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.

drive theta replication from the cis-acting oriL region in an otherwise non-replicative element in B. subtilis cells (7). Accumulation of ring-to-ring, theta type, SPP1 replication intermediates, however, has not been observed by electron microscopy examination of SPP1-infected cells (3). Upon infection, SPP1 circular and sigma molecules were detected, but branched replication intermediates have not been observed (3)(4)(5). The generation of concatemeric SPP1 DNA is at least dependent on phage-encoded (G38P, G39P, G40P, G34.1P, G35P, and G36P, see Fig. 1 below) and host-encoded (DNA PolIII, DnaG, and DNA topoisomerases) replication proteins but is independent of host-encoded components of the primosome (e.g. DnaB, DnaD, and DnaI) and recombination proteins (RecA, AddAB, and RecF) (5)(6)(7)(8)(9)(10). Furthermore, we could show that the SPP1 plating efficiency is indistinguishable on wt, priA1, and priA1 dnaB75 strains. 2 Analysis of SPP1 conditional-lethal mutants for their capacity to synthesize phage DNA has lead to the identification of two different complementation groups. Whereas mutants in genes 38, 39, and 40 show a block in DNA replication, mutants in genes 34.1 and 35 show a normal initiation but a replication arrest phenotype (5,6,10). Genes 38,39, and 40 and genes 34.1, 35, 36, and 36.1 are early transcribed genes that form part of two different operons (Fig. 1A). G36P, which is a helixdestabilizing single-stranded binding protein and shares 46% identity with B. subtilis SSB protein, does not seem to be essential under laboratory growth conditions (10). The activity of G36.1P, which shares a significant degree of identity with group I endonuclease proteins, remains to be characterized.
The initiation of theta-type replication at SPP1 oriL has been defined. In vitro studies reveal that multiple copies of G38P bound to its cognate site (AB and ab boxes, Fig. 1B) induce local unwinding of the adjacent AT-rich sequence (DUE region) present within oriL or oriR (7, see Fig. 1B). The helicase loader, G39P, specifically interacts with the replisome organizer, G38P, and with G40P⅐ATP (6,11). G40P⅐ATP is a bona fide hexameric DNA helicase with a 5Ј to 3Ј unwinding polarity (12)(13)(14). G39P directs the assembly of G40P⅐ATP to G38Pbound oriL and allows the loading and subsequent activation of the G40P⅐ATP helicase (11). G40P⅐ATP bound to the ssDNA region at the unwound AT-rich region interacts with DnaG and the subunit of DNA PolIII and then theta type replication initiates (7,12). 3 However, very little is known about the mechanism(s) involved in the recover of collapsed replication forks and in the generation of linear concatemeric SPP1 DNA.
The picture of how replication proceeds has changed over the last decade (reviewed in Refs. [15][16][17]. It is now assumed that the one or replication forks formed at the replication origin become inactivated at high frequency, for example as a result of roadblocks in or on the template strands. In bacteria, the repair of this stalled replication fork requires the RecA, RecBCD, RecF, RecG, and/or RuvABC recombination proteins that may create a D-loop that is recognized by a component of the primosome (see Refs. [15][16][17]. It has been shown that the PriA protein bound to the branched DNA molecule (recombination intermediate) directs the assembly of a new replication fork at the site through the loading of a primosome for theta type DNA replication (17). Furthermore, recent studies demonstrate that recombination intermediates formed between a linear duplex and supercoiled DNA are substrates for a DNA synthesis reaction where products longer than unit length are generated (18).
Although the presence of redundant pathways that could account for the recovery of an arrested replication fork and for the accumulation of concatemeric SPP1 DNA cannot be ruled out, we have previously shown that SPP1 replication does not require host-encoded homologous recombination proteins and components of the D-loop primosome assembly (Refs. 19 and 20, and this work). Furthermore, there are several lines of evidence that support the idea that SPP1 encodes in its genome recombination functions that may drive the restart of stalled replication forks and the switch from theta to sigma type of replication. First, plasmid transduction is markedly increased when homology between the plasmid and SPP1 is provided (9). Second, SPP1 infection triggers the appearance of long concatemeric plasmid linear molecules whose synthesis is dependent on homology, and phage-encoded functions, but independent of host-encoded recombination or primosome assembly functions (e.g. RecA, DnaB, etc.), whereas in non-infected cells synthesis of concatemeric plasmid molecules is dependent of both RecA and DnaB (21,22). It is likely, therefore, that SPP1 may encode for a system that coordinates the processing of inactivated replication forks and the subsequent fork reactivation and the synthesis of concatemeric SPP1 DNA.
G35P and G34.1P share an overall identity of 40 and 18% with the Escherichia coli recombination proteins RecT and RecE, respectively. RecT is a SSA protein and RecE is an ATP-independent 5Ј33Ј exonuclease, both encoded by the defective lambdoib Rac prophage (reviewed in Ref. 15). Unlike the homologous recombination systems of lambdoid phages (e.g. -red, P22-erf, etc.) that are required for growth in recA mutants (reviewed in Refs. 23 and 24), G35P and G34.1P are essential SPP1 replication proteins in both wt and recA host strains (3,5,10).
