elucidating a key intermediate in homologous dna strand exchange: structural characterization of the...

Elucidating a Key Intermediate in Homologous DNAStrand Exchange: Structural Characterization of theRecA–Triple-stranded DNA Complex UsingFluorescence Resonance Energy Transfer

Jie Xiao and Scott F. Singleton*

Department of Chemistry andDepartment of Biochemistryand Cell Biology, RiceUniversity, P.O. Box 1892, MS65, Houston, TX 77005, USA

The RecA protein of Escherichia coli plays essential roles in homologousrecombination and restarting stalled DNA replication forks. In vitro, theprotein mediates DNA strand exchange between single-stranded(ssDNA) and homologous double-stranded DNA (dsDNA) moleculesthat serves as a model system for the in vivo processes. To date, no high-resolution structure of the key intermediate, comprised of three DNAstrands simultaneously bound to a RecA filament (RecA–tsDNA com-plex), has been reported. We present a systematic characterization of thehelical geometries of the three DNA strands of the RecA–tsDNA complexusing fluorescence resonance energy transfer (FRET) under physiologi-cally relevant solution conditions. FRET donor and acceptor dyes wereused to label different DNA strands, and the interfluorophore distanceswere inferred from energy transfer efficiencies measured as a function ofthe base-pair separation between the two dyes. The energy transfer effi-ciencies were first measured on a control RecA–dsDNA complex, andthe calculated helical parameters (h < 5 A, Vh < 208) were consistentwith structural conclusions derived from electron microscopy (EM) andother classic biochemical methods. Measurements of the helical par-ameters for the RecA–tsDNA complex revealed that all three DNAstrands adopt extended and unwound conformations similar to those ofRecA-bound dsDNA. The structural data are consistent with thehypothesis that this complex is a late, post-strand-exchange intermediatewith the outgoing strand shifted by about three base-pairs with respectto its registry with the incoming and complementary strands. Further-more, the bases of the incoming and complementary strands are displacedaway from the helix axis toward the minor groove of the heteroduplex,and the bases of the outgoing strand lie in the major groove of the hetero-duplex. We present a model for the strand exchange intermediate inwhich homologous contacts preceding strand exchange arise in theminor groove of the substrate dsDNA.

q 2002 Elsevier Science Ltd. All rights reserved

Keywords: fluorescence resonance energy transfer; RecA; minor groove;recombination; strand exchange*Corresponding author

Introduction

The Escherichia coli RecA protein plays crucialroles in the detection of and response to genomicDNA damage. Despite its modest size (Mr 37,842;352 amino acid residues), the RecA protein dis-plays a remarkably diverse set of activities(reviews1 – 3) and carries out at least three in vivoDNA metabolic functions. First, the RecA proteincontrols the SOS response for DNA damage

0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved

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

Abbreviations used: ATPgS, adenosine 50-O-(3-thiotriphosphate); dsDNA, double-stranded DNA; EM,electron microscopy; FRET, fluorescence resonanceenergy transfer; nt, nucleotide(s); ODN,oligodeoxyribonucleotide; RecA, Escherichia coli RecAprotein; Rox, carboxy-X-rhodamine; ssDNA, single-stranded DNA; tsDNA, triple-stranded DNA.

doi:10.1016/S0022-2836(02)00462-X available online at http://www.idealibrary.com onBw

J. Mol. Biol. (2002) 320, 529–558

tolerance.4 In addition, RecA plays indispensableroles in restarting stalled replication forks,5 poss-ibly through recombinational DNA repair.6 Finally,the RecA protein mediates homologous geneticrecombination.7 The latter two roles require thepairing of a single-stranded DNA molecule(ssDNA) with a molecule of double-strandedDNA (dsDNA) containing a strand nearly identicalin sequence to that of the ssDNA.

In vitro, the RecA protein mediates DNA strandexchange between ssDNA and homologousdsDNA molecules that serves as a model for thein vivo processes. Biochemical mechanisms forstrand exchange can be separated into at least fourconceptually distinguishable phases: presynapsis,synapsis, heteroduplex formation, and branchmigration. Presynapsis is the non-sequence-specificassembly of multiple RecA monomers on theincoming ssDNA to form a multimeric right-handed helical nucleoprotein filament. Duringsynapsis, a dsDNA molecule, composed of strandswhose sequences are identical (the outgoingstrand), and complementary to that of theincoming strand, is sequence-specifically alignedwith the homologous ssDNA in the nucleoproteinfilament forming a three-stranded nucleoproteincomplex. Within the nascent three-stranded com-plex, a rapid and partial strand switch ensues,yielding a “joint molecule” containing a relativelyshort region of product, termed the heteroduplex.This third phase requires the presence of ATP butnot its hydrolysis. Under conditions permissive ofATP hydrolysis, the nascent heteroduplex DNA isextended until products are formed in a branchmigration reaction comprising the final phase.

During the second and third phases, a RecA–triple-stranded DNA (RecA–tsDNA) complex isformed that serves as the key intermediate instrand exchange. Considerable progress has beenmade over the past 20 years in determining thestructures of the RecA protein-only and protein–DNA filaments. Indeed, electron microscopy (EM)of RecA–DNA complexes trapped during sequen-tial stages of in vitro strand exchange have allowedvisualization of genetic recombination inter-mediates.8,9 In spite of this success, no structuresof RecA–tsDNA complexes that resolve proteinmonomers or DNA strands have been reported.

EM of RecA–dsDNA complexes has providedthe most consistent and highest resolution imagesof RecA nucleoprotein filaments, allowing them tobe directly visualized and a likely DNA bindinglocation to be identified.10,11 A combination of EMand classical biochemical methods has provided aconsistent and well-accepted structural modelwherein the dsDNA is stretched by ca 50% andpartially unwound to a right-handed helix with ca19 bp per turn.12 – 17 It has been proposed that thestretched and partially unwound DNA structureimposed by RecA binding is instrumental in theprocess of strand exchange18 – 22

One long-term objective of characterizing thestructures of the filament components is to reach a

molecular understanding of recombination andrecombinational repair by the construction andelaboration of three-dimensional models of theintermediates in the reactions. Moreover, by devel-oping an accurate description of the DNA structureduring strand exchange, it should be possible toelucidate the mechanistic strategy employed bythe RecA protein and its homologs. To date, eventhe basic elements of that strategy are not agreedupon universally. For example, the answer to thequestion of whether RecA mediates homology rec-ognition via major or minor-groove contacts withthe substrate dsDNA remains elusive. Althoughthe issue has been addressed in several labora-tories over the past decade, those experimentshave produced controversial results whoseinterpretation depends on the structural model ofthe RecA–tsDNA complex.18,23 Direct structuralcharacterization of the DNA constituents of theRecA–tsDNA strand exchange intermediatewould offer insight into these mechanistic issuesas well as providing valuable information on howhomologous recombination and recombinationalDNA repair are mediated in vivo.

In the work described here, we systematicallyemployed fluorescence resonance energy transfer(FRET) efficiency measurements to characterizethe helical geometry of the RecA–tsDNA complexin RecA-mediated strand exchange. The experi-ments relied on FRET between the donorfluorescein, conjugated to an amino-modifieddeoxythymidine (dTA; Figure 1(a)), and the simi-larly conjugated acceptor carboxy-X-rhodamine(Rox). Since the distance between the two dyes isa function of the base-pair separation and the heli-cal parameters (i.e. helical rise and twist angle),measuring the energy transfer efficiency betweenthe two dyes enabled us to measure the distancebetween the two dyes and, hence, the DNA’shelical geometry. We envisage several potentialadvantages of this method for DNA structuralcharacterization in nucleoprotein filaments. First, itis important to note that both key helical geometryparameters are determined simultaneously in thesame experiment. Second, because the fluorescencesignals originate from the bound DNA strands,the measurements directly characterize the localDNA structure rather than the overall filament.Finally, the non-invasive labeling of the DNAstrands with fluorophores also enables the FRETmeasurements to be made under physiologicallyrelevant solution conditions, which introduce theleast amount of perturbation into the RecA–DNAcomplex.

Prior to undertaking an analysis of the RecA–tsDNA complex, we used FRET efficiencymeasurements to characterize a RecA–dsDNAcomplex. The resulting helical geometry para-meters agreed with the well-accepted structuralmodel of RecA-bound dsDNA, providing animportant confirmation of the FRET-based method.Because the FRET measurements using oligo-nucleotides provide accurate helical geometry

530 DNA Structure During RecA-mediated Strand Exchange

information over distances comparable to theradius and pitch of the RecA nucleoprotein fila-ment, the approach described here can revealstructural information complementary to thatderived from bacteriophage DNAs with lengthscales 10–100-fold larger.

In the measurements using RecA–tsDNA com-plexes, one of the three DNA strands was labeledwith Rox, while one of the remaining strands waslabeled with fluorescein (Figure 1(b) and (c)).Hence, all three DNA strands within the strandexchange intermediate were characterized. All

N

N

O

O

HNH

OHN

OO O–

COO–

O

O N+N

OdTA dTA

O

HO OH

5

dTA =

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCCGGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTATCTTTTTTCTTCTTAGGATCCCGGGATCCTAAGAAGAAAAAAGATAGCTCCTTCGGTCCTCCGATCGTTGTCAGAAGTAAGTTGGCCGCGAATTCC RAGCTCCTTCGGTCCTCCGATCGTTGTCAGAAGTAAGTTGGCCGC TAGCRCCTTCGGTCCTCCGATCGTTGTCAGAAGTAAGTTGGCCGC TAGCTCCRTCGGTCCTCCGATCGTTGTCAGAAGTAAGTTGGCCGC

RCCTCCGATCGTTGTCAGAAGTAAGTTGGCCGC TAGCTCCTTCGGTCCRCCGATCGTTGTCAGAAGTAAGTTGGCCGC TAGCTCCTTCGGTCCTCCGARCGTTGTCAGAAGTAAGTTGGCCGC TAGCTCCTTCGGTCCTCCGATCGTTGTCAGAAGTAAGTTGGCCGC TGCCAGGTTCACTGACACTGACACTAGCATCATATCTTACTCGAT RGCCAGGTTCACTGACACTGACACTAGCATCATATCTTACTCGAT

74F 74X 74C 45R0 45R4 45R7 45R12 45R15 45R20 45Xh45h45R

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTATCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCTCCTGGCTTCCTCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCTCCTGGCTTCCTCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCTCCTGGCTTCCTCGAR

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCTCCTGGCTTCCRCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCTCCTGGCTRCCTCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCTCCRGGCTTCCTCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCTAGCCRCCTGGCTTCCTCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCC CGCCGGTTGAATGAAGACTGTTGCRAGCCTCCTGGCTTCCTCGAT

GGAATTCGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAFCTTTTTTCTTCTTAGGATCCCCCTTAAGCGCCGGTTGAATGAAGACTGTTGCTAGCCTCCTGGCTTCCTCGATAGAAAAAAGAAGAATCCTAGGG

XX

FX

FRi

ds74F

i = 0

i = 4

i = 7

i = 12

i = 15

i = 20

a

b

c

Fluorescein (F) Rox (R)

Figure 1. (a) Chemical structures of fluorescein, carboxy-X-rhodamine (Rox), and deoxythymidineamine (dTA), and(b) sequences of single-stranded DNA molecules and (c) double-stranded DNA molecules used in the study. AllssDNA and upper-strand dsDNA sequences are written from 50 to 30, while the lower, complementary strands ofdsDNA read from 30 to 50. F, fluorescein-conjugated dTA; R, Rox-conjugaed dTA.

DNA Structure During RecA-mediated Strand Exchange 531

three strands are extended and unwound to a simi-lar extent as RecA-bound dsDNA. The data areclearly in accord with recognition of homologyoccurring via minor groove contact, since the basesof the incoming and complementary strands aredisplaced away from the helix axis toward theminor groove of the heteroduplex while the basesof the outgoing strand lie in the major groove ofthe heteroduplex. The resulting structural modelcan account for much of the previously publisheddata regarding the mechanism of homolgouspairing.

Theory

Fluorescence resonance energy transfer

The efficiency of Forster energy transfer (E )between a FRET donor–acceptor pair is quanti-tatively related to the inverse sixth power of thedistance (R ) between the two dyes:

E ¼R6

0

R60 þ R6

ð1Þ

where R0 is the Forster critical distance (in A)between donor and acceptor at which the energytransfer efficiency is 50%.24 Therefore, a slightchange in the distance between the donor andacceptor leads to a large change in the energytransfer efficiency. FRET has been successfullyemployed as a “molecular ruler” in a variety ofbiomolecular systems.25 – 30 In the RecA field, FREThas been applied to a description of the kineticsteps31,32 and polarity33 of RecA-mediated strandexchange.

For a DNA double helix with a FRET donor andacceptor conjugated to opposite strands, the dis-tance between the two dyes should reflect the heli-cal parameters such as rise per base-pair (h ) andtwist angle per base-pair (Vh). (The symbols andnomenclature describing the DNA geometry arethose suggested by Olson et al.34) Indeed, it wasdemonstrated by the elegant experiments of Cleggand co-workers that FRET signals display acharacteristic modulation when the fluorophores“walk” around the helix.35 Therefore, by measuringthe energy transfer efficiency between the twodyes, the structural information characterizingthe DNA helix could be obtained. In our study,we systematically shifted the position of Roxalong one DNA strand while maintaining aconstant position of fluorescein on anotherstrand. This arrangement allows the energy trans-fer efficiency between the two dyes to be expressedas a function of the characteristic helical par-ameters. These helical parameters can be readilyobtained by fitting the experimental data to asimple DNA helix model,35 which is describedbriefly below.

Geometric model of DNA helix

For dsDNA formed from oligonucleotides, thegeometry of the DNA helix can be represented incylindrical coordinates (Figure 2). A unique,straight helical z-axis, which is perpendicular tothe base-pair plane, is assumed. The distance Rbetween donor and acceptor is expressed as a func-tion of the number of base-pairs (n ) separating thedonor (d) and acceptor (a), the helical parametersh and Vh, and a set of parameters describing thegeometry of the linker connecting the dyes to theDNA:

R2 ¼ r2a þ r2

d þ ðnh þ DhÞ2 2 2rard cosðnVh þ DVhÞ

ð2Þ

where the values of ra and rd are the lengths of thevectors pointing from the helix axis to the centersof dyes projected on the base-pair plane; Dh is thevertical distance separating the two dye centers if

donor

Z

nh + ∆hR

rd

x

x'

acceptor

ra

nΩh+∆Ω

Figure 2. Schematic representation of the DNA helixgeometry model. The central helix axis z is perpendicularto the base-pair plane. The distance R between donorand acceptor is related to the following parameters byequation (3): h (the helical rise per base-pair), Vh (helicaltwist angle per base-pair), rd (the length of the vectorextending from the z-axis to the center of the fluor-escence donor dye), ra (the length of the correspondingvector to the acceptor dye), Dh (the vertical distancebetween the two dyes if they were attached to the samebase-pair), and DV (the angle between vectors ra and rd

if the two dyes were attached to the same base-pair).Although the planes containing the vectors ra and rd areparallel, they are not necessarily coincident with thebase-pair-plane. The model does not require the twodye-labeled DNA strands to be paired (Watson-Crick orotherwise) as long as the Rox-labeled strand is character-ized by a regular helical geometry.


they were attached to the same base-pair; and DVh

is the angle between ra and rd if the two dyes wereattached to the same base-pair. For B-formdsDNA, the helical axis is through the center ofeach base-pair plane, thus the two vectors ra andrd are pointing directly from the base-pair centerto the center of the dyes. For other DNA helices,including RecA–DNA complexes, the helical axismay be displaced away from the base-pair center.Hence, the vectors ra and rd include the contri-bution of the distance from the helix axis to thebase-pair center. Importantly, this general DNAhelix geometry model does not necessarily requiretwo DNA strands to be paired as long as the Rox-labeled strand is ordered relative to the position offluorescein on the other strand. The base-pairseparation n between donor and acceptor is theonly extrinsic variable in equation (2), and can besystematically varied throughout a series of DNAmolecules. The energy transfer efficiency Ebetween donor and acceptor then can be expressedas a function of n by combining equations (1) and(2):

E ¼

R60

R60 þ r2

a þ r2d þ ðnh þ DhÞ2 2 2rard cosðnVh þ DVhÞ

3

ð3Þ

Experimental design

We designed five different but directly relatedreaction systems to facilitate the FRET measure-ments on different DNA and RecA–DNA com-plexes (Figure 3). Of the five systems, the dsDNAseries was designed to provide information aboutthe dye-linker structures, the RecA–dsDNA serieswas designed to test FRET-derived helicalgeometry parameters against those reported in theliterature, and the RecA–tsDNA-I, -C, and -Oseries were designed to extract the helical geome-tries of the incoming, complementary, and out-going strands, respectively. Each system contains afluorescein-labeled 74 nt oligodeoxyribonucleotide(ODN) and a series of six Rox-labeled 45mers(45Ri; see Figure 1). All ODNs in the 45Ri serieshave the same sequence; only the thymidine pos-ition to which Rox was attached was varied, andeach unique attachment position is indicated bythe subscript on 45Ri (Figure 1).

As an initial control, a set of partial dsDNA mol-ecules (FRi) was labeled with both fluorescein andRox (Figure 1). Each partial dsDNA FRi has 7 and22 nt overhangs at the 50 and 30 ends, respectively.The fluorescein-labeled dT in each FRi is thenucleoside in the 22 nt overhang adjacent to the 30

end of the base-paired region. In aqueous solution,this set of dsDNAs was assumed to adopt astandard B-form double-helical structure. ThisRecA-free dsDNA system was designed to isolate

the dye-related parameters for use in the otherfour RecA-involved systems.

In the second system, the same partial dsDNAFRi was used to facilitate the polymerization of theRecA protein on the middle dsDNA region. TheDNA helical parameters of the RecA–dsDNA com-plex (Figure 3) were measured using this system.

