Rad51 uses one mechanism to drive DNA strand exchange in both directions

The Rad51 protein of Saccharomyces cerevisiae, like its bacterial counterpart RecA, promotes strand exchange between circular single-stranded DNA (ssDNA) and linear double-stranded DNA (dsDNA) in vitro. However, the two proteins differ in the requirement for initiating joint molecules and in the polarity of branch migration. Whereas RecA initiates joint molecules from any type of ends on the dsDNA and branch migration proceeds exclusively in the 5'- to 3'-direction with respect to the single strand DNA substrate, initiation mediated by Rad51 requires a complementary 3' or 5' overhanging end of the linear dsDNA and branch migration proceeds in either direction. Here we report that the rates of Rad51-mediated branch migration in either the 5'- to 3'- or 3'- to 5'-directions are affected to the same extent by temperature and MgCl(2). Furthermore, branch migration in both directions is equally impeded by insertions of non-homologous sequences in the dsDNA, inserts of 6 base pairs or more being completely inhibitory. We have also found that the preference of strand exchange in the 5'- to 3'-direction does not change if RPA is replaced by Escherichia coli SSB or T4 gene 32 proteins, suggesting that the preference for the direction of strand exchange is intrinsic to Rad51. Based on these results, we conclude that Rad51-promoted branch migration in either direction occurs fundamentally by the same mechanism, quite probably by stabilizing successively formed heteroduplex base pair.     

The human Rad51 protein: polarity of strand transfer and stimulation by hRP-A.

The human Rad51 protein is homologous to the RecA protein and catalyses homologous pairing and strand transfer reactions in vitro. Using single-stranded circular and homologous linear duplex DNA, we show that hRad51 forms stable joint molecules by transfer of the 5' end of the complementary strand of the linear duplex to the ssDNA. The polarity of strand transfer is therefore 3' to 5', defined relative to the ssDNA on which hRad51 initiates filament formation. This polarity is opposite to that observed with RecA. Homologous pairing and strand transfer require stoichiometric amounts of hRad51, corresponding to one hRad51 monomer per three nucleotides of ssDNA. Joint molecules are not observed when the protein is present in limiting or excessive amounts. The human ssDNA binding-protein, hRP-A, stimulates hRad51-mediated reactions. Its effect is consistent with a role in the removal of secondary structures from ssDNA, thereby facilitating the formation of continuous Rad51 filaments.     

ATP hydrolysis and the displaced strand are two factors that determine the polarity of RecA-promoted DNA strand exchange.

When the recA protein (RecA) of Escherichia coli promotes strand exchange between single-stranded DNA (ssDNA) circles and linear double-stranded DNAs (dsDNA) with complementary 5' or 3' ends a polarity is observed. This property of RecA depends on ATP hydrolysis and the ssDNA that is displaced in the reaction since no polarity is observed in the presence of the non-hydrolyzable ATP analog, ATP gamma S, or in the presence of single-strand specific exonucleases. Based on these results a model is presented in which both the 5' and 3' complementary ends of the linear dsDNA initiate pairing with the ssDNA circle but only one end remains stably paired. According to this model, the association/dissociation of RecA in the 5' to 3' direction on the displaced strand determines the polarity of strand exchange by favoring or blocking its reinvasion into the newly formed dsDNA. Reinvasion is favored when the displaced strand is coated with RecA whereas it is blocked when it lacks RecA, remains covered by single-stranded DNA binding protein or is removed by a single-strand specific exonuclease. The requirement for ATP hydrolysis is explained if the binding of RecA to the displaced strand occurs via the dissociation and/or transfer of RecA, two functions that depend on ATP hydrolysis. The energy for strand exchange derives from the higher binding constant of RecA for the newly formed dsDNA as compared with that for ssDNA and not from ATP hydrolysis.     

RecA protein-promoted homologous pairing and strand exchange between intact and partially single-stranded duplex DNA.