Sedimentation studies of DNA synthesized by SPP1tsI20F (impaired in G34.1P, see Fig. 1A) or SPP1tsI17 (impaired in FIG. 1. Physical and genetic map of the SPP1 replication operons. A, the SPP1 genome length and its mature size are presented. The SPP1 DNA digested with EcoRI is denoted. The pac sites are indicated by upward arrows. The relevant SPP1 early promoters are shown (PE2 and PE3); the ends of the gray arrows point to the transcription start site and the direction of transcription. The location of relevant replication genes and the mapping of mutations that render SPP1 impaired in DNA synthesis are indicated. B, the SPP1 DNA sequence of oriL and oriR (top strand) is presented in a 5Ј to 3Ј polarity (thick straight line), and the relevant nucleotides are denoted. The directly repeated boxes ab and AB (framed) and the AT-rich region (framed with broken lines) are indicated. The distance between the boxes AB and ab is indicated. Points were introduced into the sequence to maximize homology. gene 35) in wt-infected cells, at the restrictive temperature, revealed that only a small percentage of the phage DNA can be recovered in a fast sedimenting form (concatemeric DNA). In both cases SPP1 particles of less than unit length (30 -35 kb in size) accumulated (5). Consistent with this DNA arrest phenotype, and considering the unidirectional movement of the SPP1 replication fork, we can assume, that any event initiated at oriL stops 30 -35 kb away, in G35P and G34.1P mutants at non-permissive temperature, suggesting that G35P and G34.1P are involved in fork reactivation and the generation of concatemeric DNA. Furthermore, this distance is consistent with the hypothesis that G38P bound at oriR might work as a roadblock that would collapse any replication fork started at oriL, being the generated replication intermediates of ϳ32 kb (22).
To understand how concatemeric DNA initiates, we began the characterization of G35P. We found that G35P binds ssDNA and forms filaments. G35P preferentially catalyzes, in an ATP-independent manner, SSA between a circular ssDNA and a homologous 3Ј tail of linear dsDNA. G35P preferentially catalyzes SSA between a 3Ј-ssDNA and the homologous ATrich region of oriL on a supercoiled molecule and specifically interacts with the G40P DNA helicase and G36P SSB proteins. The results presented provide the first evidence that the SPP1 replication/recombination protein, G35P, might direct the assembly of the hexameric replicative helicase G40P at a D-loop structure by a new primosome assembly mechanism that does not require the primosome assembly proteins of B. subtilis.
Enzymes and Reagents-Isopropyl-1-thio-␤-D-galactopyranoside and rifampicin were from Calbiochem. DNA restriction and modification enzymes were purchased from MBI or Roche Molecular Biochemicals.
DNA Manipulations and Substrates-The 6065-bp HindII-MscIcleaved M13mp18 duplex DNA fragment was gel-purified. Linear 6065-bp M13mp18 DNA molecules having an ssDNA 3Ј termini (3Јtailed dsDNA) or an ssDNA 5Ј terminus (5Ј-tailed dsDNA) were prepared by incubation of DNA with T7 gene 6 exonuclease or with E. coli ExoIII, respectively. About 50% of both substrates had an ssDNA tail with a length of ϳ140 nt, measured as the percent of substrate resistant to PvuI or BglI digestion, which are located 140 or 170 bp, respectively, from the restriction site dsDNA end. Viral M13mp18 DNA was first annealed with a 20-nt oligonucleotide containing the EcoRI restriction site and then linearized with EcoRI. The DNA ends were blunted in the presence of [␣-32 P]dATP, and the labeled viral M13mp18 linear ssDNA was purified. The 194-nt ␥-32 Plabeled EcoRI-FspI ssDNA was prepared by annealing 5Ј-end-labeled oligonucleotide (5Ј-GCAACTGTTGGGAAGGGCG-3Ј) to Rf M13mp18 and extending it by PCR up to the EcoRI site, and the product was gel-purified.
Protein Manipulations-G38P, G39P, and G40P were purified as previously described (6,12). G36P purification will be described else-where. SPP1 G35P was purified to apparent homogeneity from an E. coli BL21(DE3) strain lacking the Rac-defective prophage. Soluble G35P was precipitated by PEI. G35P was recovered from the pellet by addition of 50 ml of buffer A (50 mM Tris-HCl, pH 7.5, 5% glycerol) containing 500 mM AS. The proteins of the supernatant, free of cell debris, DNA, and E. coli RecA protein, were precipitated twice by addition of solid AS to a final concentration of 60% saturation. Dialyzed G35P was purified in three different chromatographic steps (phosphocellulose, DEAE, and Q-Sepharose). The sequence of the first 15 aminoterminal residues of the purified protein was determined. The aminoterminal sequence of the G35P polypeptide was identical to the sequence predicted from the gene 35 (10), except that the initiator methionine was not present.
The molar extinction coefficient for G35P was calculated to be 22390 M Ϫ1 cm Ϫ1 , as described previously (11). The G35P concentration was determined by using the above molar extinction coefficient and is expressed as moles of protein dimers.
Molecular Mass Determination-The native molecular mass of G35P was determined by gel filtration fast protein liquid chromatography using a Superose 12 HR 10/30 column. Chromatography was carried out in buffer A (50 mM Tris-HCl, pH 7.5; 5% glycerol) containing 100 mM NaCl and the presence or the absence of 5 mM MgCl 2 at 4°C with a flow rate of 0.5 ml/min, and the A 280 was measured. G35P (7 or 30 M) was applied onto the column. A standard curve of K av versus log 10 of relative mobility was determined as recommended by Amersham Biosciences. Protein standards were: chymotrypsinogen A, 25 kDa; BSA, 68 kDa; aldolase, 158 kDa; and catalase, 232 kDa.
The 194-nt ␥-32 P-EcoRI-FspI ssDNA or 194-bp ␣-32 P-EcoRI-FspI dsDNA fragments (1 M) were used to analyze the binding of G35P to DNA through EMSA in buffer B in the presence or the absence of MgCl 2 . The reaction was stopped and separated in a 6% ndPAGE under high ionic strength conditions as previously described (7). Gels were run for 3 h at 150 V at 4°C and dried prior to autoradiography.
