Identification of Subunit Binding Positions on a Model Fork and Displacements That Occur during Sequential Assembly of the Escherichia coli Primosome*

Background: Photo-cross-linking revealed the position of replication restart protein interactions with a fork. Results: PriA cross-links to the fork until DnaC, DnaB, and ATPγS are added. Conclusion: PriA remains stably associated with forks until the final stage of helicase assembly, when it is displaced. Significance: The approach employed should be applicable to complex protein-nucleic acid assembly reactions. When replication stalls and forks disassemble, the restart primosome is required to reload the replicative helicase so that chromosomal replication can be reinitiated. We have taken a photo-cross-linking approach, using model replication forks containing a phenyl diazirine placed at single locations, to determine the positions of primosomal protein binding and changes in interactions that occur during the assembly reaction. This approach revealed a novel mode for single-stranded DNA-binding protein (SSB)-DNA binding, in which SSB interacts with both the leading and lagging single-strand segments and the parental duplex of the fork. Cross-linking to a novel region within SSB is observed only when it is bound to forked structures. This binding mode is also followed by PriB. PriA binds to the fork, excluding SSB and PriB, interacting with the primer terminus, single-stranded leading and lagging strands and duplex in immediate proximity of the fork. SSB binds to flanking single-stranded segments distal to the fork in the presence of PriA. The addition of PriB or DnaT to a PriA-SSB-fork complex does not lead to cross-linking or displacement, suggesting that their association is through protein-protein interactions at early stages of the reaction. Upon addition of DnaC and the DnaB helicase in the presence of ATPγS, helicase is assembled, leading to contacts within the duplex region on the tracking (lagging) strand and strong contacts with the displaced leading single strand near the fork. PriA is displaced from DNA upon helicase assembly.

Not all replication forks established at the origin complete the two megabase journey to the terminus. Presumably, the helicase occasionally spontaneously disassembles because of less than absolute processivity. In addition, unrepaired lesions are sometimes encountered that can be bypassed by the action of error-prone Pol 2 Y family polymerases. The disposition of the helicase during polymerase switching has not been well studied. Unrepaired lesions can also be processed by error-free recombinational repair, a process that leads to DnaB helicase dissociation prior to re-establishment of a repaired replication fork (1). These processes require reloading of the DnaB helicase in a reaction independent of the DnaA protein.
At least two distinct mechanisms for restart of replication forks exist in E. coli. These pathways are led by association of PriA or PriC with the forked DNA structure and through interaction with intermediate protein factors lead to the reloading of the DnaB helicase (2,3). Most bacteria do not contain a PriC homolog and, presumably, are solely dependent upon PriA. Even in E. coli, PriA and PriC are not completely redundant. PriA mutants are very sick. They grow slowly and suffer from low viability as indicated by a low replating efficiency (2).
Gel shift assays suggest an ordered assembly of primosomal proteins (4,5). PriA binds first, followed by binding of PriB and then DnaT. This protein-DNA complex can support ATP-dependent loading of the DnaB helicase onto the fork by the DnaC helicase loader.
PriA binds broadly to D loops or forked structures with a primed leading strand preferentially (6,7). This specificity appears to be determined, in part, by a 3Ј-OH binding pocket within the PriA N-terminal domain (8). The C-terminal domain of PriA contains a 3Ј 3 5Ј helicase (9,10). The presence of a primer on the leading strand template apparently orients PriA such that the helicase activity is directed toward the lagging strand template instead of advancing into the unreplicated duplex of a replication fork (11). Its function is thought to be clearing of a lagging strand duplex if present adjacent to the fork to provide a single-stranded site for helicase loading. The C-terminal tail of SSB binds PriA (12).
On a D loop with a primed leading strand template, PriA yields a large footprint with interactions with the leading strand primer and its complementary strand and regions progressing into the duplex downstream of the primer (5,7). On templates with duplex lagging strands behind the fork, PriA can progress into that region using its helicase activity and generate a footprint on the lagging strand template (7).
Emphasizing its central role in primosome function, PriA also serves as a checkpoint protein, blocking the intrinsic strand displacement activity of the DNA polymerase III holoenzyme (Pol III HE) until DnaB is assembled (13,14). This is accomplished by competing with Pol III HE for binding to the leading strand primer terminus (15).