In the pairing reaction between circular gapped and fully duplex DNA, RecA protein first polymerizes on the gapped DNA to form a nucleoprotein filament. Conditions that removed the formation of secondary structure in the gapped DNA, such as addition of Escherichia coli single-stranded DNA binding protein or preincubation in 1 mM-MgCl2, optimized the binding of RecA protein and increased the formation of joint molecules. The gapped duplex formed stable joints with fully duplex DNA that had a 5' or 3' terminus complementary to the single-stranded region of the gapped molecule. However, the joints formed had distinct properties and structures depending on whether the complementary terminus was at the 5' or 3' end. Pairing between gapped DNA and fully duplex linear DNA with a 3' complementary terminus resulted in strand displacement, symmetric strand exchange and formation of complete strand exchange products. By contrast, pairing between gapped and fully duplex DNA with a 5' complementary terminus produced a joint that was restricted to the gapped region; there was no strand displacement or symmetric strand exchange. The joint formed in the latter reaction was likely a three-stranded intermediate rather than a heteroduplex with the classical Watson-Crick structure. We conclude that, as in the three-strand reaction, the process of strand exchange in the four-strand reaction is polar and progresses in a 5' to 3' direction with respect to the initiating strand. The present study provides further evidence that in both three-strand and four-strand systems the pairing and strand exchange reactions share a common mechanism.     

Homologous pairing in vitro stimulated by the recombination hotspot, Chi.

Genetic recombination in Escherichia coli is stimulated at DNA sequences known as Chi sites, 5'-GCT-GGTGG-3'. We describe the in vitro formation of homologously paired joint molecules that is dependent upon this recombination hotspot. Chi-dependent joint molecule formation requires RecA, RecBCD, and SSB proteins and a Chi site in the donor linear dsDNA. The donor dsDNA is unwound by RecBCD enzyme, and the invasive strand is generated by nicking at Chi. This Chi-dependent invading strand must contain homology to the recipient supercoiled DNA substrate at its newly formed 3' end for efficient joint molecule formation. Action at Chi generates invasive ssDNA from the 5' but not the 3' side of Chi, suggesting that the nuclease activity of RecBCD enzyme is attenuated upon encountering a Chi site. These results support the view that RecBCD enzyme action can precede RecA protein action and reconcile the seemingly opposing degradative and recombination functions of RecBCD enzyme.     

Visualizing RAD51-mediated joint molecules: implications for recombination mechanism and the effect of sequence heterology

The defining event in homologous recombination is the exchange of base-paired partners between a single-stranded (ss) DNA and a homologous duplex driven by recombinase proteins, such as human RAD51. To understand the mechanism of this essential genome maintenance event, we analyzed the structure of RAD51–DNA complexes representing strand exchange intermediates at nanometer resolution by scanning force microscopy. Joint molecules were formed between substrates with a defined ssDNA segment and homologous region on a double-stranded (ds) partner. We discovered and quantified several notable architectural features of RAD51 joint molecules. Each end of the RAD51-bound joints had a distinct structure. Using linear substrates, a 10-nt region of mispaired bases blocked extension of joint molecules in all examples observed, whereas 4 nt of heterology only partially blocked joint molecule extension. Joint molecules, including 10 nt of heterology, had paired DNA on either side of the heterologous substitution, indicating that pairing could initiate from the free 3′end of ssDNA or from a region adjacent to the ss–ds junction. RAD51 filaments covering joint ss–dsDNA regions were more stable to disassembly than filaments covering dsDNA. We discuss how distinct structural features of RAD51-bound DNA joints can play important roles as recognition sites for proteins that facilitate and control strand exchange.     

The preference for a 3' homologous end is intrinsic to RecA-promoted strand exchange

The recA protein (RecA) promotes DNA pairing and strand exchange optimally in the presence of single-stranded binding protein (SSB). Under these conditions, 3' homologous ends are essential for stable joint molecule formation between linear single-stranded DNA (ssDNA) and supercoiled DNA (i.e. 3' ends are 50-60 times more reactive than 5' ends). Linear ssDNAs with homology at the 5' end do not participate in pairing. In the absence of SSB, the strand exchange reaction is less efficient; however, linear ssDNAs with 3' end homology are still 5- to 10-fold more reactive than those with 5' end homology. The preference for a 3' homologous end in the absence of SSB suggests that this is an intrinsic property of RecA-promoted strand exchange. The preferential reactivity of 3' homologous ends is likely to be a consequence of the polarity of polymerization of RecA on ssDNA. Specifically, since RecA polymerizes in the 5'----3' direction, 3' ends are more likely to be coated with RecA and, hence, will be more reactive than 5' ends.     