Protein Affinity Chromatography-G35P or BSA proteins (6 M) were covalently cross-linked to the Affi-Gel-10 (1 ml) resin as recommended by the manufacturer (Bio-Rad). G36P, G38P, G39P, or G40P (1 M) was loaded onto the column that had been equilibrated with binding buffer C (25 mM Tris-HCl, pH 7.5, 5 mM ␤ME, 0.5 mM MgCl 2 , 5% glycerol, 0.1 mM ATP, 0.1% CHAPS) containing 50 mM NaCl. Columns were washed with 20 column volumes of buffer C containing 50 mM NaCl. Bound fractions were eluted with 5 volumes of buffer C containing 1 M NaCl. Fractions of 100 l were collected and analyzed by SDS-PAGE.
Joint Molecule Formation Assay-The assay measures the formation of a stable complex between linear 6065-bp M13mp18 dsDNA (30 M) bearing an ssDNA termini (3Ј-tailed or 5Ј-tailed dsDNA) or blunt-ended dsDNA and circular M13mp18 ssDNA (30 M). Both substrates were incubated with increasing concentrations of G35P (48 nM to 1.56 M) in buffer D (25 mM Tris-HCl, pH 7.5, 5 mM ␤ME, 5 mM MgCl 2 , 0.05 mg/ml BSA, 5% glycerol) containing 25 or 100 mM NaCl, during 10 min at 30°C. The reactions were stopped by adding EDTA, SDS, and proteinase K, and the samples were fractionated by AGE (0.8%) with EtBr (11). The gel was photographed under UV irradiation. Joint molecules were quantified by densitometry of photographic negatives.
A labeled oligonucleotide 1 to 4 (120 nM, in nt) and supercoiled homologous pCB163 (8.4 M, in bp) were incubated with 62 nM G35P in buffer D containing 100 mM NaCl at 30°C for 15 min. The reactions were stopped by adding EDTA, SDS, and proteinase K, and samples were separated by AGE. The gels were dried and analyzed by autoradiography. The signal was quantified using a PhosphorImager (Amersham Biosciences).
Electron Microscopy-Reactions (10 l) containing 3Ј-tailed dsDNA (30 M) and circular M13mp18 ssDNA (30 M) were incubated with G35P (780 nM) for 10 min at 30°C under standard strand exchange conditions in 50 mM NaCl. For the analysis of the joint molecules, upon deproteinization, cytochrome c spreading in the presence of 50% formamide and carbonate buffer on a water hypophase was used as described by Spiess and Lurz (27). The protein⅐DNA complexes, which were separated from unbound protein by Sepharose CL-4B gel filtration, were fixed with glutaraldehyde and adsorption to mica and prepared for EM as described previously (27). Single particle image analysis was performed on images of unfixed, negatively stained samples as previously described (28), using a reference-free algorithm to generate averages (29).
Isolation of G40P Loaded on Joint Molecules-Joint molecules were prepared with supercoiled plasmid pCB163, oligonucleotides 1-4, and G35P, in the presence of 1 mM ATP, as described before. After incubation during 15 min at 30°C, G40P (80 nM) was added, and the reaction was continued for another 10 min at 30°C. Aliquots of the reaction were taken and deproteinized to analyze and quantify the extent of the D-loops formed, and isolation of G40P bound to the joint molecules from unbound G40P was performed by gel filtration chromatography on a Sepharose CL-6B as previously described (11). G40P was detected in the different fractions by Western blot.

RESULTS
Characterization of G35P-Soluble G35P, which was purified to 99% homogeneity as assayed by SDS-PAGE and quantitative analysis of the amino-terminal amino-acid sequence, was assayed for its ability to act as dsDNA or ssDNA nuclease (exo-or endonuclease), to hydrolyze ATP in the presence or absence of ssDNA or dsDNA, to unwind DNA, to bind ssDNA or dsDNA, and to interact with SPP1-encoded replication proteins. From these activities tested we observed that purified G35P protein is able to bind DNA and to interact with at least G40P and G36P.
G35P consists of 287 amino acid residues corresponding to a molecular mass of 32,000 Da (10). The native molecular mass of purified G35P was estimated by size fractionation through a Superose 12 fast protein liquid chromatography gel filtration column in buffer A containing 100 mM NaCl. In the presence or in the absence of 5 mM MgCl 2 , at low G35P concentrations (7 M), G35P elutes in two peaks of similar area, one narrow peak corresponding to M r 65,000 and one broad peak corresponding to M r 250,000 -350,000. At high G35P concentrations (30 M), G35P elutes mainly in a broad peak, with M r 250,000 -350,000, in the presence or in the absence of MgCl 2 . If we assume that G35P is spherical in shape, it is likely that, under the G35P concentrations assayed, G35P is in an equilibrium between a dimer and a higher order oligomer in solution. We cannot rule out, however, that G35P is an elongated monomer with a large Stokes radius.
G35P Interacts Preferentially with ssDNA-The affinity of G35P for linear 7,250-nt M13mp18 ssDNA (1 M) was determined by filter binding assays by following the rate of complex formation as a function of G35P concentration ( Fig. 2A). G35P⅐ssDNA complex formation was not linear at low protein concentrations and appeared to be sigmoidal at high G35P concentrations. The K app of the G35P⅐ssDNA complex, which in this case is equal to protein concentration midpoint, was estimated to be ϳ 14 nM and ϳ 26 nM at pH 7.5, 30°C in buffer B in the absence and the presence of 5 mM MgCl 2 , respectively. It is likely, therefore, that, both in the absence or presence of 5 mM MgCl 2 , G35P interacts with ssDNA with a weak cooperativity. To analyze the type of complexes formed by G35P with ssDNA, EMSA was used. G35P does not bind to segments as short as the 26-nt ssDNA fragment of different sequence, indicating that there is a minimum length required for stable binding, but the formation of G35P⅐194-nt ssDNA complexes was observed by EMSA (Fig. 2B). At molar ratio of 1 G35P per 80-nt or 40-nt (1:80 or 1:40), G35P forms multiple and diffuse complexes with the 194-nt 32 P⅐ssDNA, whereas at higher molar ratios (1:20 and 1:10) one discrete complex was observed (Fig.  2B, lanes 2, 3, and 10). The K app value of the G35P⅐ssDNA complex was estimated by EMSA to be ϳ18 and ϳ36 nM at pH 7.5, 30°C in buffer B in the absence and the presence of 5 mM MgCl 2 , respectively.