PriB is an SSB homolog. Structures indicate that it is a dimer with two OB folds (16,17). PriB interacts with the isolated helicase domain of PriA free in solution, but not with full-length PriA. A DNA-PriA complex binds PriB tightly, suggesting a conformational change in PriA, revealing the PriB binding site (18). In the absence of PriA, DnaT and single-stranded DNA appear to compete for binding to PriB. On the basis of this data, a handoff model has been proposed for primosome assembly (18).
DnaB can self-assemble on model replication forks with free 5Ј-lagging strand template flaps. In a study initiated to define the minimal template required for primosome assembly on model forks to support this work, we showed that the DnaB self-loading reaction can be blocked sterically by attachment of streptavidin to a biotin near the 5Ј-end of the lagging strand flap (15). This observation suggests that the DnaB self-loading reaction proceeds by threading of the opening in the DnaB hexameric ring and that anything that gives the DNA fiber a diameter greater than the DnaB 6 central pore blocks this reaction. Consistent with this hypothesis, SSB coating the lagging strand flap also blocks the self-assembly reaction (15).
A critical deficit in our understanding pertains to identification of the sites of interaction of primosomal proteins with a replication fork in the context of the full complement of primosomal proteins. This would also reveal displacements and arrangements that occur as the complex is built up. Accordingly, we have applied a photo-cross-linking approach in an initial effort to address this issue using a series of model forks with a photo-cross-linker placed at a unique position within each fork. We chose to use a phenyldiazirine, whose utility for nucleic acid applications was largely popularized by Bartholomew and co-workers (19). Irradiation at 350 nm generates a highly reactive short-lived carbene. Carbenes can insert into the carbon-hydrogen bonds of every amino acid side chain. Thus, they do not show the selectivity exhibited by other photocross-linkers, such as 5-bromo-or 5-iodo-deoxyuridine. Thus, negative data can be interpreted with certainty. If a protein is not cross-linked to a phenyldiazirine placed at a certain position within a model fork, that protein does not bind to that position. When a carbon-hydrogen bond is not adjacent to the reactive carbene, it reacts with water to form an alcohol, rapidly quenching the reactive species and preventing false positive results.

EXPERIMENTAL PROCEDURES
Oligonucleotides and Model Forks-All oligonucleotides were obtained from Biosearch Technologies. Oligonucleotides containing amino-modified C2 dT from Glen Research were derivatized with a phenyldiazirine photo-cross-linker and HPLC-purified as described (20). Substrates used in all photocross-linking reactions were assembled from HPLC-purified oligonucleotides listed in Fig. 1A. Prior to annealing, the phenyldiazirine-containing oligonucleotide was labeled with 32 P (6000 Ci/mmol) on the 5Ј-end using T4 polynucleotide kinase. Unincorporated [␥-32 P]ATP was removed by a Microspin-G25 column (GE Healthcare).
Photo-cross-linking-Photo-cross-linking reactions were carried out in a final volume of 50 l containing 20 nM radiolabeled phenyldiazirine-containing substrate in a buffer containing 50 mM Hepes (pH 7.5), 10 mM magnesium acetate, 10 mM dithiothreitol, 20% (v/v) glycerol, 0.02% (v/v) Nonidet P-40 detergent, 100 mM potassium glutamate, and 100 M ATP␥S (unless noted otherwise). Protein-DNA solutions were incubated for 10 min at room temperature in the dark before irradiating. Reactions were irradiated at room temperature in siliconized 15-ml Pyrex tubes (filters light Ͻ 300 nm) measuring 1.25 cm inner diameter (Sigma-Aldrich #CLS982016X) covered on the top in Parafilm. Irradiation took place in a Rayonet Photochemical Reactor-200 equipped with 16 lamps (Southern New England Ultraviolet Company RPR-3500) outputting 4.5 W at 350 nm for 1 h. Reactions were irradiated 10 cm from the light source and rotated at 5 rpm.
For reactions where the 5Ј-end of the lagging strand template was sterically blocked by binding of streptavidin to incorporated biotin, 200 nM streptavidin was incubated at room temperature with the substrate for 5 min prior to the addition of other protein components. Where proteolysis is indicated, 30 g of Proteinase K was added to the reaction after irradiation. Reactions were incubated at 37°C for 1 h in the dark prior to denaturation of macromolecules.
