Distinct RPA functions promote eukaryotic DNA replication initiation and elongation

Abstract Replication protein A (RPA) binds single-stranded DNA (ssDNA) and serves critical functions in eukaryotic DNA replication, the DNA damage response, and DNA repair. During DNA replication, RPA is required for extended origin DNA unwinding and DNA synthesis. To determine the requirements for RPA during these processes, we tested ssDNA-binding proteins (SSBs) from different domains of life in reconstituted Saccharomyces cerevisiae origin unwinding and DNA replication reactions. Interestingly, Escherichia coli SSB, but not T4 bacteriophage Gp32, fully substitutes for RPA in promoting origin DNA unwinding. Using RPA mutants, we demonstrated that specific ssDNA-binding properties of RPA are required for origin unwinding but that its protein-interaction domains are dispensable. In contrast, we found that each of these auxiliary RPA domains have distinct functions at the eukaryotic replication fork. The Rfa1 OB-F domain negatively regulates lagging-strand synthesis, while the Rfa2 winged-helix domain stimulates nascent strand initiation. Together, our findings reveal a requirement for specific modes of ssDNA binding in the transition to extensive origin DNA unwinding and identify RPA domains that differentially impact replication fork function.


INTRODUCTION
Eukaryotic DNA r eplication r equir es the progr essi v e assembly of protein complexes at origins of replication ( 1 ).Origins are licensed during the G1 phase of the cell cycle, when two hexameric Mcm2-7 helicases are loaded onto each origin as a head-to-head doub le he xamer encircling double-stranded DN A (dsDN A) (2)(3)(4)(5).Upon S-phase entry, S-phase cyclin-dependent kinase (S-CDK) and Dbf4dependent kinase (DDK) along with additional replication proteins activate the helicase for DNA unwinding ( 6 ).Once activated, the helicases act separately to bidirectionally unwind the origin and adjacent DNA ( 7 ).The activated helicases and the single-stranded DN A (ssDN A) they generate r ecruit the r emainder of the DNA synthesis machinery to form a bidirectional pair of replication forks ( 8 ).
Helicase activation is the committed step of replication initiation and involves recruitment of key proteins to the helicase and origin DNA unwinding.Nine proteins, r eferr ed to as activation factors, are sufficient to activate loaded helicases in vitro ( 9 ).Eight of the activation factors coordinate the recruitment of Cdc45 and GINS to Mcm2-7 to form the CMG (Cdc45 / Mcm2-7 / GINS) complex ( 1 ).Cdc45 and GINS directly activate both the ATPase and helicase activity of Mcm2-7 ( 10 ), converting the Mcm2-7 complex to the acti v e replicati v e DNA helicase, CMG.Finally, Mcm10 further primes the helicase for DNA unwinding (11)(12)(13)(14).Experiments omitting individual activation factors reveal several steps in origin DNA unwinding ( 12 , 15 ).First, CMG formation results in a small amount of DNA melting ( ∼6-7 bp per helicase).Addition of Mcm10 stimulates CMG to unwind a limited amount of additional DNA (10-15 bp).Finally, the eukaryotic single-stranded DNA binding protein (SSB) replication protein A (RPA) is r equir ed for extensi v e DNA unwinding ( 12 ).
RPA is an essential SSB in eukaryotes that binds ss-DNA in a sequence-nonspecific manner (16)(17)(18).In addition to its r equir ement for e xtensi v e origin DNA unwinding ( 12 ) and stimulation of CMG DNA unwinding activity ( 19 ), RPA also protects ssDNA and coordinates DNA damage signaling and DNA repair ( 20 ).RPA is a heterotrimeric protein that has a conserved domain structure across eukaryotes (Figure 1 A) ( 17 ).The largest subunit, called Rfa1 in Sacchar om y ces cer evisiae , consists of four oligonucleotide / oligosaccharide-binding (OB)-fold domains, three of which bind DNA (OB-A, OB-B and OB-C) and another that functions as a protein-interaction domain (OB-F) ( 20 , 21 ).The middle sub unit, Rfa2, contrib utes a fourth DNA-binding OB-fold domain (OB-D) and a Cterminal winged-helix (WH) domain.Rfa3 is the smallest subunit, consisting of a single OB-fold (OB-E) that is r equir ed for trimerization of the complex but does not bind DNA.The three proteins form a tight, structurallyconserved trimerization core between OB-folds C, D and E (Figure 1 A) ( 22 ).