The other three systems (RecA–tsDNA-C,RecA–tsDNA-O, and RecA–tsDNA-I; Figure 3)allowed measurements of the helical parametersfor the complementary strand (C), outgoing strand(O), or incoming strand (I) of the RecA–tsDNAcomplex. The partial dsDNA FRi described abovewas used as the homologous dsDNA substrate inreactions where the Rox-labeled strand participatesas the complementary or outgoing strand (RecA–tsDNA-C and -O, respectively). For RecA–tsDNA-C, the sequence of the fluorescein-labeled strandin the partial dsDNA FRi is identical to a RecA-coated 74mer (74X). When the RecA–74X filamentpairs on FRi, the fluorescein-labeled strand is theoutgoing strand while the Rox-labeled strand(45Ri) is the complementary strand. This arrange-ment allows the structural information of thecomplementary strand to be ascertained. ForRecA–tsDNA-O, the incoming ssDNA was 74C,the sequence of which is complementary to thefluorescein-labeled strand of the dsDNA FRi.Therefore, the Rox-labeled strand (45Ri) becomesthe outgoing strand. For RecA–tsDNA-I, a full-length fluorescein single-labeled dsDNA (ds74F)was utilized. A Rox-labeled 45mer (45Ri) was usedas the incoming ssDNA to pair with ds74F. Thesequence of this incoming 45Ri is identical to a 45base sequence in the middle of the non-fluorescein-labeled ssDNA (74X) of ds74F. Thus, the fluor-escein-labeled strand becomes the complementarystrand when the RecA-mediated strand exchangetakes place. This systematic labeling of Rox oneach strand in the three RecA–tsDNA systemsenabled us to collect structural information directlyfrom the complementary strand, outgoing strand,and incoming strand of the strand exchangeintermediate.

Results

Characteristics of DNA strand exchange withfluorophore-labeled ODNs

We first examined the possibility that the conju-gation of fluorophores to the ODNs might impedethe binding of RecA or strand exchange. Differentreactions using dye-labeled DNA substrates aswell as unlabeled DNA substrates were performedunder identical conditions, and equal amounts ofeach reaction, either deproteinized or untreated,were fractionated on 3% (w/v) Metaphor agarosegels (data not shown). The untreated reactantsshowed equal amounts of involvement of thelabeled or unlabeled dsDNA substrates into thereaction complexes. No release of the outgoing


Figure 3. Fluorescent DNA reaction systems. Each DNA strand used in each system only contains one fluorescein orone Rox. The apparent multiple labeling of Rox on the 45mer serves the illustrative purpose of indicating the relativeseparations between the fluorophores. The actual set of DNA molecules used is shown in Figure 1. In the dsDNAsystem, a dual-labeled partial dsDNA FRi formed between the fluorescein-labeled 74mer and its complementary Rox-labeled 45mers 45Ri was used. In the RecA–dsDNA system, the same dsDNA FRi was used to form the complexwith the RecA protein (gray ovals). In the RecA–tsDNA-C and RecA–tsDNA-O systems, a presynaptic filament wasfirst formed between the RecA protein and a 74mer (74X or 74C). The partial dsDNA FRi was used as the homologousdsDNA substrate. In the RecA–tsDNA-I system, the presynaptic filament was formed on a series of Rox-labeled45mers, and a full-length fluorescein-labeled dsDNA Fds74 was used as the homologous substrate. The concentrationof RecA protein was sufficient to coat the full length of the 74 bp dsDNA. I, incoming strand; C, complementarystrand; O,outgoing strand.


strand was observed. After treatment to removeRecA protein, both labeled and unlabeled reactionsshowed similar extents of product formation. Inaddition, when a heterologous 45mer, either Rox-labeled or unlabeled (h45 and h45R, respectively),was used as an incoming ssDNA to pair with afull-length dsDNA (ds74 or ds74F), it was foundthat the strand exchange reaction did not occur.These results demonstrated that the labeling offluorophores did not affect the occurrence andcompletion of the reactions under our reaction con-ditions, and that the strand exchange reaction weobserved is homology-specific. Our conclusionthat strand exchange is not affected by labelingthe DNA strands with fluorescent dyes is con-sistent with published reports.31,32,36

Spectral characteristics of the reaction systems

In order to relate the energy transfer efficiencyquantitatively with the helical parameters thatgive rise to the distance between a FRET donorand acceptor, both the energy transfer efficiencyand the Forster critical distance R0 of the pair hasto be known (see equation (3)). The energy transferefficiency varies with the placement of the dye,while R0 is characteristic of the position-indepen-dent spectral properties of the dyes. This, in turn,requires a knowledge of the relative orientation ofthe two dyes (k2), the donor quantum yield in theabsence of the acceptor (Fd), and the overlapintegral J that characterizes the resonance betweenthe donor and acceptor transition dipoles.

The orientation factor k2 cannot be directlymeasured. However, rapid motion of the dyesduring their excited state lifetimes, combined withrandom static orientation of the dye transition

moments, simplifies the interpretation of energytransfer measurements by allowing the assumptionof a constant orientation factor ðk2 ¼ 2=3Þ: Thequality of the assumption can be evaluated byindependent fluorescence anisotropy measure-ments. When a dual-labeled dsDNA sample isexcited at 492 nm, only the fluorescein fluorescenceis detected at 520 nm. When it is excited at 585 nm,only Rox fluorescence is detected at 605 nm. There-fore, the fluorescence anisotropy of each fluor-escence probe can be measured independently,even in the dual-labeled duplexes. Within error,the anisotropies of both fluorescein and Rox wereindependent of the dye’s position of attachmentthroughout the length of the DNA molecule ineach system, suggesting a constant k2 value. Theaverage anisotropies of fluorescein and Rox areshown in Table 1. In general, fluorescein has amuch lower anisotropy than Rox in each system.Because the fluorescence lifetimes of both dyes aresimilar, and the dyes are attached to the samemacromolecule, fluorescein must undergo con-siderably faster rotational diffusion, or rotaterapidly within a greater cone angle, than Rox (thefundamental anisotropy of both dyes is almost0.40). Because of the lower anisotropy of fluor-escein, and the lack of any indication that eitherfluorophore is located in a preferred static orien-tation relative to the dsDNA molecule, we assumea random relative orientation of the dyes withrapid re-orientational motion and a constant valuek2 ¼ 2=3: Such an assumption has been validatedin a variety of experimental systems.26,35,37 – 47

We measured the quantum yield of fluoresceinin the absence of Rox (FF) and the spectral overlapintegral for each system. An example of thespectral overlap between fluorescein and Rox

Table 1. Spectral characteristics of fluorescein (F) and Rox (R) measured in dsDNA, and in the RecA–dsDNA andRecA–tsDNA complexes

dsDNA RecA–dsDNA RecA–tsDNA-C RecA–tsDNA-I RecA–tsDNA-O

FF 0.76 0.83 0.77 (0.77)a (0.77)a

J £ 10215 (M21 cm21 nm4) 2.0 1.8 2.0 (2.0) (2.0)R0 (A) 53.6 53.2 53.5 (53.5) (53.5)eF

492 (M21 cm21) 41,000 57,000 60,000 (60,000) (60,000)eR

585 (M21 cm21) 78,000 72,000 65,000 (65,000) (65,000)eR

492/eR585 0.053 0.059 0.056 0.053 0.064

rF(492, 520) 0.10 0.22 0.22 0.22b 0.22

rR(585, 605) 0.22 0.28 0.30 0.30 0.30

The tabulated parameters derived from spectroscopy are the fluorescein quantum yield (FF), the spectral overlap integral (J ), theForster critical distance (R0), the molar absorption coefficient (e) of each fluorophore (subscript) at a specific wavelength in nm (super-script), the ratio of the molar absorption coefficients for Rox at 492 and 585 nm (eR

492/eR585), and the anisotropy (r ) of each fluorophore

(subscript) acquired at specified excitation and emission wavelengths (superscript). The coefficients were measured by comparingtheir absorbance or excitation spectra under reaction conditions with those of the fluorescein or Rox-labeled ssDNA. Relative uncer-tainties in the parameters tabulated in the first five rows were generally less than 5% of the tabulated values as judged using standarddeviations for replicate experiments. The relative uncertainties in the eR

492/eR585 values were less than 10% in every case. The anisotropy

values were essentially independent of the relative positions of fluorescein and Rox, and the absolute uncertainties in the anisotropyvalues (as judged by one standard deviation derived from all six complexes for each system) were ^0.015.

a The spectral parameters shown in parentheses for the RecA–tsDNA-I and -O complexes were assumed to be unchanged fromthose measured for the RecA–tsDNA-C complex.

b The fluorescein anisotropy values measured for the RecA–tsDNA-I complex with Rox at postions i ¼ 0 and i ¼ 4 were signifi-cantly lower than other values and were excluded from the tabulated average. The cause and effect of the abnormal anisotropy valuesare discussed in the text.


measured on two dye-conjugated ssDNA samplesis shown in Figure 4(a). The spectral overlapbetween fluorescein and Rox is smaller than thosebetween fluorescein and tetramethylrhodamine31

or hexachlorofluorescein,32 which, in turn, resultsin a smaller R0 value. Importantly, however, asmaller R0 value reduces the sensitivity of ourFRET measurements to the cross-interactionbetween complexes in solution, and generates

well-resolved emission spectra, in which spectraldistortion is minimal. We verified that the normal-ized emission spectra of fluorescein or Rox in thepresence of the other fluorophore are identical tothose in its absence (one exception will be dis-cussed below). This suggests that the energy distri-butions of the donor or acceptor emission are notaffected by the presence of complementary accep-tor or donor, respectively. These characteristics of

0.0

0.2

0.4

0.6

0.8

1.0

0 100

1 104

2 104

3 104

4 104

5 104

6 104

7 104

8 104

450 500 550 600 650

Nor

mal

ized

Flu

ores

cenc

e

Extinction C

oefficient (M-1cm

-1)

Wavlength (nm)

0.0 100

1.0 104

2.0 104

3.0 104

4.0 104

5.0 104

6.0 104

7.0 104

8.0 104

450 500 550 600 650

Flu

ores

cenc

e

Wavelength (nm)

(a)

(b)

Figure 4. Spectral data characterizing the fluorescein–Rox reaction systems. (a) The spectral overlap between fluor-escein emission spectrum (lex ¼ 492 nm, broken line) and the absorption spectra of Rox (continuous line) in 25 mMTris–OAc (pH 7.5), 5% glycerol, 1 mM EDTA. The spectra were recorded for fluorescein-labeled dsDNA FX and Rox-labeled 45mer 45R0 (see Figure 1). (b) The emission spectra used to calculate the energy transfer efficiency for partialduplex FR15. Gray line: emission spectrum of FR15 with excitation at 492 nm. An apparent emission enhancementaround 605 nm was observed, indicating the occurrence of energy transfer between fluorescein and Rox. Broken line:normalized emission spectrum of the fluorescein-only labeled dsDNA FX with excitation at 492 nm. There is no shiftof the emission peak near 520 nm for the dual-labeled dsDNA FR15 emission spectrum compared to that of fluor-escein-only labeled FX, indicating an absence of strong electronic coupling between fluorescein and Rox. Thick blackline: sensitized emission spectrum calculated by subtraction of the fluorescein-only emission (broken line) from theemission spectrum of the dual-labeled dsDNA sample FR15 (gray line). Thin black line: emission spectrum of dsDNAFR15 with excitation at 585 nm. The (Ratio)A is determined from intensities of the sensitized emission spectrum (thickblack line) and the Rox emission spectrum (thin dark line), and the corresponding energy transfer efficiency is calcu-lated using equation (19).


the fluorophore pair employed here provide sensi-tive information on relatively short-range, localconformational changes. The calculated R0 valuesfor the reaction systems are listed in Table 1.

We also measured the extinction coefficients offluorescein and Rox at their respective excitationwavelengths, as well as the anisotropy of each dyeat different base-pair separations in each system.The results are listed in Table 1. The extinctioncoefficients of both dyes are constant, independentof their positions of conjugation to the DNAstrands, indicating similar chromophore environ-ments. These extinction coefficients are used in thecalculation of the energy transfer efficiency. Repre-sentative fluorescein and Rox emission spectranecessary to perform such calculations are illus-trated in Figure 4(b) for dsDNA. The sensitizedemission spectrum of Rox (excitation at 492 nm),corrected by removal of fluorescein backgroundemission, and the emission spectrum of Rox (directexcitation at 585 nm) were used to obtain (Ratio)A

according to equations (17) and (18). This ratio isrelated to the energy transfer efficiency by theextinction coefficients according to equation (19).

Energy transfer efficiency measurementsusing dsDNA

We first measured the energy transfer efficiencyon 0.1 mM RecA-free dsDNA with fluorescein andRox conjugated to the opposite strands. The valuesof E and R at each n are listed in Table 2. A plot ofthe energy transfer efficiencies between fluoresceinand Rox versus the number of base-pairs separatingthe two dyes is displayed in Figure 5(a) (opensquares). The data points were fitted to the E–nfunction (equation (3)), yielding a unique solutionregardless of initial parameter values. With theassumption of B-form dsDNA helical parameters(h ¼ 3:4 A and Vh ¼ 368),48 the best-fitted par-ameters were found to be: r2

a þ r2d ¼ 960 A2; 2rard ¼

140 A2; Dh ¼ 211 A; and DVh ¼ 608 (Table 3). Thepair of parameters, ra and rd, describing the radialdistances between each fluorophore and the helixaxis, can be calculated by the simultaneous sol-ution of the pair of equations for the two combinedfitting parameters, (ra

2 þ rd2) and 2rard. For the

protein-free dsDNA, the radial distances are31(^3) A and 2(^2) A (Table 4). Because the dis-tances are symmetric with respect to exchange inthe two combined fitting parameters, the helixgeometry model does not allow the unique assign-ment of individual dye-linker distances. Thesedye-linker parameters are geometrically reasonablebut are clearly different from the parametersexpected for fully extended linkers. For linkers infully extended all-anti conformations coplanarwith the thymine heterocycles, a simple calculationusing canonical bond length and bond angle datato describe the distance from the helix axis to thedyes predicts dye-linker radial distances ðra ¼ rd ¼25 AÞ and fitting parameters (r2

a þ r2d ¼ 1250 A2;

2rard ¼ 1250 A2; Dh ¼ 0 A and DVh ¼ 1808) differ-ent from those observed. One explanation for thisdifference is that one of the linkers is folded toallow the flurophore to approach the DNA moreclosely. Conformations preferring dye-DNAinteractions have been observed for both5-tetramethylrhodamine35,49,50 and cyanine-3.51 Thenet result is that one of the dyes appear closer tothe helical axis than predicted for a fully extendedlinker.

Satisfaction of a saturation binding conditionfor RecA–DNA complexes

In the four RecA-involved reaction systems,excess amounts of one non-fluorescent reactionspecies over the other fluorescent componentswere necessary to insure that greater than 90% ofthe fluorophore-labeled DNA molecules convertedto the final product. This criterion is important foraccurate FRET measurements. The FRET efficiencyis highly sensitive to the distance between thefluorescence donor and acceptor: if one member ofthe FRET pair exists in both the bound andunbound states, the apparent energy transfer effi-ciency would represent contributions from each ofthe species. Hence, FRET efficiency measurementswould be unable to characterize accurately thestructure of the single complex of interest. Forexample, in the RecA–dsDNA complex, when theRecA concentration was at its minimum stoichio-metric concentration (one monomer per 3 bp),

Table 2. Energy transfer efficiencies (E ) and distances (R ) between fluorescein and Rox at different base-pairseparations for the five reaction systems

dsDNA RecA–dsDNA RecA–tsDNA-C RecA–tsDNA-O RecA–tsDNA-I

Rox position E R (A) E R (A) E R (A) E R (A) E R (A)

0 0.80 ^ 0.02 42 ^ 6 0.37 ^ 0.03 58 ^ 7 0.17 ^ 0.01 70 ^ 4 0.43 ^ 0.01 56 ^ 34 0.94 ^ 0.04 34 ^ 26 0.74 ^ 0.01 45 ^ 1 0.45 ^ 0.02 55 ^ 4 0.28 ^ 0.01 63 ^ 3 0.46 ^ 0.01 55 ^ 27 0.96 ^ 0.03 32 ^ 20 0.54 ^ 0.01 52 ^ 2 0.33 ^ 0.02 60 ^ 5 0.22 ^ 0.01 66 ^ 4 0.32 ^ 0.01 61 ^ 212 0.76 ^ 0.05 44 ^ 12 0.20 ^ 0.01 67 ^ 3 0.13 ^ 0.00 74 ^ 2 0.09 ^ 0.00 79 ^ 4 0.13 ^ 0.01 74 ^ 615 0.56 ^ 0.01 52 ^ 2 0.10 ^ 0.00 76 ^ 1 0.07 ^ 0.01 82 ^ 6 0.07 ^ 0.01 82 ^ 10 0.07 ^ 0.00 82 ^ 520 0.24 ^ 0.03 65 ^ 10 0.02 ^ 0.01 99 ^ 25 0.03 ^ 0.01 95 ^ 36 0.02 ^ 0.00 99 ^ 4 0.06 ^ 0.01 86 ^ 12

Energy transfer efficiencies (E ) were calculated using equation (19), and distances (R ) were calculated using equation (1). Each setof data is represented as the mean of at least three independent experiments plus-or-minus one standard deviation.


only about 70% of dsDNA molecules were bound.The energy transfer efficiencies measured underthis condition were irregular and impreciselydetermined, and reasonable curve fits could not beguaranteed. In contrast, when the RecA concen-tration was raised such that over 90% of thedsDNA molecules were bound, the data werehighly reproducible. The fitting correlation coeffi-cient using the E–n function on these saturateddsDNA increased significantly. Based on these con-siderations, we measured the apparent dissociationconstants of all our RecA–DNA complexes bytreating them as simple one-step, reversible bind-ing reactions.

