The Herpes Simplex Virus Type-1 Single-strand DNA-binding Protein (ICP8) Promotes Strand Invasion*

ICP8, the herpes simplex virus type-1 single-strand DNA-binding protein, was recently shown to promote strand exchange in conjunction with the viral replicative helicase (Nimonkar, A. V., and Boehmer, P. E. (2002) J. Biol. Chem. 277, 15182–15189). Here we show that ICP8 also catalyzes strand invasion in an ATP-independent manner. Thus, ICP8 promotes the assimilation of a sin-gle-stranded donor molecule into a homologous plasmid, resulting in the formation of a displacement loop. Invasion of a homologous duplex by single-stranded DNA requires homology at either 3 (cid:1) or 5 (cid:1) end of the invading strand. The reaction is dependent on the free energy of supercoiling and alters the topology of the acceptor plasmid. Hence, strand invasion products formed by ICP8 are resistant to the action of restriction endonucleases that cleave outside of the area of pairing. The ability to catalyze strand invasion is a novel activity of ICP8 and the first demonstration of a eukaryotic viral single-strand DNA-binding protein to promote this reaction. In this regard ICP8 is functionally similar to the prototypical prokaryotic recombinase RecA and its eukaryotic homologs. This strand invasion activity of ICP8 coupled with DNA synthesis may explain the high prev-alence of branched DNA structures during viral replication. PB142, respectively. Oligonucleotides (cid:3) 32 P-labeled polynucleotide purified Sephadex G-25 (fine) columns (Roche Molecular Biochemicals). pUC18 and pUC19 were E. coli JM109 the Promega Wizard Plus DNA purification system by ethanol precipitation. P-labeled pUC18 was LB broth (5 (cid:2) Ci/ml). DNA concentra- in moles of

ICP8, the herpes simplex virus type-1 single-strand DNA-binding protein, was recently shown to promote strand exchange in conjunction with the viral replicative helicase ( Chem. 277, 15182-15189). Here we show that ICP8 also catalyzes strand invasion in an ATP-independent manner. Thus, ICP8 promotes the assimilation of a single-stranded donor molecule into a homologous plasmid, resulting in the formation of a displacement loop. Invasion of a homologous duplex by single-stranded DNA requires homology at either 3 or 5 end of the invading strand. The reaction is dependent on the free energy of supercoiling and alters the topology of the acceptor plasmid. Hence, strand invasion products formed by ICP8 are resistant to the action of restriction endonucleases that cleave outside of the area of pairing. The ability to catalyze strand invasion is a novel activity of ICP8 and the first demonstration of a eukaryotic viral single-strand DNA-binding protein to promote this reaction. In this regard ICP8 is functionally similar to the prototypical prokaryotic recombinase RecA and its eukaryotic homologs. This strand invasion activity of ICP8 coupled with DNA synthesis may explain the high prevalence of branched DNA structures during viral replication.
Herpes simplex virus type-1 (HSV-1) 1 is a double-stranded DNA virus with a genome of ϳ152 kbp (1). The HSV-1 genome undergoes a high frequency of homologous recombination in a process that is temporally linked to viral DNA replication (2). Recently, we proposed a model for HSV-1 recombination in which strand exchange is mediated by two essential DNA replication proteins and follows a single-strand annealing and helicase-mediated heteroduplex extension mechanism (3). In particular, ICP8, the viral single-strand DNA-binding protein (SSB) utilizes its helix destabilizing and reannealing activities to promote intermolecular pairing of homologous DNA. Heteroduplex DNA intermediates formed in this fashion are further processed by helicase-mediated branch migration catalyzed by the replicative DNA helicase-primase. Viral DNA replication intermediates include a high prevalence of branched structures that presumably arise due to strand invasion coupled to DNA synthesis (4). To account for this phenomenon, we examined the ability of ICP8 to promote strand invasion. Here we describe the novel finding that ICP8 promotes assimilation of single-stranded (ss) DNA into homologous supercoiled acceptor DNA, resulting in the formation of a displacement loop (D-loop).