Markers-Markers that indicate the position of individual proteins photo-cross-linked to DNA were made by performing incubations at high protein concentrations so that the equilibrium was pushed sufficiently that binary complexes formed in the absence of other proteins. They were then subjected to the same photo-cross-linking protocol described above. Forked template with a photolabel at the Ϫ6 position (unless designated otherwise) (20 nM) was incubated with either 500 nM SSB 4 , 500 nM PriA, 550 nM DnaB 6 , or 500 nM PriB 2 before irradiation. The DnaC marker was generated using a procedure whereby the McMacken lab first elucidated a previously cryptic interaction of DnaC with DnaB during the DnaB loading reaction (23). 100 nM DnaB 6 and 200 nM DnaC were incubated with forked template in the presence of 100 M ATP␥S.
Gel Electrophoresis-Reactions were resolved by 4 -20% SDS-PAGE containing 4 M urea. Gels were run at 15 W for 4 h using 25 mM Tris base, 190 mM glycine, 0.1% (w/v) SDS as the electrophoresis buffer. Gels were exposed to a Storage Phosphor Screen (GE Healthcare) overnight prior to being scanned by a PhosphorImager.

RESULTS
We employed photo-cross-linking using a series of model replication forks that contained a phenyldiazirine linked to the 5-position of dU (Fig. 1). Each substrate was photolabeled at only one position. Our intention was to use these substrates to map the position of primosomal protein binding and also to detect changes in primosomal subunit arrangements upon handoffs that occur during sequential assembly. Phenyldiazirines are particularly well suited for this application because they are highly reactive and nonselective in that they can insert into carbon-hydrogen bonds in the side chains of all amino acids (19). Thus, negative data can be interpreted with a high level of confidence. If a protein does not cross-link to a photolabeled nucleotide, the protein does not bind to that position.
Detection of a Novel Interaction between SSB and the Fork Junction-SSB plays a key role in primosome assembly and interacts, minimally, with the PriA subunit (12). Thus, we examined SSB binding first, expecting to see broad binding to all single-stranded positions within model forked templates. Consistent with this expectation, we observed one prominent cross-link (designated SSB-90-mer A in the margin of Fig. 2) at the Ϫ35 position within the 90-nt lagging strand template of the fork that corresponded, in electrophoretic mobility, to a cross-link of SSB to a single-stranded oligonucleotide of the same length (Fig. 2, compare lane 1 in A with lanes 1 and 2 in B). The presence of protein in the identified band A was confirmed by treatment with Proteinase K prior to denaturation ( Fig. 2A, lane 2). At the distal end of the duplex substrate a cross-linked band of similar mobility was observed, presumably because of SSB binding to the ends of a frayed substrate ( Fig. 2A, lane 3). A cross-link with the 35-nt leading strand primer was also observed. Because the resulting band corresponded to the mobility of a control cross-link to the primer in single-stranded form, this species was also assigned as species A and may result from SSB binding to the frayed 3Ј-primer terminus within the forked substrate (Fig. 2, compare lane 11 in A with lanes 11 and 12 in B). The same cross-linked species A was observed when the photolabel was placed at position Ϫ3, ϩ3, and Ϫ6 ( Fig. 2B).
At positions close to the fork, additional cross-linked species were observed. When a photo-cross-linker was placed 3 nt internal to the duplex at the fork junction (ϩ3 position), three bands of decreased mobility relative to the radiolabeled substrate were observed. The slowest mobility band did not contain protein, because it was unaffected by protease treatment (Fig. 2A, compare lanes 7 and 8). This band was assigned as a cross-link between the 90-nt leading and lagging strand templates. Prominent DNA-DNA cross-links were not observed at the other positions or in the absence of SSB. The carbene resulting from photoactivation of a phenyldiazirine (Fig. 1C) can rapidly insert into an oxygen-hydrogen bond of water to generate the corresponding alcohol. Presumably, this reaction is dominant in the presence of the high (ϳ55 M) concentrations of water when exposed in solution. However, upon protein binding and dehydration, cross-linking to the opposing strand of DNA would be expected to occur.
Surprisingly, two distinct SSB-DNA cross-links arose from cross-linking to the ϩ3 position ( Fig. 2A, lane 7). One was the previously identified cross-link A, and the second, of slower mobility (designated B), is novel. We consider it unlikely that the slower migrating cross-link B is the result of incomplete denaturation because the samples were subjected to a 10-min boiling step in the presence of high concentrations of SDS and 4 M urea. As an additional control, we autoclaved the sample for 60 min in the presence of the same denaturants, and the band corresponding to cross-link B persisted (Fig. 3). Thus, the new protein-DNA cross-link likely arises via a novel interaction of SSB with the fork.