The affinity of G35P for linear 7250-bp M13mp18 dsDNA (1 M) was determined by filter binding assays also in the absence or the presence of Mg 2ϩ (5 mM MgCl 2 ) ( Fig. 2A). The G35P⅐dsDNA complex retained by the filter was less than 65%. in the presence or absence of 5 mM MgCl 2 for 10 min at 30°C and then loaded onto a 6% ndPAGE. In lanes 1 and 9, no protein was added. In B, increasing amounts of G35P were added in lanes 8 to 2, and 16 to 10 (1.5 to 100 nM). In C, the amount of G35P added doubles from 4.5 to 300 nM in lanes 8 to 2, and 16 to 10. CI and CI-IV, denote the protein⅐DNA complexes, and FD denotes free DNA.
In the plateau region, where all the blunted linear dsDNA is presumably saturated with G35P, only 55-62% retention are observed, even at a G35P concentration greater than 1.5 M. Similar results were observed at 25 and 50 to mM NaCl; hence, we can assume that the protein⅐dsDNA complexes show a poor stability. Under these experimental conditions, we observed that, in the absence of Mg 2ϩ , at pH 7.5 and 30°C, G35P binds to linear 32 P⅐M13mp18 dsDNA with a K app of ϳ90 nM and 300 nM at pH 7.5, 30°C in buffer B in the absence and the presence of 5 mM MgCl 2 , respectively ( Fig. 2A). Both, in the absence or the presence of 5 mM MgCl 2 , at low G35P molar ratios (1:27) one discrete G35P⅐194-bp 32 P⅐dsDNA complex was formed (CI complex), and at larger ratios (1:13 to 1:3.3) slow moving protein⅐DNA complexes were observed (CII to CIV) (Fig. 2C, compare lanes 5 with 2 and 13 with 10). Under these conditions, the K app value of G35P⅐dsDNA complex formation was ϳ100 and ϳ300 nM in the absence or the presence of 5 mM MgCl 2 , respectively.
Hence, in the absence of Mg 2ϩ , G35P binds with 5-fold higher affinity to ssDNA over dsDNA, and this preference is even larger in the presence of 5 mM MgCl 2 . In the latter case, the protein binds with 8-fold higher affinity to ssDNA than to dsDNA.
G35P Promotes the Formation of Joint Molecules-G35P shares a 40% overall identity with the RecT protein (data not shown). RecT has been shown to promote the formation of joint molecules between a circular ssDNA and a homologous linear dsDNA having an ssDNA tail (30). To address whether G35P also catalyzes joint molecule formation, linear blunt-ended 6065-bp M13mp18 dsDNA was incubated with circular M13mp18 ssDNA and G35P in buffer D containing either 25 or 100 mM NaCl during 10 min at 30°C. The samples were deproteinized, and the DNA forms separated by AGE. As shown in Fig. 3, lanes 2 and 8, G35P fails to form any new product that migrates slower than the linear dsDNA, when the dsDNA is blunt-ended. However, when linear 6065-bp M13mp18 dsDNA, with ϳ50% of the molecules containing an ssDNA tail with an average of ϳ140-nt either at the 5Ј or the 3Ј, was incubated with circular M13mp18 ssDNA and G35P, the accumulation of a discrete new band that migrates more slowly than linear dsDNA was observed (Fig. 3). End products of the strand exchange reaction (relaxed dsDNA and linear ssDNA) as detected in the presence of B. subtilis RecA protein did not accumulate (data not shown).
In the presence of low NaCl (25 mM) annealing of the circular ssDNA and the linear dsDNA containing a 3Ј-tail accounted to ϳ37% of the total dsDNA, whereas ϳ17% of annealed molecules were observed when the 5Ј-tail dsDNA template was used with a G35P concentration of 780 nM (1 dimer per 35 nt of ssDNA) (Fig. 3, lanes 6 and 4). No further increase in the yield of the products was obtained at higher protein concentrations (data not shown). Because the length of the ssDNA tail is not uniform in both substrates, due to exonuclease treatment, we cannot rule out that the different yield observed with the 3Ј-tail over 5Ј-tail substrates is a consequence of any difference in length in both substrates. To address this question we have compared joint molecule formation with the same substrates, at low or high NaCl concentrations. At 100 mM NaCl concentration, the circular ssDNA annealed to the 3Ј-tailed dsDNA by ϳ21% and to the 5Ј-tailed dsDNA by ϳ7% (Fig. 3, lanes 12 and  10). Hence, at low NaCl, G35P shows a preference for substrates having 3Ј-tails over 5Ј-tails of ϳ2-fold, and this difference is enlarged in the presence of 100 mM NaCl. It is likely, therefore, that the protein might bind better to linear DNA having a 3Ј-tail than to linear DNA having a 5Ј-tail and that G35P shows some polarity in the reaction. Identical results are obtained when G35P is first incubated with linear dsDNA and the pre-formed complex is then incubated with circular ssDNA or when the first incubation is with ssDNA. The formation of this new product, which was dependent on incubation with G35P, was not observed when proteinase K was added at the same time that G35P or MgCl 2 was omitted or EDTA was added. Titration with MgCl 2 shows an optimum for joint molecule formation at ϳ5 mM (data not shown). The formation of this new product, which could be reversed by heating at 100°C, does not require the addition of a nucleotide cofactor, and its presence does not alter the yield of the reaction. It is likely, therefore, that G35P is free of contaminating enzymes capable of generating an ssDNA tail on the linear dsDNA substrate (e.g. exonucleases). Because the formation of SSA between tailed-dsDNA and homologous ssDNA is independent of the presence of a nucleotide cofactor, we can rule out that a DNA helicase could separate both DNA strands: the reaction lacks a nucleotide cofactor that is essential for all DNA helicases.