RPA associates with ssDNA in a highly dynamic manner.These dynamics have been studied both with isolated domains and in the context of the full-length protein ( 23 ).RPA can adopt multiple ssDNA-binding conformations with varying affinities using its four DNA-binding domains (OB-A through -D; Figure 1 A).Depending on the domains involved, RPA can bind as few as 8 nucleotides and as many as 32 nucleotides of ssDNA ( 24 , 25 ).Binding affinities of RPA and its sub-domains have been reported in the nanomolar to sub-nanomolar range ( 20 ).Although cooperati v e RPA-ssDNA binding was initially observed ( 26 ), mor e r ecent e vidence suggests RPA cooperati vity only occurs when RPA is phosphorylated during the DNA-damage response ( 27 ).Structural studies show that the different RPA DNA-binding domains bind ssDNA in different conformations.OB-A and OB-B bind linear stretches of DNA ( 28 ), wher eas, the structur e of the trimerization core shows a dramatic local bend in the ssDNA as it wraps around the OB-C and OB-D domains ( 27 , 29 ).Structural studies have yet to determine the relati v e positions of the different elements of RPA presumably due to their intrinsically dynamic behavior.
SSBs ar e pr esent in all domains of life as well as a subset of viruses, but these proteins display a wide variety of structures and ssDNA-binding affinities (re vie wed in ( 30 )).The first SSB characterized was Gp32 (Gene 32 protein), which is a monomeric 34 kDa protein that acts during T4 bacteriophage r eplication (Figur e 1 B).Gp32 consists of a single OB-fold that can bind up to 10 nucleotides of ssDNA with a K d in the 10 nanomolar range ( 31 , 32 ).The canonical bacterial SSB was initially identified in Escherichia coli (r eferr ed to here as EcSSB), and binds DNA as a homotetramer (Figure 1 C) ( 33 ).A single EcSSB tetramer has four OBfold DNA-binding domains that transition between multiple DN A-binding conformations, wra pping as few as 35 and up to 65 nucleotides around its core ( 34 ).EcSSB DNAbinding affinity is higher than Gp32 and RPA, with dissociation constants in the sub-nanomolar range ( 35 , 36 ).Unlike RPA, both Gp32 and EcSSB exhibit strong cooperati v e DNA binding ( 37 , 38 ).Despite their different ssDNAbinding properties, RPA, Gp32 and EcSSB each play an essential role during DNA replication in their cognate organism.
Much of our understanding of RPA function during eukaryotic DNA replication comes from studies of simian virus 40 (SV40) DNA replication in human cells.This viral replication fork uses the SV40 large T-antigen (LTag) as the replicati v e helicase and a subset of the human DNA synthesis machinery to replicate the SV40 genome.Studies of SV40 DNA replication in vitro have shown that neither DNA unwinding nor DNA synthesis occur in the absence of RPA (39)(40)(41).Additionally, human RPA has a unique function during SV40 DNA replication, but not DNA unwinding, that cannot be performed by other SSBs, including yeast RPA ( 42 , 43 ).Specific interactions between human RPA and the SV40 replication machinery ar e pr esumed to mediate the specificity of this function.For example, human but not yeast RPA can bind LTag to promote primosome assembly ( 44 ).Further, RPA can stimulate Pol-␣/ primase activity ( 43 , 45 , 46 ) and promote the Pol-␣/ primase to Pol ␦ polymerase switch during lagging-strand synthesis (47)(48)(49).Howe v er, RPA interactions with replication proteins have not been addressed at a fully eukaryotic DNA replication fork.
Here, we investigate the roles of RPA during eukaryotic DNA replication initiation and elongation.Because RPA has not been observed to interact with CMG helicase, we asked if DNA unwinding r equir es RPA specifically or only its ssDNA-binding activity.By testing either Gp32, EcSSB or RPA mutants in reconstituted origin unwinding assa ys, we f ound that neither m ultiple ssDN A-binding domains nor high ssDNA-binding affinity is sufficient to facilitate e xtensi v e origin DNA unwinding.Instead, a specific arrangement of DNA-binding but not pr otein-pr otein interaction domains is r equir ed for this e v ent.In contrast, using r econstituted r eplication assa ys, we f ound that, unlike origin unwinding, the replication fork requires two RPA domains in addition to its ssDNA-binding domains for normal DN A synthesis.Anal ysis of m utants individuall y lacking each of these domains provides insights into the steps in replica tion tha t they participa te in.