EXPERIMENTAL PROCEDURES
Enzymes and Reagents-Escherichia coli SSB (E-SSB), RecA, and exonuclease I were purchased from U. S. Biochemical Corp. RecA concentrations are expressed in moles of monomeric protein, while those of E-SSB are expressed in moles of tetrameric protein. DNA topoisomerase I (calf thymus) and proteinase K were purchased from Amersham Biosciences and Roche Molecular Biochemicals, respectively. Bacteriophage T4 polynucleotide kinase, E. coli RecJ, and all restriction endonucleases were obtained from New England Biolabs. ICP8 was purified as described previously (5). Its concentration, expressed in moles of monomeric protein, was determined using an extinction coefficient of 82,720 M Ϫ1 cm Ϫ1 at 280 nm calculated from its predicted amino acid sequence (6). ATP (disodium salt) and chloroquine (diphosphate salt) were purchased from Sigma. [␥-32 P]ATP (4,500 Ci/mmol) and H 3 32 PO 4 were purchased from ICN Biomedicals.
Strand Invasion Assay-Unless otherwise stated, ICP8 (0.25 M) was preincubated with oligonucleotides (10.5 nM) at 37°C for 8 min in a buffer containing 25 mM Tris acetate, pH 7.5, 10 mM magnesium acetate, 1 mM dithiothreitol, 1 mM ATP, and 100 g/ml bovine serum albumin. Strand invasion was initiated by adding pUC18 form I DNA (3.5 nM), and incubation was continued for 30 min at 30°C. RecAmediated strand invasion was performed under the same conditions using 3.5 M protein. The ability of E-SSB to promote strand invasion was also examined under the same conditions using 0.25 M protein. Reactions were quenched by the addition of termination buffer (final concentration: 2% SDS, 50 mM EDTA, and 3 g/l proteinase K) followed by incubation for 10 min at 30°C. The reaction mixtures were resolved by electrophoresis through 1% agarose-Tris acetate EDTA, pH 7.6, gels at 8 V/cm for 1.5 h. The gels were dried onto DE81 chromatography paper (Whatman), analyzed, and quantitated by storage phosphor analysis with a Amersham Biosciences Storm 820. When necessary, prior to drying, gels were stained with 0.1 g/ml ethidium bromide (EtBr) for 45 min and analyzed by ultraviolet transillumination using a Bio-Rad VersaDoc 1000 imaging system.
Chloroquine-Agarose Gel Electrophoresis-Strand invasion reaction products were extracted using the Promega Wizard DNA cleanup system and electrophoresed through 1% agarose-Tris acetate EDTA, pH 7.6, gels containing 0, 2, 4, 10, 25, 100, and 200 g/ml chloroquine at 4.5 V/cm for 3.5 h (8). Chloroquine was also included in the electrophoresis and loading buffers at the respective concentrations.

RESULTS
ICP8 Catalyzes Strand Invasion-We have examined the ability of ICP8 to catalyze strand invasion in vitro. The substrates for this reaction were acceptor DNA, consisting of form I pUC18 DNA, and donor DNA, consisting of a homologous 5Ј-32 P-labeled 100-mer (PB11). Pairing between these substrates was measured with a electrophoretic mobility shift assay of deproteinized reaction products as described under "Experimental Procedures." As shown in Fig. 1A, ICP8 results in the assimilation of the 100-mer into homologous form I acceptor DNA, presumably leading to the formation of a D-loop (lane 1). The concentration of ICP8 used was 2.5-fold in excess to that required to coat the 100-mer (9). Under these conditions, up to 15% of the form I acceptor DNA participated in D-loop formation. Strand invasion does not require a high energy co-factor, since Dloops were formed in the absence of ATP (lane 2). ICP8 and homologous acceptor DNA are indispensable for D-loop formation, since products failed to form in the absence of ICP8 (lane 3) or pUC18 (lane 4). Substitution with a heterologous acceptor DNA (pUC19) also prevented D-loop formation (lane 5). The reaction is greatly stimulated by Mg 2ϩ , since its omission resulted in drastic reduction in D-loop formation (lane 6). At a concentration of E-SSB equivalent to that of ICP8 and 4-fold in excess to that required to coat the 100-mer (assuming a site size of 40 nucleotides/tetramer), only negligible amounts of D-loops were formed (lane 7). Likewise, no significant D-loops were formed with coating concentrations of human replication protein A (RP-A) or T4 gp32 protein (data not shown). The products of ICP8-mediated strand invasion exhibited a electrophoretic mobility identical to those formed by the action of RecA (lane 8). Fig. 1, B and C, are EtBr-stained and storage phosphor images of the same gel that compare the ability of three topological forms of acceptor DNA to participate in D-loop formation. D-loops were formed with form I (lane 1) but not with forms IV or III (lanes 2 and 3) DNA, indicating that strand invasion is dependent on the free energy of supercoiling.