Visualization of the Joint Molecules Formed by G35P-To analyze whether the annealed products observed by AGE are the product of the annealing reaction of the complementary ssDNA tails present in the linear substrate with the circular ssDNA or if some exchange has taken place, G35P, at a concentration of 1 G35P molecule per 35 nt of ssDNA, was incubated with circular M13mp18 ssDNA and linear 6065-bp M13mp18 dsDNA having a 3Ј ssDNA end, the reaction was deproteinized, and the products were examined by EM. Under these conditions, about 20 -25% of total linear 3Ј-tailed M13mp18 dsDNA molecules were joined to homologous circular M13mp18 ssDNA (Fig. 4A). At short times of incubation, the complex formed by the annealing of the single-stranded region of the linear duplex and the homologous circular ssDNA molecule, leading to a sigma-shaped structure, was observed (data not shown). After 10 min of incubation, one of the DNA strands from the linear duplex DNA was displaced by the circular ssDNA (alpha-shaped structures). As revealed in Fig.  4, the short displaced strand is of ssDNA nature and a stretch of dsDNA of roughly comparable length has been formed on the circular ssDNA molecule. Neither sigma-nor alpha-shaped structures were observed when G35P was omitted (data not shown). The base pairing of the linear 3Ј-tailed DNA with the circular ssDNA did not exceed 2000 nt (Fig. 4B, data not shown), and the end products of the strand exchange reaction (relaxed dsDNA and linear ssDNA) were not observed.
G35P Forms Ring Structures on Short ssDNA Tails and Nucleoprotein Filaments with ssDNA-To visualize the type of complexes formed by G35P with the substrates used for the strand exchange reaction, the protein, at a concentration of 1 G35P molecule per 35 nt of ssDNA, was incubated with circular M13mp18 ssDNA and linear 6065-bp M13mp18 dsDNA having a 3Ј ssDNA end, and the products were directly examined by EM without deproteinization. G35P bound to ssDNA changed the compact bush-like structures typical of protein-free ssDNA to relatively open, circular structures resembling pearls of a necklace (see Fig. 5A). G35P⅐circular M13mp18 ssDNA complexed with 3Ј-tailed M13mp18 dsDNA (joint molecules) was also observed (Fig. 5B). In accordance with its preferential binding to ssDNA, G35P⅐dsDNA complexes were not observed at the protein concentration used, but the ssDNA ends of the 3Ј-tailed dsDNA were covered by doughnut-shaped oligomers of G35P (Fig. 5B, denoted by arrowheads). From the volume occupied by the G35P ring structure, it could be predicted that it is an oligomer formed by seven to eight protomers. This is consistent with the observation that G35P has a mass of 250,000 -350,000 Da in solution (see above).
Inspection of the electron micrographs of negatively stained G35P⅐ssDNA complexes revealed a striated pattern, which is suggestive of helical filaments, and these helices were determined to be left-handed by rotary shadowing (data not shown). To analyze these structures, 8867 short segments were selected from these filament images. Analysis of these segments showed that there was a one-start helix with a variable pitch, a range greater than 85-105, and a mean of ϳ95 (Fig. 5C). The filaments have a clear polarity, but the periodicities arising from subunits along this continuous helix must be extremely weak, because they were not observed.
The observation that the filaments were formed only in the presence of ssDNA suggests that these filaments contain ssDNA. Unfortunately, no information is available about the location of the ssDNA within this protein filament.

G35P Promotes SSA at the Unwound AT-rich Region on a Supercoiled DNA Substrate and Loads G40P at This Region-
Previously, it has been reported that (i) the DNA helix at the tandemly repeated AT-rich region present at replication origins is thermodynamically instable and (ii) supercoiling of DNA, in the absence of the initiator proteins, induces localized unwinding (DUE) at the same sequence opened by the initiator protein (31). The ability of G35P to catalyze the assimilation of ssDNA, into the homologous supercoiled plasmid-borne SPP1 replication origin oriL (SPP1-DUE) or into a region with lower prediction of unwinding by supercoiling, was assayed. Four different oligonucleotides (1 to 4) have been synthesized (Fig.  6A). The 32 P-labeled 230-nt-long oligonucleotides 1 and 4 have a region of homology of 40 nt, at either the 5Ј-or 3Ј-ends, respectively, with the AT-rich region of oriL; whereas the 32 Plabeled 215-nt-long oligonucleotides 2 and 3 show only a short region of homology (10 nt) with the AT-rich region of oriL at the 5Ј-or 3Ј-ends, respectively (Fig. 6A). G35P has a similar affinity for the four ssDNA segments as measured by EMSA (data not shown).
G35P, at a concentration of 1 G35P per 135 bp of supercoiled dsDNA, was incubated with the labeled oligonucleotides and then with the homologous supercoiled plasmid-borne SPP1-DUE. The samples were deproteinized and analyzed by AGE. G35P preferentially promoted the formation of a stable complex between oligonucleotide 4 and the plasmid DNA that co-migrated with supercoiled plasmid DNA on an agarose gel (Fig. 6B). The amount of SSA in three independent experiments was quantified and found to be 13%, 3%, 6%, and 24% for oligonucleotides 1 through 4, respectively. The oligonucleotides 4 and 3, which have a region homologous to the AT-rich region of oriL at their 3Ј-end, base-paired to one strand of the supercoiled plasmid molecule with 2-fold higher efficiency than that of their respective counterparts. The reaction required an Mg 2ϩ cation but did not require a nucleotide cofactor. Similarly, to join the molecule formation, the optimum Mg 2ϩ concentration was 5 mM (data not shown) and the protein showed some preference for pairing the oligonucleotide that enters the SPP1-DUE region in a 3Ј orientation (Fig. 6B, lanes 4 and 16). The percentage of D-loops formed was independent of the order of incubation of the DNA substrates with G35P. The migration distance of the reaction product on the agarose gel was identical to that of the D-loop product formed by B. subtilis RecA. SSA was not observed when the plasmid molecule was relaxed with DNase I prior incubation with G35P (data not shown).