coding sequence for Rfa1 residues 131-621, and 3xFlag-Rfa2 WH contains coding sequence for Rfa2 residues 1-202.The respecti v e plasmids were integrated into yRH101 to generate yAP05 (CBP-Rfa1 OB-F , Rfa2, Rfa3); yAP17 (CBP-Rfa1, 3x-Flag-Rfa2 WH , Rfa3); yAP18 (CBP-Rfa1 OB-F , 3x-Flag-Rfa2 WH , Rfa3).Other RPA mutants wer e expr essed in bacteria.To make RPA AB , the coding sequence for CBP-OB-A-OB-B was ordered as a gBlock (IDT) and cloned into the pET3aTr backbone for expression in E. coli .RPA ABAB was generated by PCR amplifying OB-A and OB-B and the linker that joins to the N-terminal boundary of OB-C (OB-AB + linker) followed by Gibson assembly into the RPA AB plasmid.To express the trimerization core, the p11d-tscRPA-30MxeHis6 plasmid ( 50 ) was modified by truncating the coding sequences for Rfa1 and Rfa2 to express only OB-C (Rfa1 442-621 ) and OB-D (Rfa2 232-182 ) respecti v ely.The resulting plasmid contains an inducible Rfa1-OB-C, Rfa2-OB-D and Rfa3 coding sequences with an intein, chitin-binding domain and a 6xHis tag at the Rfa2 C-terminus.All mutations were confirmed by Sanger sequencing.Additional details on expression constructs and yeast strains can be found in Supplementary Tables S2 and S3.

Protein expression and purification
Purified E. coli SSB (Sigma-Aldrich S3917) and T4 Gene 32 Protein (Gp32; Roche 10972983001) were obtained from commercial vendors.Sacchar om y ces cer evisiae RPA, RP A OB-F , RP A WH and RP A OB-F WH were purified from yeast and RP A AB , RP A ABAB and RP A Tri-C were purified from E. coli as described below.RPA purified from both yeast and bacteria have been shown to support in vitro DNA replication ( 9 , 14 ) Wild-type RPA (CBP-Rfa1, Rfa2 and Rfa3) was expressed and purified from yAE31 using calmodulin and Hi-Trap Heparin columns as described in ( 9 ).RPA OB-F was expressed and purified from yAP05 using the same protocol.Briefly, 8 L of logarithmic culture were alpha-factor arr ested and RPA expr ession was induced with galactose for 3.5 hours.Cells were harvested and ground into po w der which was then thawed into Buffer C (25 mM Tris-HCl pH 7.2, 10% glycerol, 1 mM DTT) with 500 mM NaCl.Lysate was clarified by ultracentrifugation (200,000 x g for 90 min), then supplemented with 2 mM CaCl 2 and bound to a 1 ml calmodulin-affinity column.RPA was eluted with Buffer C supplemented with 200 mM NaCl, 2 mM EDTA and 2 mM EGTA.RPA-containing fractions were pooled, dialyzed against Buffer C with 50 mM NaCl and 1 mM EDTA, and applied to a 1 ml HiTrap Heparin column (Cytiva) equilibrated in the same buffer.RPA was eluted with a 30 column volume (CV) gradient from 50 mm to 1 M NaCl in Buffer C + 1mM EDTA.RPA-containing fractions were collected, snap-frozen on liquid nitrogen and stored at −80 • C until use.
For purification of RPA WH , lysate from 8L of yAP17 (CBP-Rfa1, 3xFlag-Rfa2-WH and Rfa3) was pr epar ed as described for RPA and then bound to 1ml Flag resin (Sigma) equilibrated in Buffer C with 100 mM NaCl, washed and eluted with Buffer C supplemented with 100 mM NaCl and 0.3 mg / ml 3xFlag peptide.RPA-containing fractions were pooled, concentrated and applied to S200 Increase 10 / 300 column (Cytiva) equilibrated in Buffer C with 200 mM NaCl and 1 mM EDTA.Stoichiometric RPA complex es wer e snap frozen and stored a t −80 • C .RPA OB-F WH was expressed and purified from 16 L of yAP18 using the approach for WT RPA, except a flag affinity step (as described for RPA WH ) was added between the calmodulin and heparin columns to enrich for complexes that contained both mutant subunits.