Substrates in which the photolabel was placed in singlestranded regions near the fork (Ϫ3 in the lagging strand tem-plate and Ϫ6 within the leading strand template) also generated the slower migrating SSB-DNA cross-link B, but not DNA-DNA cross-links ( Fig. 2A, lanes 5 and 9). Remarkably, only cross-link B is generated when SSB cross-links to the Ϫ6 position within a 10-nt gap in the substrate. This species is dependent upon an opposing single strand at the fork. If the 45-nt single-stranded half of the lagging strand template is omitted, species A is generated (Fig. 3). Thus, the unique region of SSB that interacts with the first few nucleotides of the fork duplex is also involved in binding adjacent single-strands. We eliminated the possibility that band B resulted from heterogeneity or something peculiar to the photolabeled oligonucleotide used to construct the corresponding forked substrates by demonstrating that generation of species B is dependent upon a forked structure. Only species A is generated when a cross-link occurs between SSB and a single-stranded oligonucleotide (Fig. 2B).
PriA Binds Exclusively to the Fork Junction-We next examined interactions of PriA with the fork, initially in the absence of SSB. PriA cross-linked to the Ϫ3 and the ϩ3 positions, but not the Ϫ35 position within the lagging strand template (Fig. 4). On the leading strand, cross-links to both the Ϫ6 position and the penultimate nucleotide (Ϫ12) within the primer occurred. No interaction with the distal end of the duplex was observed. Thus, its interactions are limited to the vicinity of the fork in contrast to the broader binding observed in the absence of SSB and other proteins (5-7).
As with SSB, template-template cross-links were observed within the duplex ϩ3 region of the fork junction, as judged by their insensitivity to protease digestion (Fig. 4). Unlike SSB, PriA binding also led to DNA-DNA cross-links within the single-stranded region of the fork junction, at the Ϫ3 and Ϫ6 positions of the lagging and leading strand templates, respectively. The mobility of markers generated by incubation of high con-  APRIL 24, 2015 • VOLUME 290 • NUMBER 17

JOURNAL OF BIOLOGICAL CHEMISTRY 10831
centrations of PriA with single-stranded oligonucleotides corresponded to the mobility of the PriA cross-links observed on forked substrate (data not shown).
When PriA and SSB were added sequentially to the forks, PriA was observed to largely exclude SSB from the regions to which PriA binds (Ϫ3 and Ϫ6) and to the primer terminus (Ϫ12) (Fig. 5). No binding or greatly reduced binding of SSB was observed at these positions. In Fig. 5, no cross-link of PriA to the duplex region of the fork in the presence of SSB was observed. This result was variable: PriA cross-linked to the duplex in a similar experiment shown in Fig. 6 (lane 3). In either case, PriA binding is excluding SSB from the fork. At positions to which PriA does not bind (Ϫ35 position and distal end of duplex template) prominent SSB cross-links are observed (Fig.  5). PriA can rapidly access a fork in the presence of high concentrations of SSB during the strand displacement reaction catalyzed by Pol III HE in the absence of helicase (14). However, in our static fork model, exclusion of SSB by PriA is dependent upon addition of PriA first. If SSB is added first, it bound to all photolabeled positions and excluded PriA (Fig. 6).
