Biochimie (1991) 73,363-370 © Soci6t6 f,,,,~aise de biochimie et biologie mol6culaire / Elsevier, Paris

363

Production of triple-stranded recombination intermediates by RecA protein, in vitro

BJ Rao, B Jwang, M Dutreix*

The Department of Human Genetics aJ:d Molecular Biophysics & Biochemistry, Yale University School of Medicine, New Haven, CT 06510. USA

(Received 8 February ! 991; accepted 25 February 1991 )

Summary m During tile directional strand exchange that is promoted by RecA protein between linear duplex DNA and circular single-stranded DNA, a triple-stranded DNA intermediate was formed and persisted even after the completion of strand transfer fol- lowed by deproteinization. In the deproteinized three-stranded DNA complexes, the sequestered linear third strand resisted digestion by E coli exonuclease !. In relation to polarity of strand exchange which defines the proximal and distal ends of the duplex DNA, when homology was restricted to the distal region of duplex substrate, the joints formed efficiently and were stable even upon complete deproteinization. Enzymatic probing of deproteinized distal joints with nuclease P1 revealed that the joints consist of long three-stranded structures that at neutral pH lack significant single-stranded character in any of the three strands. Instead of circular single-stranded DNA, when a linear single strand is recombined with partially homologous duplex DNA, in the presence of SSB, the formation of homologous joints by RecA protein, is significantly more efficient at distal end than at the proximal. Taken together, these observations suggest that with any single-stranded DNA (circular or linear), RecA protein efficiently promotes the formation of distal joints, from which, however, authentic strand exchange may not occur. Moreover, these joints might represent an intermediate which is trapped into a stable triple stranded state.

RecA protein / recombination / triple-stranded intermediates / strand exchange

Introduct ion

Homologous pairing and strand exchange promoted by RecA protein involves two right-handed helices: the naked duplex DNA and the RecA nucleoprotein filament containing single-stranded DNA [1, 2]. Howard-Flanders et al [3] proposed that extensive three-stranded, or even four-stranded helices are created within RecA nucleoprotein filament as inter- mediates in strand exchange. Observations by elec- tronmicroscopy support that view [4-6]. Recent biochemical studies detected intermediates in which the three strands of DNA were seen to have held to- gether even in the absence of RecA protein scaffold [7]. Therefore, it is plausible that during strand exchange the two helices, RecA nucleoprotein fila- ment and the naked duplex DNA, merge into a common three-stranded helix of some unknown length, rotation of which along the long axis serves to

*Correspondence and reprints. Present address: Groupe d'Etude 'Mutag6n~:se et Canc6rog6n~:se' Laboratoire d'Enzy- mologie, CNRS, F-91198 Gif-sur-Yvette, France

transfer one strand from the original duplex to the nucleoprotein filament through a spooling motion [8].

Strand exchange itself is directional 5' to 3' with respect to the circular plus strand or the displaced linear homolog which defines a proximal and a distal end in the duplex DNA [2, 9, 10]. However, the pairing event which precedes strand exchange is not polarized [1 1]. When the homology is limited to the distal region of the duplex DNA, polarity of strand transfer prevents the joints from undergoing true strand exchange reaction. Conceivably, the distal joints represent an intermediate in which the strands of DNA are aligned into a precursor state that precedes strand exchange. In this paper, we describe protein-free triple stranded DNA molecules formed during two different stages of RecA reaction, namely, pairing at the distal homologous end and strand exchange through fully homologous substrates.

Mater ia ls and M e t h o d s

Materials

RecA protein and E coli single-stranded DNA binding protein (SSB) were purified according to published protocols 112. 131.

364 BJ Rao et al

Restriction endonucleases Ncol, Kpnl, XhoI, SacI, NarI, MluI and BamHl. were purchased from New England Biolabs. SI nuclease was purchased from Pharmacia LKB. PI nuclease was from United States Biochemical Corporation. Proteinase K was from EM laboratories. Exonuclease I was a gift from S Kushner.