The apparent dissociation constant (Kd) for eachcomplex was measured by fitting fluorometrictitrations, and the necessary concentration of eachreaction component was determined as follows. Inthe RecA–dsDNA, RecA–tsDNA-C, and RecA–

tsDNA-O reactions, the concentration of the partialdsDNA FRi was set at 0.1 mM while the concen-tration of the other reactant was adjusted toachieve the desired extent of product formation.For RecA–dsDNA, the RecA concentration was4 mM. For RecA–tsDNA-C and RecA–tsDNA-O,the concentration of the RecA–ssDNA filamentwas 0.73 mM. In RecA–tsDNA-I, we used a differ-ent strategy. The reaction occurs between a RecA-bound Rox-labeled ssDNA (45Ri) and a fluor-escein-labeled full-length dsDNA (ds74F), andboth of the species emit fluorescence. In order toreduce the complexity inherent in calculating theenergy transfer efficiency between the FRET pairwhen one dye is in excess, the concentrations ofboth species were maintained at 0.5 mM to achievethe 90% reaction completion. Taken together, theagarose gel mobility shift assays and thethermodynamic analysis using fluorescence

Figure 5 (legend opposite)

0.0

0.2

0.4

0.6

0.8

1.0

-5 5 10 15 20 25

Ene

rgy

Tran

sfer

Effi

cien

cy

Base Pair Separation

0.0

0.1

0.2

0.3

0.4

0.5

-5 0 5 10 15 20 25

Ene

rgy

Tran

sfer

Effi

cien

cy


(a)

(b)


titrations confirm that greater than 90% of theobserved FRET signals originated from the speciesof interest.

DNA helical parameters for RecA-bound dsDNA

The energy transfer efficiencies for the set ofRecA-bound dsDNA molecules were measured inthe second reaction system. Excess RecA protein(4 mM), saturated with ATPgS (0.5 mM), was incu-bated with 0.1 mM partial dsDNA FRi for one hourat 37 8C and the emission spectra of fluoresceinand Rox were recorded. This amount of RecA–ATPgS was used to insure that greater than 90%of the dsDNA was bound by the RecA protein.

The energy transfer efficiencies were calculatedand plotted against base-pair separation (Figure5(a), filled circles). The data were fitted with theE–n function as described in equation (3). Thedye-linker parameter Dh obtained from B-formdsDNA above was fixed, leaving the other helicalparameters variable. The fitting yielded a uniquesolution regardless of initial parameter values. Thefitted helical parameters are: h ¼ 4:9ð^0:1Þ A; Vh ¼17ð^4Þ8 (Table 3). The original dsDNA FRi is pre-sumed to adopt B-form conformation in solutionwith a helical rise of 3.4 A and a helical twistangle of 368. As such, the fitted helical parametersfor the RecA-bound dsDNA reveal an apparentconformational change. The dsDNA molecule was

Figure 5. Energy transfer effi-ciency (E ) between fluorescein andRox at different base-pair separ-ations (n ) in (a) the B-form dsDNAand RecA–dsDNA, (b) and (c) theRecA–tsDNA-C and -O, and (d)the RecA–tsDNA-I reaction sys-tems. In each case, the symbols rep-resent the actual data points andthe continuous lines are the best-fitcurves according to non-linearleast-squares analysis usingequation (3). (a) The critical Forsterdistances R0 were set at 53.6 A fordsDNA (open squares, gray line)and 53.2 A for RecA–dsDNA (filledcircles, black line). For B-formdsDNA, the helical pitch per base-pair h at 3.4 A and helical twistangle per base-pair Vh at 368 wereused to obtain the fitted dye-linkingparameters: r2

a þ r2d ¼ 960 A2;

2rard ¼ 140 A2, Dh ¼ 210.6 A, andDV ¼ 608. For RecA–dsDNA, thevertical distance between the twodyes was taken from the B-formdsDNA as Dh ¼ 210.6 A. The fittedparameters are: r2

a þ r2d ¼ 2000 A2;

2rard ¼ 280 A2, h ¼ 4.9 A, Vh ¼17.38, and DV ¼ 128. (b) For theRecA–tsDNA-C (filled squares, redline) and RecA–tsDNA-O (filledcircles, green line) complexes, thebase registry was taken to beDn ¼ 0. The critical Forster distanceR0 was set at 53.5 A, and the verticaldistance between the two dyes wastaken from the B-form dsDNA asDh ¼ 210.6 A. The fitted para-meters for RecA–tsDNA-C are: r2

a þr2

d ¼ 3400 A2; 2rard ¼ 370 A2, h ¼4.3 A, Vh ¼ 18.68, and DV ¼ 2908;for RecA–tsDNA-O, the para-meters are: r2

a þ r2d ¼ 4290 A2;

2rard ¼ 480 A2, h ¼ 4.2 A, Vh ¼ 31.18, and DV ¼ 2008. (c) For the RecA–tsDNA-C (red) and RecA–tsDNA-O (green)complexes, the base registry was taken to be Dn ¼ 23. The fitted parameters for RecA–tsDNA-C are: r2

a þ r2d ¼

3590 A2; 2rard ¼ 830 A2, h ¼ 4.9 A, Vh ¼ 13.38, and DV ¼ 308; for RecA–tsDNA-O, the parameters are: r2a þ r2

d ¼4400 A2; 2rard ¼ 550 A2, h ¼ 5.2 A, Vh ¼ 21.98, and DV ¼ 3608. (d) For the RecA–tsDNA-I (filled triangles, blue line),the best fit helical parameters are: r2

a þ r2d ¼ 4800 A2; 2rard ¼ 1900 A2, h ¼ 1.9 A, Vh ¼ 11.08, and DV ¼ 3408.

0.0

0.1

0.2

0.3

0.4

0.5

-5 0 5 10 15 20 25

Ene

rgy

Tran

sfer

Effi

cien

cy


0.0

0.1

0.2

0.3

0.4

0.5

-5 0 5 10 15 20 25

Ene

rgy

Tran

sfer

Effi

cien

cy


(c)

(d)


stretched by more than 40% of its original B-formlength and was unwound by 198. Within experi-mental uncertainty, the helical rise and twistangle are in quantitative agreement with thevalues of h ¼ 5:2ð^0:18Þ A12 and Vh ¼ 19.4(^1.4)814

measured previously. This successful applicationof quantitative interfluorophore distance measure-ments demonstrates that the FRET method is validand provides an accurate complement to EMmethods.

DNA helical parameters for RecA–tsDNAcomplexes

The energy transfer efficiencies between fluor-escein and Rox in the RecA-mediated strandexchange products were measured for the threeRecA–tsDNA complexes. For RecA–tsDNA-Cand -O, the same dual-labeled partial dsDNA FRi

described above was used as the homologousdsDNA substrate. During strand exchange, onestrand of the homologous dsDNA FRi serves asthe complementary strand while the other oneserves as the outgoing strand depending on thesequence of the incoming RecA–ssDNA filament.In these two cases, the fluorophores are located onthe outgoing and complementary strands. In orderto analyze the data, the base-pair registry (Dn )between the two strands was required to deter-

mine the effective separation of the fluorophores(n ). For the dsDNA and RecA–dsDNA structures,we assumed that the strands were paired in thecomplexes and that Dn ¼ 0: We initially assumedthat bases on the two strands of FRi remained inthe same base-pair plane in the RecA–tsDNA com-plex. Thus, the base-pair registry between fluor-escein and Rox remained the same as depicted inFigure 5(a) ðDn ¼ 0Þ: Based on this assumption,the measured energy transfer efficiency betweenthe two dyes for both complexes was plottedagainst their original base-pair separations, andthe E–n function (equation (3)) was used to fit theexperimental data (Figure 5(b), red squares forRecA–tsDNA-C and green circles for RecA–tsDNA-O). The best-fit helical parameters are: h ¼4:3 A; Vh ¼ 198 for the complementary strand andh ¼ 4:2 A; and Vh ¼ 318 for the outgoing strand(Table 3; Dn ¼ 0). Despite the similarity of the heli-cal rise parameters for the two strands, the helicaltwist angle of the outgoing strand is significantlydifferent from that of the complementary strand inthe RecA–tsDNA complex. This would lead to theconclusion that the originally base-paired comple-mentary and outgoing strands are not interwoundin this strand exchange intermediate. However,the helical rise parameters for both characterizedstrands of the RecA–tsDNA complex were sub-stantially different from that of the RecA–dsDNA

Table 3. Comparison of best-fit helical and dye-linker parameters for the five reaction systems

Dn h (A) Vh (8) ra2 þ rd

2 (A2) 2rard (A2) DV (8)

dsDNA 960 ^ 210 140 ^ 140 60 ^ 35

RecA–dsDNA 4.9 ^ 0.1 þ17.3 ^ 4.1 2000 ^ 90 280 ^ 80 12 ^ 27

RecA–tsDNA-I 1.9 ^ 3.7 þ11.0 ^ 3.3 4800 ^ 1600 1900 ^ 1700 340 ^ 5

25 6.0 ^ 0.2 þ16.5 ^ 1.7 4200 ^ 90 1000 ^ 60 72 ^ 624 5.6 ^ 0.1 þ19.1 ^ 1.7 4360 ^ 60 720 ^ 60 38 ^ 623 5.2 ^ 0.1 þ21.9 ^ 1.7 4400 ^ 50 550 ^ 60 360 ^ 722 4.8 ^ 0.1 þ25.1 ^ 1.5 4380 ^ 50 470 ^ 60 310 ^ 921 4.5 ^ 0.1 þ28.4 ^ 1.4 4330 ^ 50 450 ^ 60 250 ^ 13

RecA–tsDNA-O 0 4.2 ^ 0.1 þ31.1 ^ 1.5 4290 ^ 50 480 ^ 60 200 ^ 151 4.0 ^ 0.1 þ33.0 ^ 1.4 4250 ^ 50 530 ^ 60 150 ^ 152 3.7 ^ 0.1 þ34.4 ^ 1.5 4200 ^ 50 590 ^ 60 110 ^ 163 3.6 ^ 0.1 þ35.5 ^ 1.5 4160 ^ 50 640 ^ 60 60 ^ 174 3.4 ^ 0.1 þ36.8 ^ 1.8 4100 ^ 50 700 ^ 80 15 ^ 125 3.5 ^ 0.1 þ43.6 ^ 1.0 4100 ^ 60 1500 ^ 300 260 ^ 20

25 5.3 ^ 2.1 þ10.7 ^ 2.8 3430 ^ 1460 1500 ^ 320 80 ^ 5524 5.1 ^ 1.7 þ11.9 ^ 3.8 3560 ^ 1000 1120 ^ 530 60 ^ 4223 4.9 ^ 1.3 þ13.3 ^ 4.9 3590 ^ 730 830 ^ 580 30 ^ 2822 4.7 ^ 1.0 þ15.0 ^ 6.3 3550 ^ 580 610 ^ 540 5 ^ 1521 4.6 ^ 0.8 þ17.0 ^ 8.1 3500 ^ 500 450 ^ 470 330 ^ 27

RecA–tsDNA-C 0 4.3 ^ 0.7 þ18.6 ^ 10.8 3390 ^ 510 370 ^ 470 290 ^ 641 4.6 ^ 0.1 þ41.9 ^ 3.9 3100 ^ 80 390 ^ 270 90 ^ 362 4.3 ^ 0.1 þ41.9 ^ 3.9 3100 ^ 80 670 ^ 290 30 ^ 163 4.1 ^ 0.1 þ44.5 ^ 1.0 3000 ^ 90 900 ^ 300 340 ^ 114 4.0 ^ 0.1 þ44.8 ^ 0.7 3000 ^ 90 1100 ^ 300 300 ^ 95 3.8 ^ 0.1 þ45.0 ^ 0.6 2900 ^ 90 1260 ^ 300 250 ^ 8

The parameters are the best-fit solutions describing each reaction system’s E–n data set using equation (3). Each parameter rep-resents the mean value of at least three independent experiments plus-or-minus one standard deviation. For dsDNA, the helical riseper base-pair h and helical twist angle per base-pair Vh are fixed at 3.4 A and 368, respectively. The fitted parameter Dh (210.6 A)was used in the other four RecA-involved reaction systems in order to obtain the best fits of the other five parameters. For theRecA–tsDNA-O and -C series, the vertical base registry (Dn ) was varied from 25(50 end shift) to þ5(30 end shift), and the correspond-ing fitted parameters are shown.


complex. This observation led us to question thevalidity of the assumption regarding the base-pairregistry ðDn ¼ 0Þ:

A recent paper reported that the outgoing strandinside a post-strand-exchange RecA–tsDNA fila-ment may be shifted towards its 50 end three tofour bases with respect to the corresponding basesin the heteroduplex.23 Based on this prediction, werefitted our experimental data by systematicallyshifting the outgoing strand towards either its 50

end (Dn , 0) or its 30 end (Dn . 0). The best-fithelical parameters for each shift are listed in Table3. Based on the non-linear least-squares analysis,using reduced x2 as the fitting criterion, a registryshift from 21 to 25 produced quality fits withindistinguishable fitting goodness. We were unableto establish a quantitative criterion to select anabsolute registry shift from 21 to 25 bases. How-ever, the helical rise and twist angle parameters

for both the complementary and outgoing strandswere indistinguishable from one another or fromthose of RecA-bound dsDNA (Table 3) when Dn ¼23 (Figure 5(c)). This result is consistent with thepublished report.23 Moreover, the internal consist-ency of strands’ helical parameters would mini-mize topological stress in the RecA–tsDNAcomplex. One important implication of the 50

registry shift of the outgoing strand is the same asthat mentioned above based on the assumptionDn ¼ 0: Namely, the complementary and outgoingstrands are not paired in this strand exchangeintermediate.

In the RecA–tsDNA-I system (Figure 5(d), bluetriangles), a full-length fluorescein-labeled dsDNA(ds74F) was utilized. A set of Rox-labeled 45mers(45Ri) was used as the incoming ssDNA to pairwith ds74F. The sequence of this incoming ssDNAis identical to a 45 base sequence in the middle ofthe non-labeled strand of ds74F. Consequently,Rox labels the incoming strand. The best-fit helicalparameters to the E–n data were found to be:h ¼ 1.9 A, and Vh ¼ 118 (Table 3). The apparenthelical rise per residue is much smaller than thatof B-form dsDNA. Moreover, the residualuncertainty in the fit parameters is large(s(h) ¼ ^3.7 A). Further analysis of the spectraldata revealed that the fluorescein and Roxemission spectra for RecA–45R0–ds74F andRecA–45R4–ds74F show positions of maximumintensity that are blue-shifted about 1 nm com-pared to those spectra obtained for each fluoro-phore in isolation. The spectral maxima of thedyes at every other position in every other complexare consistent. Such spectral changes contra-indicate the application of equation (1) to the datafor complexes RecA–45R0 –ds74F and RecA–45R4 –ds74F.26,27 The Forster theory is based onweak electronic coupling between donor andacceptor fluorophores.24 This phenomenum indi-cates that there may be strong electronic coupling52

between the fluorescein and the Rox at positionsi ¼ 0 and i ¼ 4 for the RecA–tsDNA-I complex.Interestingly, the anisotropies of the fluorescein inthese two complexes were significantly lower(0.16 ^ 0.01) than those from the other RecA–tsDNA complexes (0.22 ^ 0.01) in the same system.While the nature of the relationships between theputative strong electronic coupling, the spectralchanges, and the reduced anisotropy values isbeyond the scope of this paper, the unique photo-physical characteristics of the first two membersof the RecA–tsDNA-I series undoubtedly result inthe low and imprecise h value.

Such strong electronic coupling between the twodyes may be caused by the particular junctionstructure in this system. In the RecA–tsDNA-Icomplex, the incoming ssDNA is shorter than thedsDNA substrate: there are 7 and 22 bp flankinghomologous sequences in the dsDNA substrate.After strand exchange, the outgoing strand cannotbe completely displaced because of the flankingbase-paired regions. The junction between the

Table 4. Summary of experimental helical geometryparameters and dye-linker structural parameters forB-DNA and the RecA-bound dsDNA and tsDNAcomplexes

h (A)a Vh (8)a rd (A)b ra (A)b

A. Individual analysisa

B-dsDNA (3.4) (þ36) 31 ^ 3 2 ^ 2RecA–dsDNA 4.9 ^ 0.1 þ17 ^ 4 44 ^ 1 3 ^ 1RecA–tsDNA-Cc 4.9 ^ 1.3 þ13 ^ 5 59 ^ 6 7 ^ 5RecA–tsDNA-I 1.9 ^ 3.7 þ11 ^ 3 68 ^ 13 14 ^ 14RecA–tsDNA-Oc 5.2 ^ 0.1 þ22 ^ 2 66 ^ 1 4 ^ 1

B. Global analysisb

RecA–tsDNA-Cc 4.8 ^ 0.4 þ15 ^ 4 60 ^ 2 7 ^ 2RecA–tsDNA-I 4.8 ^ 0.4 þ15 ^ 4 53 ^ 2 5 ^ 2RecA–tsDNA-Oc 4.8 ^ 0.4 þ15 ^ 4 68 ^ 2 5 ^ 2

a The helical rise (h ) and helical twist (Vh) per nucleotidewere obtained as described in the text and the legend to Figure5. The values for the protein-free dsDNA were taken as thecanonical values (in parentheses), while the value for theRecA–DNA complexes are from Table 3. The uncertainties arerepresented by one standard deviation as described for Table 3.The r parameters describe the radial distances of the donor (sub-script d) and acceptor (subscript a) fluorophores from the helixaxis as described in the legend for Figure 2. The values of rwere calculated from the two combined fitting parameters,(ra

2 þ rd2) and 2rard (Table 3), by solution of two simultaneous

equations as described in the text. The uncertainties in the rvalues were calculated by propagation of the standarddeviations in the combined fitting parameters (Table 3).

b The parameters were obtained by global data analysis inwhich h and Vh were adjustable parameters fitted simul-taneously to the data for all three strands while the respectivedye-linker structure parameters (e.g. ra

2 þ rd2 and 2rard) were sep-

arately fitted to each strand’s E–n data (Figure 6) as described inthe text and the legend to Figure 6. The uncertainties for the twohelical geometry parameters and the two combined dye-linkerstructure parameters were estimated from a reduced x2 analysislike that described in Figure 7. The values of r were calculatedfrom the two combined fitting parameters, ra

2 þ rd2 and 2rard, by

solution of two simultaneous equations as described in the text.The uncertainties in the r-values were calculated by propagationof the estimated errors in the combined fitting parameters.

c The helical parameters listed for the complementary andoutgoing strands assume a base registry shift of Dn ¼ 23 forthe outgoing strand (see the text).


paired and the displaced segments of the outgoingstrand may create an environment in which thefluorescein on the complementary strand and theRox at position zero or four on the incoming strandare brought close together, and they interactstrongly as a result. This idea was tested by a con-trol experiment, in which a fluorescein-labeled par-tial dsDNA (FX) was paired with the RecA–45Ri

filament. Because the outgoing strand now has thesame length as the incoming strand, a D-loop junc-tion should not be formed. Indeed, the spectralshifts and abnormally low anisotropies were nolonger observed (data not shown). This idea couldbe exploited if we were able to use the alternativepartial dsDNA FX for the RecA–tsDNA-I complex.Unfortunately, because the apparent dissociationconstant between this partial dsDNA FX and theRecA–45Ri filament is relatively high (0.03 mM),we were not able to observe 90% complex for-mation at reasonable concentrations of both reac-tants. In addition, we were unable to choose twoalternative thymidine positions for Rox-labelingon the incoming 45mer. Based on these obser-vations, we conclude that problems associatedonly with the data collected for RecA–tsDNA-Iwith Rox at i ¼ 0 and i ¼ 4 positions led to thesmall values and relatively large errors of the heli-cal parameters obtained for this complex.

