-labeled DNA was obtained byaveraging their fluorescence signals over one second(ten frames of 0.1 second duration). Histograms ofsingle-molecule FRET values were constructed fromhundreds of molecules for each condition. a, Cy3-Rep(labeled on single Cys at positions 43, 473 and 333) bind-ing to DNA with Cy5 at the partial duplex junction.b, Cy3-Rep (labeled on a single Cys at positions 43, 473and 333) binding to DNA with Cy5 at the 30 end of thessDNA tail. Each histogram shows two peaks; the peakat zero FRET is attributed to donor-only cases (likelyCy3 labeled Rep bound to DNA without an active Cy5).Negative FRET values are caused by noise and back-ground subtraction and the fact that the acceptor signalis at the background level. The non-zero peaks of FRETefficiencies show clear differences dependent upon thesite of Cy3 (acceptor) labeling on Rep protein as well asthe site of Cy5 (donor) labeling on the DNA. The relativeFRET values are consistent with the results of the bulksolution measurements shown in Figure 3.

Figure 5. The relative changes in FRET values amonglabeling sites are independent of the FRET pair used. Totest the effect of the particular FRET pair on the robust-ness of the dependence of the FRET values on the pos-ition of the Rep protein labeling sites, we tested thefollowing different FRET pairs. (i) One of three differentdonors (Cy3, Alexa555, or tetramethylrhodamine) wasattached to the protein with Cy5 as the acceptor on theDNA. (ii) The Cy3 donor and Cy5 acceptor locationswere switched (i.e. Cy3 was placed on the DNA andCy5 on the protein). Average FRET values were obtainedby fitting the non-zero FRET peaks as shown in Figure 4using Gaussian functions. The scatter of the FRET valuesfor each protein labeling site is significantly smaller thanbetween sites, independent of the FRET pair used.


protein could be influenced by the local environ-ment of each labeling site and this could lead toapparent variations in FRET efficiencies among thedifferent sites, for instance due to effects involvingdifferences in k2: In order to test the robustness ofthe binding orientation information, we switchedthe locations of the donor and acceptor fluoro-phores (i.e. placed the Cy3 donor on the DNA andthe Cy5 acceptor on the Rep), and also replacedthe Cy3 donor with other fluorophores possessingsimilar spectral properties (e.g. Alexa555 and tetra-methylrhodamine). Although the absolute valuesof the FRET efficiencies, E; varied slightly amongthe different fluorophore combinations, each pro-tein labeling site displayed a distinct clustering ofthe values of E; independent of the fluorophore(Figure 5). In addition, the monotonic decrease inFRET when the labeling site is moved from residue43 to 473 and then to residue 333 is observed for allfluorophore combinations (Figure 5). This resultstrongly supports the conclusion that the differ-ences in FRET observed for the different labelingsites reflect the relative changes in distancebetween sites on the protein and the DNA due tothe unique binding orientation of a Rep monomeron the DNA, rather than effects due to specificchanges in the properties of the fluorophoresthemselves that are dependent on the labeling site.

Triangulation shows that Rep bound to apartial DNA duplex is in the closed form

To obtain more detailed information on the bind-ing geometry between Rep and DNA, we performedadditional single-molecule FRET experiments. Inthese experiments, Rep monomers were labeledwith Cy3 at each of the eight labeling sites on theprotein, as shown in Figure 1b (eight distinctRep proteins with one cysteine at each of the eightpositions) and the DNA was labeled with Cy5 atthe ss/duplex DNA junction. The resulting aver-age FRET values, widths of the FRET peaks in thehistograms as well as the fluorescence quantum

yields of each Cy3 donor and fluorescence polariz-ation anisotropy are presented in Table 1.

Using these data, we wished to determinewhether the Rep monomer was bound to the ss/duplex DNA principally in either the open orclosed conformation. To estimate the location ofthe partial duplex junction we performed the fol-lowing triangulation procedure.

(1) To account for the flexible fluorophore lin-ker, we assumed that the donor fluorophore onthe protein is displaced from the correspondingCys residue by 5 A in the direction movingaway from the protein center. Residue locationswere obtained from the structural coordinatefile (Protein Data Bank Id: 1uaa).

(2) We used a biased random walk of the junc-tion location ~r to minimize the function P ¼P

i ðEi 2 Ecal;iÞ2 where Ei is the experimentally

determined FRET efficiency for residue i and

Ecal;i ¼ 1 þl~r 2 ~ril

R0;i

� �6" #21

is the calculated FRET efficiency between theacceptor fluorophore (Cy5) at the junctionlocation ð~rÞ and the donor fluorophore (Cy3) atthe residue i on the protein ð~riÞ: R0;i is the Forsterradius calculated for residue i; which rangedbetween 61 and 65 A (see Materials andMethods). Three-dimensional biased randomwalks of ~r with a step size of 0.5 A were per-formed using MATLAB to minimize P, startingfrom random initial positions chosen within a200 A3 volume surrounding the protein.Approximately, half of the trajectories weretrapped in local minima and only the trajectoriesthat resulted in the global minimum were con-sidered. Averaging ten such runs yielded an esti-mate for the location of the acceptor, and thusalso for the junction.