Preparation of DNA subso'ates

Unlabeled circular single-stranded DNA and superhelical I3HIDNA from Ml3Gori l , MI3 and G4 phages were prepared as described [14, 151. Preparations of circular single-stranded DNA contained less than 10% linear molecules as judged by agarose gel electrophoresis. For some experiments, circular single-stranded DNA was linearized by Narl or Mlul restriction enzyme cleavage of a short duplex region generated by annea- ling two 30-mers at the specific sites. Annealing was done by incubation of 500 laM MI3Goril circular DNA and 87 BM oligonucieotides in 100 mM NaCI, 10 mM Tris-HCI (pH 8.0), lbr 61) min at 65°C lbllowed by slow cooling for at least 7 h to reach the final temperature of 37°C. Digestion with Mlul was done for 2 h, with Narl tbr 4 h after a two-fold dilution of the sample and addition of 10 mM MgCI2, 2 mM DTT and 25 mM Tris-HCI (pH 7.5). The preparation of linear single-stranded DNA was treated with 100 Bg/ml proteinase K and 1% SDS, and purified on a sepharose 2B column.

Standard reaction conditions

Presynaptic filaments were formed by incubating 10 BM circu- lar single-stranded DNA with 5 BM RecA protein and 0.83 btM SSB at 37°C for 12 rain in a reaction mixture containing 33 mM Tris-HC! (pH 7.5), 12 mM MgCI2, 2 raM DTT, !.2 mM ATP, 8 mM phosphocreatine, 10 unit~A~:i creatine phos- phokinase and 100 Bg/ml BSA (nuclease free, Bethesda Research Laboratories). Pairing and strand exchange reactions were initiated by adding linear IaHIDNA at the concentrations indicated in the figure legends.

Assay fi,r joint molecules and strand exchange

The formation of joint molecules was measured by retention of linear double-stranded [3HIDNA on nitrocellulose filters. Joint molecv!es (D-loops) formed during the reaction were measured by diluting 10 lai aliquots at indicated time points into 200 BI of 20 mM EDTA, !% SDS on ice, followed 5 min later by the addition of 5 ml of chilled 1.5 M NaCI, 0.15 M sodium citrate (pH 7.5). The extent of strand exchange was measured by determining the fraction of 3H-labeled duplex DNA that became sensitive to nuclease S 1 [11] or oy detecting the ap- pearance of a characteristic product during gel electrophoresis.

At indicated time points, 25 B! aliquots were quenched into I% SDS and 20 mM EDTA. Proteinase K was added to a concentration of 100 Bg/ml and samples were incubated at 37°C for an additional 20 min before loading directly on a gel. Gel electrophoresis was done in 0.8% agarose using 40 mM- Tris acetate (pH 7.8) and 2 mM EDTA at 120 V/60 mA for 10 h at room temperature. The gel was dried and examined by autoradiography after an overnight exposure.

Deproteinization of joint molecules and isolation by gel filtration

After the formation of joint molecules was initiated by the addition of linear duplex DNA to an otherwise complete re-

action mixture, the reaction was carded out at 37°C for 15 min and stopped by adding EDTA to 20 mM, SDS to 0.5% and proteinase K to 100 Bg/ml, followed by incubation at 37°C for an additional 20 rain. The sample was filtered through a column of Sepharose 2B at 4-10°C to recover deproteinized joint molecules. During gel filtration, typically bed volumes of 3.5 ml and 7 ml were used for the sample volumes of 250 Ial and 500 ~tl respectively. Fractioas consisting of 2 drops were collected, and the void volume samples were located by count- ing 2 lal aliquots from the fractions. The joint molecules were free of bound proteins as monitored by their lack of retention on nitrocellulose filters that had been treated with alkali [ 16] as well as by ELISA analysis [17]. The concentration of joint molecules was expressed as the moles of nucleotide residues in the duplex DNA part of the molecule.

Specific assay to score the finished strand exchange products

The pairing reaction was initiated by adding to a standard re- action mixture 3H-labeled linear MI3 duplex DNA (10 ~tM) which was also 32p-labeled at both 5' ends. At appropriate time intervals, aliquots of 100 B! were withdrawn into 4.1 I.tl of 0.5 M EDTA (pH 8.0) and 1 ~! of 20 mg/ml proteinase K. These were stored on ice, and all the samples were sub- sequently incubated at 37°C for 20 min following which the samples were extracted with equal volumes of phenol, and later chloroform. The samples were precipitated in ethanol, the precipitates were washed once in chilled 75% ethanol, dried and subsequently dissolved in 150 I.tl of 10 mM Tris.HC! (pH 7.5) and 0.5 mM EDTA. Each sample (65 Bi) was digested with 1.2 B! lambda exonuclease (1 unit/~tl) in 2 mM MgCI2 at 37°C for 15 min, then digestions were stopped by adding 10 mM EDTA (pH 8.0) and 0.2% SDS. Aliquots were with- drawn from the samples to measure the fraction of 32p label that remained resistant to lambda exonuclease digestion and also to score the fraction of such complexes which are filter- retainable by a regular D-loop assay.