DnaB helicase self-assembles in the absence of SSB, binding to the fork junction and making a prominent contact with the displaced strand. In the absence of steric blocks or SSB, bacterial replicative helicases can self-assemble by threading over the free 5Ј-end of the lagging strand template (15). When self-assembly was allowed to occur in the presence of ATP, no crosslinks were observed (Fig. 7). This was expected because the active helicase should split the two template strands and disso-   1-6 were treated prior to gel loading as described under "Experimental Procedures." Lanes 9 -12 were autoclaved for 60 min in gel loading buffer prior to gel loading. The "no flap" designation in lane 6 refers to an experiment performed on a substrate containing the photolabel at position Ϫ6 on the leading strand template but lacking the single-stranded 45-nucleotide 5Ј-flap of the lagging strand template (refer to Fig. 1). ciate. If nucleotide was omitted, only weak cross-linking was observed. However, in the presence of 100 M ATP␥S, prominent cross-links were observed, indicating that this nonhydrolyzable ATP analog stabilizes DnaB interaction with a fork junction but does not support progression of the helicase to separate the two template strands. FIGURE 5. PriA remains stably bound to the replication fork junction in the presence of SSB. 20 nM forked substrate was combined with 500 nM PriA followed by 500 nM SSB 4 (ϳ30 s later). Lanes that indicate ϩ protease were subjected to treatment with Proteinase K postirradiation to identify proteincontaining bands. FIGURE 6. The order of addition between SSB and PriA determines which protein is excluded at the fork juncture. 20 nM forked substrate was combined with 500 nM PriA and/or 500 nM SSB 4 . subunit 1 indicates the subunit added to the reaction first, and subunit 2 indicates the subunit added to the reaction second. Where no protein is listed for subunit 2, only subunit 1 was present in the reaction. Lanes that indicate ϩ protease were subjected to treatment with Proteinase K postirradiation. We exploited the DnaB self-loading reaction that occurs in the presence of ATP␥S to determine contacts made by DnaB in the absence of other primosomal proteins. Cross-links were observed on the lagging strand template that goes through the center of the hexameric helicase ring (Ϫ3 and ϩ3 positions) (Fig. 8, lanes 5 and 7). Interestingly, a strong cross-link was observed between the displaced leading strand template and the helicase at the Ϫ6 position (Fig. 8, lane 9). Template-template cross-links are also observed in the presence of bound protein.
Self-assembly of DnaB is blocked by attachment of streptavidin to the 5Ј-end of model replication forks (15). We conducted a parallel set of reactions in the presence of streptavidin attached to a biotin incorporated into the penultimate nucleotide at the 5Ј-end of the lagging strand template flap (Fig. 8,  lanes 13-17). The presence of streptavidin eliminated DnaB cross-linking. Interestingly, streptavidin tethered only nine nucleotides away from the Ϫ35 label fails to yield detectable cross-links (Fig. 8, lane 17). Given its monomeric molecular mass (13.2 kDa), we would have expected a streptavidin-90-mer cross-link to migrate with a mobility between the free 90-mer and SSB-90-mer (A). Thus, proximity is insufficient to yield cross-links. Cross-linking apparently requires direct association to compete effectively with the rapid insertion of a carbene into an oxygen-hydrogen bond of water.
Loading of DnaB by the Primosomal Pathway Can Be Detected by Photo-cross-linking-Using streptavidin linkage to the 5Ј-end of the lagging strand template to block self-assembly renders the DnaB loading reaction dependent upon primosomal proteins (15). Using the Ϫ6 photolabel in the displaced leading strand as a probe, we conducted a DnaB assembly reaction in the presence of the full complement of primosomal proteins in the presence of ATP␥S to stabilize DnaB interactions with the fork. We observe an efficient cross-link of DnaB (Fig.  9A, lane 3). Omission of any of the primosomal proteins results in loss of the cross-link, consistent with the biochemical dependences established by functional assays (Fig. 9A, lanes 4 -9) (15,24,25). The faint band with approximately the same mobility as a DnaB-DNA cross-link observed in the reaction in which PriA was omitted (Fig. 9A, lane 4) corresponds to a third minor cross-link observed between SSB and DNA at positions near the fork (compare lane 2). The presence of PriA (lanes 5-9) excludes SSB from the fork and eliminates this minor band.
Surveying the full complement of photo-cross-linking substrates employed in this study, we observe that DnaB crosslinks are restricted to ϩ3 on lagging strand template and Ϫ6 on the displaced strand. The Ϫ3 cross-link observed in the DnaB self-assembly reaction does not occur (Fig. 9B, lanes 7, 9, and  11). We note that some PriA and SSB cross-links are still detected, even in the presence of DnaB. These bands are light and presumably due to incomplete DnaB loading.

DnaB Is Not Present at the Fork in the Presence of High Inhibitory Concentrations of DnaC-DnaC is the helicase loader (26).
In multiple functional studies, DnaC titrations have yielded sharp optima with activity decreasing sharply at DnaC concentrations beyond the optimum (26). Formally, it is possible that this inhibition is due to DnaC association with DnaB on the fork, stalling its activity. Alternatively, inhibition could be due to unloading of DnaB from the fork or failure of complexes to form. We employed cross-linking to determine the status of DnaB at the fork in the presence of high inhibitory concentra-   (lanes 3-14).
tions of DnaC. We observe an optimal concentration of DnaC for DnaB cross-linking that parallels its optimum in functional assays based on the helicase activity of DnaC (Fig. 10, A, lanes 3  and 16, and B). At high DnaC concentrations, DnaB is not present at the fork, consistent with either removal or inhibition of the assembly reaction (Fig. 10A, lanes 5 and 18).