For RPA AB and RPA ABAB , the appropriate expression plasmid was transformed into Rosetta(DE3)pLysS cells, expanded to 2 l and induced with 1 mM IPTG during midlog phase and incubated at 16 • C overnight.Cells were pelleted and lysed by sonication in C / 500 (25 mM Tris-HCl pH 7.2, 10% glycerol, 1 mM DTT, 500 mM NaCl and 1 mM EDTA).Lysates were then purified by calmodulinaffinity and heparin columns as described for WT RPA.
For RPA Tri-C , p11d-scTriC was transformed into Rosetta(DE3)pLysS and expressed and purified as in ( 14 ), except after the Ni-NTA and chitin columns, the eluate, which had a significant excess of Rfa2, was pooled and passed over a heparin column.The monomeric Rfa2 sa tura ted the heparin column and Tri-C was concentrated in the flow-through.The flow-through was collected and Tri-C was further purified by binding to a 1 ml MonoQ column (Cytiva) equilibrated in Buffer C with 50 mM NaCl and 1 mM EDTA, then eluted with a 30 CV gradient from 50 mM to 1 M NaCl.Fractions containing trimeric RPA were further purified further by size exclusion on S75 Increase column (Cytiva) in Buffer C with 200 mM NaCl and 1 mM EDTA, snap-frozen and stored at −80 • C.

Fluorescence polarization assays
For RPA / EcSSB / Gp32 DNA-binding, serial 2-log dilutions of the respecti v e SSB were prepared in 30 mM HEPES pH 7.5, 30 mM NaCl, 0.25 mM EDTA, 10% glycerol, 0.01% NP-40, 1 mM DTT.Proteins were then mixed with 6-FAMlabeled oligo-dT30 (IDT) to a final concentration of 2.5 nM supplemented with 0.5 M NaCl ( 54 ).The protein-DNA mix was incubated for at least 30 min at room temperatur e to r each equilibrium ( 55 ).Thr ee technical r eplicates were plated in a black 384-well nonbinding microplate (Greiner Bio-One) and read in a SpectraMax ID5 plate reader (Molecular Dynamics).Anisotropy values from the technical r eplicates wer e averaged and corr ected by subtracting the value from a buffer control that contained no pr otein.Values fr om thr ee independent experiments wer e plotted in GraphPad PRISM and fit to the Hill equation.For Pol-␣/ primase recruitment experiments, 6-FAMlabeled oligodT60 was pre-incubated with 2.5 nM of the respecti v e RPA protein in the absence of supplemental NaCl.Then, Pol-␣/ primase was titrated into RPA-bound DNA mixture and incubated for 30 min before reading samples as described above.All samples were corrected for by subtracting a control well with no added Pol-␣/ primase.

EcSSB, but not Gp32, substitutes for RPA during origin DNA unwinding
To study the role of RPA during eukaryotic DNA replication origin unwinding, we compared purified yeast RPA, E. coli SSB (EcSSB), and T4 bacteriophage Gp32 (Figure 1 D).The ssDNA-binding properties of these proteins have been w ell characterized, how e v er, it was important that we measur ed the r elati v e ssDNA-binding affinities for our protein preparations.To estimate DNA binding affinities, we incubated varying concentrations of each protein with a fixed concentration of a fluorescently-labeled 30-nucleotide  dT oligonucleotide (6FAM-dT 30 ), then measured fluorescence polarization.To ensure equilibrium binding, all reactions were performed in high-salt conditions ( 54 ) and incubated for 30 min, after which we observed no further changes in anisotropy.Consistent with previously reported measurements ( 31 , 32 , 35 , 56 ), Gp32 showed the lowest binding affinity, followed by RPA, then EcSSB (Figure 1 E and Supplemental Table S1).Importantly, all subsequent assa ys are perf ormed under sa tura ting concentra tions for each SSB.