P! mwlease digestion

Deproteinized joints were incubated at 37°C in 30 mM Tris,HCl (pH 7.5), 0.5 mM EDTA (pH 8), 4 mM zinc acetate and 50 mM NaCI with various concentrations of nuclease PI. Digestions were stopped by adding 30 mM EDTA (pH 8) and 100 mM NaOH, followed by the analysis on alkaline gels.

Results

A stable DNA intermediate that still retains all three strands after the completion o f strand exchange

W h e n R e c A pro te in pairs a c i rcular s ing le s t rand wi th l inear dup lex D N A and br ings about c o m p l e t e strand exchange , the f inal products that can be scored af ter depro te in iz ing the mix tu re consis t o f a n i cked c i rcu lar he te roduplex and a d i sp l aced l inear strand. O n e o f the mode l s o f the reac t ion env i sages the e x c h a n g e o f s t rands wi th in a t r ip le-s t randed D N A nuc leopro te in c o m p l e x [3]. I f this were true, the f inal D N A products o f the s trand e x c h a n g e migh t be de tec ted in a pre- cursor in te rmedia te , n a m e l y th ree-s t randed D N A complex . The re fo re we dev i sed the f o l l o w i n g exper-

Production of triple-stranded recombination intermediates by RecA protein 365

iment to monitor the association of the linear single strand with the deproteinized heteroduplex DNA product in which strand exchange has gone to com- pletion (fig 1A). The protocol is based on the known specificity of lambda exonuclease which does not digest a 5' label in a nick, but readily digests the 5' label from the ends of either linear duplex or single- stranded DNA [18]. In a reaction between circular single-stranded DNA and (5'-32p)-labeled duplex DNA, the completion of strand exchange could be monitored by the appearence of nicked circular heteroduplex [32P]DNA as the sole species resistant to lambda-exonuc!ease digestion after the removal of RecA protein. Consequently, the stable association of the third strand with the nicked circular heteroduplex [32p] product, ie the three-stranded intermediate, gets retained by a nitrocellulose filter in high salt due to the associated single-stranded DNA.

The pairing reaction was initiated between M13 circular single-stranded DNA and uniformly labeled linear duplex M13 [3H]DNA in which both 5' ends were also 32P-labeled. The reaction showed complete strand exchange within 30 min as measured by S1 assay (fig 1B). As reported earlier [19], the joint mol-

ecules reached a maximum well before the completion of strand exchange and thereafter decayed slowly indicating the slow displacement of the third ~trand. During the same reaction, larger aliquots were with- drawn at different time intervals to prepare the de- proteinized samples of the strand exchange inter- mediates (See Materials and Methods). The depro- teinized intermediates were then treated with suf- ficient lambda-exonuclease to digest all the terminal (5'-32p) label of the sensitive species, namely un- reacted linear duplex DNA, unfinished intermediates and the displaced linear strand. Control lambda- exonuclease digestions involving uniformly 3H- labeled linear duplex DNA with 32p at 5' ends and the same DNA after heat denaturation released about 95% to 100% 32p label and about 20% 3H-label into acid- soluble form (data not shown).

The progressive appearance of the 32P-labeled nicked circular heteroduplex DNA was quantitated as a function of time by measuring the fraction of total [32P]DNA that remained insoluble in trichioroacetic acid after lambda exonuclease treatment. As shown in figure 1, this time course closely matched the strand exchange reaction as monitored by S 1 nuclease assay.

B

lOO

90

s' e [3HI c-) 3~ F,

I / I Homologous Pairing, Completion of Completion of

Initiation of Strand Switching Strand Switching Strand Displacement & Displacement

80

70

60 "-~

.~ 50

40

30

5' HETERODUPLEX DNA

20

I i I I I i I 0 10 20 30 40 50 60 70 80

Time (min)

Fig 1. Strand exchange between circular DNA and homologous linear duplex DNA goes through a three-stranded DNA inter- mediate. Panel A: Strand exchange reaction pathway. (0): 5'-32p label sensitive to lambda exonuclease digestion; (e): 5'-32p label resistant to lambda exonuclease digestion. Panel B: Time course of formation of three-stranded DNA intermediates during strand exchange reaction. The samples were deproteinized at different times of strand exchange reaction. Aliquots were withdrawn from these DNA samples to measure D-loops (o--o), heteroduplex DNA by S I assay (~:,), 32p label resistant to lambda exonuclease digestion (~--~) and 32p-label retained in the D-loop assay after lambda exonuclease digestion (x-x).

