High-affinity DNA-binding Domains of Replication Protein A (RPA) Direct SMARCAL1-dependent Replication Fork Remodeling*

Background: Replication protein A (RPA) inhibits SMARCAL1 translocation on some substrates but activates it on others. Results: High-affinity RPA DNA-binding domains (DBDs) are critical to confer substrate specificity to SMARCAL1. Conclusion: The orientation of the RPA DBDs at the replication fork controls SMARCAL1 substrate specificity. Significance: The results provide insight into the regulation of DNA translocases by ssDNA-binding proteins. SMARCAL1 catalyzes replication fork remodeling to maintain genome stability. It is recruited to replication forks via an interaction with replication protein A (RPA), the major ssDNA-binding protein in eukaryotic cells. In addition to directing its localization, RPA also activates SMARCAL1 on some fork substrates but inhibits it on others, thereby conferring substrate specificity to SMARCAL1 fork-remodeling reactions. We investigated the mechanism by which RPA regulates SMARCAL1. Our results indicate that although an interaction between SMARCAL1 and RPA is essential for SMARCAL1 activation, the location of the interacting surface on RPA is not. Counterintuitively, high-affinity DNA binding of RPA DNA-binding domain (DBD) A and DBD-B near the fork junction makes it easier for SMARCAL1 to remodel the fork, which requires removing RPA. We also found that RPA DBD-C and DBD-D are not required for SMARCAL1 regulation. Thus, the orientation of the high-affinity RPA DBDs at forks dictates SMARCAL1 substrate specificity.

Replication is a fundamental process in all organisms and, in humans, involves the accurate and complete duplication of over 6 billion bp of DNA in each cell division cycle. Replication is challenged by DNA lesions, interference from transcription, and difficult-to-replicate sequences, all of which cause replication forks to stall (1). Stalled forks activate the ATR checkpoint kinase, which phosphorylates hundreds of downstream targets to coordinate the replication stress response and promote replication restart and repair (2). SMARCAL1 is one ATR substrate that travels with the replication fork (3). SMARCAL1 interacts with replication protein A (RPA), 3 the major ssDNA-binding protein (SSB) in human cells, and this interaction is required for SMARCAL1 localization (4 -8). SMARCAL1-deficient cells are hypersensitive to replication stress (4 -8) and accumulate DNA double-strand breaks catalyzed by the MUS81 nuclease (3). Too much SMARCAL1 activity through either overexpression or lack of restraining phosphorylation by ATR also causes an accumulation of double-strand breaks due to fork cleavage by an SLX4-dependent endonuclease (4,9). In humans, inherited loss-of-function mutations in SMARCAL1 cause Schimke immuno-osseous dysplasia, a disease characterized by renal failure, immune deficiencies, cancer susceptibility, and growth defects (10 -12).
Biochemically, SMARCAL1 is a DNA translocase that can evict RPA off DNA and anneal complementary strands (13). SMARCAL1 also performs branch migration and fork-remodeling activities (3). Fork remodeling is a proposed mechanism of replication fork stabilization in which the stalled fork is regressed to form a four-way junction called a "chicken foot" (14). This remodeling may promote repair by placing the DNA lesion that stalled the polymerase back into the context of dsDNA. It might also promote lesion bypass through a template-switching mechanism or simply be a mechanism for fork stabilization (14). Once the damage has been bypassed or repaired, the chicken foot is "restored" to a normal fork structure to resume DNA synthesis. RPA confers substrate specificity to SMARCAL1, directing it to regress stalled forks caused by leading-strand lesions and to restore normal forks with laggingstrand ssDNA (15), consistent with a function for SMARCAL1 in promoting genome stability by catalyzing fork remodeling. However, the mechanism by which RPA selectively stimulates SMARCAL1 on certain substrates but inhibits it on others is unknown.
RPA is a heterotrimeric protein with four DNA-binding domains (DBDs) that bind to DNA with a specific orientation (see Fig. 1A). These four DBDs do not bind DNA equivalently, with DBD-A having the highest affinity, followed by DBD-B, DBD-C, and DBD-D (16 -18). RPA binds DNA, with DBD-A and DBD-B making the initial contacts with 8 nucleotides of DNA (16), and an additional 20 nucleotides are protected when DBD-C and DBD-D bind (17,19,20). This 28 -30-nucleotide (nt) binding mode causes both the DNA and RPA to undergo major conformational changes and modifies the accessibility of protein-interacting surfaces on RPA (17,19,20).