ICP8 Promotes Complete Assimilation of a 100-mer into a Homologous Plasmid-To determine whether ICP8 can facilitate complete invasion of the donor 100-mer into the acceptor DNA, we examined the susceptibility of the resulting D-loops to digestion with ssDNA exonucleases. D-loops were resistant to digestion with exonuclease I (3Ј-5Ј exonuclease) as well as RecJ (5Ј-3Ј exonuclease) when analyzed by electrophoresis through a 1% agarose gel ( Fig. 2A). Quantitative analysis indicates that the amount of 100-mer protected from both exonuclease I and RecJ digestion is equal to the amount of 100-mer participating in D-loop formation (ϳ5%). It should be noted that exonuclease digestion products (5Ј-32 P-labeled dGMP and d(pGpT) for RecJ and exonuclease I, respectively) migrate more slowly through 1% agarose gels than the 100-mer. Fig. 2B confirms that the concentrations of exonucleases used were sufficient to completely degrade the donor 100-mer. Since the 32 P label of the invading 100-mer is at the 5Ј end, resistance to RecJ demonstrates that the 5Ј end of the 100-mer was stably assimilated into the acceptor DNA ( Fig. 2A, lane 3). To ascertain that the 3Ј end of the 100-mer was fully incorporated into the donor DNA, exonuclease I digestion products were also resolved by denaturing polyacrylamide gel electrophoresis. Fig. 2C shows that the 100-mer involved in D-loops was not shortened by treatment with exonuclease I, thereby indicating that the 3Ј end of the 100-mer was also stably integrated into the acceptor DNA (compare lanes 3 and 4). The amount of 100-mer that was resistant to degradation in Fig. 2C, lane 4 (ϳ5%), corresponds to the amount of 100-mer participating in D-loop formation (ϳ5%) (Fig. 2A, lane 1).
ICP8-mediated D-loop Formation Requires a Homologous End-D-loop formation was examined with a variety of oligonucleotide donors that were either completely homologous to the acceptor plasmid (PB11, 100-mer and PB9, 22-mer) or possessed heterology at their 3Ј (PB136, 100-mer), 5Ј (PB137, 104-mer), or both (PB142, 68-mer) ends. Fig. 3 shows that long oligonucleotides possessing heterology at either 3Ј (PB136) or 5Ј (PB137) ends were nevertheless capable of participating in D-loop formation compared with the standard 100-mer (PB11) (compare lanes 2, 4, and 6). Similar results were obtained with shorter oligonucleotides possessing heterology at either 3Ј or 5Ј ends (data not shown). However, oligonucleotide PB142 (68mer) possessing a 23-nucleotide heterology at both ends failed to participate in D-loop formation (lane 8). A control oligonucleotide (PB9) with an equivalent length of homology as PB142 (22 nucleotides) was efficiently assimilated into the acceptor plasmid, indicating that the failure of PB142 to form D-loops was not due to insufficient homology but rather due to the lack of homology at the ends (lane 10).