Next, we tested the RPA r equir ements for origin activation using an in vitro origin DNA-unwinding assay using purified budding yeast proteins ( 12 ).We detected e xtensi v e origin-dependent DNA unwinding using a topologicallyconstrained circular ARS1 -containing DNA template, onto which we loaded and activated Mcm2-7 in the presence of a topoisomerase (Figure 2 A).The topoisomerase relie v es supercoiling generated by DNA duplex unwinding.Reactions are terminated by rapid protein denaturation, removing all pr otein fr om the DNA.This tr eatment r esults in unwound DNA re-annealing in the absence of topoisomerase, genera ting nega ti v e supercoils tha t migra te faster in an agarose gel compared to the relaxed plasmid (Figure 2 A).The topological changes observed are DDK-dependent, consistent with this kinase being r equir ed for assembly of the CMG helicase (Supplemental Figure S1 and Figure 2 B , C , lane 5).As previously observed ( 12 ), the extensive origin DNA unwinding measured in this assay r equir es RPA (Figur e 2 B , C , lanes 3 versus 4).We note that this assay detects e xtensi v e origin unwinding, but not the initial, RPA-independent origin melting that results from the initial steps of helicase activation.
If e xtensi v e origin unwinding only r equir es the ssDNAbinding activity of RPA, then other SSBs should support eukaryotic DNA unwinding.When we tested Gp32 in the DNA-unwinding assay, we did not observe levels of DNA unwinding above that detected in the absence of any SSB (Figure 2 B, lanes 6-10).Even increasing the concentration of Gp32 into the micromolar range did not r estor e DNA unwinding activity, eliminating the possibility that Gp32's lower ssDNA-binding affinity explains the lack of DNA unwinding (Figure 2 B).The inability of Gp32 to support origin unwinding suggests having a protein coat the exposed ssDNA is not sufficient for this process.Interestingly, substituting EcSSB in the unwinding assay supported robust DNA unwinding (Figure 2 C, lanes 6-8).We observed similar le v els of unwinding between RPA and EcSSB when provided in reactions at the same concentration (Figure 2 C, compare lanes 4 and 6).Thus, EcSSB shares the properties of RPA that ar e r equir ed for eukaryotic replication origin unwinding.

Multiple RPA ssDNA-binding domains are required to promote origin unwinding
The ability of EcSSB, but not Gp32, to substitute for RPA during origin unwinding has two interesting implications.First, simple ssDNA binding is not sufficient to promote extensi v e origin unwinding, suggesting other ssDNA-binding properties are important.Second, because EcSSB functions as well as RPA, specific interactions with the CMG helicase are not required for this e v ent.Unlike the monomeric Gp32, RPA and EcSSB are both large, multimeric complexes containing a total of four DNA-binding domains, bind a larger ssDNA footprint than Gp32, and engage in dynamic associations with ssDNA ( 20 , 34 ).One or more of these shared features of RPA and EcSSB must be r equir ed f or eukary otic origin unwinding.Since Gp32 contains only a single DNA-binding domain, we first tested whether the number of DNA-binding domains explained the different ability of Gp32 and RPA / EcSSB to facilitate origin unwinding.To this end, we generated RPA mutants containing two DNA-binding domains.RPA can be divided into two subdomains, each containing two DNA-binding domains that bind ssDNA with different properties.RPA AB is composed of OB-A and OB-B domains, which are adjacent to one another in Rfa1 and together bind 8 nts of ssDNA in an approximately linear conformation (Figures 1 A and 3 A) ( 28 ).In contrast, the RPA trimerization core (RPA Tri-C ) is a trimer made up OBdomains from three separate subunits (OB-C / D / E; only OB-C and OB-D bind ssDNA) that exerts a local bend in the 19 nt of bound ssDNA (Figures 1 A and 3 A) ( 27 ).These different DNA-binding properties contribute to different affinity for and dynamics on ssDNA.
We purified the RPA AB and RPA Tri-C subdomains and tested their ability to bind DNA and stimulate origin unwinding.When we tested these RPA mutants in fluorescence polarization assays, both RPA AB and RPA Tri-C displayed lower ssDNA-binding affinity compared to fulllength RP A. RP A AB showed ssDNA binding comparable to that of Gp32, whereas RPA Tri-C DNA-binding affinity was between that of RPA and RPA AB (Figure 3 C and Supplemental Table S1).The relative affinities are consistent with recent studies on RPA ssDNA-binding dynamics ( 23 , 25 ).When we tested each two-ssDNA-binding-domain mutant in the origin unwinding assay, neither RPA AB nor RPA Tri-C stim ulated DN A unwinding.As with Gp32, increasing the concentration of RPA AB or RPA Tri-C to compensate for their lower ssDNA affinity did not r estor e origin DNA unwinding (Figure 3 D and Supplemental Figure S2).We conclude that an ssDNA-binding protein with two DNAbinding domains is not sufficient to promote origin DNA unwinding.