366 BJ R a o et al

The autoradiographic analyses of the agarose gel revealed that as expected, the lambda exonuclease removed the terminal 32p-label from all the species of unfinished strand exchange intermediates, from the unreacted duplex substrates and also from the dis- placed end of the linear single-stranded DNA. Throughout the reaction, only nicked circular hetero- duplex generated by the completion of strand exchange, was resistant to lambda-exonuclease (data not shown). We looked for DNA complexes that contained all three strands by measuring the percent- age of lambda-exonuclease resistant label that was retained by nitrocellulose filters in the D-loop assay. During the first 30 min, nearly all the 32p label that was present in completed nicked circular heteroduplex DNA was also retained by nitrocellulose filters by D- loop assay. As diagrammed in figure I A, this in- dicates that even after the removal of RecA protein by proteinase K, the third strand remained stably as- sociated with the nascent heteroduplex DNA in a novel complex. Controls indicated that retention was not due to incomplete digestion of RecA protein by proteinase K. During the second phase of the reaction which encompassed the subsequent 50-60 rain, the third strand of the three-stranded complexes slowly got displaced as evidenced by the ,,radual loss of D- loops in the 32P-labeled lambda-exonuclease resistant complexes (fig i B).

The third strand of the three-stranded DNA conq~lexes formed qfter the completion of strand e.whange is resistant to exonuc/ease ! digestion

Exonuclease 1 is a single-strand DNA specilic nu- clease which acts in 3' to 5' direction [20] and hence was used for analyzing the accessibility of the third strand from the distal end of the three-stranded complex.

The three-stranded DNA complexes were obtained fi'om a strand exchange reaction between M l3 circular single-stranded DNA and 3H-M13 linear duplex which was also labeled with 32p at both 5' ends. The complexes were deproteinized and characterized for the 'three-strandedness' by using lambda exonuclease digestion protocol described earlier in this paper. An aliquot of the sample revealed that about 91% of linear duplex substrate molecules were converted into nicked circular heteroduplex product, all of which were also present in three-stranded DNA complexes as monitored by the D-loop assay after lambda-exo- nuclease treatment.

Digestion with exonuclease I of the deproteinized DNA complexes ceased when about 15% of the third strand was made acid soluble (fig 2) indicating that the "displaced' strand remains partially protected after completion of strand exchange. In contrast, when the sample was heated at 70°C for l0 rain, the third strand

80

70

60

o

._g

I t 8° 7O

• 60

50 i::3

Cl 40

so ~

20 ~ _ ,. ~

10

. . . . ~ 0

40

30

20

10

0 0 10 20 30 ' 40

Reaction Time (rain)

Fig 2. The third strand of the three-stranded DNA complexes is resistant to exonuclease I digestion. Three- stranded DNA complexes (14.8 laM; 62 lal) were digested with 0.046 unit of exonuclease ! at 37°C in a buffer contain- ing 20 mM Tris.HCi (pH 8.0), 0.4 mM EDTA, 3 mM MgCI2, 2 mM dithiothreitoi and 100 lag bovine serum albu- mine/ml. At different time intervals, 8 lal portions were withdrawn into 100 lai of ! !% trichloroacetic acid with 20 lag of carrier DNA. The samples were stored on ice for 20 min and subsequently centrifuged in a microfuge for 15 min. The percentage of total 3H CPM rendered acid- soluble were multiplied by 2 to express the digestion in terms of the percentage of strand digested, and plotted as a function of time (A). A parallel reaction included the three- stranded complexes which were heated at 70°C for 10 min, slow-cooled to 37°C and subsequently digested with exo- nuclease I (, ~), Control digestion reactions were performed with 7.4 laM 3H-labeled MI3 linear duplex DNA (A) and also with the same concentration of heat denatured SH- labeled M 13 linear duplex DNA [e].

from the complexes was released to a form com- pletely sensitive to exonuclease I digestion. In addi- tion, with the same heat treatment, the digestion of duplex [SH]DNA by exonuclease I did not increase above the usual background level of 0.5% (data not shown). A parallel reaction with heat denatured [3H] duplex DNA served as a positive control to represent the complete digestion by exonuclease I under the same conditions.