Although the helical rise h was not determinedwith high precision for the RecA–tsDNA-I series,both h and Vh are similar to those parametersobtained for RecA–tsDNA-C and -O with a regis-try shift Dn ¼ 23. Moreover, each helical geometryparameter for the three systems was indistinguish-able (accounting for experimental uncertainty)from the weighted mean value of each parameter(derived from all three RecA–tsDNA data sets),h ¼ 5.2(^0.1) A and Vh ¼ 18(^3)8. Based on theconformational similarity among the three strands,we used a global data analysis in which the helicalrise and twist angle were adjustable parametersfitted simultaneously to the data for all three

strands (the data points corresponding to the firsttwo i values for the RecA–tsDNA-I complex hav-ing been omitted), while the respective dye-linkerparameters were separately fitted to each strand’sE–n data (Figure 6). This algorithm provided thebest helical rise and twist angle for the simul-taneous description of all three strands. We foundthat the best fit parameters, h ¼ 4.8(^0.4) A andVh ¼ 15(^4)8, for a three-base registry shift (Table4) were in good quantitative agreement with theweighted mean parameters. These data are consist-ent with the conclusion that, within uncertaintylimits, all three strands inside the RecA filamentshare an identical helical geometry.

Probing the potential effect ofATPgS hydrolysis

Prior to undertaking a more careful structuralinterpretation of these results, we wished toaccount for the uncertainty in the helical par-ameters. We envisaged two main sources of uncer-tainty: those arising from experimental conditionsand those arising from the analytical methodology.For the former set, possible structural pertur-bations from ADP, the nucleotide product ofATPgS hydrolysis, were investigated. Because theRecA protein filament is known to have differentstructures and DNA-binding properties in the pre-sence of ATPgS and ADP,53 the presence of a sub-stoichiometric amount of ADP could impact theobserved FRET efficiencies. The same set of reac-tions described above was performed under thesame conditions using the non-hydrolyzable ATPanalog ADP-AlF4 in place of ATPgS. The measuredenergy transfer efficiency using ADP–AlF4 weresystematically about 20% lower than thosemeasured in the presence of ATPgS (data notshown). However, the trends in the E–n dataremained the same. Most significantly, curve-fittingusing equation (3) generates essentially identicalDNA helical parameters with slightly different

Figure 6. Global non-linear least-squares analysis of E–n data forthree RecA–tsDNA reaction sys-tems using the same helical riseand twist angle for all three strands,while leaving the dye-labeling par-ameters variable within each series.The data symbols and best-fitcurves are color-coded as describedfor Figure 5. The first two datapoints of the RecA–tsDNA-I serieswere excluded in this global fittingfor reasons discussed in the text. Aunique and common helix structure(h ¼ 4.8(^0.4) A, Vh ¼ 15(^4)8)were found to fit all three strandssimultaneously.

0.0

0.1

0.2

0.3

0.4

0.5

-5 0 5 10 15 20 25

Ene

rgy

Tran

sfer

Effi

cien

cy



dye-linker parameters for each system. Thisdemonstrated that the existence of ADP resultingfrom ATPgS hydrolysis has negligible impact onthe structure of the DNA helix in our reactiontime scale.

Evaluation of potential analytical sources ofuncertainty in the helical parameters

As described above, a second potential source ofuncertainty is the analytical methodologyemployed. Within this context, we identified threeparticular elements warranting further investi-gation. First, calculation of the Forster critical dis-tance R0 required the assumption that theorientation factor is constant, k2 ¼ 2=3: Second,curve-fitting produced the most robust fits uponthe assumption of a constant dye-linker parameter,Dh ¼ 210:6 A (determined from the protein-freeB-DNA data). Finally, non-linear least-squaresanalysis of the E–n function using multiple adjust-able parameters could lead to unreliable parametervalues. Each of these issues was evaluated in turnto evaluate the accuracy of the helical parameters.

Although the assumption of a constant k2 valuecorresponding to randomly oriented fluorophorehas been validated in a number of FRET systems,we estimated the maximum and minimum valuesof k2 in our system to explore how much uncer-tainty was introduced by assuming a k2 value of2/3. The k2 factor cannot be measured experimen-tally, but the upper (kmax

2 ) and lower (kmin2 ) limits of

k2 can be obtained from the measured limitinganisotropy of the donor and acceptor and the cal-culated axial depolarization factors using a pub-lished procedure.54 The kmax

2 and kmin2 for the

fluorescein and Rox in the RecA-bound dsDNAwere calculated at 3.5 and 0.064 as described inMaterials and Methods, which provided a widerange for R0: 0.68R0 , R0 , 1.32R0.

The calculation of kmax2 and kmin

2 using equation(10) provides a worst-case estimation, whichusually overestimates the effects of k2 on the calcu-lated distance.29 For fluorophores with mixedextents of polarization, r0 , 0.3, the error in dis-tance is thought to be below 10%.29 Although therange of possible Forster distances is wide, theuncertainty in the determination of the averagedistance between fluorescein and Rox will be smal-ler than suggested by the analysis of the orien-tation factor k2 value. Indeed, employing eitherR0,min or R0,max in place of R0 in our analysis led todramatically increased error ranges of the fittedparameters and significantly decreased goodnessof fitting in all five systems. Combined with theindependent anisotropy analysis presented above,the loss of fitting robustness indicates that the truek2 is substantially different from kmin

2 and kmax2 , and

validates our use of k2 ¼ 2=3:In all five systems, the same E–n function was

used to fit the relationship between the measuredenergy transfer efficiency and the DNA structure.The dsDNA system was first fitted by assuming a

helical rise of 3.4 A and a twist angle at 368. Thedye-labeling parameter, Dh ¼ 210:6 A; which isthe difference between the vertical positions of thetwo dyes if they were attached to the same base-pair, was kept constant in the fitting of the otherfour systems. Although this is based on the reason-able assumption that the dye-linker structuresremain essentially unaltered across all five sys-tems, its impact on the fitting results was exploredby allowing Dh to vary. The same curve-fitting pro-cedure, using the E–n function, was performed oneach RecA–DNA complex while the value of Dhwas systematically varied from 25 to 215 A. Therange was chosen based on the nucleotide pos-itions labeled by the fluorophores and the possiblegeometric arrangement of the dye linkers. It wasfound that the fitted parameters in each systemonly vary about 10 to 15% when Dh is varied from25 A to 215 A (data not shown). This range hasalready been accounted for by the standard devi-ation weighting procedure of the E–n functionfitting in each system. Therefore, we drew the con-clusion that potential variability of the dye-linkerparameter Dh exerted a minimal impact on ourfitting results.

The most informative results derived from theanalysis of FRET efficiencies described above arethe h and Vh helical descriptors; however, theseare only two of five adjustable parameters necess-ary to fit the helical geometry model to the E–ndata. A key issue in the interpretation of the helicalparameters is their robustness as quantitativestructural descriptors. Non-linear least-squaresanalysis of the data sets using the helical geometrymodel allowed this issue to be explored. Asdescribed in Materials and Methods, the values ofall five parameters were simultaneously varied foreach reaction system to minimize the reduced x2.The parameter space around all five parametersfor each of the three RecA–DNA complexes wassearched using a manual grid search. The resultsof this analysis are displayed as two-dimensionalcross-sectional plots of x2 versus h and Vh values(Figure 7(a) and (b)). These plots demonstrate thequadratic nature of the x2 surface near the opti-mum for each dataset’s parameters and verifiesthat the fitting routines were not trapped in a localminimum. Moreover, the steepness of the surfacesnear the optima provides an estimate of howreliably determined each best-fit parameter is. Forthe helical rise per residue, the h parameter valuesdescribing the RecA–dsDNA, RecA–tsDNA-C,and RecA–tsDNA-O complexes are unequivocallydetermined. In contrast, the h value of the RecA–tsDNA-I complex is much less well determined, afact that recapitulates the statistical imprecision infitting this parameter (Table 3). For the helicaltwist angle, the Vh parameter values are approxi-mately equally well defined for all four reactionsystems. The plots in Figure 7 allow us to assignconservative round upper estimates to the uncer-tainties in the best-fit h and Vh values as ^0.4 Aand ^48, respectively. The same analytical method


applied to the parameters derived from the globalanalysis (x2 plots not shown) led to the same uncer-tainty estimates.

Upon consideration of all three sources ofexperimental uncertainty described above, it isapparent that each source yields uncertainty levelsthat are adequately accounted for by the statisticalmeasures of precision previously tabulated (Table4). Hence, we used these latter values as reasonableestimates of the level of experimental uncertaintyin each helical parameter.

Discussion

We have presented a systematic implementationof FRET to characterize the structure of the RecA–tsDNA intermediate in RecA-mediated strandexchange. The FRET pair, fluorescein and Rox,was chosen to label a series of DNA molecules,and the energy transfer efficiency between themwas measured in different DNA structures. The

energy transfer efficiency is quantitatively relatedto the distance between the two dyes, which iscorrespondingly related to the particular DNAstructure. By solving the relationship between theenergy transfer efficiency and the DNA structure,we were able to extract structural information,such as the helical rise and twist angle, for eachparticular RecA–DNA complex. The data are con-sistent with the conclusion that all the DNAstrands in both RecA–dsDNA and RecA–tsDNAare similarly stretched and unwound (h < 5 A andVh < 208) relative to B-DNA. As discussed below,the data can be further analyzed to resolve keystructural features of the strand exchangeintermediates.

DNA structure inferred frominterfluorophore distances

The systematic application of FRET describedabove provides an overlapping set of interfluoro-phore distances, which are partly determined by

Figure 7. Reduced x2 valuesplotted as a function of the fit par-ameters (a) h and (b) Vh for theRecA–dsDNA (black), RecA–tsDNA-I (blue), RecA–tsDNA-C(red), and RecA–tsDNA-O (green)complexes. For the helical rise perresidue, the h values describing theRecA–dsDNA, RecA–tsDNA-C,and RecA–tsDNA-O complexes areunequivocally determined. In con-trast, the h-value of the RecA–tsDNA-I complex is much less welldetermined, a fact that recapitulatesthe statistical imprecision in fittingthis parameter (see Table 3). Forthe helical twist angle, the Vh

values are approximately equallywell defined for all four reactionsystems.

0

5

10

15

20

0 1 2 3 4 5 6

χχχχ 2

Helical Rise / Residue (Å)

0

5

10

15

20

25

0 5 10 15 20 25 30 35 40

χχχχ 2

Helical Twist / Residue (˚)

(a)

(b)


the structures of the DNA strands to which thedyes are conjugated. The interfluorophore dis-tances are also sensitive to the distance betweenthe dye and the conjugated thymine heterocycle,and the extraction of structural informationbeyond helical geometry parameters requires aknowledge of the conformation of the dye linkerarms. Unfortunately, the dyes are conjugated tothe DNA via flexible alkyl chains in order to pro-vide sufficient rotational freedom to justify theapproximation k2 < 2/3 and alleviate the other-wise requisite knowledge of the orientations of thetransition dipoles of the two dyes.26

In order to resolve this dilemma and allow theuse of interfluorophore distances as exact struc-tural constraints, the geometries of dye-DNA con-jugates have been resolved for the dye pairsfluorescein/cyanine-351 and fluorescein/5-tetramethylrhodamine.35,49,50 Whereas the conju-gates employed here involve covalent attachmentto thymine moieties at internal nucleotidepositions, in both cited cases the dyes are coupledto the DNA via deoxyribose moieties at terminalpositions. Nevertheless, a consideration of the pre-vious works is instructive and relevant. In general,NMR structure determination,51 crystallographicstructure analysis,50 computational modeling,50,55

and fluorescence anisotropy measurements46,49 incombination with interfluorophore distancemodeling50,51 have led to the following con-clusions: the linker to the negatively chargedfluorescein dye adopts a conformation that ismaximally extended, while the linker to theother dye (either cyanine-3 or 5-tetramethyl-rhodamine) adopts conformations that allow thepositively charged polycyclic moiety to stack onthe DNA helix end.30

While the helix geometry model does not allowthe unique assignment of dye-linker conformationsor lengths, we can extract the pair of radial dis-tances, ra and rd, from the linear combination ofthe combined fitting parameters, (ra

2 þ rd2) and

2rard. The measurement of FRET efficiencies usingthe protein-free control B-DNA molecule revealedthat the distance from the helix center to one dyeis only about 2 A, while the distance to the otherfluorophore is about 31 A (Table 4). This obser-vation is consistent with published data (seeabove), suggesting that tetramethylrhodamine(analogous to Rox) is likely to coil back close tothe DNA helix while fluorescein is maximallyextended laterally from the DNA helix. Corrobora-tive evidence is derived from the anisotropymeasurements demonstrating that Rox adopts amore rigid conformation than fluorescein (Table1). Because Rox resides in the major groove (Roxis attached to C5 of thymine) and the distancebetween Rox and the central helix axis of B-DNAis shorter than the radius of a typical B-DNAbase-pair (<5 A),56 Rox most likely adopts a pos-ition near the floor of the major groove50 within<2 A of the helix axis. Hence, it is reasonable toassign tentatively the shorter (ra) and longer (rd)

dye-linker distance to Rox and fluorescein,respectively.

Interestingly, this general case of one long dye-linker and one short dye-linker was observed inall the reaction systems (Table 4). However, foreach of the RecA–DNA complexes, the value forone of the dye-linker distances was significantlylonger than can be accounted for by an extendedalkyl linker. This result may arise due to (i) asystematic problem with the fitting procedure,(ii) an altered DNA structure near one of the dyes,or (iii) a combination of both. If the anomalouslylong dye-linker distance is a result of a systematicproblem, the meaning of the distances and theirapplication would be suspect. It is important tonote that the helical geometry parameters, notbeing correlated with the radial distance par-ameters (see above), are not invalidated by thispossibility. On the other hand, if the apparentanomaly is the result of an altered DNA structure,then the dye-linker distance values would be use-ful. We argue that this latter case is the more likelyscenario.

Four experimental observations argue againstthe possibility of a systematic error in the data col-lection and analysis. First, the spectral character-istics (Table 1) and FRET efficiencies (Table 2) forthe RecA–DNA complexes are reasonable andhighly reproducible. Second, the dye-linker dis-tance values in the case of the protein-free DNAare physically reasonable and in agreement withprevious determinations. Third, the helicalgeometry parameters, h and Vh, for the RecA–dsDNA complex, also obtained by curve-fitting,are in good agreement with those obtained byother methods. Finally, similar values for the dye-linker parameters are obtained for the RecA–tsDNA complexes whether the data are analyzedindividually or globally (Table 4).

One important alternative to a systematic experi-mental error is the possibility that the dye-linkerradial distance accurately reflects an altered DNAgeometry near the fluorophore. Three sets of factssuggest that this is a possibility. First, the experi-mental design places the thymidine residue towhich fluorescein is conjugated one nucleotideposition away from the end of the designated45 nt region of homology (Figure 3). This designinadvertantly places the site of fluorescein at apotentially flexible junction between structuraldomains. Second, there is evidence from restrictionendonuclease reactivity,57 affinity modification,58

and computational modeling23 that the junctionsbetween a RecA–tsDNA structure and dsDNA (orRecA–dsDNA) may adopt an irregular structure.Finally, there is evidence that a fluorescein labelcan interact non-covalently with RecA protein in anucleoprotein filament.59 In addition, our ownobservations of anomalous spectrofluorometricbehavior of fluorescein in two of the RecA–tsDNA-I complexes are consistent with an unusualjunction structure (see above). Taken together,these data suggest that the fluorescein located at a


point of DNA structural transition may be posi-tioned further than expected from the helical axisat a site favored by protein-dye interactions.