Because both RPA and EcSSB include four DNAbinding domains, we asked if the higher avidity or larger binding site of this number of ssDNA-binding domains is r equir ed to promote origin unwinding.To test this possibility, we made an artificial RPA mutant containing two copies each of the OB-A and OB-B domains in a single polypeptide (RPA ABAB ).The two copies of OB-A and OB-B are connected by the nati v e linker that separates OB-B from OB-C (Figure 4 A, B).Structural evidence shows the OB-A and OB-B domains bind 8 nt of ssDNA ( 28 ), sug-gesting RPA ABAB can bind at least 16 nt and perhaps more due to the fle xib le linker between the two AB domains.This RPA mutant showed ssDNA binding that is comparable to wild-type RPA (Figure 4 C and Supplemental Table S1).When tested in DNA-unwinding assays, however, RPA ABAB failed to promote e xtensi v e origin DNA unwinding (Figure 4 D, lanes 6-8).These data support the conclusion that neither ssDNA-binding affinity nor number of DNA-binding domains explain the SSB r equir ement during origin unwinding.
To test whether origin unwinding r equir es the endogenous, multimeric context of RPA, we generated an RPA mutant, RPA OB-F WH , that contains all four DNA-binding domains but lacks the known RPA protein-interaction domains, Rfa1-OB-F and Rfa2-WH (Figure 4 A-B).These domains interact with DNA repair and replication proteins, such as Pol-␣/ primase and Dna2, but are not involved in DNA binding ( 21 , 47 , 57-63 ).Indeed, we found this RPA m utant bound DN A with near wild-type affinity (Figure 4 C and Supplemental Table S1).In contrast with the RPA ABAB m utant, RPA OB-F WH supported robust DN A unwinding a t all concentra tions tested (Figure 4 D, lanes 4 versus 9-11).To gether, these RPA m utant results indica te tha t extensi v e DNA unwinding at replication origins r equir es the specific arrangement of DNA-binding domains in RPA.Combined with our finding that EcSSB functions during origin unwinding, these studies support a model in which DNA bending and / or dynamic changes in ssDNA-binding modes ar e r equir ed for origin DNA unwinding.

Replication elongation r equir es the RPA OB-F and WH domains
We next asked whether ssDNA-binding proteins that support origin unwinding also support DNA synthesis.To monitor replica tion initia tion and elonga tion, we utilized an in vitro reconstituted DNA replication assay on linearized ARS1 -containing plasmid DNA (Figure 5 A) ( 9 , 14 ).Nascent DNA strands are monitored by incorporation of radiolabeled dNTPs followed by separation by alkaline gel electrophoresis.Omission of the flap endonuclease (Fen1) and DNA ligase from the reactions allows distinction of long leading-strand synthesis products (3000-6000 nts) from shorter lagging-strand products (100-500 nts) (Figure 5 ) ( 9 ).
We found that DNA synthesis has distinct RPA r equir ements compared to DNA unwinding.As previously reported ( 9), we saw no DNA replication without RPA, and addition of RPA led to robust, DDK-dependent leadingand lagging-strand synthesis (Figure 5 B-D, compare lanes 1-4).Consistent with its defect in promoting origin unwinding, addition of Gp32 did not support any DNA synthesis (Supplemental Figure S3A).When EcSSB was substituted for RPA, howe v er, we observ ed a distinct pattern of replication products.When supplied at the equivalent concentration of RPA (75 nM), EcSSB supported weak DNA-synthesis activity (Figure 5 B, lane 4).Thus, although this EcSSB concentration supports robust DNA unwinding (Figure 2 ), it showed a strong defect in initia ting and / or elonga ting nascent DNA.Increasing the Ec-SSB concentration resulted in a corresponding increase in DNA synthesis, howe v er, the distribution of leading-and lagging-strand products was strikingly different than that observed with RPA.With EcSSB, we consistently observed a much larger fraction of long products (Figure 5 B, lanes 5-6), suggesting that lagging-strand DNA synthesis is defecti v e.This could be due to fewer lagging-strand synthesis products or to longer Okazaki fragments that comigrate with the leading-strand products.In either scenario, RPA performs one or more essential functions at the eukaryotic replication fork that cannot be complemented by EcSSB.