The regulation of DNA translocases by SSBs is an evolutionarily conserved feature of SMARCAL1-related enzymes. Escherichia coli RecG and T4 UvsW are regulated by interactions with their SSBs (15,21,22). Furthermore, RPA regulates the specificity of many other reactions during replication and repair while being rapidly placed on and taken off DNA (17,23,24). The mechanisms underlying this fundamental aspect of DNA biology remain largely unknown. Although RPA can increase the local enzyme concentration to improve reaction rates, this mechanism does not explain how it generates substrate specificity for enzymes such as for SMARCAL1 (15).
In a previous study, we found that RPA regulation of SMARCAL1 is not due to a change in the ability of SMARCAL1 to bind DNA or hydrolyze ATP in the presence of RPA on the different substrates (15). In this study, we examined how RPA directs SMARCAL1 activity on specific substrates by testing two models. In the first model, we hypothesized that the location of the SMARCAL1-binding surface on RPA may be critical for generating substrate specificity. In the second model, we hypothesized that how RPA binds to the ssDNA in relation to the fork junction is important to create an optimal DNA-protein substrate for SMARCAL1. Our data support the second model and provide insights into how SSBs like RPA can generate substrate specificity for DNA translocases.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-HEK293T and U2OS cells were cultured in DMEM supplemented with 7.5% fetal bovine serum. HEK293T transfections were performed using PEI (Polysciences, Inc.), and U2OS transfections were done using FuGENE HD (Roche Applied Science).
DNA Substrate Purification-The oligonucleotides used to create the fork-remodeling substrates were described previously (15). The oligonucleotides used for creating the fork regression substrates with 8-nt gaps are as follows: lead74 (to create an 8-nt lead gap substrate), CGTCGCAGCGACGATC-GCACGTCGCGAACAACTTCAGCTGATAGACACGTGG-CAATTGCCTACATGTATCCTCA; and lag74 (to create an 8-nt lag gap substrate), TGAGGATACATGTAGGCAA-TTGCCACGTGTCTATCAGCTGAAGTTGTTCGCGACG-TGCGATCGTCGCTGCGACG.
The substrates were purified as described (15) with the exception of the vacuum centrifugation step. Instead, an Amicon Ultra 0.5-ml centrifugal filter (Ultracel, 10,000 nominal molecular weight limit, Millipore) was utilized following the manufacturer's specifications to concentrate the substrate solution.
Electrophoretic Mobility Shift Assays for RPA Binding-32 P-Labeled substrate (3 nM) was incubated with different concentrations of RPA in binding buffer containing 40 mM Tris (pH 7.5), 100 mM KCl, 5 mM MgCl 2 , 100 g/ml BSA, and 2 mM DTT for 20 min at room temperature. The reactions were separated by electrophoresis on an 8% gel (19:1 polyacrylamide, 1ϫ TBE) at 80 V for 80 min. Gels were dried and quantified using a Molecular Imager FX system (Bio-Rad).
Fork-remodeling Assays-The fork-remodeling assays were performed as described (15). The fork restoration reactions for experiments with RPA mutants DBD-C and FAB were carried out using 1 nM SMARCAL1. The regression reactions and other restoration reactions were carried out using 3 nM SMARCAL1.
Immunofluorescence and Antibodies-For the co-localization experiments, U2OS cells transfected with vectors encoding GFP-SMARCAL1 were plated on coverslips, fixed, and stained using antibodies as described (4).

RESULTS
The Location of the SMARCAL1-interacting Domain on RPA Does Not Dictate Substrate Specificity-RPA activates SMARCAL1 on some fork substrates and inhibits it on others (15). Normal replication forks containing RPA bound to the lagging-strand template are poor substrates for SMARCAL1catalyzed fork regression, whereas stalled forks with RPA bound to the leading-strand template are good substrates (Fig.  1B). Conversely, RPA bound to the chicken foot structures that are created by regression of stalled forks are poor substrates for SMARCAL1-catalyzed fork restoration, whereas chicken foot structures with RPA bound to the nascent leading strand are good substrates. Single-molecule studies indicate that SMARCAL1 moves farther on good substrates than on poor ones (15); however, how RPA dictates this specificity is unknown.
We hypothesized that the selective stimulation of SMARCAL1 is dictated by the orientation of RPA on DNA. Because the four RPA DBDs bind with a 5Ј to 3Ј polarity (17,19,27), the orientation of RPA with respect to the fork junction is different on the stalled fork and the normal fork (Fig. 1B). This could translate into a difference in the location of the RPA32C domain that contacts SMARCAL1 and thereby alters whether RPA stimulates or inhibits SMARCAL1 (Fig. 1, B and C).