Resistance to Nuclease PI of all the three strands in the distal joints

Strand exchange itself is directional 5' to 3' with respect to the incoming plus strand in the RecA

P r o d u c t i o n o f t r ip le -s t randed r e c o m b i n a t i o n in te rmedia tes b y R e c A prote in 367

filament or its displaced linear homolog [2, 9, 10]. However, the synaptic pairing that precedes the strand exchange is not strongly polarized but rather can occur at the proximal 5' end of plus strand or the distal 3' end of a linear duplex molecule (fig 3) or anywhere in between. Since the distal joints lack displacable 5' end in the region of homology, directional strand exchange does not ensue. Therefore, we surmised that such a joint molecule might represent an intermediate which is trapped into a stable three-stranded state that can not be further resolved into the strand exchange products, namely heteroduplex DNA and displaced strand. To test this hypothesis, we did the following structural characterization of the deproteinized distal joints.

The distal joints were formed between circular single-stranded DNA and 3H-labeled duplex DNA, deproteinized and purified by gel filtration (See Materials and Methods). The deproteinize¢t joint molecules, after gel filtration, contained about 70% of 3H duplex molecules as stable joints. The joints were completely free of bound protein as monitored by ELISA (data not given). In the following experiments, we separately assessed the sensitivity to nuclease P1 of each strand in distal joints. Nuclease P I is similar to S 1 nuclease in its specificity for cleavage of single strands, but has the advantage of acting at neutral pH [21, 22]. It has been used to detect single-stranded regions in replication intermediates [23] and to detect three-stranded intermediates formed by RecA protein during strand exchange [16].

5'

STRAND EXCHANGE BY FULL Y HOMOLOGOUS MOLECULES

1)

( ) s ' /" / (~) 3'

(+) ~3' p r o x i m a l - - - a,. d i s t a l

SUBSTRATES W ITH RESTRICTED REGIONS OF HOMOL OGY

. . ,, ~ d~=.-=~--------~_

d e

1+1 1+1

f g

Fig 3. The formation of joints in relation to the polarity of strand exchange when either circular single-stranded or linear single-stranded DNA recombines with a linear duplex DNA in the presence of RecA protein. Heavy stippled lines denote the regions of heterology. When the homology was limited as shown, the joints formed were either proximal (d, f) or distal (e, g).

The distal homologous end of the parental chimeric duplex DNA was uniquely labeled with 32p either at the 3' end of the plus strand or at the 5' end of the minus strand (see the legend to fig 4B, C). About 75% of 32p radioactivity was in stable deproteinized joints and migrated as a distinct band during electrophoresis in a non-denaturing agarose gel (fig 4A). The distal joints and the heat denatured control samples were titrated with nuclease P1 under identical conditions, followed by the electrophoretic analysis of the di- gested products on denaturing alkaline agarose gels. The autoradiographic analysis clearly revealed that, in the distal joint, both the strands contributed by the parental duplex molecule were resistant to nuclease P I, while under the same conditions of digestion, the denatured single-stranded DNA controls were sensi- tive (fig 4B, C). The control joints incubated without nuclease P l, remained stable, without any detectable loss in the total level of joint molecules in the samples (data not shown). It is important to note that the elec- trophoretic analysis used is a kind of footprinting. Samples were denatured prior to electrophoresis and analyzed on denaturing gels. Cleavage by PI nuclease anywhere in each polynucleotide strand should have caused terminal label to migrate as a smaller species.

The circular single-stranded DNA in the RecA filament contributes the third strand that is part of the distal joint. To study the action of nuclease PI on this strand, the joints were formed by pairing uniformly 3H-labeled circular single-stranded DNA with duplex DNA which was 32p-labeled at 5' ends. The pairing reaction was done with an excess of linear duplex DNA such that all of the single-stranded [3HIDNA was incorporated into joints (details in legend to fig 4D). The time course of digestion on deproteinized distal joints by nuclease PI revealed that about 30% of labeled single-stranded DNA was resistant under conditions which caused complete digestion of control circular single-stranded [3HIDNA from RecA filaments, which had gone through the same protocol of deproteinization and gel filtration (fig 4D). In ad- dition, the single-stranded [3HIDNA in the deprotein- ized joint molecules became fully sensitive to nuclease P I after heat denaturation of these joint molecules.

The fraction of 3H-labeled circular DNA that re- sisted nuclease P I was equivalent to 2000 nucleotide residues. Thus the experiments with nuclease PI demonstrate that a long deproteinized distal joint lacks single-strandedness in any of the three strands in- volved.