When an ssDNA segment complementary to another region of the supercoiled plasmid DNA was used, 3-5% joint molecule formations were observed, similar to the yield with ssDNA segments 2 and 3, which only have 10-nt annealing to the AT-rich region (data not shown). It is therefore likely that G35P promotes SSA between a pre-existing unwound region (e.g. AT-rich region of oriL, DUE region) and that the region of homology needs to have a minimum length (see Ref. 20 and this work).
We have previously shown that (i) the loading of G40P at the oriL region is dependent on the replisome organizer, G38P, and the helicase loader G39P (11)  to ssDNA regions with no potential secondary structures and to ssDNA as short as 10 -12 nt in length (13). 3 In both E. coli and B. subtilis, the loading of the replicative helicase onto collapsed replication forks is a PriA-dependent step (8,17). Because SPP1 replication is independent of PriA, DnaB, DnaD, and DnaI primosomal proteins (Ref. 4 and this work), it can be assumed that D-loops catalyzed by G35P might have some role on the loading of G40P into ssDNA. To address whether G35P creates an ssDNA region of sufficient length to which the replicative helicase G40P can be loaded and, therefore, the replisome established at this region, joint molecules promoted by G35P between supercoiled plasmid pCB163 and oligonucleotides 1-4, in the presence of 1 mM ATP, were prepared. Then G40P (80 nM) was added, and the reaction was incubated for another 10 min at 30°C. The free G40P was separated from G40P bound to the joint molecules by gel filtration chromatography, and the amount of G40P loaded on the different D-loops formed was quantified. Accordingly to the yield of D-loops formed with the different ssDNA segments, different yields in the loading of G40P could be observed. When D-loops were formed with oligonucleotides 2 and 3 and supercoiled plasmid pCB163, no G40P could be detected in the DNA-containing fractions, whereas when oligonucleotides 1 and 4 were used about 6 and 15% of total input G40P, respectively, was present and associated with the joint molecules (data not shown). It is likely, therefore, that in both cases the region of ssDNA created by the annealing of the invading oligonucleotide to the complementary strand promoted by G35P has a length that is sufficient to load G40P at the D-loop region. Alternatively, the loading of G40P is stimulated by a direct protein-protein interaction, and G35P is sufficient for loading G40P at the D-loop, as suggested by the high efficiency in the loading of G40P at joint molecules (compare the percentage of input G40P present in the D-loops with the percentage of D-loops formed with oligonucleotides 1 and 4).
G35P Interacts with G40P and G36P-G35P preferentially catalyzes strand invasion on a pre-existing unwound region; i.e. as the AT-rich region at the origins of replication of SPP1 (see above). To address whether G35P could recruit any replication protein at this site, we analyzed whether G35P physically interacts with four (G36P, G38P, G39P, and G40P) of the six SPP1-encoded replication products (except for the G34.1P and G36.1P endonucleases). We immobilized the proteins G35P or BSA (6 M), as a control, on an Affi-Gel 10 matrix. G36P, G38P, G39P, and G40P were then loaded (1 M) separately onto the matrix. As shown in Table I, the DNA helicase protein, G40P, and the SSB, G36P, were retained by G35P bound to the column, whereas the replisome organizer, G38P, the helicase loader, G39P, or BSA were not retained by the same column. None of the proteins bind to the BSA column. It is likely, therefore, that G35P physically interacts with both G36P and G40P.

DISCUSSION
G35P Is a Bona Fide ATP-independent SSA Protein-The characterization of G35P revealed a significant biochemical similarity with evolutionarily distinct families of ATP-inde-pendent SSA proteins that form rings, filaments, or both. Many of these families appear to be primarily of bacteriophage origin, namely the RecT/Red␤, Erf, and Rad52 families (see Ref. 32). The RecT/Red␤ family comprises the Rac-deficient prophage RecT, phage ␤ protein (-␤), and SPP1 G35P proteins, the Erf family comprises the phage P22 Erf protein (P22-Erf), and the Rad52 family comprises the Rad52 eukaryotic protein and putative proteins of phage origin (see Ref. 32). A biochemical similarity with ATP-independent SSA complex proteins of eukaryotic origin, namely the hXRCC3⅐hRad51C/hRad51L2 and hXRCC2⅐hRad51D/hRad51L3 (33)(34)(35), was also observed. One common feature in these families of proteins is their ability to form rings composed by a divergent number of subunits with different requirements. Although ring formation seems to be independent of Mg 2ϩ for the eukaryotic proteins hRad52 and hXRCC2⅐hRad51D (33,35,36) and G35P (this work), Mg 2ϩ seems to be essential for ring formation for RecT and -␤ (37,38). In the case of G35P, ring formation seems to be concentrationdependent, whereas this possibility has not been analyzed for the eukaryotic ATP-independent strand-annealing proteins. Furthermore, the broad peak obtained with G35P suggests a heterogeneity in the number of subunits that compose the ring. This is consistent with the polymorphism observed with the -␤ protein when examined by electron microscopy (38) and suggests a dynamic behavior in all this family of proteins.