The failure of EcSSB to support normal leading-and lagging-strand replication suggests that replication r equir es direct contacts between RPA and the replisome.Interestingl y, w hen we tested the RPA OB-F WH mutant that lacks both of RPA's protein-interaction domains but binds ssDNA with wild-type affinity (Figure 4 ), we observed results that were very similar to experiments with EcSSB (Figure 5 C).These results suggest that the OB-F and WH domains control important aspects of DNA replication.
To investigate this connection further, we generated deletion mutants lacking either the OB-F or the WH domain.Like RPA OB-F WH , these mutants both support robust origin unwinding at all concentrations tested (Supplemental Figure S3B, C).Strikingly, both RPA OB-F and RPA WH have strong replication defects when supplied at the same concentration as RPA (Figure 5 D, E, lanes 2 versus 4), showing that they each are critical for efficient replication initiation or elongation.Increasing the concentration of either mutant protein increased nucleotide incorporation but resulted in abnormal patterns of leading-and lagging-strand replication products.Higher concentrations of RPA OB-F led to an accumulation of short products comigr ating with lagging-str and products (Figure 5 D, lanes 4-6).In contrast, increasing amounts of RPA WH produced a preponderance of longer replication products that lacked the shorter Okazaki fragment lengths.This pattern was similar to that in reactions with EcSSB or with RPA OB-F WH (Figure 5 E, lanes 4-6).These results show that the two RPA pr otein-interaction domains contr ol different pr operties of DNA replication initiation or elongation.
Defects in priming would more strikingly affect laggingstrand synthesis because of the need to constantly re-initiate DNA synthesis on this strand.Because RPA affects multiple Pol-␣/ primase activities, we considered that altering RPA-Pol-␣/ primase interactions may explain the defects in replication product size distribution.To test this, we first increased the concentration of Pol-␣/ primase in the reconstituted replication assay.Because reactions with the RPA mutants had low amounts of DNA synthesis in our standard assay conditions (Figure 5 D-E, lane 2 versus 4-6), we used a higher concentration of RPA (150nM) in these experiments.
Increasing Pol-␣/ primase concentration resulted in distinct changes in replication products for the different RPAs.In reactions with wild-type RPA, additional Pol-␣/ primase resulted in a higher prevalence of short products relati v e to the long leading-strand products.Additionally, the laggingstrand products were slightly shorter (Figure 6 A, lanes 3-4, and Supplemental Figure S 4).This trend resembles the defect we have observed with the RPA OB-F mutant, with an increase in short products at the expense of the longer  products.Inter estingly, in r eactions with RPA OB-F , increasing Pol-␣/ primase also shortened the length of the lagging-strand products but did not strikingly alter the leading-and lagging-strand distribution (Figure 6 A, lanes 5-6 and Supplemental Figure S4).This finding is consistent with negati v e role for the OB-F domain in Pol-␣/ primase function.In contrast, reactions with RPA WH showed partial restoration of short lagging-strand products without any major changes in the longer leading-strand products (Figure 6 A, lanes 7-8), a result that is consistent with a positi v e regulatory role of this domain.Together these results are consistent with the RPA OB-F and WH domains negati v ely and positi v ely regulating the priming of DNA synthesis, respecti v ely.
Next, we asked whether the RPA mutants affect Pol-␣/ primase binding to ssDNA.To this end, we measured Pol-␣/ primase binding on naked ssDNA or ssDNA preincubated with an equimolar amount of RPA using fluorescence polarization (Figure 6 B).By using a longer oligonucleotide (6FAM-dT60), we can observe binding of RPA and Pol-␣/ primase on the same ssDNA molecule.When we titrated Pol-␣/ primase, we observed a concentrationdependent increase in anisotropy.The maximum anisotropy was notably higher on naked DNA than on RPA / ssDNA complex es (Figur e 6 B), consistent with RPA r educing Pol-␣/ primase association with the oligo-dT60 template.When we pre-incubated ssDNA with RPA WH , there was no significant difference from pre-incubation with WT RPA.Howe v er, w hen ssDN A was pre-incubated with RPA OB-F , Pol-␣/ primase binding matched that of the naked DNA (Figure 6 B).These results show that the removal of the OB-F domain allows more Pol-␣/ primase to bind to an ssDNA templa te coa ted with RPA, supporting a nega ti v e regulatory role of this domain.This observation is consistent with our observation of increased lagging-strand synthesis in replication reactions with RPA OB-F (Figures 5 D and 6 A).Overall, our RPA mutants have revealed two distinct activities that are crucial for appropriate replication fork assembly or function.