We first assumed that the Rep monomer wasbound in the closed conformation and used thecrystal structure of the closed form of Rep to pre-dict the position of the Cy5 on the DNA substrate.For the closed form of Rep, triangulations usingall eight constraints (two from 1A, one from 1B,three from 2A and two from 2B) yielded the Cy5location shown in Figure 6. The model is welldefined because the standard deviation about themean position is only 0.9 A for ten runs. The sametriangulation process but using only six constraints(excluding the two constraints from the fluoro-phores within the 2B domain) gave a mean Cy5acceptor location that was displaced only by 5.5 A.If instead, we assume that Rep was bound to theDNA in the open conformation, then the predictedCy5 acceptor location estimated using all eightconstraints differs by 66 A from that estimatedusing the six none-2B constraints. These resultsare therefore consistent with the Rep monomer

Table 1. Single molecule FRET analysis of Rep–DNAcomplex

Cy3labelingsite

FRET to Cy5at the junction Width

Quantumyield (%)

Fluorescenceanisotropy

43 0.33 0.19 39 0.1997 0.38 0.3 37 0.3233 0.31 0.27 47 0.27310 0.55 0.24 48 0.24316 0.58 0.3 40 0.3333 0.68 0.17 41 0.17473 0.47 0.31 31 0.31486 0.71 0.23 31 0.23

FRET histograms for the eight different protein sites as shownin Figure 5 were fitted using Gaussian functions and the valuesfor the center and width are listed. Also shown are the valuesfor the quantum yield and fluorescence anisotropy of Cy3attached to the protein.


being bound to the DNA in the closed form andinconsistent with the open form.

Based on these measurements and calculations,it appears that Rep monomer exists in a confor-mation more similar to the closed form than to the

open form when it is bound to a DNA unwindingsubstrate in solution in vitro. We also note thatthese conclusions are insensitive to the valuesused for parameters such as linker length (5 Aversus 10 A) and R0 values except for slight changesin the numeric values above. However, we cannotrule out minority populations (,10%) of the openconformation or other conformations that maydiffer from the crystal structures, since the triangu-lation procedure is not sensitive to the presence ofsuch small populations.

Discussion

Quantitative FRET measurements can be used tobuild models of macromolecular complexesthrough triangulation procedures.19 Single-mol-ecule measurements are useful for the quantitativedetermination of FRET efficiencies because mul-tiple species may exist in bulk solution measure-ments, which can complicate the analysis.17,20,21

Furthermore, gel FRET methods14 that measureFRET after electrophoretic separation of the desiredcomplexes are not practical if the complex dis-sociates on a time scale shorter than that of theelectrophoresis experiment.

In order to observe single-molecule fluorescencesignals through extended periods of time, it isnecessary to immobilize the molecules while pre-serving their biological activity. Previously, wehave shown that the use of a PEG surface coatingefficiently prevents non-specific interactions ofDNA and proteins with quartz surfaces, therebyreproducing bulk solution kinetics.2 Here, we havedirectly shown that fluorescently labeled proteinsdo not adsorb to the PEG surface even thoughthey strongly bind non-specifically to untreated orBSA-coated quartz surfaces. DNA immobilizationdoes not appear to affect Rep’s ability to bind tothe DNA with a proper orientation because therelative FRET efficiencies among the different label-ing sites determined from the single-moleculestudies are fully consistent with bulk solutionmeasurements. Cysteine engineering and fluor-escent labeling did not significantly affect the enzy-matic activities of Rep as we have tested these bothin vivo and in vitro. Our data also provide directevidence that at the low nanomolar Rep proteinconcentrations used, most of the binding events tothe (dT)20-tailed DNA reflect individual Rep mono-mers rather than multiple monomers binding tothe same DNA. This is consistent with previousstudies, which indicated that an occluded regionon a ssDNA by high-affinity binding of Rep isapproximately 15 nucleotides22 and our single-molecule studies which indicated that only amonomer binds to the DNA at low nanomolarconcentrations.2

In the use of FRET measurements to estimatedistances between fluorophores bound to biomole-cules, the most significant unknown is the orienta-tional factor, k2: In the triangulation procedures