Pairing reaction at proximal and distal ends of duplexes using linear single-stranded DNA

Since the ends of DNA chains are important in hom- ologous recombination in which RecA protein and

368 B J R a o e t al

B C D Label at 3'-end of plus strand Label at 5'-end of minus strand 100

Dena~red cent_re!_ _ J o i n t s _ . Denatured control Joints Sample

1 2 , ~1 0~ ~=, 45 01 05 15 45 0 01 05 15 45 01 05 1,5 45 Pl-nucleaseun~tsx 10 3 90

• lO 80 2

: :~ ' o 70

..... ~ o 60 CL

J ' , i~i . ~. 40

I= ~ ~ 30

20

10

0 0

- - - - - O O 0 O O

I ! I

10 20 3O

Time (min i

Fig 4. Insensitivity to PI nuclease of the two strands contributed to a distal joint by the original duplex DNA. MI3 circular single-stranded [3H]DNA (10 ~tM) was paired with 32p end-labeled chimeric duplex large fragment (8 ~tM) derived from the Kpnl/BamHI double digestion of Ml3Goril form I DNA, the distal end of which was singly T-labeled by Klenow fragment of E coli DNA polymerase I or 5'-labeled by T4 polynucleotide kinase. This was achieved by labeling both ends of full length M 13 Gori duplex DNA after the first cut with BamHI, followed by the second cut with KpnI and purification of the large frag- ment. (A) The joint molecules were analyzed by electrophoresis on 0.8% nondenaturing agarose gel. Lane 1: joint molecules; Lane 2: standard duplex substrate. (B and C) Aliquots (40 l.tl) of the singly 32P-labeled joint molecules (3 I~M) were treated with various levels of nuclease P1 (0.1, 0.5, 1.5 or 4.5 10-3 units). Parallel control digestions were performed under identical conditions after heat denaturing the joint molecules at 95°C for 2 min, followed by quick chilling. All the samples were de- natured with 0.1 M NaOH before analyzing by electrophoresis on 1% Seakem GTG agarose (FMC Bioproducts) under alkaline gel conditions as described [27]. (D) The joint molecules (300 l.ti) were digested with nuclease PI (5 10-2 units/nmol). At different time points, 50 ~tl aliquots were withdrawn into 100 gl of 15% trichloroacetic acid with 20 ~g of carder DNA. The samples were stored on ice for 20 min and subsequently centrifuged in a microcentrifuge for 15 min. The percentage of 3H M 13 plus strand that remained acid-precipitable was plotted as a function of time (A). Under identical conditions, a control digestion was performed after heat-denaturing the joints at 95°C for 2 min, followed by quick chilling (A). Another positive control included the digestion of M I3 circular single-stranded [3H]DNA (2.1 I.tM) obtained after the standard protocol of presynapsis with RecA protein and SSB, followed by deproteinization (o). As a negative control, MI3 linear duplex [3H]DNA (5.22 ~M) was digested with nuclease P I in a parallel incubation (0).

similar prokaryotic proteins play a central role, the way in which such proteins process ends is important [24]. To study this question we wanted to analyze whether distal joints, the novel synaptic intermediates in which all the three strands are trapped into long triplex structures, are also formed in a reaction between single-stranded DNA having free ends (linear single strands) and a linear duplex where the hom- ology was limited to the distal end (fig 3). In order to score the stable joints, we deproteinized the reaction samples and analyzed by agarose gel electrophoresis. As additional controls for comparison, we analyzed the proximal and the distal joints formed by circular single-stranded DNA as well as the proximal joints formed by linear single-stranded DNA. All these four sets of pairing reactions were done under the same conditions using RecA and SSB proteins. In addition, the length of homology at either ends was the same (2000 base pairs). In this electrophoretic system, the

joint molecules showed retarded mobility as com- pared to that of unreacted duplex substrates. When the single-stranded DNA had no free ends (circular single strands), the formation of synaptic joints was efficient both at proximal as well as at distal ends of the duplex (fig 5). However, with the linear single-stranded DNA, the proximal joints were barely detectable, while the distal joints formed as efficiently as that seen with circular single-stranded DNA. These results suggest that in the presence of SSB, the predominant synaptic intermediate formed by RecA protein between linear single strands and duplex DNA is a distal joint, which as described earlier in this paper, might include a long triplex structure.