To determine whether the location of the protein-interacting domain on RPA is critical for directing SMARCAL1 to the right substrate, we utilized a SMARCAL1 mutant protein (SMARCAL1-RPA70BD) that interacts with the RPA70N domain instead of the RPA32C domain (28). This mutant substituted an RPA70N-binding motif for the RPA32C-binding motif at the N terminus of SMARCAL1 ( Fig. 2A). SMARCAL1-RPA70BD bound RPA with similar affinity as wild-type SMARCAL1 (28) and co-localized with RPA in DNA damage foci induced by hydroxyurea (Fig. 2B) (4). In contrast, the SMARCAL1-⌬N protein, lacking any RPA-interacting surface, did not localize to stalled forks (Fig. 2B) (4). As expected, the DNA-binding ability of SMARCAL1-RPA70BD in the presence of RPA was indistinguishable from that of the wild-type protein (Fig. 2C). Because the ⌬N protein did not interact with RPA, the DNA-binding ability of SMARCAL1-⌬N in the presence of RPA was slightly reduced, as reported previously (15).
We then tested if the regulation of SMARCAL1 by RPA (15) is retained when SMARCAL1 binds RPA70N instead of RPA32C. RPA modestly stimulated both wild-type SMARCAL1-and SMARCAL1-RPA70BD-catalyzed fork regression when the ssDNA was on the leading-strand template (Fig. 3, A and B). RPA inhibited SMARCAL1-⌬N-mediated regression as described previously (15), suggesting that RPA acts as a block to SMARCAL1 when there is no interaction between RPA and SMARCAL1. Similarly, RPA stimulated wild-type SMARCAL1 and SMARCAL1-RPA70BD to catalyze fork restoration to the same extent (Fig. 3, C and D). There was some stimulation of SMARCAL1-⌬N by RPA on the restoration substrate as described previously (15), suggesting an effect of RPA on the DNA substrate that is independent of a physical interaction with SMARCAL1. Also, RPA inhibited fork regression of SMARCAL1-RPA70BD when bound on the lagging-strand template (Fig. 4), similar to its inhibition of wild-type SMARCAL1 and SMARCAL1-⌬N (15), suggesting that inhibition on some substrates is due to a physical block because RPA must be removed during the reaction. This also further confirms that inhibition by RPA is independent of an RPA-SMARCAL1 interaction. Overall, these results support the conclusion that the regulation of SMARCAL1 by RPA is independent of how the two proteins interact because there is no difference in the regulation of wild-type SMARCAL1 and SMARCAL1-RPA70BD.  . C, a model fork substrate was prebound with RPA (so that 100% of the substrate was bound) and incubated with increasing concentrations of the indicated SMARCAL1 protein for 30 min at room temperature. The products were then resolved on a 5% polyacrylamide gel. NPC, no-SMARCAL1 protein control.
High-affinity RPA-DNA Binding Is Required to Stimulate SMARCAL1-We next hypothesized that the DNA-binding polarity of RPA might be important to regulate SMARCAL1 due to some change imparted on the DNA substrate. In contrast to the highly schematized images in our models, the DNA bound to RPA adopts specific conformations (17,19,20,29), which would make the substrate presented to SMARCAL1 different depending on the orientation of the RPA DBDs. To investigate whether the affinity of RPA-DNA binding is important for stimulating SMARCAL1, we first utilized RPA mutants with altered affinity for DNA due to mutations in either DBD-A or DBD-B (25,26,30). These mutations decrease but do not eliminate the affinity of these domains for ssDNA. Wild-type and mutant RPA proteins were added to the DNA substrates at concentrations sufficient to obtain full DNA binding (Fig. 5A). The addition of excess wild-type RPA to mimic the higher concentrations of the RPA mutants needed to obtain fully bound substrates did not affect its regulation of SMARCAL1 (data not shown) as reported previously (15).