Figure 6. Position of the partial duplex junction rela-tive to the Rep structure. Two views, rotated 90 degreesrelative to each other, are shown of the closed form ofthe Rep monomer bound to ssDNA (from the Rep–ssDNA crystal structure) indicating the location of theCy5 acceptor (red dot) that is attached to the ss/dsDNAjunction of the partial DNA duplex, as estimated fromthe triangulation procedures described in the text.


used to determine whether the 2B domain is in theopen or closed conformation, we assumed thatk2 ¼ 2=3: This is clearly a crude approximationsince the fluorophores displayed relatively largefluorescence anisotropy and the assumption ofk2 ¼ 2=3 is strictly valid only if both fluorophoresare rotating freely relative to the host moleculewithin their fluorescence lifetimes. Nevertheless,dipole orientations are not expected to be a majorconcern in this study that used eight different con-straints since such over-sampling would reducethe contributions of potential errors that arisefrom the orientational effects. The fact that weobtained consistent results for various fluorophorecombinations also suggests that the informationobtained here is robust regardless of how a particu-lar fluorophore interacts with the DNA or protein.

Our studies indicate that a Rep monomer bindsto ssDNA with a definite polarity in solution. Pre-vious fluorescence studies have shown that a Repmonomer binds to ssDNA with a directional biasbut the binding orientation could not be deter-mined since only a single probe was used.23 Allthree protein–DNA crystal structures (two forRep,6 one for PcrA7) indicate the same bindingorientation for the ssDNA. Our ensemble FRETdata show that there is an orientational bias inDNA binding consistent with the crystal struc-tures. Single-molecule FRET data show that thebias is very strong and, within our resolution, theprotein binds to the DNA with a definitive polarity.Similar measurements using single-stranded DNAin bulk solution indicate that the same orienta-tional bias is maintained (data not shown); hencethe binding orientation is unlikely to be influencedby the presence of the duplex.

Crystal structures of the Rep protein bound tossDNA ((dT)16) show the Rep monomer in two dis-tinct conformations, referred to as closed andopen.6 The two forms differ primarily by a largereorientation (1308 swiveling) of the 2B sub-domainabout a hinge region connecting it to the 2A sub-domain. Rep binding to ssDNA increases the sensi-tivity of the hinge region to trypsin cleavage,suggesting that some movement of the 2B sub-domain is coupled to ssDNA binding.24 The openform of Rep has its sub-domains arranged in asimilar configuration as is observed in the crystalstructure of the apo PcrA monomer (withoutDNA),8 whereas the closed form of Rep moreclosely resembles the form of PcrA observed inthe crystal structure of a complex with a partialduplex DNA.7 The studies reported here weredesigned to test if a Rep monomer binds to a par-tial duplex DNA in a unique conformation (closedor open) or whether it can sample manyconformations.

The triangulation procedures that were based onsix or eight distance constraints allowed us todeduce the average location of the fluorophoreattached to the DNA junction with respect to thesite-specific fluorophores on the bound Rep pro-tein. The estimated location of the partial duplex

junction is in the general vicinity of the junctionseen in the PcrA structure, although the agreementis not exact. If the partial duplex junction bound toRep were positioned as in the PcrA crystal struc-ture, this would predict a near 100% FRET signalwhen Cy3 donors are attached to residues 310, 316and 333, whereas we observed much smallervalues of FRET. This difference suggests that aRep monomer does not bind precisely at the junc-tion for the ssDNA tail lengths used in our experi-ments. This might arise from the difference in thelengths of the ssDNA tails used in the two studies.In contrast to the (dT)20 tail used in our studies, theDNA in the PcrA structure had only a (dT)7 ssDNAtail, which would be expected to constrain the pro-tein to bind closer to the partial duplex junction. Incontrast, Rep does not show specificity for bindingto the ss/duplex DNA junction1 and thus Repwould be expected to sample the multiple bindingsites available along the ssDNA tail. If this is thecase, then the junction location estimated in ourstudies should be viewed as an average over thesemultiple binding sites.

Performing the triangulation procedures for bothclosed and open forms as they appear in the Repcrystal structures allowed us to deduce that theconformation of a Rep monomer bound to theDNA substrate is most similar to the closed form.This represents the first determination of the orien-tation of the 2B domain for a DNA bound SF1 heli-case in solution. Both the Rep and PcrA crystalswere formed at much higher salt concentrations(100–150 mM NaCl) than those that favor DNAunwinding in vitro, and because crystal packingcould affect the orientation of the highly flexible2B domain, it was not a priori clear which confor-mation would be preferred under solution con-ditions that support DNA unwinding. However, itis important to note that a Rep monomer is unableto initiate1 or sustain DNA unwinding in vitro.2

Rather, Rep must oligomerize in order to initiateDNA unwinding in vitro.1 Since we used the samesolution conditions as in the previous Rep–DNAunwinding experiments,1,2 the 2B domain orien-tation that we determined here corresponds to theconformation of a Rep monomer bound to a partialduplex DNA in the absence of ATP. This would bethe typical initial state that exists in single-turnoverDNA unwinding kinetics experiments when theDNA substrate is in excess over the Rep protein.However, under these conditions, single-turnoverDNA unwinding experiments do not detect evenpartial unwinding of duplex DNA.1 Therefore, theclosed form of a Rep monomer bound to a partialduplex DNA, which largely resembles the PcrAcrystal structure, likely represents an inhibitedform of the enzyme.