Based on the evidence and reasoning presentedabove, we conclude that the radial distance par-ameters, ra and rd, are valid and meaningful. More-over, for each complex, we assign the longerdistance to fluorescein and the shorter distance toRox in analogy with the assignments for protein-free dsDNA (Table 4). While the assignment of theradial distances to specific fluorophores is not aprerequisite for a description of the helicalgeometry of the DNA nor the subsequent struc-tural interpretation, as mentioned in the openingparagraph of this section, a knowledge of thelocation of the fluorophores relative to the DNAprovides a necessary foundation for drawing con-clusions about the relative positions of the DNAstrands. We reiterate, however, that accurate helicalgeometry data is obtained directly from the par-ameters, h and Vh, and does not depend on thefluorophore dye-linker structures.

Structural features of RecA-bound dsDNA

After confirming that interfluorophore energytransfer measurements using this system of fluor-escein- and Rox-labeled DNA strands could satis-factorily describe the canonical B-DNA geometry,the energy transfer efficiency was measured onthe same series of dsDNA molecules after additionof excess RecA protein. For each of the six dsDNAmolecules, an increase of the fluorescein emissionpeak and decrease of Rox emission were observed,indicating that the dsDNA was stretched and/orunwound by the binding of RecA protein. Thefitted helical rise for the RecA-bound dsDNA is4.9(^0.1) A and the helical twist angle is 17(^4)8.The significant difference in the structure of thisRecA-bound dsDNA in comparison to standardB-form dsDNA revealed the simultaneous exten-sion and unwinding of the DNA molecule by thebinding of the RecA protein. As discussed below,the values of these two helical geometry par-ameters determined simultaneously from FRETefficiency measurements are in excellent agreementwith those parameter values measured indi-vidually using other methods. Furthermore, theanalyses of interfluorophore distances in the con-text of the helical geometry model provide animportant modification to a widely acceptedmodel of the RecA–dsDNA nucleoproteinfilament.

Over the past 20 years, considerable progress hasbeen made in determining the structures of theRecA protein as well as those of the RecA–ssDNAand the RecA–dsDNA filaments. EM providedthe first images of the RecA nucleoproteinfilament.60 It was observed that the cofactor had adrastic effect on the observed pitch of the helicalfilament. A “collapsed” filament with a helicalpitch of ca 75 A occurs in the absence of a cofactoror with ADP. In contrast, an “extended” filament

(95 A pitch) is observed with non-hydrolyzableanalogs of ATP. The extended conformation is theactive form of the RecA nucleoprotein filament,since the quarternary complex formed betweenRecA protein, ssDNA, ATP and Mg2þ in vitro isthe species that pairs and exchanges homologousDNA strands.61,62 The crystal structure of a RecA–ADP complex provided the first high resolutionview (2.3 A) of the protein-only filament.63 How-ever, the conformation of the filament in the crystalis probably that of the inactive collapsed filamentsince the pitch (83 A) is not as extended as thatseen when DNA is bound.64 The active confor-mation is achieved only in the presence of bothDNA (either single or double-stranded) and ATP(or a slowly or non-hydrolyzable analog, such asATPgS or ADP–AlF4

2, respectively).65 In theensuing discussion, only complexes, and theirrelated structural parameters, formed in the pre-sence of ATPgS (or ADP–AlF4

2) will be addressed.EM of RecA–dsDNA complexes have provided

the most consistent and highest resolution imagesof RecA nucleoprotein filaments. Early EM imagesrevealed that the contour lengths of RecA–dsDNA filaments were 50–60% longer than thatof the underlying protein-free replicative formDNAs from bacteriophage,12,16,66 and allowed thedetermination of a helical rise per base-pair,h ¼ 5.2(^0.18) A.12 This value is identical withinexperimental uncertainty to that derived herein(4.9(^0.1) A) from FRET efficiencies for fluoro-phore pairs separated by no more than 20 bp.

While EM has allowed RecA nucleoprotein fila-ments to be directly visualized and a likely DNAbinding location to be identified,10,11 it has not pro-vided for the direct characterization of the DNAstrands buried inside. A key limitation in thisregard is the relative masses of the filament com-ponents: the two DNA strands in a RecA–dsDNAcomplex account for less than 5% of the total massof the complex.10,11 Nevertheless, a comparison ofthe number of turns of the overall nucleoproteinfilament images with the number of base-pairs inthe substrate dsDNA revealed that the averagepitch of the dsDNA was 18.6 bp (Vh ¼ 19.48).13

Using an assay to monitor changes in the topo-logical linking number of supercoiled DNA,Radding and co-workers first demonstrated aRecA-induced reduction in the twist of substratedsDNA.67,68 Subsequently, several laboratorieshave used related assays to quantify the extent ofunwinding, reporting helical repeats of18.6(^1.3),14 17(^1),15 21(^5),16 and 17.9(^0.3)bp.17 Thus, in spite of a number of critical assump-tions and limitations in the topologicalassays,14,15,17,69 – 71 a remarkable agreement with theprediction derived from EM images has beenachieved. The helical repeat of the RecA-bounddsDNA derived from the FRET efficiency analysis(360/(17(^4)) ¼ 21 ^ 5.0 bp/turn) is in excellentagreement with this collection of literature values.

The quantitative agreement between theFRET-derived helical geometry parameters for


RecA-bound dsDNA and the published character-izations of RecA–DNA filaments provides animportant validation of the FRET efficiencymeasurements for the quantitative elucidation ofRecA–DNA complex structures. Moreover,because the helical geometry parameters weredetermined from interfluorophore separations ofno more than 20 nt, these parameters provideimportant confirmation that the average helicalparameters derived from EM and topologicalmeasurements are reflected by the local DNAstructure. Hence, the combination of global andlocal structural analyses corroborate the hypothesisthat the DNA structure is uniform throughout thenucleoprotein filament.

Functional implications of the RecA-bounddsDNA structure

The apparent local structural uniformity of theDNA is important in the context of RecA–DNAinteractions. As discussed previously,10 the factthat the DNA adopts the same helical geometry asthe RecA filament suggests that the observed stoi-chiometry of 3 bp per RecA monomer arises fromstructural coincidence of the protein and DNAwithin the helical filament. The conclusion thatcontacts between the protein and DNA within thecomplex would be equivalent for every RecAmonomer is supported by the uniformity of thelocal structure of the DNA.11

Published results suggest that the RecA proteininteracts with dsDNA along its minor groove.16,72,73

One possible result of the stretched and under-wound DNA structure imposed by RecA bindingis to widen the minor groove and move its floorcloser to the cylindrical surface contour of theDNA molecule.18 Such structural changes wouldlikely provide for more facile interaction betweenthe interior surface of the RecA protein filamentand the exposed floor of the minor groove ofDNA, as described for TATA box-bindingprotein19,74 and architectural proteins75 bindingtheir cognate dsDNA sequences. Although it isgenerally accepted that RecA binds DNA withouthigh sequence specificity, the protein demonstratessome sequence preferences in its binding to single-stranded DNA.76 – 81 The putative existence of directcontacts between RecA protein side-chains andDNA functional groups is supported by the afore-mentioned structural coincidence of the proteinfilament with the bound DNA. Given the openinterior of the RecA filament,63 the formation ofsuch interactions would require the base-pairs ofDNA to be displaced away from the central helixaxis.

Based on the conclusions described above for theB-DNA dye-linker conformations, Rox most likelysits on the major groove edge of the thymine-adenine base-pair with a distance of 2 A to thebase-pair center. Moreover, if the base-pairs aredisplaced away from the helix axis, the axis mustbe located within the major groove.56 This require-

ment results from the fact that the DNA strands inthe RecA–dsDNA complex are right-handed (asindicated by the positive sign of the helical twistangle). Therefore, the maximum distance from thehelix axis to the base-pair center is the sum of thedistance from the helix center to Rox (ra ¼ 3 A)and the distance from Rox to the base-pair center(2 A). This approximation yields a distance of5(^3) A between the helix axis and each base-paircenter. Hence, the base-pair centers are likely dis-placed away from the helix axis and a hollowhelix center may be present. Consistent with thishypothesis, the application of a geometric modelrequiring coincidence of the base-pair centers andhelix axis generates an unreasonable fit (notshown).

The interfluorophore distance data presentedhere are consistent with the widely acceptedmodel of stretched and underwound RecA-bounddsDNA, but the data further suggest that thebase-pairs are not coaxially stacked. The possibilitythat the base-pairs are radially displaced from thehelix axis is not incompatible with the helical geo-metry parameters, h < 5 A and Vh < 208. Egelmanand co-workers have shown that the radial dis-tance between the base-pair centers and thefilament axis is a function of the local separationbetween phosphate groups along each DNAstrand.10,11 In turn, the helical path length alongthe phosphate backbone is constrained by stereo-chemical forces and has a maximum value of7.6 A. Moreover, as the models in Figure 4 ofEgelman & Yu11 indicate, relatively small changesin the local separation between phosphate groups(e.g. 5.5 ! 7.2 A) are correlated with substantialchanges in helical radius of the DNA (e.g.6 ! 15 A). Importantly, DNA structure modelsentirely consistent with the combination of con-straints imposed by the measured helical geometryparameters, stereochemical backbone limitations,and significant radial displacement of the base-pairs have been predicted for stretched dsDNA19

as well as for dsDNA in RecA nucleoproteinfilaments.10,11 Our data are consistent with theDNA structural models presented in Figure 4(b)of ref. 11 and in Figure 6 (“At” family) of ref. 19by the previous authors. The implications of sucha DNA structure for the RecA–tsDNA strandexchange intermediate are discussed further below.

Structural features of three DNA strands in thestrand-exchange intermediate

We used the same method to measure the energytransfer efficiencies as a function of base-pairseparation for each strand in the RecA–tsDNAcomplex representing the strand exchange inter-mediate. In each complex, Rox was labeled on onestrand of the three strands and its position wasvaried throughout the strand. This enabled us toisolate the helical geometry information for thatparticular strand relative to the fluorescein-labeledstrand.


For the complementary and outgoing strands,the requirement that the corresponding bases onthe complementary and outgoing strands sharethe same base-pair plane results in significantlydifferent fitted twist angles for the two strands. Inturn, this difference excludes the possibility thatthe two strands are interwound in a cohelical man-ner. This inference is challenged by the followingproblems. First, a separation of the two strandswould require that the topological tension betweenthe originally interwound strands be released. Yet,it is well known that the RecA protein does notdemonstrate helicase activity,82 and under our reac-tion conditions, ATP hydrolysis is absent. Second,if the outgoing strand is not cohelical with theRecA filament or the complementary strand, theprotein monomers must make discontinuous con-tacts with one of the two strands. To date, the sortof structural discontinuities in the protein filamentrequired for such contacts have not been reported.Lack of a satisfactory solution to these twoproblems led us to reconsider the structuralassumption.

These issues were readily resolved by shiftingthe outgoing strand three bases toward its 50 end,as predicted in a recent report.23 For a three baseregistry change between the outgoing and comple-mentary strands, the best-fit helical parameters areh ¼ 4.9(^1.3) A, Vh ¼ 13(^5)8 for the complemen-tary strand, and h ¼ 5.2(^0.1) A, Vh ¼ 22(^2)8 forthe outgoing strand. The conformation of the twostrands now exhibits marked similarity. When con-sidered together with the helical parameters of theincoming strand (h ¼ 1.9(^3.7) A, Vh ¼ 11(^3.3)8),we noted that all three sets of helical parametersfall into the category of the RecA-bound extendedand unwound DNA strands (weighted mean par-ameters, h ¼ 5.2(^0.1) A and Vh ¼ 18(^3)8).

Indeed, the data sets of all three RecA–tsDNAcomplexes can be fit simultaneously by a commonhelical geometry (Figure 6). The best-fit commonhelical parameters for a registry shift of threebases for the outgoing strand, h ¼ 4.8(^0.4) A,and Vh ¼ 15(^4)8, are consistent with the weightedmean parameters. The global analysis results areconsistent with a structural model in which allthree RecA-bound DNA strands are cohelicalwithin the RecA filament. This is the first directevidence that all three DNA strands in the RecA–tsDNA complex are simultaneously extended andunwound and share a common helical geometrywith the RecA-bound dsDNA (Table 4).

EM has been invaluable in imaging RecAprotein-only and protein-DNA filaments. Micro-graphs of RecA–DNA complexes trapped duringsequential stages of in vitro strand exchange haveallowed visualization of genetic recombinationintermediates.8,9 Moreover, it has been proposedthat the stretched and partially unwound DNAstructure imposed by RecA binding is instrumentalin the process of strand exchange.9,18,20,83 – 87 In spiteof this success, no structures of RecA–tsDNA com-plexes to the resolution of the protein monomer

have been reported, in part because the conditionsof the reaction do not allow continuous coverageof the DNA by RecA and the necessary sample fix-ation results in compaction of the complexes.8,9

Thus, hypotheses regarding the molecular mechan-ism of strand exchange have been largely evalu-ated in the absence of direct structural data to thispoint.

Radding and co-workers used the topologicalassay described above to demonstrate that there isextensive unwinding of the substrate dsDNAduring the early stages of DNA strand exchangeinitiated by a RecA–ssDNA nucleoproteinfilament.67,68 Several groups have corroborated andextended the early results.88 – 90 More recently, twolaboratories have employed related assays to quan-tify the extent of unwinding of the substratedsDNA, reporting helical repeats of ca 1991 and2792 bp per turn in the region of homologous pair-ing. The helicity determined herein for all threestrands of the RecA–tsDNA complex (360/(18(^3)) ¼ 20(^3.3) bp/turn) agrees closely withthe smaller reported value.91 Although many limi-tations in the use of topological assays have beendiscussed (see above), it is not clear whether thelarger reported value92 is a result of a systematicdifference in those authors’ experiments (e.g. theinclusion of ADP in the reaction buffer) or maysimply reflect a large inherent uncertainty in suchan analysis.16 We conclude that all three strands inthe RecA–tsDNA complex share a common heli-city and that the average helical twist parameterfor those strands (Vh ¼ 18 ^ 3) is indistiguishablefrom that of the better characterized RecA–dsDNA complex.

To the best of our knowledge, no laboratory haspublished a measurement of the helical rise orpitch for any of the DNA strands in a RecA–tsDNA complex. However, Malkov and co-workersconcluded that the distribution of complemetaryand outgoing strand breaks resulting from theAuger radiodecay of a site-specific 125I label wereconsistent with an extended helical rise for bothstrands.23 The average helical rise between nucleo-tides of each of the three strands (h ¼ 5.2(^0.1) A)confirms the expectation that the extension of theDNA in the RecA–tsDNA complex is indistin-guishable from that observed for RecA-bounddsDNA.

Implications for the structure of the strandexchange intermediate

Using agarose gel mobility shift assays, wedemonstrated that the conjugation of the dye-linker structures to DNA on the major grooveedge of thymine has no effect on the efficiency ofstrand exchange. Moreover, the assay results allowus to conclude that the strand exchange inter-mediate retains all three DNA strands bound tothe protein filament. This is consistent with theconclusion drawn from previous studies that nostrand exchange products are released from the


filament in the absence of ATP hydrolysis.32,62,93 – 96

Of course, the gel assay result is also consistentwith the observation of strong FRET signals whenusing the fluorophore-conjugated outgoing strandas the excited state energy donor or acceptor.Finally, the analysis described above demonstratesthat an outgoing strand registry shift of threenucleotides is required for reasonable, internallyconsistent structural parameters. Taken together,these conclusions support only one model for theRecA–tsDNA complex we have probed. Namely,that the RecA–tsDNA complex must represent anadvanced, post-strand-exchange intermediatewherein the outgoing strand has been unpairedfrom the original dsDNA but remains bound tothe nucleoprotein filament. Importantly, the combi-nation of inter-strand distances (inferred frominterfluorophore distances; Table 2) and registryshift of the outgoing strand exclude the possibilitythat a triplex DNA, involving at least one set ofnovel, non-Watson–Crick hydrogen-bond inter-actions between the bases of an intact duplex andthose of a third strand, represents this RecA–tsDNA complex. Elements of the structural modelare consistent with recent data relying on a varietyof biochemical methods including chemicalmodification,97,92 affinity modification,98 – 100 top-ology testing,91,101 fluorescence,31,32 gel mobilityshift,102 and radioprobing.23

The next interesting question to address involvesthe structural organization of the three strandswithin the RecA filament. Given the commonhelical parameters of all three strands and theirstructural coincidence with the protein filament,the simplest assumption for the relative positionsof the three strands is that they have a commonhelix axis and are interwound with each other.This arrangement presents minimal topologicalproblems and does not require any energy inputfrom ATP hydrolysis. However, the determinationof the location of the three strands in relation tothe helix center requires a knowledge of the confor-mation of the dye linker arms.

Based on the conclusions regarding the B-DNAdye-linker conformations and the approximationsdescribed for the RecA–dsDNA complex, thelength of the smaller dye-linker radial projection(ra) likely provides information about the positionof Rox and the strand to which it is conjugated.An analysis of the radial distances between eachRox and the helix center (ra; Table 4) indicate thatthe Rox of the outgoing, complementary, andincoming strands are located 4(^1), 7(^5), and14(^14) A, respectively, from the common helixaxis. There is considerable uncertainty in two ofthe three ra values and it is not possible to dis-tinguish the three strands’ radial positions withconfidence. Nevertheless, when taken togetherwith the magnitudes of the ra parameters, theresults of the global analysis suggest that ra . 0for all three strands (Table 4). As discussed above,the position of Rox is likely within 2 A of thethymine residue to which it is conjugated. Hence,

the maximum distance from the helix center to thebase center of a strand is the sum of the distancefrom the helix center to Rox and the distance fromRox to the base center (2 A). We conlcude that thedistances from the helix center to each of the threestrands likely fall within the approximate range,5–10 A. Thus, the radial positions of Rox labels oneach strand strongly suggest that each strand isdisplaced away from the common helical axis.This finding is in qualitative agreement with thestructure for the RecA–dsDNA complex proposedabove.