We then tested if the RPA proteins with mutated DBD-A or DBD-B stimulated SMARCAL1 in the fork-remodeling assays. Because the DBD-A and DBD-B mutations should make it easier for RPA to be removed from the DNA during the reaction, we expected SMARCAL1 to remodel substrates with these mutants more rapidly. However, both the DBD-A and DBD-B mutants displayed a marked decrease in their ability to stimulate SMARCAL1, particularly in the fork restoration assay (Fig.  5, B-D). There was still some stimulation compared with the no-RPA control because the mutations did not eliminate binding completely. An RPA protein with mutations in both DBD-A and DBD-B could not be tested because that mutant could not bind the substrates under the conditions used in this assay. Nevertheless, these results indicate that mutations in the highaffinity DBDs that should make it easier for SMARCAL1 to evict RPA actually interfere with the ability of RPA to stimulate SMARCAL1 translocation. However, the RPA DBD-A and DBD-B mutants did not inhibit SMARCAL1 as much as wildtype RPA when bound to the poor substrates (Fig. 5E), suggest-  High-affinity RPA Binding Regulates SMARCAL1 FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7

JOURNAL OF BIOLOGICAL CHEMISTRY 4113
ing that the inhibitory action of RPA is due to a blocking effect on enzyme movement.
RPA DBD-A and DBD-B Are Sufficient to Regulate SMARCAL1-RPA is removed from ssDNA in two or more steps, with DBD-D and DBD-C coming off first, followed by DBD-B and DBD-A (17,30,31). Thus, during the fork-remodeling reactions, there should be an intermediate state where DBD-A and DBD-B are bound to DNA, but DBD-C and DBD-D are not. This could allow SMARCAL1 to access the ssDNA that was bound by DBD-C and DBD-D, and thus allow SMARCAL1 to translocate more efficiently depending on the orientation of the DBDs. To test this idea and if the order of removal of the subunits of RPA off DNA is important, we designed fork substrates with 8-nt ssDNA gap (Fig. 6A), sufficient to allow only DBD-A and DBD-B binding to DNA (18,(31)(32)(33). Thus, any effect of DBD-C and DBD-D binding the DNA was eliminated. Binding of only RPA DBD-A and DBD-B to ssDNA on the leading-strand template was sufficient to stimulate SMARCAL1-catalyzed regression (Fig. 6B). RPA did not stimulate SMARCAL1 when bound to an 8-nt gap on the lagging-strand template (data not shown). The activity of SMARCAL1 on the 8-nt gap substrates was slightly less than on the 30-nt gap substrates. Overall, these experiments suggest that the DNA binding and removal of RPA DBD-C and DBD-D are not essential for regulation of SMARCAL1 and that only DBD-A and DBD-B are critical.
To further test if the DNA binding of DBD-C and DBD-D is dispensable for regulation of SMARCAL1, we utilized an RPA DBD-C mutant that abolished binding of DBD-C and DBD-D to DNA (17,25,26,34) even on a 30-nt ssDNA substrate (Fig. 7A). Consistent with the results from the 8-nt ssDNA regression substrates, the RPA DBD-C mutant stimulated SMARCAL1 activity to the same extent as wild-type RPA (Fig.  7B), suggesting that DBD-A and DBD-B binding to DNA is sufficient to selectively stimulate SMARCAL1 activity. Because the DBD-C mutant bound DNA with an overall affinity similar to the DBD-A and DBD-B mutants (26), this result also demonstrates that the decrease in the ability of those mutants to stimulate SMARCAL1 is not due to the overall decrease in the DNA affinity of the mutant RPA proteins.
Finally, to confirm that the two high-affinity RPA DBDs are indeed sufficient to stimulate SMARCAL1 when attached to a SMARCAL1-interacting domain, we utilized FAB-RPA, an RPA truncation mutant that has only DBD-A and DBD-B, and the RPA70N protein-interacting domain (26). SMARCAL1 did not interact with the FAB mutant and hence would be expected to behave similarly to SMARCAL1-⌬N. However, SMARCAL1-RPA70BD, the SMARCAL1 mutant that interacted with RPA70N, should have a high level of stimulation with FAB-RPA if the high-affinity DBDs plus a SMARCAL1-interacting surface are sufficient for stimulation. Indeed, wild-type SMARCAL1 behaved like SMARCAL1-⌬N in the presence of FAB-RPA, but SMARCAL1-RPA70BD was stimulated by FAB-RPA to the same extent as wild-type SMARCAL1 by full-length RPA (Fig. 7, C and D). Thus, an RPA truncation mutant with only the high-affinity DBDs stimulates a SMARCAL1 mutant that interacts with it.

DISCUSSION
RPA is an essential regulator of DNA metabolism, serving as both a platform for protein recruitment and a director of enzymatic activities (17,29). RPA interacts with many DNA translocases and can either stimulate or inhibit enzyme function. However, the mechanism by which it regulates enzymes in addition to increasing their local concentration is largely unknown. Our results reveal insights into the mechanism by which RPA regulates SMARCAL1, a fork-remodeling enzyme, and provide a paradigm for understanding the dynamic nature of RPA-directed DNA metabolism.