Recent studies have shown that a Rep deletionmutant that lacks the 2B domain can still functionas a helicase, both in vitro and in vivo,25 hence the2B sub-domain is clearly not required for Rep heli-case function. In fact, the 2B deletion mutant ofRep is more efficient as a helicase in vitro than is


the full length wild-type Rep,25 suggesting furtherthat the 2B domain has an inhibitory effect onunwinding. An interesting hypothesis then is thatthe interaction of an additional Rep monomer(s)with the inhibited Rep monomer bound to DNAin the closed form may be necessary to activatethe Rep protein to initiate DNA unwinding.25

Future experiments using doubly labeled Rep pro-teins may be able to test this hypothesis directly.

Conclusions

We have developed reliable procedures formeasuring interactions between the E. coli Repprotein and DNA at the single molecule level. Theorientation of an SF1 monomer with respect to theDNA backbone polarity has been determined insolution for the first time and the orientation ofthe flexible 2B domain of Rep has been deducedto be in the closed form when bound to a partialduplex DNA. Future studies of more labeling sitesand of doubly labeled proteins at different stepsduring the ATP hydrolysis cycle should be able toelucidate overall structure changes of the Rep–DNA complex along the reaction pathway.

We anticipate that the approaches outlined herecan be used to probe many outstanding questionsabout helicase mechanisms in general, and Rephelicase in particular. Why do many helicases,including Rep, require oligomerization (dimer, tri-mer, hexamer,…) to function? What determinesthe directionality in DNA unwinding and trans-location? How does the helicase couple confor-mational changes induced by ATP hydrolysis foruse in DNA unwinding and translocation? Futureexperiments using fluorescently labeled helicasescan be designed to detect conformational changesof the enzyme and DNA unwinding simul-taneously from the same molecule, thereby directlyprobing the heart of the structure–functionrelationship.

Materials and Methods

Buffers

Lysis buffer is 50 mM Tris (pH 8.0 at 4 8C), 200 mMNaCl, 15% (v/v) glycerol and 20% (w/v) sucrose. BufferA (for Ni-NTA column) is 50 mM Tris–HCl (pH 8.0 at4 8C), 150 mM NaCl, 25% glycerol. Buffer B (Ni-NTAelution buffer) is buffer A plus 100 mM imidazole. BufferC (for ssDNA cellulose column) is 50 mM Tris (pH 7.5 at4 8C), 100 mM NaCl, 2 mM EDTA, 20% glycerol. BufferC-1 M (ssDNA–cellulose elution buffer) is buffer C plus1 M NaCl. Buffer C-3 M (ssDNA–cellulose wash) is buf-fer C plus 3 M NaCl. The final purified protein wasstored in 30 mM Tris (pH 7.5 at 4 8C), 600 mM NaCl,1.2 mM EDTA, 50% glycerol at 220 8C.

Site-directed mutagenesis

Site-directed mutagenesis was performed using the

“QuikChange” technique (Stratagene) on pGG236, aplasmid containing the rep gene under control of the tacpromoter. All five native cysteine residues in the Repprotein were replaced with the indicated amino acids(C18L, C43S, C167V, C178A, C612A) using this approach.Cysteine residues at positions 18 and 167 were changedto leucine and valine (corresponding to the residuesfound in E. coli UvrD and B. stearothermophilus PcrA),respectively, because protein solubility was lowered con-siderably when they were replaced with serine. None ofthe original cysteine residues in Rep are conservedamong Rep, UvrD and PcrA. After each round of muta-genesis, the DNA sequence for the entire Rep gene wasconfirmed and Rep activity was tested in vivo using aplaque assay for fX174 phage25 in an E. coli strainCK11Drep/pIWcI that lacks Rep gene.26 The open read-ing frame (ORF) of the cysteine-free mutant was digestedwith restriction enzymes (Nde I and Xho I), and ligatedinto the plasmid pET28a (Novagen) that carries anN-terminal hexa-histidine tag to generate an expressionvector. Using this plasmid, pRepNoCys that carries thecysteine-free Rep gene as a background, we made plas-mids encoding mutant rep genes encoding only a singlecysteine residue at eight different positions, pRepS43C,pRepA333C, pRepS316C, pRepA310C, pRepS233C,pRepA97C, pRepA473C, pRepS486C, thereby generatingeight different Rep mutants containing only one cysteineresidue each.