The Binding of PriB to Model Forks Mimics the Characteristics of SSB Binding-PriB is a homolog of SSB and will bind DNA directly by itself (17). We observe a pattern of binding that mimics that observed with SSB with the generation of a novel cross-linked species with slower mobility when PriB is photocross-linked to templates containing photolabel near the replication fork (Fig. 11A). Cross-linking of PriB to a single-stranded region 35 nt in front of the fork yields a single cross-link (designated species A to be consistent with the convention we adapted for SSB) that is identical whether cross-linked to a model fork or corresponding photolabeled single strand (Fig.  11A, lanes 1 and 3). Near the fork (Ϫ3 position in lagging strand, Ϫ6 position in leading strand, or in ϩ3 position in fork duplex), a slower migrating species B is dominant (Fig. 11A,  lanes 6, 11, and 8, respectively). Cross-linking to the single-stranded form of the photolabeled oligonucleotides used to construct the model forks yields only cross-linked species A (Fig. 11A, lanes 7, 13, and 10, respectively). DNA-DNA crosslinks also occur when PriB is bound as demonstrated by protease resistance (Fig. 11A, lanes 9 and 12). PriB also cross-links weakly to the leading strand primer terminus (Fig. 11A, lane  14).
PriA Is Not Displaced by PriB-Based on protein requirements to achieve strong binding by gel shift assays, it has been proposed that PriA, PriB, and DnaT act sequentially in the assembly of a primosome (4,5). A handoff model has been proposed for primosome assembly (18). We used our newly developed cross-linking tool to determine the presence of PriA at the fork at intermediate stages of primosomal assembly.
If we add PriA to model forks and then titrate in PriB, we observe strong binding and cross-linking of PriB to the Ϫ35 position far from the fork where PriA does not bind (Fig. 11B,  lanes 2 and 3). When positions near the fork are probed (Ϫ3 on the lagging strand, ϩ3 in the fork duplex, Ϫ6 on the leading stand, and Ϫ12 at the leading strand primer terminus), PriA is not displaced (Fig. 11B, lanes 9, 12, 15, and 18, respectively). Furthermore, the intensity of PriB cross-links near the fork is very weak compared with PriB binding to the same forked substrate alone (Fig. 11, A and B).
If we serially add PriA and SSB (like in Fig. 5) and then add PriB, we do not observe PriB effectively competing with SSB at the Ϫ35 single-stranded position upstream from the fork (Fig.  11C, lane 5). PriA interaction with positions near the fork are not diminished either (Fig. 11D, lanes 7, 9, 11, and 13). If the experiment conducted in Fig. 11D is repeated with adding DnaT and PriB simultaneously, the same result is obtained (Fig.  11E), except PriA interactions also occurred more broadly at the Ϫ35 position and the distal end of the duplex.

DISCUSSION
Previous studies have established an ordered assembly of replication restart primosomal proteins and have suggested a handoff mechanism in the loading of the DnaB helicase and restarting stalled E. coli replication forks (4,5,18). Footprinting and binding experiments have provided information regarding the regions in which single restart primosomal proteins can bind (5)(6)(7)(8)11). However, little information is available regarding the positions to which restart primosomal proteins bind in the context of their functional partners and how binding positions change during the competitions imposed during complex assembly and helicase loading. We chose to address this issue to gain a deeper understanding of the mechanism of restart primosome function using highly reactive photo-cross-linking moieties attached to specific positions within a model replication fork.