Homologous-pairing activity of the Bacillus subtilis bacteriophage SPP1 replication protein G35P

Genetic evidence suggests that the SPP1-encoded gene 35 product (G35P) is essential for phage DNA replication. Purified G35P binds single-strand DNA (ssDNA) and double-strand (dsDNA) and specifically interacts with SPP1-encoded replicative DNA helicase G40P and SSB protein G36P. G35P promotes joint molecule formation between a circular ssDNA and a homologous linear dsDNA with an ssDNA tail. Joint molecule formation requires a metal ion but is independent of a nucleotide cofactor. Joint molecules formed during these reactions contain a displaced linear ssDNA strand. Electron microscopic analysis shows that G35P forms a multimeric ring structure in ssDNA tails of dsDNA molecules and left-handed filaments on ssDNA. G35P promotes strand annealing at the AT-rich region of SPP1 oriL on a supercoiled template. These results altogether are consistent with the hypothesis that the homologous pairing catalyzed by G35P is an integral part of SPP1 DNA replication. The loading of G40P at a d-loop (ori DNA or at any stalled replication fork) by G35P could lead to replication fork reactivation.     

Single-stranded DNA binding protein and DNA helicase of bacteriophage T7 mediate homologous DNA strand exchange.

Two proteins encoded by bacteriophage T7, the gene 2.5 single-stranded DNA binding protein and the gene 4 helicase, mediate homologous DNA strand exchange. Gene 2.5 protein stimulates homologous base pairing of two DNA molecules containing complementary single-stranded regions. The formation of a joint molecule consisting of circular, single-stranded M13 DNA, annealed to homologous linear, duplex DNA having 3'- or 5'-single-stranded termini of approximately 100 nucleotides requires stoichiometric amounts of gene 2.5 protein. In the presence of gene 4 helicase, strand transfer proceeds at a rate of > 120 nucleotides/s in a polar 5' to 3' direction with respect to the invading strand, resulting in the production of circular duplex M13 DNA. Strand transfer is coupled to the hydrolysis of a nucleoside 5'-triphosphate. The reaction is dependent on specific interactions between gene 2.5 protein and gene 4 protein.     

Kinetic studies of DNA cleavage reactions catalyzed by an ATP-dependent deoxyribonuclease on a 27-MHz quartz-crystal microbalance

Catalytic DNA cleavage reactions by an ATP-dependent deoxyribonuclease (DNase) from Micrococcus luteus were monitored directly with a DNA-immobilized 27-MHz quartz-crystal microbalance (QCM). The 27-MHz QCM is a very sensitive mass-measuring device in aqueous solution, as the frequency decreases linearly with increasing mass on the electrode at a nanogram level. Three steps in ATP-dependent DNA hydrolysis reactions, including (1) binding of DNase to the end of double-stranded DNA (dsDNA) on the QCM electrode (mass increase), (2) degradation of one strand of dsDNA in the 3' --> 5' direction depending on ATP (mass decrease), and (3) release of the enzyme from the nonhydrolyzed 5'-free-ssDNA (mass decrease), could be monitored stepwise from the time dependencies of QCM frequency changes. Kinetic parameters for each step were obtained as follows. The binding constant (K(a)) of DNase to the dsDNA was determined as (28 +/- 2) x 10(6) M(-)(1) (k(on) = (8.0 +/- 0.3) x 10(3) M (-)(1) s(-)(1) and k(off) = (0.29 +/-0.01) x 10(-)(3) s(-)(1)), and it decreased to (0.79 +/- 0.16) x 10(6) M(-)(1) (k'(on) = (2.3 +/- 0.2) x 10(3) M (-)(1) s(-)(1) and k'(off) = (2.9 +/- 0.1) x 10(-)(3) s(-)(1)) for the completely nonhydrolyzed 5'-free ssDNA. This is the reason the DNase bound to the dsDNA substrate can easily release from the nonhydrolyzed 5'-free-ssDNA after the complete hydrolysis of the 3' --> 5' direction of the complementary ssDNA. K(a) values depended on the DNA structures on the QCM, and the order of these values was as follows: the dsDNA having a 4-base-mismatched base-pair end (3) > the dsDNA having a 5' 15-base overhanging end (2) > the dsDNA having a blunt end (1) > the ssDNA having a 3'-free end (4) > the ssDNA having a 5'-free end (5). Thus, DNase hardly recognized the free 5' end of ssDNA. Michaelis-Menten parameters (K(m) for ATP and k(cat)) of the hydrolysis process also could be obtained, and the order of k(cat)/K(m) was as follows: the dsDNA having a blunt end (1) approximately the dsDNA having a 4-base-mismatched base-pair end (3) > the ssDNA having a free 3' end (4) > the ssDNA having a free 5' end (5). Thus, DNase could not recognize and not hydrolyze the free 5' end of ssDNA. The DNA hydrolysis reaction could be driven by dATP and GTP (purine base) as well as ATP, whereas the cleavage efficiency was very low driven with UTP, CTP (pyrimidine base), ADP, and AMP.     