Protein purification and labeling

The plasmids encoding each Rep mutant wereeach used to transform BL21(DE3) (Novagen, 69864-3)and colonies were selected on agar plates containing30 mg/ml kanamycin (Sigma, St Louis, MO). Singlecolonies were picked and grown overnight at 37 8C inLB broth with kanamycin (30 mg/ml). This culture(1 ml) was used to inoculate 100 ml of fresh LB medium,and incubated at 30 8C. The growth temperature waslowered to 30 8C to improve solubility of the Rep protein.Cell growth was monitored by optical density at 600 nm.When OD600 ¼ 0:8; expression of the Rep protein wasinduced by the addition of 300 mM IPTG (isopropyl-b-D-thiogalalctopyranoside). Following induction, the cellswere grown for an additional three hours and harvestedby centrifugation. The following purification andlabeling of the Rep mutants were performed at 4 8C.Cell paste (1 g) was resuspended in 5 ml of lysis buffer.PMSF was added to 0.1 mM and lysozyme to 0.4 mg/ml.After one hour of gentle stirring, the cell suspensionwas subjected to pulsed sonication on ice for threeminutes (70% duty cycle, power setting of 7). The celllyzate was centrifuged at 14,000 rpm for one hour. Thesupernatant was mixed with 500 ml of Ni-NTA agarosecolumn (Qiagen) for one hour by gentle stirring at 4 8C.The protein-bound Ni-NTA agarose was then pouredinto an empty column. The column was washed with15 column volumes of buffer A. The column was treatedwith Tris(2-carboxyethyl)phosphine (TCEP) (100 mM) toreduce any disulfide bonds that might have formed.The column was then washed with 15–20 columnvolumes of buffer A. The appropriate fluorophore (Cy3-maleimide or Cy5-maleimide (Amersham Biosciences,Piscataway, NJ) or Alexa555-maleimide and tetra-methylrhodamine-maleimide (Molecular Probes, Eugene,OR)) was then added to the Ni-NTA column at a tenfoldmolar excess over the protein concentration in buffer Aand incubated overnight at 4 8C, while slowly rotatingthe sample container. Free fluorophores were removed


by washing the column with buffer A (10–15 columnvolumes). The labeled protein was then eluted from thecolumn with buffer A containing 100 mM immidazole.The eluted protein was loaded onto a single-strandedDNA–cellulose column (USB, Cleveland, Ohio) pre-equilibriated with buffer A and eluted with elution buf-fer (buffer C-1M). The concentrations of Rep and eachfluorophore in the final preparation were determinedspectrophotometrically by recording spectra from240 nm to 700 nm. Spectral readings revealed that thefractions eluted from the Ni-NTA column containedsome free fluorophores (5–10%), which were removedby the ss-DNA–cellulose column. The final labeling effi-ciency was generally above 90%. The protein was storedin 30 mM Tris (pH 7.5 at 4 8C), 600 mM NaCl, 1.2 mMEDTA, 50% glycerol at 220 8C.

Oligonucleotides

Sequences used for binding studies were 50-Cy5(orCy3)-GCCTCGCTGCCGTCGCCA-30-Biotin and 50-TGGCGACGGCAGCGAGGC(T)20-30 for a partial duplexDNA with the fluorophore at the junction and 50-GCCTCGCTGCCGTCGCCA-30-Biotin and 50-TGGCGACGGCAGCGAGGC(T)19 -Cy5(or Cy3)-T-30 for a partial duplexDNA with the fluorophore at the end of the 30 end ofthe single-stranded tail. Bulk unwinding test was doneusing 50-Cy5-GCCTCGCTGCCGTCGCCA-30-Biotin and50-TGGCGACGGCAGCGAGGC-Cy3-(T)20-30. Oligonu-cleotides were synthesized and purified as described.1

Cy3 and Cy5 were incorporated in phosphoramiditeforms and biotin was added as BiotinTEG CPG (allthree from Glen Research, Sterling, VA) during auto-mated synthesis. The DNA was gel-purified andannealed in 10–30 mM concentrations in 20 mM Tris–HCl (pH 8.0), 500 mM NaCl.