Due to the high-affinity binding of RPA to ssDNA, RPA is almost always present when ssDNA forms during DNA replication or repair. Enzymes that translocate on DNA, such as helicases, translocases, and nucleases, must work in the presence of RPA (15,(35)(36)(37)(38)(39). Regulation of DNA enzymes by SSBs is a conserved property in all organisms. For example, both E. coli SSB and T4 phage gp32 regulate the fork-remodeling proteins RecG and UvsW, respectively (15,21,22). However, differences in the structures of the prokaryotic, viral, and eukaryotic SSBs suggest differences in mechanisms.
In some cases, a change in enzyme recruitment or ATPase activity in the presence of the corresponding SSB is important (40 -42). However, the regulation of many helicases and translocases such as SMARCAL1 cannot be explained by either mechanism. Also, in more complex examples, RPA stimulates enzymes such as FANCJ and RECQ1 in a strand-specific manner (43). The basis of such regulation is largely not understood.
In the case of SMARCAL1, RPA selectively stimulates regression of a stalled fork and inhibits regression of a normal fork substrate (Fig. 1C). It also stimulates SMARCAL1 to restore normal forks while inhibiting its restoration activity when the product looks like a stalled fork. We hypothesized that the polarity of RPA on DNA, which controls the location of the DBDs and the location of the SMARCAL1-interacting surface with respect to the fork junction, conferred substrate specificity. Specifically, we developed two models. The first predicted that the location of the SMARCAL1-interacting surface on RPA is critical for regulation. This model was based on the idea that the RPA protein-interacting domain might be needed as an anchoring point for SMARCAL1 to remove RPA from DNA, with the low-affinity DBD-C and DBD-D being removed first (17,29,44). However, our results are inconsistent with this model because a SMARCAL1 mutant that interacts with RPA70N is still regulated by RPA in the same way as wild-type SMARCAL1 that interacts with RPA32C. Also, the crystal structure of RPA bound to DNA and small-angle x-ray scattering and NMR data suggest that RPA bends DNA and that its protein-interacting domains are connected to the DBDs with flexible tethers (19,20). Thus, the protein-interacting domains could be close in space and incapable of providing firm anchoring points.
In our second model, we hypothesized that the way in which RPA binds replication forks, which is different between the normal fork and the stalled fork due to its asymmetry, is crucial for regulating SMARCAL1. This model predicts that alterations in the RPA DBDs would impact regulation. Indeed, mutations in either of the two high-affinity RPA DBDs caused a marked decrease in the ability of RPA to stimulate SMARCAL1 activity. Furthermore, an RPA protein with only a SMARCAL1-interacting domain and the two high-affinity DBDs is sufficient to activate SMARCAL1. These results are consistent with observations that RPA with mutations in the high-affinity DBDs is defective in DNA repair but can support DNA replication (25).
In summary, these results indicate that the placement of the high-affinity RPA DBDs on DNA substrates dictates SMARCAL1 substrate specificity. Thus, although fork remodeling by SMARCAL1 must happen concurrently with displacement of RPA, some (but not all) mutations that decrease its DNA affinity actually make it more difficult for SMARCAL1 to translocate on the RPA-bound substrate. As might be expected, RPA mutants with lower affinity for DNA make the poor SMARCAL1 substrates better.
RPA binding to DNA is a highly dynamic process, and RPA can diffuse along ssDNA (29,44,45). This diffusion of RPA along DNA may destabilize secondary structures to allow access of other proteins to ssDNA (29,45). RPA diffusion can melt DNA hairpins efficiently when the hairpin is located at the 3Ј-end. However, when the hairpin is located 5Ј to RPA, very little strand melting is observed (45). The high-affinity DBD-A and DBD-B domains are sufficient to induce hairpin destabilization (45). RPA can also promote unwinding of a duplex DNA arm of a synthetic fork in an orientation-specific manner (46). We speculate that RPA might stimulate SMARCAL1 in part by promoting the transient destabilization of the nascent-parental DNA duplex at the fork. This would explain why RPA selectively stimulates SMARCAL1 on some substrates but not others. Alternatively, the selective stimulation of SMARCAL1 could be because RPA induces specific DNA conformations that are more or less conducive to SMARCAL1 DNA translocation depending on RPA orientation with respect to the fork junction. Ultimately, high-resolution structures of enzyme-RPA-DNA complexes will be needed to distinguish between these ideas.