From a consideration of physical constraintsimposed by the RecA filament and stereochemicalconstraints imposed by the covalent bonds in thephosphate backbone (see above), it has been esti-mated that the phosphate backbone of a RecA-bound DNA strand will possess a helical radiusbetween 6 and 17 A.11 This limitation coincideswith the probable radial distances from the helixcenter to the DNA strands described above. There-fore, the base-pairs composed from the incomingand complementary strands are likely located onthe surface of a cylinder with a radius of 5–10 A,with the base-pairs displaced into the minorgroove of the right-handed double-helix (seeabove). Moreover, the outgoing strand must belocated in the major groove of the incoming-complementary heteroduplex at a similar radialdisplacement (Figure 8(b)). Notably, this arrange-ment of strands would leave the minor groove ofthe complementary-incoming heteroduplex DNAavailable for protein-DNA interactions. Previouswork showed that the RecA protein binds dsDNAalong its minor groove.16,72,73 Because the hetero-duplex DNA in the RecA–tsDNA complex exhibitsan extended and unwound conformation indistin-guishable from that of RecA-bound dsDNA in ourFRET work, it is likely that RecA protein alsobinds to the heteroduplex along its minor groove.This inference is based on the theory that similarRecA–dsDNA interactions will produce similarconformational changes in the bound DNAmolecule and is consistent with the idea that theRecA–dsDNA is similar to the product of strandexchange.76

Implications for mechanistic aspects ofhomology recognition

In the context of the post-strand-exchangeRecA–tsDNA complex described here, the locationof the outgoing strand within the major groove ofthe complementary-incoming heteroduplex necess-arily implies that homology recognition wasinitiated in the minor groove of the substratedsDNA. The question of whether RecA mediateshomology recognition via major or minor-groovecontacts with the substrate dsDNA has beenaddressed in several laboratories over the pastdecade. To a large extent, those experiments haveproduced controversial results. Indeed, thehypothesis that homology is recognized via the


major groove has been supported by compu-tational modeling85 and radioprobing experi-ments,23 while the alternative hypothesis,recognition of homology from the minor groove,has been supported by chemical modification,97,103

atomic mutagenesis,104 ligand displacement,105 andaffinity modification98 – 100 experiments as well ascomputational modeling.18 Our data are clearly inaccord with recognition of homology occurring viaminor groove contacts. Moreover, the accompany-ing structural model can account for much of thepreviously published data.

The experimental evidence related to the issue ofwhether homology recognition occurs from themajor or minor groove of the substrate dsDNAhas been reviewed recently in the context of argu-ments favoring the former23 and latter18,98,100

hypotheses. In particular, in the context of theirelegant experiments, Malkov et al. were able torationalize much of the published data in terms ofa model favoring major-groove homology recog-nition. These authors’ own results were derivedfrom a powerful biochemical approach to the hom-ology recognition issue using the novel techniqueof radioprobing with the Auger radiodecay of125I.106 Based on the distribution of DNA strandbreaks localized near a site-specific [5-125I]dC resi-

due, Malkov and co-workers were able to deter-mine relative distances between the radioactiveiodine atom and various nucleotide positions onthe outgoing and incoming strands. In the contextof a model wherein the heteroduplex DNA isfound near the central filament axis and the out-going strand forms an extended helix on the sur-face of the heteroduplex, two of the strongestpieces of evidence for the minor groove location ofthe outgoing strand were the observations of (1)“a low overall frequency of breaks of the outgoingstrand compared to the complementary strand”,and (2) “a flat pattern and 30 shift of the cleavageof the outgoing strand”.23 According to thoseauthors, this evidence was best reconciled with aRecA–tsDNA structure wherein the outgoingstrand is displaced into the minor groove of themore central heteroduplex, indicating that hom-ology must have been detected in the major grooveof the substrate duplex.

The first condition requires that the distancebetween the iodine atom at C5 of a complementarystrand cytosine and C10 of the outgoing strandnucleotide is around twofold longer than thecorresponding distance to C10 of the comple-mentary strand nucleotide (Figure 8(a)). Thesecond condition requires that the contour length

Figure 8. Models depicting possible arrangements of bases corresponding to the three DNA strands in the RecA–tsDNA complex. The gray circles, with radii corresponding to 17 A (see the text) represent cross-sections of theRecA–tsDNA nucleoprotein filament. The bases of the incoming (I, blue), complementary (C, red), outgoing (O,green) strands are shown. (a) Model for the post-strand-exchange heteroduplex and outgoing strand accounting formajor groove homology recognition as proposed by Malkov and co-workers.23 The heteroduplex DNA (comprised ofthe incoming and complementary strands) is located in the center of the nucleoprotein cylinder (white cross) withthe outgoing strand in the minor groove of the heteroduplex DNA near the surface of the cylinder. The indicated dis-tances between an 125I atom attached to C5 of cytosine (incoming strand) and the C10 atoms of the complementaryand outgoing strands correspond to those proposed to account for previous radioprobing data (see the text and Figure4(d) of Malkov et al.23). (b) Model for the post-strand-exchange heteroduplex and outgoing strand consistent with thisFRET work. The broken circles represent the radial distances of 5 A and 10 A (see the text). The heteroduplex DNA isnear the surface of the cylinder while the outgoing strand is interwound in the major groove of the heteroduplex.This position of the outgoing strand is consistent with minor groove homology recognition, in contrast to the proposalof Malkov et al.23 Nevertheless, the FRET-derived model adequately accounts for the two distance requirementsderived from those authors’ radioprobing experiments.


of the complementary and outgoing strands besubstantially different. The pair of evidentiarydata presented by Malkov et al. is not sufficient foruniquely supporting the location of the outgoingstrand in the major groove of the heteroduplex. Infact, a model in which the bases of the hetero-duplex dsDNA, rather than the bases of the out-going strand, are displaced away from the centralfilament axis can adequately account for both keyobservations by Malkov et al. (Figure 8(b)). Giventhe magnitudes of the ra parameters describedhere (Table 4), we conclude that the bases of theheteroduplex DNA are displaced away from thecentral helix axis and that the outgoing strandforms an extended helix intertwined along themajor groove of the heteroduplex. Such a groovelocation is consistent with the conclusions of theDervan and Adzuma groups, and suggests thathomology recognition leading to the observedpost-strand-exchange intermediate occurred viaminor groove contacts with the substrate dsDNA.

While these structural conclusions are consistentwith the results of many other researchers in thefield as well as internally consistent with our owndata, we emphasize that the mechanistic con-clusion is inferred from the steady-state structureof an intermediate following homology recognitionand strand exchange. Ultimately, the detection ofhomology and exchange of DNA strands are activemolecular processes whose mechanistic detailsremain elusive. The results presented here providea structural framework upon which to base sub-sequent mechanistic hypotheses. Moreover, theremarkable generality of real-time fluorescence-based assays may offer an approach to tease apartthe intricate mechanistic details of homology rec-ognition and strand exchange mediated by theRecA protein.

Summary

We have presented a systematic analysis of FRETefficiencies to infer interfluorophore distancesdescribing the structures of the DNA strandswithin RecA–DNA complexes. The helicalgeometry parameters of the RecA–dsDNA com-plex, h ¼ 4.9(^0.1) A and Vh ¼ 17(^4)8, are con-sistent with structural conclusions derived fromEM and other classic biochemical methods. Hence,the direct FRET measurements are accurate andsuitable for the solution-phase elucidation ofRecA–DNA complex structures. In addition, theinterfluorophore distance measurements unexpect-edly suggest that the DNA base-pairs are displacedaway from the central helix axis, providing for anopen helix center. The key structural conclusionsregarding the RecA–tsDNA complex can be sum-marized as: (1) all three strands are extended andunwound to a similar extent relative to B-DNA,and all show h and Vh-values that are similar toRecA-bound dsDNA; (2) the outgoing strand isshifted by about three nucleotide positions with

respect to its registry with the complementary andincoming strands; (3) the complex represents alate, post-strand-exchange intermediate; (4) thebases of the incoming and complementary strandsare displaced away from the helix axis toward theminor groove of the heteroduplex; and (5) thebases of the outgoing strand lie in the major grooveof the heteroduplex. Taken together, the datacorroborate an important role for interactions withthe minor groove of dsDNA, both between proteinand DNA for RecA-bound dsDNA and hetero-duplex DNA and between the nucleoprotein fila-ment and the substrate dsDNA in the recognitionof sequence homology.

Materials and Methods

RecA protein

E. coli RecA was purified as described107 and stored inaqueous buffer (25 mM Tris–HCl (pH 7.5), 1 mM DTT,5% (v/v) glycerol) at 280 8C. The protein concentrationwas determined spectrophotometrically using an extinc-tion coefficient at 279 nm of 0.59 mg21 ml21.108

Oligodeoxyribonucleotides

ODNs containing single fluorescent labels were syn-thesized and HPLC-purified by Integrated DNA Tech-nologies. ODNs without fluorescent labels weresynthesized on an automated nucleic acids synthesis sys-tem (PerSeptive Biosystems; model 8905) and purifiedusing 12% (w/v) denaturing polyacrylamide gel electro-phoresis. The DNA bands were viewed by short-wave-length UV shadowing and eluted from gel slices using aSchleicher & Schuell Elutrap in 50 mM Tris–borate (pH8.0), 2 mM EDTA. The DNA was then recovered byethanol precipitation followed by dialysis against Milli-Q water, the pH of which was adjusted to pH 7.5. Thedialyzed ODN samples were then lyophilized and storedat 280 8C. The concentrations of the ODNs were deter-mined using their extinction coefficients at 260 nm calcu-lated by the nearest-neighbor method from nucleosidemonomer and dimers.109 The absorbance contributionsof labeled fluorophores at 260 nm were subtracted usingthe extinction coefficients of the fluorophores at 260 nm(fluorescein: 19,361 M21 cm21; Rox: 38,065 M21 cm21).110

All concentrations of ODNs were expressed inmolecules.

Double-stranded DNA was prepared by mixing equalmolar concentrations of the two complementary strandsin 10 mM Tris–HCl (pH 7.5), 50 mM NaCl, 1 mMEDTA. The mixtures were incubated at 80 8C for fourminutes followed by a slow cooling back down to roomtemperature. All duplexes were analyzed on 12% nativepolyacrylamide gel to insure that no excess singlestrands were present.

Reaction conditions

For the reaction systems investigated, reactions andpre-reaction incubations were carried out at 37 8C inRx6 buffer (25 mM Tris–HOAc (pH 7.5), 2 mM DTT,6 mM Mg(OAc)2) or Rx2 buffer (25 mM Tris–HOAc (pH7.5), 2 mM DTT, 2 mM Mg(OAc)2) unless otherwisespecified.


Agarose gel mobility shift assay

Reactions using either dye-labeled DNA substrates orunlabeled DNA substrates performed under identicalreaction conditions were incubated in Rx6 buffer for onehour, then equal amounts of each reaction, eitheruntreated or treated with 0.1 mg/ml proteinase K and0.5% (w/v) SDS at 37 8C for 25 minutes to removeRecA, were loaded on a 3% (w/v) Metaphor agarose gel(FMC Bioproducts) in chilled 50 mM Tris–borate (pH7.5), 2 mM EDTA, 2 mM Mg(OAc)2. The gel was rununder “Rapid Run” conditions (as described in themanufacturer’s protocol), at 4 8C for two hours. The gelwas stained with SybrGold (Molecular Probes) at roomtemperature for 20 minutes and viewed under an UVlamp.

Steady-state fluorescence measurements

All fluorescence data were collected on a SLM 8100spectrofluorometer (SLM-Aminco, Illinois) with a3 mm £ 3 mm cuvette. The excitation and emissionbeams were polarized with Glan–Thompson polarizingprisms set at the magic angle values (excitation 08 andemission 54.78) to eliminate polarization artifacts. Thefluorescein and Rox emission spectra were collectedwith excitation wavelengths of 492 nm and 585 nm,respectively, unless otherwise noted. In both cases, thebandpass was set at 2 nm for the excitation mono-chromator and 4 nm for the emission monochromator.Scan rate was set at 0.95 nm/second with 1 nm per step.Correction for the inner filter effect was not necessarysince the absorbance values at the excitation wavelengthwere less than 0.05. All spectra were corrected for back-ground using an identical control reaction withoutfluorophores. Instrumental response was also correctedby comparing to the emission spectra of a standardfluorophore using the following formula:27,29

Scorrsamðlex; lÞ ¼

½Sobssamðlex; lÞ2 Sobs

bufðlex; lÞScorrstd ðlex; lÞ

Sobsstd ðlex; lÞ

ð4Þ

where S(lex,l) is a wavelength l dependent emissionspectrum with excitation at lex. The subscripts indicatethe spectrum is taken on the sample interested (sam),buffer (buf) or the standard fluorophore (std) used tocorrect for instrumental response. The superscripts indi-cate the spectrum is observed on the SLM8100 fluor-ometer (obs) or is the spectrum after correction (corr).Fluorescein (F-1300; Molecular probes) in 0.1 M NaOHwas used as the standard fluorophore. The correct emis-sion spectrum of fluorescein was provided by MolecularProbes.

For the dsDNA system, the emission spectra of fluor-escein and Rox were recorded using 0.1 mM duplex inRx2 buffer at 37 8C. RecA protein (4.0 mM final concen-tration) and ATPgS (0.2 mM final concentration) werethen added to the cuvette, the mixture was incubated at37 8C for one hour, and the emission spectra of fluor-escein and Rox were taken again for the RecA–dsDNAsystem. For the RecA–tsDNA-C and -O series, 0.73 mMincoming 74mer (74X or 74C) was first incubated with18.0 mM RecA protein in Rx2 buffer supplemented with0.5 mM ATPgS at 37 8C for ten minutes, then the dual-labeled partial dsDNA (FRi) was added at a final concen-tration of 0.1 mM. Aliquots of a 500 mM Mg(OAc)2 stockwere also added together with the dsDNA to bring thefinal concentration of Mg2þ to 6 mM (i.e. Rx6 buffer con-ditions). This concentration of Mg2þ (6 mM) was chosen

to balance the effects of the precipitation of RecA proteinat high Mg2þ concentration and the ability of Mg2þ tofacilitate completion of the strand exchange (data notshown). The solutions were incubated at 37 8C for onehour before the emission spectra of fluorescein and Roxwere collected. For the RecA–tsDNA-I series, 0.5 mMssDNA Ri was incubated with 18.5 mM RecA protein inRx2 buffer at 37 8C for ten minutes, and fluorescein-labeled full-length dsDNA ds74F was added to a finalconcentration of 0.5 mM along with aliquots ofMg(OAc)2 to a final Mg2þ concentration of 6 mM (i.e.Rx6 buffer conditions). The solutions were incubated forone hour at 37 8C before the emission spectra of fluor-escein and Rox were recorded.

For the reactions performed using ADP-AlF4 insteadof ATPgS, the following reaction conditions were used.For RecA–dsDNA, 4.0 mM RecA protein and 0.1 mM par-tial dsDNA FRi were incubated at 37 8C for one hour inRx2 buffer supplemented with 5 mM ADP, 0.4 mMAl(NO)3 and 10 mM NaF before the emission spectra offluorescein and Rox were collected. For the RecA–tsDNA systems, the incoming ssDNA was first incubatedat 37 8C for ten minutes with an appropriate amount ofthe RecA protein in Rx2 buffer supplemented with5 mM ADP, 0.4 mM Al(NO)3 and 10 mM NaF, and thenthe homologous dsDNA substrate was added. Aliquotsof 500 mM Mg(OAc)2 stock were also added togetherwith the dsDNA to bring the final concentration ofMg2þ to 6 mM (Rx6 buffer conditions).

Steady-state fluorescence anisotropy measurements

All steady-state fluorescence anisotropy measure-ments were performed in L format using Glan–Thompson polarizers placed in the excitation andemission channels of the SLM 8100 spectrofluorometer.The fluorescence anisotropy of the sample was calcu-lated using the following equation:29

r ¼IVV 2 GIVH

IVV þ 2GIVHð5Þ

where I is the fluorescence intensity and the first andsecond subscripts refer to vertical (V) polarization of theexcitation and vertical (V) or horizontal (H) polarizationof the emitted light, respectively. The factor G ¼ IHV/IHH

corrects for the different sensitivity of the emissionmonochromator for vertically and horizontally polarizedlight. For each sample, the anisotropy was measured tentimes with one second measuring times, and the valueswere averaged. Fluorescein and Rox anisotropy datawere measured with emission wavelengths of 520 nmand 605 nm, respectively.

Determination of the limiting anisotropy

The limiting anisotropy ri was measured by construct-ing Perrin plots in which the steady-state polarizationwas measured for various viscosities29,111

1

r¼

1

riþ

1

ri

A

hð6Þ

where r is the measured sample anisotropy; A is a con-stant which is related to the lifetime of the fluorophore,the volume of the rotating unit and the temperature atthe measurement; and h is the viscosity of the sample.The solvent viscosity of a fluorescein-only labeled or aRox-only labeled sample was varied at 30 8C by addingaliquots of 62% (w/v) stock sucrose solution. Published


viscosity values for aqueous sucrose solutions at 30 8Cwere used.112 The emission spectra of the fluorophoreswere not altered as the sucrose concentration wasincreased. Perrin plots were constructed, and the limit-ing anisotropies of both dyes were obtained by extra-polating the fitted curve to reach the y-interceptaccording to the Perrin equation.

Determination of the quantum yield of fluorescein

The quantum yield of the donor fluorescein FF wasdetermined for a fluorescein-labeled DNA sample in allreaction systems by the comparative method:27,29

FF ¼FstdAstd

ÐSFðlex; lÞdl

AF

ÐSstdðlex; lÞdl

ð7Þ

where Fstd is the quantum yield of a standard fluoro-phore,

ÐSstdðlex; lÞdl and

ÐSFðlex; lÞdl are the areas

under the corrected emission spectra of the standardfluorophore and fluorescein. Astd and AF are the absor-bance of the standard fluorophore and the fluoresceinsample at a given excitation wavelength using the sameconcentration at which their fluorescence intensitieswere measured.