ATPase assay

Steady-state ATPase activities of Rep mutants and thewild-type were measured using the EnzChek phosphateassay (cat. no. E-6646) (Molecular Probes, Eugene, OR).The kit enables a continuous spectrophotometric moni-toring of reactions that generate inorganic phosphate.Each reaction is carried out by mixing DNA substrate((dT)70, 1 mM), MESG (2-amino-6-mercapto-7-methylpur-ine), buffer U, 1.5 mM ATP, purine nucleoside phos-phorylase, and 10–20 nM mutant Rep protein. ATP wasadded last after incubating the mixture at room tempera-ture for ten minutes. OD at 360 nm (OD360) was moni-tored for five minutes after addition of ATP. Enzymaticconversion of MESG results in a spectrophotometricshift in the peak of maximum absorbance from 330 nmto 360 nm. The number of ATP molecules hydrolyzedper Rep per unit time is calculated from the initial linearincrease in OD360.

Ensemble fluorescence measurements

Ensemble solution measurements were performedusing an Eclipse fluorometer (Varian) in buffer U(10 mM Tris–HCl (pH 7.4), 6 mM NaCl, 1.7 mM Mg2þ,10% glycerol, 1% (v/v) beta-mercaptoethanol, 0.1 mg/ml BSA) in a total volume of 1 ml. For FRET measure-ments, donor-labeled DNA was first added to 5 nM con-centration, then acceptor-labeled Rep protein was addedat 1 nM (or smaller) increments until saturation. Emis-sion spectra were measured upon excitation at 532 nm

and corrected using the pre-determined spectralresponse function of the fluorometer. The apparent equi-librium-binding constant KD was estimated by fitting theFRET versus [Rep] curve to the following equation,Eapp ¼ E0 þ ðEmax 2 E0Þ½Rep�=ð½Rep� þ KDÞ where Eapp isthe apparent FRET signal, E0 is the FRET value in theabsence of protein and Emax is the FRET value at saturat-ing protein concentrations. Fluorescence anisotropy ofprotein-conjugated fluorophores was measured using5 nM solution of labeled proteins. Fluorescence aniso-tropy of Cy5-labeled DNA was measured in 20 nM sol-ution. The quantum yields of protein-conjugated Cy3were measured in 5 nM concentration using rhodamine101 as a standard. Unwinding of 18mer duplex with(dT)20 tail was measured via FRET with the DNA concen-tration of 1 nM and 0.5 mM ATP.

Determination of R0

R0 is defined according to:27

R0 ¼9ðln 10ÞFDk2JðnÞ

128p5NAn4

� �1=6

;

where FD is the quantum yield of the donor, JðnÞ is thespectral overlap between the donor’s emission and theacceptor’s absorption, NA is the Avogadro’s number, nis the index of refraction of the medium, and k2 is deter-mined by the relative orientation of the two dipolemoments. We assumed k2 ¼ 2/3 but this is only anapproximation because the fluorescence anisotropy ran-ged from 0.2 to 0.3 for protein-conjugated Cy3 and was0.3 for the Cy5 attached to DNA, indicating limitedrotational motion within the radiative lifetime. All par-ameters except k2 were experimentally measured andused to calculate R0 for each pair of fluorophores. In par-ticular, FD was measured for each Rep mutant singlylabeled by Cy3 and ranged from 0.27 to 0.47, and corre-sponding R0 against Cy5 at the partial duplex junctionranged from 61 A to 65 A.

Surface and sample cell preparations

To minimize non-specific adsorption of proteins andmaximize specific binding of biotinylated DNA mol-ecules, we prepared PEG surfaces that contain a smallfraction (,1%) of biotinylated PEG molecules using thefollowing procedure. First, the glass coverslips andquartz slides were soaked in a 1% (v/v) solution of anamino-silane reagent (Vectabond, Vector Labs.) inacetone and then soaked for three hours with a PEGsolution, containing 25% (w/w) M-PEG-SPA Mr 5000(Nektar Therapeutics) and 0.25% (w/w) biotin-PEG-SPA,Mr 3400 (Nektar Therapeutics) in 0.1 M sodium bicar-bonate (pH 8.3). Once the surfaces were coated, weformed a sample cell by putting the PEG-coated side ofthe coverslip over the PEG-coated side of the quartzslide separated by a 100 mm thick spacer (3M doublesided tape). Two 0.75 mm diameter holes were drilledinto the quartz slide to form the inlet and outlet.Remaining boundaries between the sample cell and out-side were sealed using epoxy. This technique has theadvantage that many experimental conditions can beexplored by flowing different solutions through thesame sample cell. Small holes minimize evaporationduring prolonged measurements and reduce oxygenuptake by the solution thereby reducing photobleachingeffects. Once the sample cells are constructed they can


be kept in a dry environment for up to one week with-out degradation.