Discussion

The notion that RecA protein aligns homologous sequences in multistranded DNA structures (triple and

Production of triple-stranded recombination intermediates by RecA protein 369

SS.DNA Linear l inear Circular Circular

Homologous, 5' 3' 5' 3'

SSB + + + +

Time (rain) 5 10 20 30 5 10 20 30 0 5 10 20 30 5 10 20 30

Fig 5. The formation of joint molecules at 5' vs 3' ends as measured by gel electrophoresis. Reactions were performed under standard conditions described in Materials and Methods. At various times, samples were treated with 1% SDS and 100 l.tg/ml proteinase K, and analyzed by electro- phoresis in agarose gels.

four-stranded) appeared soon after the discovery of its ability to promote homologous pairing and strand exchange [3]. Pairing between molecules that lack free ends [25], model building [4], electron micro- scopy [4--6], and physical characterization of the intermediates [7] strongly supported the notion. In this paper we describe results which reveal that the three- stranded DNA intermediates are formed during RecA pairing and strand exchange reactions. In vitro, the products of homologous recombination between circular single-stranded DNA and fully homologous linear duplexes are linear single-strand DNA and nicked circular duplex DNA. Here, we detected an intermediate in which the products of the reaction were associated into a three-stranded complex which, even after deproteinization by a variety of methods, could withstand isolation procedures involving gel filtration and sucrose density gradient centrifugation (data not given). In this complex, any lingering attach- ment of the plus strand to the heteroduplex DNA may involve different interactions than those in the original Watson-Crick duplex DNA. On the other hand, the three-stranded complexes could simply reflect a struc- ture in which the displaced plus strand was physically wound around the new heteroduplex DNA in loose solenoidal turns but not otherwise chemically linked. However, the resistance of the displaced plus strand to exonuclease I suggests that the nascent displaced strand is not a free single strand, but is more inti- mately associated with the heteroduplex DNA. The deproteinized three-stranded DNA intermediate reflects the occurrence of a more complex nucleo- protein structure that was created by RecA protein

during strand exchange reaction. Consistent with this mechanism, when the molecules were paired at the distal region of homology in the duplex DNA, we detected yet another three-stranded synaptic inter- mediate. Since the strand exchange is directional in 5' to 3' on the plus strand (see fig 3), distal joints do not exchange strands but represent a trapped synaptic intermediate which is a precursor to strand exchange. Enzymatic probing of deproteinized distal joints suggests that these consist of long triplex structures that at neutral pH lack single-stranded character in any of the three strands. The deproteinized distal joints were about as long as the region of available homology, as indicated by the protection of several kb of the circular single strand from digestion by Pl nuclease. Moreover, terminal 5' and 3' labels at the distal homologous end of the duplex resisted a single cut by Pl nuclease anywhere along their length. Distal joints might represent a stable triplex DNA which could serve a useful role in some DNA recombination and repair events that may not require any extensive strand exchange.

This speculation provided an added impetus to verify whether such a synaptic joint is also formed in a reaction where single-stranded DNA has free ends (linear) as compared to the one without ends (circu- lar). Moreover, it is likely that during recombination and repair in vivo, ends of the single-stranded DNA might be important in regulating the way enzymes process the intermediates to the final products. We assessed the relative frequency of stable pairing at the proximal as well as the distal ends between linear single strands and partially homologous linear duplexes by gel electrophoresis. Reactions with circu- lar single-strands provided the positive controls. In these experiments, joints at the distal end of the duplex DNA, which corresponds to the 3' end of a linear single strand in the RecA filament, were formed more efficiently than those at the proximal end at least by a factor of three (quantitative data not shown). Therefore, in the presence of SSB, of which there are several thousand copies in the cell, pairing is favored at the 3' end from which strand exchange can not proceed. By inference, we believe that these synaptic intermediates formed between the 3' end of a linear single strand and the distal homology of the duplex DNA, might also encompass an extensive triplex DNA, whose, as yet undiscovered structure and mode of processing, could perhaps lead to a novel pathway of recombination. Unlike the artificial triplexes that are formed in vitro between polypurine and polypyri- midine strands [26], the triplexes formed by recombi- nation enzymes should sequester the two like-strands in a parallel orientation and presumably exhibit little or no sequence dependence. The putative triplex DNA structure, an intermediate in homologous recombi- nation, evokes a strong curiosity.

370 BJ Rao et al

Acknowledgments

We are grateful to Dr C Radding for helpful discussions and critical supervision. M Dutreix was supported in part by the Centre National de la Recherche Scientifique. and The Association pour la Recherche sur le Cancer. We thank C Wong for technical assistance and L Romanik and M Pierre for preparing the manuscript.