Determination of Forster critical distance R0

The Forster critical distance R0 is characteristic of theposition-independent spectral property of the dyes.This, in turn, requires the measurements of the donorquantum yield in the absence of the acceptor (Fd), theoverlap integral J that characterizes the resonancebetween the donor and acceptor dipoles:27

R60 ¼ 8:875 £ 1025 k

2JFd

n4ð8Þ

where k2 is the orientation factor, which has a value of2/3 when the donor and the acceptor rotate rapidly in ashort time compared to the donor lifetime. Fd is thedonor quantum yield in the absence of acceptor and n isthe refractive index of the medium that is normallytaken as 1.4 for aqueous buffer. The overlap integral Jcharacterizes the resonance between the donor andacceptor dipoles and is evaluated by integration of themutual area of the overlap between the donor emissionspectrum Sd(lex,l) and the acceptor absorption spectrumea(l) using the following equation:27

J ¼

ÐSdðlex; lÞeal

4

dlÐSdðlex; lÞdl

ð9Þ

Determination of the maximum and minimum valueof R0

When both depolarization factors are positive, k2max

and k2min can be calculated from the following equation:

k2max ¼

2

3ð1 þ dx

d þ dxa þ 3dx

ddxaÞ

k2min ¼

2

31 2

1

2ðdx

d þ dxaÞ

ð10Þ

where dxd and dx

a are the axial depolarization factors forthe donor and acceptor, respectively. The axial depolariz-ation factors dx can be calculated from the limiting aniso-tropies (ri) and fundamental anisotropies (r0) of the

donor and acceptor:

dx ¼

ffiffiffiffiffiffiri

r0

rð11Þ

The value of dx represents the depolarization factor dueto segmental motion of the donor ðdx

dÞ or acceptor ðdxaÞ;

but not the depolarization due to overall rotational diffu-sion of the protein. The overall rotational diffusion is notimportant because it does not change the donor–acceptor orientation.

After the limiting anisotropy of the fluorophores areobtained, maximum and minimum values of the Forstercritical distance R0 can be calculated using the followingformula:

R0;min ¼3

2k2

min

16

R0

R0;max ¼3

2k2

max

16

R0 ð12Þ

RecA–DNA complex dissociationconstant measurements

For the RecA–dsDNA complex, the concentration of afluorescein-labeled partial dsDNA FX was kept constantat 0.1 mM in 400 ml of Rx2 buffer at 37 8C, while small ali-quots of RecA stock solution were added. The RecA finalconcentration was varied from 1 mM to 9 mM. For theRecA–tsDNA-C and -O systems, a RecA-74F presynapticfilament was first formed between 0.1 mM 74F ssDNAand 3 mM RecA protein in 400 ml of Rx6 buffer and incu-bated at 37 8C for ten minutes, then small aliquots of con-centrated partial plain dsDNA (XX) stock of 9.85 mMwere added. The final concentration of XX was variedfrom 0.02 mM to 0.6 mM. For the RecA–tsDNA-I system,a Rox-labeled ssDNA 45R20 was used to form theRecA–45R20 presynaptic filament. In 400 ml of Rx6 buffer,0.1 mM 45R20 ssDNA and 3 mM RecA were incubated at37 8C for ten minutes, then small aliquots of concentratedplain dsDNA (ds74) stock of 7.6 mM were added. Thefinal concentration of ds74 was varied from 0.02 mM to0.5 mM. Between each addition of titrant, the reactionswere allowed to equilibrate for 15 to 20 minutes beforethe emission spectrum and anisotropy of the fluorophorewere recorded. Fluorescein emission data were collectedfrom 510 to 530 nm, and Rox emission data were col-lected from 595 to 615 nm. Each spectrum representedan average of three repetitions. The fluorescence inten-sity was calculated by integrating the area under thespectral curve after the spectrum has been corrected foraddition dilution and buffer background. A RecA–DNAbinding isotherm was then constructed by plotting thefluorescence intensities at different titration pointsagainst the titrant’s concentration. The dissociation con-stant was determined by fitting the experimental data tothe theoretical curve using the following equationderived from the mass balance equations for the rapid,one-step reversible association of molecules A and B:

Ft ¼ F0

þDFn

2B0A þ

B0

nþ Kd 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA þ

B0

nþ Kd

2

24AB0

n

s24

35

ð13Þ


In this equation, Ft is the total fluorescence intensity; F0 isthe initial fluorescence when no titrant has been added;DF is the total fluorescence change defined by the differ-ence between the end fluorescence and initial fluor-escence; A is the concentration of the titrant; B0 is theconcentration of the substrate being titrated; n is the stoi-chiometry number between the titrant and the substrate;and Kd is the apparent dissociation constant betweentitrant R and substrate B0.

For the RecA–dsDNA complex, the concentration ofthe dsDNA B0 is expressed as the following:

B0 ¼74CdsDNA

nð14Þ

where CdsDNA is the concentration of the dsDNAexpressed in molecules, 74 is the length of the dsDNA,and n is the stoichiometry number, which indicateshow many nucleotides were bound by each RecAmonomer.

For the RecA–tsDNA complexes, B0 is the concen-tration of the dual-labeled partial dsDNA FRi inmolecules, A is the concentration of the preformedRecA–ssDNA filament in molecules and n equals onebecause presumably each RecA–ssDNA filament onlybinds one homologous dsDNA molecule. Because RecAprotein has a high affinity for ssDNA,113 and the dis-sociation of RecA from linear ssDNA is blocked in thepresence of ATPgS,114 we used stoichiometric amountsof RecA protein to form the presynaptic filament withssDNA and counted it as one species in the reaction.

After the apparent dissociation constant for eachsystem was obtained, a binding isotherm plot was con-structed. For RecA–dsDNA, RecA–tsDNA-C and -O,the following equation is used to determine the concen-tration of the non-fluorescent species required to keep atleast 90% of the fluorescent species involved in thereaction:

A ¼ uB0 þu

1 2 uKd ð15Þ

where A is the concentration of the species in excess, B0

is the concentration of the limiting fluorescent specieskept at a fixed concentration, and u is the desiredfractional completion of the reaction (u $ 0.9 for allreactions), and Kd is the apparent dissociation constantdetermined from equation (13).

For RecA–tsDNA-I, the following equation was used,which is a simplified version of equation (15) when Aequals B0:

A ¼u

ð1 2 uÞ2Kd ð16Þ

Anisotropy measurements were also used to deter-mine the apparent dissociation constants. The anisotropyof the fluorophore at each titration point was plottedagainst titrant concentration and the same curve-fittingwas performed using equation (13) except that thefluorescence terms have been changed to anisotropy.

Energy transfer efficiency measurements

The energy transfer efficiencies between fluoresceinand Rox were calculated by measuring the (Ratio)A,which is a normalized measurement of the enhancement

of the fluorescence from the acceptor Rox due to FRET:26

ðRatioÞA ¼

ð610

600

½SFRðl492; lÞ2 aSFðl492; lÞdlð610

600

SFRðl585; lÞdl

ð17Þ

where S(l,l) is a wavelength l dependent emission spec-trum with excitation wavelength indicated at the sub-script of the first l. The subscripts FR or F of each S(l,l)indicates the emission spectrum is taken on the samplecontaining both fluorescein and Rox (FR) or just fluor-escein (F). The denominator of the equation is the fluor-escence intensity of the acceptor Rox at 605 nm due todirect excitation at 585 nm in a reaction sample contain-ing both donor and acceptor. The numerator is the fluor-escence intensity of the same sample at Rox emissionpeak 605 nm with excitation at 492 nm after the fluor-escein emission background at 605 nm has been sub-tracted by a normalizing factor a, which is expressed inthe following formula:

a ¼

ð525

515

SFRðl492; lÞdlð525

515

SFðl492; lÞdl

ð18Þ

The denominator of equation (18) is the fluorescenceintensity of fluorescein at 520 nm with excitation at492 nm in the absence of Rox. The numerator is the fluor-escence intensity of fluorescein at 520 nm with excitationat 492 nm in the presence of Rox.

This (Ratio)A is linearly dependent on the energytransfer efficiency E and the donor fluorescein labelingfraction dþ on the oligonucleotides:

E ¼ðRatioÞAe

585R 2 e492

R

e492F dþ

ð19Þ

where e is the molar absorption coefficient of the fluoro-phore (indicated by the subscripts) at a given wave-length (indicated by the superscripts). (Ratio)A

normalizes the measured sensitized FRET signal for theconcentration and for the apparent quantum yield of theacceptor. In all experiments, the donor fluorescein label-ing percentage dþ was verified to be 1 by comparing thelabeled dye concentration with the oligonucleotide con-centrations, which were calculated using their extinctioncoefficients at their absorbance maxima, respectively.

FRET efficiency versus base-pair separationdata analysis

For each reaction system, the energy transfer effi-ciency, E, was plotted as a function of the correspondingbase-pair separation, n. To determine values of ra

2 þ rd2,

2rard, h, Vh, DV, and Dh that best fit the data for each sys-tem, non-linear least squares fitting to the E–n function(equation (3)) was used with five (or four) of the par-ameters varied to minimize x2. All curve fitting and sub-sequent analysis was performed using proFit 5.5.1(Quantum Soft, Zurich, Switzerland). For each data set,an initial Monte Carlo search was performed with atleast 2 £ 106 iterations to search parameter spacerandomly for the best starting values. These startingparameters were then used as initial values inLevenberg–Marquardt fits to generate the best-fitparameters. For dsDNA, a B-form conformation wasassumed (h ¼ 3.4 A, Vh ¼ 368 per base-pair), and the


other four parameters were variable. For RecA–dsDNAand RecA–tsDNA complexes, the dye-labeling par-ameter Dh generated from B-form dsDNA was usedwhile the other five parameters, ra

2 þ rd2, 2rard, h, Vh and

DV were left variable for best fitting. All the resultsreported here are the best solutions regardless of initialvalues varied over a wide range.

Acknowledgments

The authors gratefully acknowledge the financial sup-port of the NIH (grant GM58114) and the Robert A.Welch Foundation (grant C-1374). We are grateful toR. A. Simonette, M. D. Berger, and Dr A. I. Roca fortechnical assistance, and to our Rice University col-leagues Professors K. S. Matthews, E. P. Nikonowicz,and Y. Shamoo for stimulating discussions. We arealso indebted to Professor E. H. Egleman (Universityof Virginia) for suggesting the experiments withADP–AlF4.

References

1. Radding, C. M. (1982). Homologous pairing andstrand exchange in genetic recombination. Annu.Rev. Genet. 16, 405–437.

2. Kowalczykowski, S. C. (1991). Biochemistry of gen-etic recombination: energetics and mechanism ofDNA strand exchange. Annu. Rev. Biophys. Biophys.Chem. 20, 539–575.

3. Roca, A. I. & Cox, M. M. (1997). RecA protein: struc-ture, function, and role in recombinational DNArepair. Prog. Nucl. Acids Res. Mol. Biol. 56, 129–223.

4. Reagan, M. S., Pittenger, C., Siede, W. & Friedberg,E. C. (1995). Characterization of a mutant strain ofSaccharomyces cerevisiae with a deletion of theRAD27 gene, a structural homolog of the RAD2nucleotide excision repair gene. J. Bacteriol. 177,364–371.

5. Courcelle, J. & Hanawalt, P. C. (2001). Participationof recombination proteins in rescue of arrestedreplication forks in UV-irradiated Escherichia colineed not involve recombination. Proc. Natl Acad.Sci. USA, 98, 8196–8202.

6. Cox, M. M., Goodman, M. F., Kreuzer, K. N.,Sherratt, D. J., Sandler, S. J. & Marians, K. J. (2000).The importance of repairing stalled replicationforks. Nature, 404, 37–41.

7. Bianco, P. R., Tracy, R. B. & Kowalczykowski, S. C.(1998). DNA strand exchange proteins: a biochemi-cal and physical comparison. Front. Biosci. 3,D570–D603.

8. Stasiak, A., Stasiak, A. Z. & Koller, T. (1984). Visual-ization of RecA–DNA complexes involved in con-secutive stages of an in vitro strand exchangereaction. Cold Spring Harb. Symp. Quant. Biol. 49,561–570.

9. Stasiak, A. & Egelman, E. H. (1986). RecA–DNAhelical complexes in genetic recombination.Biochem. Soc. Trans. 14, 218–220.

10. Egelman, E. H. & Stasiak, A. (1986). Structure ofhelical RecA–DNA complexes. Complexes formedin the presence of ATP-gamma-S or ATP. J. Mol.Biol. 191, 677–697.

11. Egelman, E. H. & Yu, X. (1989). The location of DNAin RecA–DNA helical filaments. Science, 245,404–407.

12. Stasiak, A., Di Capua, E. & Koller, T. (1981).Elongation of duplex DNA by recA protein. J. Mol.Biol. 151, 557–564.

13. Di Capua, E., Engel, A., Stasiak, A. & Koller, T.(1982). Characterization of complexes betweenrecA protein and duplex DNA by electronmicroscopy. J. Mol. Biol. 157, 87–103.

14. Stasiak, A. & Di Capua, E. (1982). The helicity ofDNA in complexes with recA protein. Nature, 299,185–186.

15. Chrysogelos, S., Register, J. C., III & Griffith, J.(1983). The structure of recA protein–DNA fila-ments. 2 recA protein monomers unwind 17 basepairs of DNA by 11.5 degrees/base pair in the pre-sence of adenosine 50-O-(3-thiotriphosphate). J. Biol.Chem. 258, 12624–12631.

16. Dombroski, D. F., Scraba, D. G., Bradley, R. D. &Morgan, A. R. (1983). Studies of the interaction ofRecA protein with DNA. Nucl. Acids Res. 11,7487–7504.

17. Pugh, B. F., Schutte, B. C. & Cox, M. M. (1989).Extent of duplex DNA underwinding induced byRecA protein binding in the presence of ATP. J. Mol.Biol. 205, 487–492.

18. Bertucat, G., Lavery, R. & Prevost, C. (1999). A mol-ecular model for RecA-promoted strand exchangevia parallel triple-stranded helices. Biophys. J. 77,1562–1576.

19. Kosikov, K. M., Gorin, A. A., Zhurkin, V. B. &Olson, W. K. (1999). DNA stretching and com-pression: large-scale simulations of double helicalstructures. J. Mol. Biol. 289, 1301–1326.

20. Egelman, E. H. (2001). Does a stretched DNA struc-ture dictate the helical geometry of RecA-likefilaments? J. Mol. Biol. 309, 539–542.

21. Shibata, T., Nishinaka, T., Mikawa, T., Aihara, H.,Kurumizaka, H., Yokoyama, S. & Ito, Y. (2001).Homologous genetic recombination as an intrinsicdynamic property of a DNA structure induced byRecA/Rad51-family proteins: a possible advantageof DNA over RNA as genomic material. Proc. NatlAcad. Sci. USA, 98, 8425–8432.

22. Yu, X., Jacobs, S. A., West, S. C., Ogawa, T. &Egelman, E. H. (2001). Domain structure anddynamics in the helical filaments formed by RecAand Rad51 on DNA. Proc. Natl Acad. Sci. USA, 98,8419–8424.

23. Malkov, V. A., Panyutin, I. G., Neumann, R. D.,Zhurkin, V. B. & Camerini-Otero, R. D. (2000).Radioprobing of a RecA-three-stranded DNA com-plex with iodine 125: evidence for recognition ofhomology in the major groove of the target duplex.J. Mol. Biol. 299, 629–640.

24. Forster, T. (1965). Modern Quantum Chemistry(Sinanoglu, O., ed.), vol. 3, Academic Press, New York.

25. Stryer, L. (1978). Fluorescence energy transfer as aspectroscopic ruler. Annu. Rev. Biochem. 47, 819–846.

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

27. Wu, P. & Brand, L. (1994). Resonance energy trans-fer: methods and applications. Anal. Biochem. 218,1–13.

28. Yang, M. & Millar, D. P. (1997). Fluorescence reson-ance energy transfer as a probe of DNA structureand function. Methods Enzymol. 278, 417–444.


29. Lakowicz, J. R. (1999). Principles of FluorescenceSpectroscopy, 2nd edit., Kluwer Academic/PlenumPublishers, New York.

30. Hillisch, A., Lorenz, M. & Diekmann, S. (2001).Recent advances in FRET: distance determinationin protein–DNA complexes. Curr. Opin. Struct.Biol. 11, 201–207.

31. Bazemore, L. R., Takahashi, M. & Radding, C. M.(1997). Kinetic analysis of pairing and strandexchange catalyzed by RecA. Detection by fluor-escence energy transfer. J. Biol. Chem. 272,14672–14682.

32. Gumbs, O. H. & Shaner, S. L. (1998). Three mechan-istic steps detected by FRET after presynaptic fila-ment formation in homologous recombination.ATP hydrolysis required for release of oligonucleo-tide heteroduplex product from RecA. Biochemistry,37, 11692–11706.

33. Gupta, R. C., Golub, E. I., Wold, M. S. & Radding,C. M. (1998). Polarity of DNA strand exchange pro-moted by recombination proteins of the RecAfamily. Proc. Natl Acad. Sci. USA, 95, 9843–9848.

34. Olson, W. K., Bansal, M., Burley, S. K., Dickerson,R. E., Gerstein, M., Harvey, S. C. et al. (2001).Nomenclature. J. Mol. Biol. 313, 229–237.

35. Clegg, R. M., Murchie, A. I., Zechel, A. & Lilley,D. M. (1993). Observing the helical geometry ofdouble-stranded DNA in solution by fluorescenceresonance energy transfer. Proc. Natl Acad. Sci.USA, 90, 2994–2998.