Single-molecule measurements

A wide-field total internal reflection fluorescencemicroscope based on an inverted microscope (IX70,Olympus) was used to image an area of50 mm £ 100 mm to an intensified CCD camera (iPentamax,Roper, operating at ten frames per second). Moleculeswere excited using a doubled Nd:YAG laser (532 nm,Crystalaser, power 10–20 mW at the sample plane)through a quartz prism placed over a quartz slide via athin layer of immersion oil. The incident angle of thelaser was controlled to achieve the total internal reflec-tion at the interface between the quartz slide and aqu-eous imaging buffer. Fluorescently labeled DNAmolecules are attached to this interface and the exci-tation intensity decays exponentially from the interfaceso that background fluorescence arising from the fluoro-phores in solution can be minimized. The laser light waslinearly polarized for the data presented but signalswere not dependent on the laser polarization, indicatingrelatively free rotation of the fluorophores in the labora-tory frame. Fluorescence signal was collected using awater immersion objective (Olympus; 60 £ , 1.2 numeri-cal aperture). After rejecting the scattered laser lightusing a long pass interference filter at 550 nm (Chroma),the imaging area was defined using a vertical slit (width3 mm) located at the imaging plane of the microscopejust outside the left side port. The emission is sub-sequently collimated using a 12 cm focal length achro-mat lens (Oriel), is split by a long pass extendedreflection dichroic mirror at 635 nm (Chroma), is recom-bined using an identical dichroic mirror after reflectingoff a mirror each, and is finally imaged onto the CCDusing a 24 cm focal length achromat lens (Oriel) therebyachieving 2 £ magnification. Donor and acceptorimages are laterally displaced by tilting the mirrors sothat each image occupies one half of the camera. Anindependent calibration of the detection system usingimmobilized, fluorescent beads allows the constructionof a mathematical mapping between the two imagingchannels. Each fluorescent spot in the donor channelhas a corresponding one in the acceptor channel, com-ing from the same spot in the sample. Ideally, one chan-nel would map onto the other by a pure translation butaberrations introduced by the optical system make themapping algorithm more complex. In a typical samplewe can observe about 200 molecules in the imagingarea. All single-molecule measurements were performedin buffer U with an oxygen scavenger system (0.1 mg/ml glucose oxidase, 0.02 mg/ml catalase, 1% b-mercap-toethanol and 0.4% (w/w) b-D-glucose) to increase thephoto-stability of the fluorophores. All data wereacquired using software written in Visual Cþþ(Microsoft)and analyzed using programs written in IDL (Researchsystems) to obtain time records of donor and acceptorfluorescence intensities of single molecules as a functionof time.

Acknowledgements

We thank the National Institutes of Health(T.H. (GM65367) and T.M.L. (GM45948)) and the

National Sciences Foundation (T.H.) for support ofthis work. We also thank Dr Katherine Brendzafor performing single-turnover DNA unwindingexperiments with some of the Rep mutants.

References

1. Cheng, W., Hsieh, J., Brendza, K. M. & Lohman,T. M. (2001). E. coli Rep oligomers are required toinitiate DNA unwinding in vitro. J. Mol. Biol. 310,327–350.

2. Ha, T., Rasnik, I., Cheng, W., Babcock, H. P., Gauss,G., Lohman, T. M. & Chu, S. (2002). Initiation andreinitiation of DNA unwinding by the Escherichiacoli Rep helicase. Nature, 419, 638–641.

3. Lohman, T. M. & Bjornson, K. P. (1996). Mechanismsof helicase-catalyzed DNA unwinding. Annu. Rev.Biochem. 65, 169–214.

4. Weiss, S. (1999). Fluorescence spectroscopy of singlebiomolecules. Science, 283, 1676–1683.

5. Moerner, W. E. & Orrit, M. (1999). Illuminating singlemolecules in condensed matter. Science, 283,1670–1676.

6. Korolev, S., Hsieh, J., Gauss, G. H., Lohman, T. M. &Waksman, G. (1997). Major domain swivelingrevealed by the crystal structures of complexes ofE. coli Rep helicase bound to single-stranded DNAand ATP. Cell, 90, 635–647.

7. Velankar, S. S., Soultanas, P., Dillingham, M. S.,Subramanya, H. S. & Wigley, D. B. (1999). Crystalstructures of complexes of PcrA DNA helicase witha DNA substrate indicate an inchworm mechanism.Cell, 97, 75–84.

8. Subramanya, H. S., Bird, L. E., Brannigan, J. A. &Wigley, D. B. (1996). Crystal structure of a Dexx boxDNA helicase. Nature, 384, 379–383.

9. Maluf, N. K., Fischer, C. J. & Lohman, T. M. (2003). Adimer of Escherichia coli UvrD is the active form ofthe helicase in vitro. J. Mol. Biol. 325, 913–935.