-␤ protein seems to be the only protein of this family that does not form filamentous structures on ssDNA (38). The filamentous structures with ssDNA observed under the electron microscope for hRad52, hXRCC2⅐hRad51D complex, and the hXRCCc3⅐hRad51C complex do not seem to have a helical structure and can be composed of stacked rings packed in an edge to edge manner (33)(34)(35)(36). However, the filaments observed for G35P with ssDNA are helical and have a clear polarity (Fig.  5C). Furthermore, -␤ protein seems to be the only protein of this family to filament on dsDNA, and filament formation required the dsDNA substrate to have an ssDNA tail to start filamentation, where the protein first assembles as a ring structure (38). Similarly to -␤, G35P, binds to dsDNA with an ssDNA tail forming a ring structure with the tails (Fig. 5B), but accordingly to its preferential binding to ssDNA, no protein was bound to the dsDNA region of the DNA substrate. The ssDNA tails on the linear dsDNA used in Fig. 5B were obtained by exonuclease treatment and are no longer then ϳ140 nt; therefore, it is likely that, depending on the length of the ssDNA, G35P forms rings or filaments with a helical structure with ssDNA.
G35P and RecT catalyze joint molecule formation between a linear dsDNA having an ssDNA tail and its homologous circular ssDNA, where some strand displacement is observed. RecT preferentially catalyzes this reaction at a low concentration of Mg 2ϩ (Ͻ0.35 mM MgCl 2 ) (30,39), with similar efficiency with substrates having a 3Ј-or a 5Ј-tail (39); G35P, however, performs limited strand exchange at high Mg 2ϩ concentrations (5 mM) with 2-to 3-fold higher efficiency with substrates having a 5Ј-tail. G35P (Fig. 6), RecT (39), hRad52 (33), hXRCC2⅐hRad51D complex (34), and the hXRCC3⅐hRad51C complex (35) have been shown to catalyze DNA strand invasion between an oligonucleotide and a homologous supercoiled plasmid DNA, but the requirements for this reaction seem to be different. Although RecT does not require Mg 2ϩ for D-loop formation, it is required for D-loops formed with G35P and the hXRCC2⅐ hRad51D and hXRCC3-hRad51C complexes, but the optimum Mg 2ϩ concentrations are in each case different (34,35). The differences in the Mg 2ϩ requirements correlate with their preferential binding for dsDNA or ssDNA and suggest that,

G35P Promotes DNA Pairing and Strand Transfer
although the products of the reaction are the same, mechanistically this group of proteins may act in a different way. Many proteins of these family such as RecT (39) and hRad52 (33) bind preferentially to dsDNA at no or low Mg 2ϩ concentrations, and this is the optimum Mg 2ϩ concentration required for D-loop formation. However, incubation of the protein first with dsDNA inhibits the joint molecule reaction and suggests that these proteins need to contact dsDNA but in a prefixed order. G35P, however, performs D-loop formation and limited strand exchange at high Mg 2ϩ concentrations, where the protein does not bind to dsDNA (Fig. 2). Although all these proteins have been considered so far as recombination proteins, the functional similarities between the SPP1 replication protein, G35P, and phage-encoded -␤, P22erf, and RecT suggest that they could work in the processing of inactivated replication forks and the generation of concatemeric linear substrates that are encapsidated into empty procapsids. Furthermore, the functional similarity between the phage-encoded homologous-pairing proteins and hRad52 and the hXrcc2⅐hRad51D and hXrcc3⅐hRad51C complexes, despite lacking obvious sequence homology, suggest that the human proteins may also act in the assembly of replication forks after stalling. In fact, yeast Rad52 has been shown to play a major role in BIR, a process that has been shown to be rad51-independent (40). In the case of the phage recombination systems, a limited exonuclease activity associated with the SSA protein may play an active role in the annealing process (41), although the presence of an exonuclease in not obvious in the human recombination proteins.

G35P Might Direct the Assembly of a New Replication Fork:
A Model-Genetic evidence suggests that SPP1 mutants impaired in the putative 5Ј to 3Ј exonuclease G34.1P or homologous-pairing G35P accumulate less than unit-length DNA molecules (30 -35 kb in length) under non-permissive conditions (5). Why would the failure to use the recombination proteins G34.1P or G35P alone lead to a DNA arrest phenotype? Previously, it was shown that such a defect cannot be overcome by any of the host recombination nor by the DnaA-and PriA-DnaB-DnaD-DnaI-dependent assembly pathways (see the introduction). Here we show that G35P catalyzes SSA and promotes D-loop formation, which are features associated with bona fide homologous-pairing proteins. Because SPP1 replication is independent of host-encoded primosomal components, and G35P physically interacts with G40P and G36P, we assume that G35P may be involved in recombination-dependent replication and it might help in the re-establishment of a stalled replication fork at oriR or at any other stalled region.