36. Gourves, A. S., Defais, M. & Johnson, N. P. (2001).Equilibrium binding of single-stranded DNA to thesecondary DNA bindig site of the bacterial recombi-nase RecA. J. Biol. Chem., 276, 9613–9619.

37. Wang, S., Martin, E., Cimino, J., Omann, G. &Glaser, M. (1988). Distribution of phospholipidsaround gramicidin and D-beta-hydroxybutyratedehydrogenase as measured by resonance energytransfer. Biochemistry, 27, 2033–2039.

38. Shahrokh, Z., Verkman, A. S. & Shohet, S. B. (1991).Distance between skeletal protein 4.1 and theerythrocyte membrane bilayer measured by reson-ance energy transfer. J. Biol. Chem. 266,12082–12089.

39. Chapman, E. R., Alexander, K., Vorherr, T., Carafoli,E. & Storm, D. R. (1992). Fluorescence energytransfer analysis of calmodulin–peptide complexes.Biochemistry, 31, 12819–12825.

40. Clegg, R. M., Murchie, A. I., Zechel, A., Carlberg,C., Diekmann, S. & Lilley, D. M. (1992). Fluor-escence resonance energy transfer analysis of thestructure of the four-way DNA junction.Biochemistry, 31, 4846–4856.

41. Clegg, R. M., Murchie, A. I. & Lilley, D. M. (1993).The four-way DNA junction: a fluorescence reson-ance energy transfer study. Braz. J. Med. Biol. Res.26, 405–416.

42. Johnson, D. A., Leathers, V. L., Martinez, A. M.,Walsh, D. A. & Fletcher, W. H. (1993). Fluorescenceresonance energy transfer within a heterochromaticcAMP-dependent protein kinase holoenzymeunder equilibrium conditions: new insights intothe conformational changes that result in cAMP-dependent activation. Biochemistry, 32, 6402–6410.

43. Morrison, L. E. & Stols, L. M. (1993). Sensitive fluor-escence-based thermodynamic and kineticmeasurements of DNA hybridization in solution.Biochemistry, 32, 3095–3104.

44. Clegg, R. M., Murchie, A. I. & Lilley, D. M. (1994).The solution structure of the four-way DNA junc-tion at low-salt conditions: a fluorescence resonanceenergy transfer analysis. Biophys. J. 66, 99–109.

45. Cai, K. & Schirch, V. (1996). Structural studies onfolding intermediates of serine hydroxymethyl-transferase using fluorescence resonance energytransfer. J. Biol. Chem. 271, 27311–27320.

46. Stuhmeier, F., Welch, J. B., Murchie, A. I., Lilley,D. M. & Clegg, R. M. (1997). Global structure ofthree-way DNA junctions with and withoutadditional unpaired bases: a fluorescence resonanceenergy transfer analysis. Biochemistry, 36,13530–13538.

47. Vamosi, G. & Clegg, R. M. (1998). The helix-coiltransition of DNA duplexes and hairpins observedby multiple fluorescence parameters. Biochemistry,37, 14300–14316.

48. Watson, J. D. & Crick, F. H. C. (1953). Molecularstructure of nucleic acids. Nature, 171, 737–738.

49. Vamosi, G., Gohlke, C. & Clegg, R. M. (1996). Fluor-escence characteristics of 5-carboxytetramethyl-rhodamine linked covalently to the 50 end ofoligonucleotides: multiple conformers of single-stranded and double-stranded dye–DNA com-plexes. Biophys. J. 71, 972–994.

50. Lorenz, M., Hillisch, A., Goodman, S. D. &Diekmann, S. (1999). Global structure similarities ofintact and nicked DNA complexed with IHFmeasured in solution by fluorescence resonanceenergy transfer. Nucl. Acids Res. 27, 4619–4625.

51. Norman, D. G., Grainger, R. J., Uhrin, D. & Lilley,D. M. (2000). Location of cyanine-3 on double-stranded DNA: importance for fluorescence reson-ance energy transfer studies. Biochemistry, 39,6317–6324.

52. Dexter, D. L. (1953). A theory of sensitized lumines-cence in solids. J. Chem. Phys. 21, 836–850.

53. Yu, X. & Egelman, E. H. (1992). Direct visualizationof dynamics and co-operative conformationalchanges within RecA filaments that appear to beassociated with the hydrolysis of adenosine 50-O-(3-thiotriphosphate). J. Mol. Biol. 225, 193–216.

54. Dale, R. E., Eisinger, J. & Blumberg, W. E. (1979).The orientational freedom of molecular probes. Theorientation factor in intramolecular energy transfer.Biophys. J. 26, 161–193.

55. Stuhmeier, F., Hillisch, A., Clegg, R. M. & Diek-mann, S. (2000). Practical aspects of fluorescenceresonance energy transfer (FRET) and its appli-cations in nucleic acid biochemistry. In DNA–Protein Interactions (Travers, A. & Buckle, M., eds),pp. 77–94, Oxford University Press, Oxford.

56. Saenger, W. (1983). Principles of Nucleic Acid Struc-ture Springer Advanced Text in Chemistry (Cantor,C. R., ed.), Springer, New York.

57. Hsieh, P., Camerini-Otero, C. S. & Camerini-Otero,R. D. (1992). The synapsis event in the homologouspairing of DNAs: RecA recognizes and pairs lessthan one helical repeat of DNA. Proc. Natl Acad.Sci. USA, 89, 6492–6496.

58. Singleton, J. W. (1994). Affinity cleavage of double-stranded DNA directed by a RecA nucleoproteinfilament containing an oligonucleotide–EDTA–Fe(II). Master of Science, California Institute ofTechnology, CA

59. Volodin, A. A., Smirnova, H. A. & Bocharova, T. N.(1994). Efficient interaction of recA protein with


fluorescent dye-labeled oligonucleotides. FEBSLetters, 349, 65–68.

60. Stasiak, A. & Egelman, E. H. (1994). Structure andfunction of RecA–DNA complexes. Experientia, 50,192–203.

61. Honigberg, S. M., Gonda, D. K., Flory, J. & Radding,C. M. (1985). The pairing activity of stable nucleo-protein filaments made from recA protein, single-stranded DNA, and adenosine 50-(gamma-thio)tri-phosphate. J. Biol. Chem. 260, 11845–11851.

62. Menetski, J. P., Bear, D. G. & Kowalczykowski, S. C.(1990). Stable DNA heteroduplex formation cata-lyzed by the Escherichia coli RecA protein in theabsence of ATP hydrolysis. Proc. Natl Acad. Sci.USA, 87, 21–25.

63. Story, R. M., Weber, I. T. & Steitz, T. A. (1992). Thestructure of the E. coli recA protein monomer andpolymer. Nature, 355, 318–325.

64. Yu, X. & Egelman, E. H. (1992). Structural datasuggest that the active and inactive forms of theRecA filament are not simply interconvertible.J. Mol. Biol. 227, 334–346.

65. Egelman, E. H. & Stasiak, A. (1993). Electronmicroscopy of RecA–DNA complexes: two differ-ent states. Their functional significance and relationto solved crystal structure. Micron, 24, 309–324.

66. Dunn, K., Chrysogelos, S. & Griffith, J. (1982).Electron microscopic visualization of recA–DNAfilaments: evidence for a cyclic extension of duplexDNA. Cell, 28, 757–765.

67. Cunningham, R. P., Shibata, T., DasGupta, C. &Radding, C. M. (1979). Single strands induce recAprotein to unwind duplex DNA for homologouspairing. Nature, 281, 191–195.

68. Shibata, T., Cunningham, R. P. & Radding, C. M.(1981). Homologous pairing in genetic recombina-tion. Purification and characterization of Escherichiacoli recA protein. J. Biol. Chem. 256, 7557–7564.

69. Keller, W. (1975). Determination of the number ofsuperhelical turns in simian virus 40 DNA by gelelectrophoresis. Proc. Natl Acad. Sci. USA, 72,4876–4880.

70. Shure, M. & Vinograd, J. (1976). The number ofsuperhelical turns in native virion SV40 DNA andminicol DNA determined by the band countingmethod. Cell, 8, 215–226.

71. Sperrazza, J. M., Register, J. C., III & Griffith, J.(1984). Electron microscopy can be used to measureDNA supertwisting. Gene, 31, 17–22.

72. Di Capua, E. & Muller, B. (1987). The accessibility ofDNA to dimethylsulfate in complexes with recAprotein. EMBO J. 6, 2493–2498.

73. Tuite, E., Sehlstedt, U., Hagmar, P., Norden, B. &Takahashi, M. (1997). Effects of minor and majorgroove-binding drugs and intercalators on theDNA association of minor groove-binding proteinsRecA and deoxyribonuclease I detected by flowlinear dichroism. Eur. J. Biochem. 243, 482–492.

74. Lebrun, A., Shakked, Z. & Lavery, R. (1997). LocalDNA stretching mimics the distortion caused bythe TATA box-binding protein. Proc. Natl Acad. Sci.USA, 94, 2993–2998.

75. Bewley, C. A., Gronenborn, A. M. & Clore, G. M.(1998). Minor groove-binding architectural proteins:structure, function, and DNA recognition. Annu.Rev. Biophys. Biomol. Struct. 27, 105–131.

76. Mazin, A. V. & Kowalczykowski, S. C. (1996). Thespecificity of the secondary DNA binding site ofRecA protein defines its role in DNA strand

exchange. Proc. Natl Acad. Sci. USA, 93,10673–10678.

77. Tracy, R. B. & Kowalczykowski, S. C. (1996). In vitroselection of preferred DNA pairing sequences bythe Escherichia coli RecA protein. Genes Dev. 10,1890–1903.

78. Wittung, P., Bazemore, L. R., Takahashi, M.,Norden, B. & Radding, C. (1996). Second-siteRecA–DNA interactions: lack of identical recog-nition. Biochemistry, 35, 15349–15355.

79. Wittung, P., Ellouze, C., Maraboeuf, F., Takahashi,M. & Norden, B. (1997). Thermochemical andkinetic evidence for nucleotide-sequence-dependentRecA–DNA interactions. Eur. J. Biochem. 245,715–719.

80. Bar-Ziv, R. & Libchaber, A. (2001). Effects of DNAsequence and structure on binding of RecA tosingle-stranded DNA. Proc. Natl Acad. Sci. USA, 98,9068–9073.

81. Volodin, A. A. & Camerini-Otero, R. D. (2002).Influence of DNA sequence on the positioning ofRecA monomers in RecA–DNA cofilaments. J. Biol.Chem. 277, 1614–1618.

82. MacFarland, K. J., Shan, Q., Inman, R. B. & Cox,M. M. (1997). RecA as a motor protein. Testingmodels for the role of ATP hydrolysis in DNAstrand exchange. J. Biol. Chem. 272, 17675–17685.

83. Howard-Flanders, P., West, S. C. & Stasiak, A.(1984). Role of RecA protein spiral filaments ingenetic recombination. Nature, 309, 215–219.

84. Ogawa, T., Shinohara, A., Nabetani, A., Ikeya, T.,Yu, X., Egelman, E. H. & Ogawa, H. (1993). RecA-like recombination proteins in eukaryotes: functionsand structures of RAD51 genes. Cold Spring Harb.Symp. Quant. Biol. 58, 567–576.

85. Zhurkin, V. B., Raghunathan, G., Ulyanov, N. B.,Camerini-Otero, R. D. & Jernigan, R. L. (1994). Aparallel DNA triplex as a model for the inter-mediate in homologous recombination. J. Mol. Biol.239, 181–200.

86. Camerini-Otero, R. D. & Hsieh, P. (1995). Homo-logous recombination proteins in prokaryotes andeukaryotes. Annu. Rev. Genet. 29, 509–552.

87. Nishinaka, T., Ito, Y., Yokoyama, S. & Shibata, T.(1997). An extended DNA structure through deox-yribose-base stacking induced by RecA protein.Proc. Natl Acad. Sci. USA, 94, 6623–6628.

88. Schutte, B. C. & Cox, M. M. (1988). Homology-dependent underwinding of duplex DNA in recAprotein generated paranemic complexes.Biochemistry, 27, 7886–7894.

89. Conley, E. C. & West, S. C. (1990). Underwinding ofDNA associated with duplex–duplex pairing byRecA protein. J. Biol. Chem. 265, 10156–10163.

90. Rould, E., Muniyappa, K. & Radding, C. M. (1992).Unwinding of heterologous DNA by RecA proteinduring the search for homologous sequences. J. Mol.Biol. 226, 127–139.

91. Kiianitsa, K. & Stasiak, A. (1997). Helical repeat ofDNA in the region of homologous pairing. Proc.Natl Acad. Sci. USA, 94, 7837–7840.

92. Voloshin, O. N. & Camerini-Otero, R. D. (1997). Theduplex DNA is very underwound in the three-stranded RecA protein-mediated synaptic complex.Genes Cells, 2, 303–314.

93. Kim, J. I., Cox, M. M. & Inman, R. B. (1992). On therole of ATP hydrolysis in RecA protein-mediatedDNA strand exchange. II. Four-strand exchanges.J. Biol. Chem. 267, 16444–16449.


94. Kim, J. I., Cox, M. M. & Inman, R. B. (1992). On therole of ATP hydrolysis in RecA protein-mediatedDNA strand exchange. I. Bypassing a shortheterologous insert in one DNA substrate. J. Biol.Chem. 267, 16438–16443.

95. Cox, M. M. (1994). Why does RecA protein hydro-lyse ATP? Trends Biochem. Sci. 19, 217–222.

96. Bedale, W. A. & Cox, M. (1996). Evidence for thecoupling of ATP hydrolysis to the final (extension)phase of RecA protein-mediated DNA strandexchange. J. Biol. Chem. 271, 5725–5732.

97. Adzuma, K. (1992). Stable synapsis of homologousDNA molecules mediated by the Escherichia coliRecA protein involves local exchange of DNAstrands. Genes Dev. 6, 1679–1694.

98. Baliga, R., Singleton, J. W. & Dervan, P. B. (1995).RecA–oligonucleotide filaments bind in the minorgroove of double-stranded DNA. Proc. Natl Acad.Sci. USA, 92, 10393–10397.

99. Podyminogin, M. A., Meyer, R. B. & Gamper, H. B.(1995). Sequence-specific covalent modification ofDNA by cross-linking oligonucleotides. Catalysisby RecA and implication for the mechanism ofsynaptic joint formation. Biochemistry, 34,13098–13108.

100. Zhou, X. & Adzuma, K. (1997). DNA strandexchange mediated by the Escherichia coli RecAprotein initiates in the minor groove of double-stranded DNA. Biochemistry, 36, 4650–4661.

101. Volodin, A. A., Smirnova, H. A. & Bocharova, T. N.(1997). Periodicity in recA protein–DNA com-plexes. FEBS Letters, 407, 325–328.

102. Mazin, A. V. & Kowalczykowski, S. C. (1998). Thefunction of the secondary DNA-binding site ofRecA protein during DNA strand exchange. EMBOJ. 17, 1161–1168.

103. Chiu, S. K., Rao, B. J., Story, R. M. & Radding, C. M.(1993). Interactions of three strands in joints madeby RecA protein. Biochemistry, 32, 13146–13155.

104. Jain, S. K., Inman, R. B. & Cox, M. M. (1992).Putative three-stranded DNA pairing intermediatein recA protein-mediated DNA strand exchange:

no role for guanine N-7. J. Biol. Chem. 267,4215–4222.

105. Kumar, K. A. & Muniyappa, K. (1992). Use of struc-ture-directed DNA ligands to probe the binding ofrecA protein to narrow and wide grooves of DNAand on its ability to promote homologous pairing.J. Biol. Chem. 267, 24824–24832.

106. Martin, R. F. & Haseltine, W. A. (1981). Range ofradiochemical damage to DNA with decay ofiodine-125. Science, 213, 896–898.

107. Berger, M. D., Lee, A. M., Simonette, R. A., Jackson,B. E., Roca, A. I. & Singleton, S. F. (2001). Designand evaluation of a tryptophanless RecA proteinwith wild type activity. Biochem. Biophys. Res.Commun. 286, 1195–1203.

108. Craig, N. L. & Roberts, J. W. (1981). Function ofnucleoside triphosphate and polynucleotide inEscherichia coli recA protein-directed cleavage ofphage lambda repressor. J. Biol. Chem. 256,8039–8044.

109. Borer, P. N. (1975). Handbook of Biochemistry andMolecular Biology, Nucleic Acids (Fasman, G. D.,ed.), 3rd edit., Volume I, CRC Press, Cleveland, OH.

110. Sjoback, R., Nygren, J. & Kubista, M. (1998). Charac-terization of fluorescein–oligonucleotide conjugatesand measurement of local electrostatic potential.Biopolymers, 46, 445–453.

111. Perrin, F. (1929). La fluorescence des solutions.Induction moleculaire. Polarisation et duree d’emis-sion. Photochimi. 10, 169–275.

112. Fasman, G. D. (1975). CRC Handbook of Biochemistryand Molecular Biology, 3rd edit., pp. 589, CRC Press,Cleveland, OH.

113. Menetski, J. P., Varghese, A. & Kowalczykowski,S. C. (1988). Properties of the high-affinity single-stranded DNA binding state of the Escherichia colirecA protein. Biochemistry, 27, 1205–1212.

114. Shan, Q., Bork, J. M., Webb, B. L., Inman, R. B. &Cox, M. M. (1997). RecA protein filaments: end-dependent dissociation from ssDNA and stabiliz-ation by RecO and RecR proteins. J. Mol. Biol. 265,519–540.

Edited by I. Tinoco

(Received 16 November 2001; received in revised form 30 April 2002; accepted 2 May 2002)


elucidating a key intermediate in homologous dna strand exchange: structural characterization of the...

Documents