10. Dillingham, M. S., Wigley, D. B. & Webb, M. R.(2002). Direct measurement of single-stranded DNAtranslocation by PcrA helicase using the fluorescentbase analogue 2-aminopurine. Biochemistry, 41,643–651.

11. Dillingham, M. S., Wigley, D. B. & Webb, M. R.(2000). Demonstration of unidirectional single-stranded DNA translocation by PcrA helicase:measurement of step size and translocation speed.Biochemistry, 39, 205–212.

12. Jezewska, M. J., Rajendran, S., Bujalowska, D. &Bujalowski, W. (1998). Does single-stranded DNApass through the inner channel of the protein hex-amer in the complex with the Escherichia coli DnabHelicase—fluorescence energy transfer studies.J. Biol. Chem. 273, 10515–10529.

13. Scott, J. F., Eisenberg, S., Bertsch, L. L. & Kornberg,A. (1977). A mechanism of duplex DNA replicationrevealed by enzymatic studies of phage phi X174:catalytic strand separation in advance of replication.Proc. Natl Acad. Sci. USA, 74, 193–197.

14. Ramirez-Carrozzi, V. & Kerppola, T. (2001). Gel-based fluorescence resonance energy transfer (gel-FRET) analysis of nucleoprotein complex architecture.Methods, 25, 31–43.

15. Sofia, S. J., Premnath, V. & Merrill, E. W. (1998).Poly(ethylen oxide) grafted to silicon surfaces: graft-ing density and protein adsorption. Macromolecules,31, 5059.


16. Prime, K. L. & Whitesides, G. M. (1991). Self-assembled organic monolayers: model systems forstudying adsorption of proteins at surfaces. Science,252, 1164.

17. Deniz, A. A., Dahan, M., Grunwell, J. R., Ha, T. J.,Faulhaber, A. E., Chemla, D. S. et al. (1999). Single-pair fluorescence resonance energy transfer on freelydiffusing molecules: observation of Forster distancedependence and subpopulations. Proc. Natl Acad.Sci. USA, 96, 3670–3675.

18. Ha, T. J., Ting, A. Y., Liang, J., Caldwell, W. B., Deniz,A. A., Chemla, D. S. et al. (1999). Single-moleculefluorescence spectroscopy of enzyme conformationaldynamics and cleavage mechanism. Proc. Natl Acad.Sci. USA, 96, 893–898.

19. Mekler, V., Kortkhonjia, E., Mukhopadhyay, J.,Knight, J., Revyakin, A. & Kapanidis, A. N. (2002).Structural organization of bacterial RNA polymeraseholoenzyme and the RNA polymerase–promoteropen complex. Cell, 108, 599–614.

20. Schuler, B., Lipman, E. A. & Eaton, W. A. (2002).Probing the free-energy surface for protein foldingwith single- molecule fluorescence spectroscopy.Nature, 419, 743–747.

21. Rothwell, P. J., Berger, S., Kensch, O., Felekyan, S.,Antonik, M., Wohrl, B. M. et al. (2003). Multipara-meter single-molecule fluorescence spectroscopy

reveals heterogeneity of HIV-1 reverse transcriptase:primer/template complexes. Proc. Natl Acad. Sci.USA, 100, 1655–1660.

22. Chao, K. & Lohman, T. M. (1991). DNA-induceddimerization of the Escherichia coli Rep helicase.J. Mol. Biol. 221, 1165–1181.

23. Bjornson, K. P., Hsieh, J., Amaratunga, M. &Lohman, T. M. (1998). Kinetic mechanism for thesequential binding of two single-stranded oligodeoxy-nucleotides to the Escherichia coli Rep helicase dimer.Biochemistry, 37, 891–899.

24. Chao, K. & Lohman, T. M. (1990). DNA andnucleotide-induced conformational changes in theEscherichia coli Rep and helicase II (UvrD) proteins.J. Biol. Chem. 265, 1067–1076.

25. Cheng, W., Brendza, K. M., Gauss, G. H., Korolev, S.,Waksman, G. & Lohman, T. M. (2002). The 2Bdomain of the Escherichia coli Rep protein is notrequired for DNA helicase activity. Proc. Natl Acad.Sci. USA, 99, 16006–16011.

26. Colasant, J. & Denhardt, D. T. (1987). The Escherichiacoli rep mutation. X. Consequences of increased anddecreased Rep protein levels. Mol. Gen. Genet. 209,382–390.

27. Clegg, R. M. (1992). Fluorescence resonance energytransfer and nucleic acids. Methods Enzymol. 211,353–388.

Edited by D. E. Draper

(Received 30 September 2003; received in revised form 8 December 2003; accepted 10 December 2003)


dna-binding orientation and domain conformation of the e. coli

Documents