ICP8-promoted Strand Invasion Perturbs the Topology of the Supercoiled Acceptor Plasmid-Upon electrophoresis through a 1% agarose gel, D-loops and acceptor DNA have a similar mobility (Fig. 4A, lanes 2 and 3). To resolve differences in topology between these two species, we examined their migration through 1% agarose gels containing varying concentrations of chloroquine. Intercalation of chloroquine into negatively supercoiled DNA leads to DNA unwinding (10). This at first reduces the superhelicity (relaxation) of the plasmid, concomitantly reducing its electrophoretic mobility (compare lanes 2 of Fig. 4, A-D). However, as more chloroquine intercalates into the DNA, the relaxed plasmid is converted into positively supercoiled DNA, increasing its electrophoretic mobility (compare lanes 2 of Fig. 4, D-G). Chloroquine does not reduce the electrophoretic mobility of D-loops, indicating that they are not unwound by the intercalating agent. On the contrary, D-loops migrate faster with increasing chloroquine concentrations (compare lanes 3 of Fig. 4, A-G). This suggests that assimilation of the 100-mer alters the superhelical density by overwinding the acceptor plasmid. Electrophoresis in the presence of chloroquine also resolves an additional species. We believe that this species is form X DNA, which is presumably generated due to ICP8-mediated unwinding of the acceptor DNA involved in D-loops (Fig. 4G, lane 3). This observation is similar to that hypothesized for RecA where extensive RecA filament formation extending from the D-loop into the duplex portion of the plasmid causes unwinding and accumulation of positive supercoils in the plasmid (11,12).
Since D-loops exhibit a modified topology (Fig. 4), we predicted that they would show altered sensitivity to restriction endonucleases. D-loops were digested with the following enzymes: HindIII and EcoRI that cleave in the region of pairing at co-ordinates 399 and 450, respectively, AlwNI and AflIII that cleave outside the region of pairing at co-ordinates 1217 and 806, respectively. The EtBr-stained image serves as an internal control to show that the acceptor DNA (pUC18) was linearized by all four enzymes (Fig. 5A, lanes 1, 2, 4, and 5). Storage phosphor analysis of the same gel shows that acceptor DNA involved in D-loops was cleaved by HindIII and EcoRI but not by AlwNI or AflIII (Fig. 5B, lanes 1, 2, 4, and 5). In addition, D-loops were resistant to cleavage by BsrBI and DraI both of which have three recognition sites outside of the region of pairing at coordinates 498, 739, and 2540 and 1565, 1584, and 2276, respectively (Fig. 5, C and D). This indicates that the donor 100-mer undergoes Watson-Crick base pairing with the complementary strand of the acceptor DNA, thereby retaining the recognition sites for the enzymes encompassing the region of pairing. On the other hand, invasion of the 100-mer, resulting in a change in topology, disrupts the recognition sequences of those enzymes that cleave distally to the site of pairing, leading to resistance to cleavage. Two additional enzymes whose recognition sites are within the region of pairing, SmaI (433) and BamHI (429), cleaved D-loops, while two other enzymes whose recognition sites are outside of the region of pairing, BsrFI (1779) and AatII (2617), failed to cleave the D-loops (data not shown). It should be noted that cleavage by restriction endonucleases in the region of pairing was observed with deproteinized D-loops as well as with ongoing D-loop reactions (i.e. in the presence of ICP8), indicating that triple helical structures, which would presumably interfere with cleavage, are not the stable end products.

DISCUSSION
Using a classical in vitro assay, we have demonstrated that ICP8 promotes strand invasion. This reaction is distinct to its helix destabilizing and reannealing activities and to its ability to promote strand exchange in conjunction with the viral replicative helicase (3). To our knowledge, ICP8 is the first eukaryotic viral SSB shown to mediate this reaction. It has previously been pointed out that several thermodynamic parameters of ICP8 are closer to those of known recombinases (e.g. E. coli RecA and T4 UvsX protein) than they are to other SSBs (e.g. E-SSB, RP-A, and T4 gp32) (9). Thus, like bona fide recombinases, ICP8 stretches ssDNA, possibly to facilitate the search for homology, forms complexes with ssDNA that are stable at high salt concentrations and exhibits both weaker and lower cooperativity ssDNA binding than other SSBs. Although ICP8 promotes D-loop formation, its ability to do so is relatively weak (up to 15% product formation) when compared with the prototypical E. coli RecA recombinase (ϳ40% product formation under our reaction conditions that lack an ATP regenerating system). However, the efficiency of the reaction promoted by ICP8 is comparable with that achieved by eukaryotic RecA counterparts such as yeast Rad51, which only promotes ϳ2% product formation (13).