We initiated our study by monitoring cross-linking of SSB to forks. We anticipated that this would provide a convenient control and baseline using a well understood protein. As anticipated, binding of SSB to single-stranded DNA yielded a well defined single band presumably between the OB fold and the photolabel placed at a unique position within the singlestranded DNA probe. The cross-linked species exhibited approximately the same mobility whether the photolabel was located at either end or central to the oligonucleotide. To our surprise, when experiments were extended to model forks, three cross-links were observed within the double-stranded region of the fork, close to the fork junction, but not at the distal end of the duplex (Fig. 2). One species corresponded to that observed on single-stranded DNA. Another was assigned as a DNA-DNA cross-link because the band remained after protease digestion. DNA-DNA cross-links are not observed on forks in the absence of protein. This is likely because the reactive carbene generated rapidly reacts with water. When bound in a protein binding pocket, the DNA in the vicinity of the photolabel would be expected to be relatively dehydrated, enabling We attempted to generate a DnaT-DNA cross-linked marker but were unsuccessful using DnaT concentrations up to 500 nM. We note that a recent report (35) shows a DnaT-DNA interaction, but only at very high concentrations (Ͼ2.8 M).
reaction of the carbene with nearby protein or nucleic acid residues.
The third band contains protein and is novel, not occurring with the single-stranded DNA controls. Because the mobility is significantly slower on SDS gels, it must result from cross-linking to a position within SSB that is closer to the ends than the species observed with single-stranded DNA. This would result in a more asymmetric, slower migrating band. The same novel band is observed in the single-stranded regions adjacent to the fork, but not at the distal ends of the single-stranded flaps. This novel cross-link in the single-stranded regions requires a forked structure. If the opposing flap is deleted, only the species characteristic of binding single-stranded DNA is observed. Thus, SSB binds both single-and double-stranded regions at the fork differently than single-stranded DNA.
Our findings suggest that a single SSB tetramer can interact with a replication fork, including both single-strands and the flanking duplex. The spacing between the photolabels (6 nt on the lagging strand and 9 nt on the leading strand template) are well within the binding site size for an SSB tetramer (27). The requirement for single strands on both the leading and lagging strand side of a fork for the novel cross-link to occur suggests binding of both by the same SSB tetramer. If true, this opens the possibility that SSB could interact with not only the lagging strand template but the leading strand, perhaps nestled between Pol III HE and the DnaB helicase. This would allow the leading strand half of the Pol III HE to enjoy the same stabilization and face the same challenges as the lagging strand half. It has been commonly presumed, without evidence, that SSB binding occurs only on the lagging strand. Additionally, or alternatively, this interaction could support an important intermediate in repair or recombinational reactions in which SSB participates (28). If the interaction of SSB with a fork is sufficiently tight, it could help direct localization of limiting SSB to the fork through cooperative interactions.
PriB is a paralog of SSB, containing an OB fold and binding single-stranded DNA (17,29). Like SSB, it yields a single crosslinked species on denaturing gels when linked to a singlestranded probe and three distinct species when cross-linked at or near a fork. One species is the same as observed on singlestranded DNA, another is a DNA-DNA cross-link, and the third represents a slower migrating species resulting from a novel interaction with forks, both in the flanking duplex and single-stranded regions (Fig. 11).
Structures are available for PriB bound to single-stranded DNA. Binding takes place through an OB fold in a slightly different mode that uses different interactions (16,30). Unlike SSB, PriB binds both single-and double-stranded DNA (31). However, our studies indicate the interaction with a fork is different from general double-stranded DNA because the novel cross-linked species are only observed near the fork. It is possible that the binding observed near the distal end of the duplex arises from fraying, because no DNA-DNA cross-link is observed. No structure is available for PriB binding doublestranded DNA, but two models have been proposed (30). In one, the duplex has been modeled to interact with the OB-fold, through some contacts used to bind single strands plus others. In another model, a different binding site mimics binding of double-stranded DNA by the E. coli histone-like HU protein but still shares some interactions with the OB fold (30). Our findings could be explained by either model, provided novel contacts occur that are closer to the ends of the protein than the OB-fold single-stranded DNA contacts.
PriA, in the absence of other proteins, cross-links only to regions close to the fork, including the primer terminus and both flaps and the duplex. Interactions are not observed at the distal single-and double-stranded regions. These observations are consistent with the known binding properties of PriA (5)(6)(7)(8). DNA-DNA contacts are also observed at the fork, not only in the double-stranded region, but in the flanking single-strands. This indicates that the PriA complex brings the two opposing single-strands into close proximity in a dehydrated environment.
When SSB is added to established PriA-fork complexes, SSB is excluded from the fork but cross-links to the distal singlestranded regions of the flap and duplex, presumably by fraying it. A contact between PriA and the primer terminus remains, consistent with there being a specific binding pocket within PriA for the primer terminus. PriA binds the C terminus of SSB (12).