Heteroduplex formation by human Rad51 protein: effects of DNA end-structure, hRP-A and hRad52.

Purified human Rad51 protein (hRad51) catalyses ATP-dependent homologous pairing and strand transfer reactions, characteristic of a central role in homologous recombination and double-strand break repair. Using single-stranded circular and partially homologous linear duplex DNA, we found that the length of heteroduplex DNA formed by hRad51 was limited to approximately 1.3 kb, significantly less than that observed with Escherichia coli RecA and Saccharomyces cerevisiae Rad51 protein. Joint molecule formation required the presence of a 3' or 5'-overhang on the duplex DNA substrate and initiated preferentially at the 5'-end of the complementaryx strand. These results are consistent with a preference for strand transfer in the 3'-5' direction relative to the single-stranded DNA. The human single-strand DNA-binding protein, hRP-A, stimulated hRad51-mediated joint molecule formation by removing secondary structures from single-stranded DNA, a role similar to that played by E. coli single-strand DNA-binding protein in RecA-mediated strand exchange reactions. Indeed, E. coli single-strand DNA-binding protein could substitute for hRP-A in hRad51-mediated reactions. Joint molecule formation by hRad51 was stimulated or inhibited by hRad52, dependent upon the reaction conditions. The inhibitory effect could be overcome by the presence of hRP-A or excess heterologous DNA.     

Saccharomyces cerevisiae Mer3 helicase stimulates 3'-5' heteroduplex extension by Rad51; implications for crossover control in meiotic recombination

Crossover and noncrossover recombinants can form by two different pathways during meiotic recombination in Saccharomyces cerevisiae. The MER3 gene is known to affect selectively crossover, but not noncrossover, recombination. The Mer3 protein is a DNA helicase that unwinds duplex DNA in the 3' to 5' direction. To define the underlying molecular steps of meiotic recombination, we investigated the role of Mer3 helicase in DNA strand exchange promoted by Rad51 protein. We found that Mer3 helicase does not function as an initiator of DNA pairing events but, rather, it stimulates DNA heteroduplex extension in the 3' --> 5' direction relative to the incoming (or displaced) single-stranded DNA. Conversely, Mer3 helicase blocks DNA heteroduplex extension in the 5' --> 3' direction. Our results support the idea that Mer3 helicase stabilizes nascent joint molecules via DNA heteroduplex extension to permit capture of the second processed end of a double-stranded DNA break, a step which is required for crossover recombinant product formation.     

The effects on strand exchange of 5' versus 3' ends of single-stranded DNA in RecA nucleoprotein filaments

Since the ends of DNA chains are thought to be important in homologous recombination, the way in which RecA protein and similar recombination enzymes process ends is important. We analyzed the effects of ends both on the formation of joints, and the progression of strand exchange. When the only homologous end was provided by a single strand, there was no significant difference between the formation of joints at a 5' end or a 3' end; but in agreement with the report of Konforti & Davis, Escherichia coli single-stranded DNA binding protein (SSB) selectively inhibited the activity of 5' ends. Complete strand exchange, assessed by study of linear single-stranded and double-stranded substrates, took place only in the 5' to 3' direction relative to DNA in the nucleoprotein filament. These observations pose a paradox: in the presence of SSB, of which there are about 800 tetramers per cell, the formation of homologous joints by RecA protein is favored at a 3' end, from which, however, authentic strand exchange appears not to occur. Since observations reported here and elsewhere show that joints have different properties when formed at a 5' versus a 3' end, we suggest that they may be processed differently in vivo.     

Bacillus subtilis bacteriophage SPP1-encoded gene 34.1 product is a recombination-dependent DNA replication protein

SPP1-encoded replication and recombination proteins, involved in the early steps of the initiation of concatemeric DNA synthesis, have been analyzed. Dimeric G34.1P exonuclease degrades, with a 5' to 3' polarity and in a Mg2+-dependent reaction, preferentially linear double-stranded (ds) DNA rather than single-stranded (ss) DNA. Binding of the replisome organizer, G38P, to its cognate sites (oriDNA) halts the 5' to 3' exonucleolytic activity of G34.1P on dsDNA. The G35P recombinase increases the affinity of G34.1P for dsDNA, and stimulates G34.1P activity on dsDNA, but not on ssDNA. Then, filamented G35P promotes limited strand exchange with a homologous sequence. The ssDNA binding protein, G36P, protects ssDNA from the G34.1P exonuclease activity and stimulates G35P-catalyzed strand exchange. The data presented suggest a model for the role of G34.1P during initiation of sigma replication: G38P bound to oriDNA might halt replication fork progression, and G35P, G34.1P and G36P in concert might lead to the re-establishment of a unidirectional recombination-dependent replication that accounts for the direction of DNA packaging.     