An important feature of ICP8-mediated strand invasion is that it does not require ATP. This distinguishes ICP8 from the RecA-type recombinases (e.g. E. coli RecA and eukaryotic Rad51) that require ATP binding to catalyze strand invasion (14,15). ICP8 presumably utilizes the energy stored in the negatively supercoiled form I DNA to drive the reaction. D-loop formation was minimal in the absence of Mg 2ϩ , whereas concentrations ranging from 5 to 50 mM Mg 2ϩ greatly stimulated the reaction. The low efficiency of D-loop formation in the absence of Mg 2ϩ can be rationalized by the fact that ICP8 is more proficient at helix destabilization in the absence of Mg 2ϩ and may therefore lead to dissociation of D-loops under these conditions (3). In addition, stimulation by Mg 2ϩ may be related to the fact that ICP8-mediated pairing is Mg 2ϩ -dependent (3).
Our data indicate that ICP8 mediates complete assimilation of a donor 100-mer into the acceptor plasmid, which, to our knowledge, has not been demonstrated with other strand invasion proteins such as RecA or Rad51. The stability of D-loops formed by ICP8 following deproteinization indicates that these structures entail plectonemic DNA interactions. The results obtained with oligonucleotides possessing heterology at 5Ј or 3Ј ends provides evidence that ICP8-mediated D-loop formation can initiate at either end. Moreover, donor DNA with heterology at both ends did not form D-loops, thereby indicating the need for a homologous end. It may be possible for a substrate with double heterology to form paranemic D-loops. However, such intermediates are unstable and would dissociate upon deproteinization. The lack of bias toward an end is in contrast to either RecA or Rad51, which preferentially lead to assimilation of 3Ј and 5Ј ends, respectively (13).
Strand invasion by ICP8 alters the topology of the acceptor plasmid as evidenced by anomalous migration in chloroquinecontaining gels and resistance to restriction endonucleases that cleave outside the area of pairing. ICP8 also generates form X DNA, which has previously been described for RecAmediated D-loop formation (11,12). Generation of form X DNA requires protein-induced unwinding of the acceptor DNA, starting at the point of assimilation of the donor DNA. This generates compensatory positive supercoils allowing such structures to be resolved from normal D-loops by agents (e.g. chloroquine) that alter the writhing number.
The exact mechanism for strand invasion remains unclear. Two competing mechanisms have been proposed (reviewed in Ref. 16). According to the first mechanism (R-form hypothesis), the donor nucleoprotein filament forms a canonical DNA triple helix involving non-Watson-Crick base pairing. The second mechanism (base-flipping model) states that the donor nucleoprotein filament induces base-flipping in one strand of the acceptor DNA, thereby permitting a homology search between the protein bound oligonucleotide and the flipped bases based on Watson-Crick interactions. Our data suggest that ICP8mediated D-loop formation occurs by the later model. This conclusion is based on the susceptibility of D-loops to restriction endonucleases that cleave in the region of pairing, whereas they would otherwise be resistant if the D-loops were to possess triple helical character (17).
Another protein that has been shown to promote strand invasion is E. coli RecT (18). Although RecT-and ICP8-mediated strand invasion occur under similar conditions and at similar protein to ssDNA ratios, the ICP8-mediated reaction is greatly stimulated by Mg 2ϩ , while the RecT-mediated reaction is Mg 2ϩ -independent and inhibited with increasing Mg 2ϩ . The major difference between these two proteins, however, is that ICP8 functions as an SSB, while the RecT-mediated reaction is dependent on its interaction with both ss and duplex DNA (1,18).
During its replicative cycle, multiple concatemeric HSV-1 genomes are generated by rolling circle replication. Highly branched networks of DNA are prevalent at later times during replication (4). These may arise from intra-and interconcatemeric recombination. The strand invasion activity of ICP8 may be pivotal in the formation of such structures. Thus, we envisage that ICP8 mediates the invasion of ssDNA into homologous regions. Invading 3Ј-terminal strands would presumably prime DNA synthesis and lead to the formation of replication forks that would ultimately result in the formation of branched DNA structures. This process may be initiated by double-strand DNA breaks that arise due to a variety of reasons (e.g. collapsed replication forks, endonuclease G cleavage at viral a sequences (19), and DNA damaging agents). ssDNA required for strand invasion may be generated by the helix-destabilizing activity of ICP8, possibly in conjunction with the viral replicative helicase or by exonucleolytic processing of broken DNA ends (3).