For SSB to be excluded from the fork, a specific order of addition in which PriA is added first is required. If SSB is added first, PriA is excluded from the fork (Fig. 6). However, PriA rapidly arrests strand displacement by Pol IIII HE in the absence of DnaB helicase in the presence of SSB (14). Perhaps PriA can bind to a smaller site than SSB, giving it earlier entry, or it interacts with Pol IIII HE, facilitating its entry.
The DnaB helicase can self-assemble on forks by threading itself onto the free end of a 5Ј-flap in the absence of steric blocks provided by SSB or attached proteins (15). This provides an avenue to study the positional binding specificity of this helicase in the absence of other proteins. To obtain a stable interaction, we must provide ATP␥S (Fig. 7). In the absence of nucleotide or in the presence of ATP, no DnaB-fork cross-links are observed. In the presence of ATP, DnaB certainly loads but proceeds with its helicase activity splitting the strands and dissociating. ATP␥S stabilizes binding but will not support helicase action. Cross-linking of DnaB is observed in the duplex region and the single-stranded flaps, all limited to the region of the fork junction (Fig. 8). This technique has proven useful in studying protein-protein interactions at the fork. For example, the affinity of the subunit of the Pol III HE for DnaB is enhanced when the latter is assembled onto a DNA fork in the presence of ATP␥S. 3 Addition of streptavidin to a biotin attached to the 5Ј flap blocks the DnaB self-loading reaction and makes it dependent on the full complement of primosomal proteins (15). We observe a full dependence upon primosomal proteins for DnaB loading in the photo-cross-linking reaction as well. Omission of any single primosomal protein results in the loss of DnaB assembly in the presence of ATP␥S. Like PriA, we see more precise positioning of DnaB in the presence of functional protein partners. When loaded by the primosomal pathway, the contact with the single-stranded portion of the lagging strand is lost because the helicase presumably progresses into the duplex for one helicase step, without hydrolysis.
We do not observe significant DnaC cross-links with forked substrates in complete reactions in the presence of ATP␥S and DnaB. On single-stranded DNA, DnaC binds DNA in the presence of ATP␥S and excludes DnaB from binding (23). We have also confirmed a DnaC-ss DNA interaction in our experiments (Fig. 10A, lane 13). Thus, the DnaC-ATP␥S-DNA complex observed (23) likely occurs transiently, but the reaction proceeds further on a forked structure without ATP hydrolysis.
We observe a strong DnaB cross-link with a photolabel attached to the displaced strand, 6 nucleotides in front of the fork (Fig. 8). Because we observe no cross-link between streptavidin and the photolabel attached to the Ϫ35 position, we know that proximity is not sufficient to support a cross-link. From our knowledge of the chemistry of the competing reactions, we would expect sequestration in a binding pocket that leads to dehydration to be required to enable cross-link formation. Reports of contacts with the displaced strand by helicases have been inferred from functional data. Bujalowski et al. (32) have observed that a leading strand arm of at least 14 nucleotides is essential for processive action of the helicase. They observed that ATP is required for the helicase to engage the arm. Our results show that hydrolysis is not necessary, because ATP␥S will suffice. Raney and co-workers (33) have observed that at least a 6-nucleotide leading strand tail is required for efficient action of the Dda helicase. Trakselis and co-workers (34) detected a contact with the displaced strand from an archaeal MCM helicases and showed that interaction made important contributions to stability and function.
We do not observe competition by PriB for the fork. PriB is excluded from the fork but binds to distal single-stranded regions. In the presence of both SSB and PriA, no PriB (or DnaT or DnaC) binding is observed to any of the photolabeled positions. We know both proteins are part of the complex because supershifts in labeled forks occur on native gels in their presence (4,5). Thus, the proposed handoff mechanism (18), although it likely occurs, probably awaits the presence of all protein partners with all interactions occurring rapidly and transiently before dissociation upon loading DnaB. Thus, our observations are consistent with a model in which PriA binds to forks, probably as a multimer binding both single-stranded flaps at the replication fork and the primer terminus (Fig. 12). SSB binds to flanking single-stranded regions and forms a specific contact with PriA, leading to precision in its positioning. The addition of PriB and DnaT does not change any of the existing contacts nor create new ones, so their association is peripheral. Upon addition of DnaC and ATP, DnaB loading occurs, probably with undetected additional rapid transient contacts before the remaining primosomal proteins dissociate. Upon reassociation of dimeric Pol III HE on the leading strand, replication can resume.