We hypothesize that, after the initial phase of initiation of theta type SPP1 replication at oriL (6,7), the progression of the replication fork might be stalled when the replication fork encounters G38P bound at oriR (roadblock) in the absence of an overt DNA damage (see Fig. 7) or at any region in the presence of a DNA damage. The stalled replication fork breaks, and the broken fork is rescued by a process dependent on phage-encoded G35P and G34.1P functions. After Skalka (23,24), Formosa and Alberts (42), Viret et al. (18), and Kuzminov (15), we propose that G34.1P exonuclease may degrade the 5Ј-end of the linear dsDNA of a collapsed replication fork at oriR or at any FIG. 7. Roadblock as a model for the shift from theta to sigma replication. A, G38P bound to oriL or oriR blocks replication fork progression. After stalling, a nick in the leading strand (bottom part) will be processed by the putative 5Ј33Ј exonuclease, G34.1P, to generate a 3Ј-ssDNA tail on which G35P will polymerize, whereas a nick in the lagging strand (top part), will not require further processing, and G35P will polymerize on it. B, model for SPP1 initiation of sigma type DNA replication. G38P recognizes AB boxes of the SPP1 replication origin (oriL or oriR), and binds to them in an ATP-independent fashion, opening the adjacent AT-rich region, where G36P binds with high affinity. A G35P⅐ssDNA filament pairs with the leading strand of the unwound region. By direct G35P⅐G36P interactions, a remodeling of both proteins can take place, so that the G39P⅐G40P⅐ATP complex can be loaded in the unwound region by the ATP-dependent ssDNA binding capacity of G40P, as well as the interactions between G35P and G40P, and G39P and G38P, respectively. G38P⅐G39P, which forms a heterodimer, dissociates G39P from the G39P⅐G40P⅐ATP complex, and releases G38P from the origin. G40P helps the assembly of DnaG and perhaps DNA PolIII at the AT-rich region. The 3Ј-OH end of the paired strand could be used to prime the leading strand and DnaG could provide the primer for lagging strand synthesis. The arrow indicates the direction of helicase movement. other stalled region generating 3Ј overhangs to which G35P binds (Fig. 7A). G35P-mediated joint molecule formation could provide a 3Ј-end to anneal at oriR on a second supercoiled SPP1 molecule (Fig. 7B). Recently it has been shown that the RecE/ RecT and -␣/-␤ pairs physically interact, and homologous pairing was favored with respect to the exonuclease activity (41). Because RecT, -␤, and G35P belong to the same family of SSA proteins (32), it is likely that G35P physically interacts with G34.1P.
We envisage two pathways, one that is G38P-dependent and one G38P-independent, for the assembly of the hexameric replicative helicase G40P at a replication fork or at a re-established fork. In the case of the G38P-dependent pathway, G35P, by protein-protein interaction, might cause a local remodeling of G36P, bound to the G38P-promoted locally melted AT-rich region on oriR or oriL and catalyze SSA (Fig. 7B). G35P alone or in combination with G38P, again by protein-protein interaction, stimulates the loading of the G39P⅐G40P⅐ATP complex at the ssDNA region. The interaction of G38P with G39P remodels the G39P⅐G40P⅐ATP inactive complex and releases G40P⅐ATP of the G39P⅐G40P⅐ATP complex. Then, the interaction of G35P with G40P⅐ATP stabilizes the later at the open complex (Fig. 7B). In the case of the G38P-independent pathway, G35P would catalyze SSA at any pre-existing unwound homologous ssDNA region to which G36P is bound and cause a local remodeling of G36P. G35P would stimulate the loading of the G39P⅐G40P⅐ATP complex or free G40P⅐ATP at the unwound region by G35P⅐G40P⅐ATP interaction. However, a mechanism for activation/remodeling of the G40P hexameric DNA helicase has to be envisaged if G35P stimulates loading of the inactive G39P⅐G40P⅐ATP complex. Because G39P synthesis is halted and its relative amount drops after min 18 of post-infection infection, whereas G40P remains constant (43), we assume that G35P might load free G40P⅐ATP at the unwound region. Then, in both replication assembly pathways, G40P⅐ATP bound to ssDNA directs the assembly of DnaG and DNA PolIII in an ATP-independent manner (12). 3 There is no indication as to whether the 3Ј-OH end of the G35P-promoted SSA may act as a primer to initiate concatemeric DNA synthesis (sigma replication) on the invaded supercoiled template or whether DnaG is responsible also for the synthesis of the leader strand primers.
This model is consistent with the observations that: (i) SPP1 replication begins at a unique origin and proceeds unidirectionally, but two SPP1 replication origins have been mapped (3-6), (ii) G38P binds to oriL with higher affinity than to oriR, and theta thype replication initiates at oriL (7), (iii) SPP1 replication is unidirectional, and a replication fork that starts at oriL and might stop at oriR will duplicate a 32-kb DNA segment (22) (Fig. 1), (iv) a SPP1 duplex circle with a multiunit linear appendage has been observed (3), and (v) G39P synthesis halts at late times after infection, whereas G40P synthesis remains constant during the entire lytic cycle (43). It is likely, therefore, that the stalled fork at the DSB could provide the means for the switch from theta to sigma type replication, and replication restart is dependent on the G35P and G34.1P functions.
The features presented in this model could have mechanistic implications for understanding the replication of HSV virus or phage (reviewed in Refs. 44 and 45). HSV type 1 genome and its replication have several features in common with phage SPP1: (i) both have two distinct replication origins (SPP1, oriL and oriR; HSV type 1, ori S and ori L ) of similar sequence, (ii) after initiation by the theta replication mode, replication switches to the rolling circle mode, and (iii) molecules consisting of duplex circles with multiunit linear appendages have been observed by EM (3, 46, 47). Replication of bacteriophage also initiates by the theta replication mode, and at late times of infection, sigma type replication starts (45). theta replication initiates bidirectionally but for the switch from theta to sigma type of replication, unidirectional theta replication, which is driven by limiting amounts of DnaA, is required (48). In the case of phage concatemeric replication, it was suggested that the -O protein bound to the -ori might create a physical barrier that permits only one round of unidirectional theta replication both in vivo and in vitro (49,50). The broken fork could be rescued either by the concerted action of Red products (-␣ and -␤ proteins, "functional counterpart" of SPP1 G35P, and G34.1P) or by the host recombination and repair machinery (see Refs. 23,24,and 51). This is consistent with the fact that the red genes are not essential, although involved in replication (23,51).
The proposed model also indicates that generation of the sigma type of replication will not be random but will occur preferentially at origins of replication or places where preexisting unwound regions exist. This hypothesis is consistent with the fact that BIR events also are initiated non-randomly and occur adjacent to an autonomous replicating sequence (52).