Electron microscopic visualization of the RecA protein-mediated pairing and branch migration phases of DNA strand exchange

The RecA protein of Escherichia coli will drive the pairing and exchange of strands between homologous DNA molecules in a reaction stimulated by single-stranded binding protein. Here, reactions utilizing three homologous DNA pairs which can undergo both paranemic and plectonemic joining were examined by electron microscopy: supertwisted double-stranded (ds) DNA and linear single-stranded (ss) DNA, linear dsDNA and circular ssDNA, and linear dsDNA and colinear ssDNA. Several major observations were: (i) with RecA protein bound to the DNA, plectonemic joints were ultrastructurally indistinguishable from paranemic joints; (ii) complexes which appeared to be joined both paranemically and plectonemically were present in these reactions in roughly equal numbers; and (iii) in complexes undergoing strand exchange, both DNA partners were often enveloped within a RecA protein filament consisting of hundreds of RecA protein monomers and several kilobases of DNA. These observations suggest that, following RecA protein-ssDNA filament formation, strand exchange proceeds by a pathway that can be divided structurally into three phases: pairing, envelopment/exchange, and release of the products.     

The efficiency of strand invasion by Escherichia coli RecA is dependent upon the length and polarity of ssDNA tails

RecA protein is essential for homologous recombination and the repair of DNA double-strand breaks in Escherichia coli. The protein binds DNA to form nucleoprotein filaments that promote joint molecule formation and strand exchange in vitro. RecA polymerises on ssDNA in the 5'-3' direction and catalyses strand exchange and branch migration with a 5'-3' polarity. It has been reported previously, using D-loop assays, in which ssDNA (containing a heterologous block at one end) invades supercoiled duplex DNA that 3'-homologous ends are reactive, whereas 5'-ends are inactive. This polarity bias was thought to be due to the polarity of RecA filament formation, which results in the 3'-ends being coated in RecA, whereas 5'-ends remain naked. Using a range of duplex substrates containing ssDNA tails of various lengths and polarities, we now demonstrate that when no heterologous block is imposed, 5'-ends are just as reactive as 3'-ends. Moreover, using short-tailed substrates, we find that 5'-ends form more stable D-loops than 3'-ends. This bias may be a consequence of the instability of short 3'-joints. With more physiological substrates containing long ssDNA tails, we find that RecA shows no intrinsic preference for 5' or 3'-ends and that both form D-loop complexes with high efficiency.     

The direction of RecA protein assembly onto single strand DNA is the same as the direction of strand assimilation during strand exchange

The RecA protein of Escherichia coli optimally promotes DNA strand exchange reactions in the presence of the single strand DNA-binding protein of E. coli (SSB protein). Under these conditions, assembly of RecA protein onto single-stranded DNA (ssDNA) occurs in three steps. First, the ssDNA is rapidly covered by SSB protein. The binding of RecA protein is then initiated by nucleation of a short tract of RecA protein onto the ssDNA. Finally, cooperative polymerization of additional RecA protein accompanied by displacement of SSB protein results in a ssDNA-RecA protein filament (Griffith, J. D., Harris, L. D., and Register, J. C. (1984) Cold Spring Harbor Symp. Quant. Biol. 49, 553-559). We report here that RecA protein assembly onto circular ssDNA yields RecA protein-covered circles in which greater than 85% are completely covered by RecA protein with no remaining SSB protein-covered segments (as detected by electron microscopy). However, when linear ssDNA is used, 90% of the filaments contain a short segment at one end complexed with SSB protein. This suggests that RecA protein assembly is unidirectional. Visualization of the assembly of RecA protein onto either long ssDNA tails (containing either 5' or 3' termini) or ssDNA gaps generated in double strand DNA allowed us to determine that the RecA protein polymerizes in the 5' to 3' direction on ssDNA and preferentially nucleates at ssDNA-double strand DNA junctions containing 5' termini.     