References

I Radding CM (1988) Homologous pairing and strand exchange promoted by Escherichia coli RecA protein, in: Genetic Recombimaion (Kucherlapati R, Smith GR, eds) American Society for Microbiology, Washington, DC, 193-229

2 Cox MM, Lehman IR (1981) Directionality and polarity in recA protein-promoted branch migration. Proc Nati Acad Sci USA 78, 6018-6022

3 Howard-Flanders P, West SC, Stasiak A (1984) Role of RecA protein spiral filaments in genetic recombination. Nature 309, 2 ! 5-220

4 Stasiak A, Stasiak AZ, Koller T (1984) Visualization of RecA-DNA complexes involved in consecutive stages of an in vitro strand exchange reaction. Cold Spring Harbor Syrup Quant Biology 49, 560-571

5 Register III JC, Christiansen G, Griffith J (1987) Electron microscopic visualization of the RecA protein-mediated pairing and branch migration phases of DNA strand exchange. J Biol Chem 262, 12812- ! 2820

6 Um!auf SW, Cox MM, lnman RB (1990) Triple-helical DNApairing intermediates formed by RecA protein. J Biol Chem 265, ! 6898-16912

7 Hsieh P, Camerini-Otero CS, Camerini-Otero RD (1990) Pairing of homologous DNA sequences by proteins: evidence tbr three-stranded DNA. Genes and Dev 4, 1951-1963

8 Honigberg SM, Radding CM (1988) The mechanisms of winding and unwinding helices in recombination: Torsional stress associated with strand transfer promoted by RecA protein. Cell 54, 525-532

9 West SC, Cassuto E, Howard-Flanders P (1981) Heteroduplex formation by RecA protein: Polarity of strand exchanges. Proc Natl Acad Sci USA 78, 6149-6 ! 53

i0 Kahn R, Cunningham RP, DasGupta C, Radding CM (1981) Polarity of heteroduplex formation promoted by Escherichia coli RecA protein. Proc Natl Acad Sci USA 78. 4786-4790

11 Wu AM, Kahn R, DasGupta C, Radding CM (1982) Formation of nascent heteroduplex structures by RecA protein and DNA. Cell 30, 37-44

12 Shibata T, Cunningham RP, Radding CM (1981) Homologous pairing in genetic recombination: purification and characterization of E coli RecA protein. J Biol Chem 256, 7557-7564

13 Lohman TM, Green JM, Beyer RD (1986) Large-scale overproduction and rapid purification of the Escherichia coli SSB gene product. Expression of the SSB gene under lambda PL control. Biochemistry 25, 21-25

14 Cunningham RP, Wu AM, Shibata T, DasGupta C, Radding CM (1981 ) Homologous pairing and topological linkage of DNA molecules by combined action of E coli RecA protein and topoisomerase I. Cell 24, 213-223

15 DasGupta C, Shibata T, Cunningham RP, Radding CM (1980) The topology of homologous pairing promoted by RecA protein. Cell 22, 437-446

16 Rao BJ, Jwang B, Radding CM (1990) RecA protein reinitiates strand exchange on isolated protein-free DNA intermediates. An ADP-resistant process. J Mol Biol 213, 789-809

17 Buchanan D, Karmarck M, Ruddle NH (1981) Develop- ment of a protein A enzyme immunoassay for use in screening hybridomas. J Immunol Methods 42, 179-185

18 Sriprakash KS, Lundh N, Moo-On-Huh M, Radding CM (1975) The specificity of lambda exonuclease. J Biol Chem 250, 5438-5445

19 Chow SA, Rao BJ, Radding CM (1988) Reversibility of strand invasion promoted by RecA protein and its inhi- bition by Escherichia coli single-stranded DNA-binding protein or phage T4 gene 32 protein. J Biol Chem 263, 200-209

20 Lehman IR, Nussbaum AL (1984) The deoxyribo- nucleases of Escherichia coli. V. On the specificity of exonuclease ! (phosphodiesterase). J Biol Chem 239, 2628-2636

21 Fujimoto M, Kuninaka A, Yoshino H (1974) Substrate specificity of nuclease P. Agric Biol Chem 38, ! 555-1561

22 Kowalski D (1984) Changes in site specificity of single- strand specific endonucleases on supercoiled PM2 DNA with temperature and ionic environment. Nucleic Acids Res 12, 7071-7086

23 Bramhill D, Kornberg A (1988) Duplex opening by dnaA protein at novel sequences in initiation of replication at the origin of the E coil chromosome. Cell 52, 743-755

24 Smith GR (1988) Homologous recombination in proca- ryotes. Microbiol Rev 52, 1-28

25 Bianchi M, DasGupta C, Radding CM (1983) Synapsis and the formation of paranemic joints by E coli RecA protein. Cell 34, 931-939

26 Htun H, Dahiberg JE (1988) Single strands, triple strands and kinks in H-DNA. Science 241, 1791-1796