Effect of the Streptococcus pneumoniae MmsA protein on the RecA protein-promoted three-strand exchange reaction. Implications for the mechanism of transformational recombination

Streptococcus pneumoniae is a naturally transformable bacterium that is able to incorporate DNA from its environment into its own chromosome. This process, known as transformational recombination, is dependent in part on the mmsA gene, which encodes a protein having a sequence that is 40% identical to that of the Escherichia coli RecG protein, a junction-specific DNA helicase believed to be involved in the branch migration of recombinational intermediates. We have developed an expression system for the MmsA protein and have purified the MmsA protein to more than 99% homogeneity. The MmsA protein has DNA-dependent ATP hydrolysis and DNA junction-helicase activities that are similar to those of the E. coli RecG protein. The effect of the MmsA protein on the S. pneumoniae RecA protein-promoted three-strand exchange reaction was also investigated. In the standard direction (circular single-stranded (ss) DNA + linear double-stranded (ds) DNA --> linear ssDNA + nicked circular dsDNA), the MmsA protein appears to promote the branch migration of partially exchanged intermediates in a direction opposite of the RecA protein, resulting in a nearly complete inhibition of the overall strand exchange reaction. In the reverse direction (linear ssDNA + nicked circular dsDNA --> circular ssDNA + linear dsDNA), however, the MmsA protein appears to facilitate the conversion of partially exchanged intermediates into fully exchanged products, leading to a pronounced stimulation of the overall reaction. These results are discussed in terms of the molecular mechanism of transformational recombination.     

Recognition and processing of double-stranded DNA by ExoX,a distributive 3′–5′ exonuclease

Members of the DnaQ superfamily are major 3′–5′ exonucleases that degrade either only single-stranded DNA (ssDNA) or both ssDNA and double-stranded DNA (dsDNA). However, the mechanism by which dsDNA is recognized and digested remains unclear. Exonuclease X (ExoX) is a distributive DnaQ exonuclease that cleaves both ssDNA and dsDNA substrates. Here, we report the crystal structures of Escherichia coli ExoX in complex with three different dsDNA substrates: 3′ overhanging dsDNA, blunt-ended dsDNA and 3′ recessed mismatch-containing dsDNA. In these structures, ExoX binds to dsDNA via both a conserved substrate strand-interacting site and a previously uncharacterized complementary strand-interacting motif. When ExoX complexes with blunt-ended dsDNA or 5′ overhanging dsDNA, a ‘wedge’ composed of Leu12 and Gln13 penetrates between the first two base pairs to break the 3′ terminal base pair and facilitates precise feeding of the 3′ terminus of the substrate strand into the ExoX cleavage active site. Site-directed mutagenesis showed that the complementary strand-binding site and the wedge of ExoX are dsDNA specific. Together with the results of structural comparisons, our data support a mechanism by which normal and mismatched dsDNA are recognized and digested by E. coli ExoX. The crystal structures also provide insight into the structural framework of the different substrate specificities of the DnaQ family members.     

DNA strand exchange catalyzed by proteins from vaccinia virus-infected cells.

Vaccinia virus infection induces expression of a protein which can catalyze joint molecule formation between a single-stranded circular DNA and a homologous linear duplex. The kinetics of appearance of the enzyme parallels that of vaccinia virus DNA polymerase and suggests it is an early viral gene product. Extracts were prepared from vaccinia virus-infected HeLa cells, and the strand exchange assay was used to follow purification of this activity through five chromatographic steps. The most highly purified fraction contained three major polypeptides of 110 +/- 10, 52 +/- 5, and 32 +/- 3 kDa. The purified protein requires Mg2+ for activity, and this requirement cannot be satisfied by Mn2+ or Ca2+. One end of the linear duplex substrate must share homology with the single-stranded circle, although this homology requirement is not very high, as 10% base substitutions had no effect on the overall efficiency of pairing. As with many other eukaryotic strand exchange proteins, there was no requirement for ATP, and ATP analogs were not inhibitors. Electron microscopy was used to show that the joint molecules formed in these reactions were composed of a partially duplex circle of DNA bearing a displaced single-strand and a duplex linear tail. The recovery of these structures shows that the enzyme catalyzes true strand exchange. There is also a unique polarity to the strand exchange reaction. The enzyme pairs the 3' end of the duplex minus strand with the plus-stranded homolog, thus extending hybrid DNA in a 3'-to-5' direction with respect to the minus strand. Which viral gene (if any) encodes the enzyme is not yet known, but analysis of temperature-sensitive mutants shows that activity does not require the D5R gene product. Curiously, v-SEP appears to copurify with vaccinia virus DNA polymerase, although the activities can be partially resolved on phosphocellulose columns.