DISCUSSION
By substituting other SSBs or mutant versions of RPA into differ ent r eplication assays, we have gained important insights into the function of RPA during eukaryotic replication initiation and elongation.We found that RPA r equir es two protein-interaction domains, OB-F and WH, to coordinate DNA synthesis at the eukaryotic replication fork.Interestingly, these domains are dispensable for origin unwinding.Instead, we find that a particular arrangement of ssDNA binding domains found in RPA and EcSSB, but not Gp32, is r equir ed for this e v ent.

RPA function during helicase activation and unwinding
Pre vious wor k has shown e vidence of CMG acti vation and translocation on ssDNA in the absence of RPA ( 12 ).These data suggest that the purpose of RPA is to simply keep the str ands separ ated after CMG passes.Howe v er, our results indica te tha t pre v enting strand re-annealing is not the sole purpose of RPA during origin unwinding (Figures 2 -4 ).
We find that, e v en at v ery high concentrations, Gp32 does not support e xtensi v e DNA unwinding in our assays.Additionally, Gp32 fails to support any DNA synthesis (Supplemental Figure S3), suggesting that the DNA is insufficiently unwound to support replisome assembly.This is despite the fact that Gp32 participates in rapid cooperati v e ss-DNA binding ( 64 ) with a slow of f-ra te.Importantly, singlemolecule evidence indicates that Gp32 binding pre v ents reannealing of ssDNA ( 65 ), and Gp32 is known to stabilize ssDNA regions to allow primer binding in plasmid DNA ( 66 ).In addition, the RPA ABAB mutant has the same ss-DNA binding affinity as RPA, yet fails to promote e xtensi v e origin DNA unwinding, further supporting our conclusion that RPA contributes to more than pre v enting strand reannealing during the transition to e xtensi v e origin DNA unwinding.Based on these findings, we argue that RPA must improve the efficiency of CMG activation or translocation on ssDNA.
Ther e ar e se v eral features of the three proteins that stimulated e xtensi v e origin unwinding (RPA, EcSSB and RPA OB-F WH ) that are not shared by the small monomeric Gp32 or the RPA mutants tested.First, multiple dynamic ssDNA-binding modes contribute to RPA ( 23 , 25 ) and Ec-SSB ( 36 ) function.These multiple ssDNA-binding modes are important for binding various DNA structures, such as single-strand gaps, bubbles and ss / dsDNA junctions ( 67 ).Second, not only are RPA and EcSSB larger than Gp32 by molecular weight, their structures also contain clustered OB-fold domains (Figure 1 A-C).Finally, structural evidence shows that both RPA and EcSSB can induce ssDNAbending (Figure 1 A and C, ( 27 , 29 , 33 )).The single binding domain of Gp32 cannot reproduce the ssDNA-binding dynamics , clustered OB-folds , or ssDNA bending exhibited by EcSSB and RPA.
A combination of these three properties could contribute to two important r equir ements for origin activation.First, RPA ssDNA-binding dynamics may improve the efficiency of CMG activation by supporting the DNA remodeling steps r equir ed as the helicase transitions from encircling ds-DN A to encircling ssDN A. For example, RPA binding to partially extruded DNA could facilitate strand extrusion (Supplemental Figure S5B).Secondly, because the Mcm2-7 helicases are loaded in a head-to-head conformation ( 2 , 4 ), they must pass each other on opposite strands of DNA befor e unwinding DNA bidir ectionally from the replication origin.RPA could promote this process by eliminating steric barriers, through its physical size, effects on ss-DNA conformation, or both (Supplemental Figure S5C).Although CMG translocation on ssDNA is detectable in the absence of RPA ( 12 ), the efficiency of this transition is low.We propose that RPA increases the efficiency of this transition as is observed in DNA unwinding assays (Figure 2 and ( 12)).
Finally, RPA binding to the ssDNA strand excluded from CMG could stimulate helicase processivity in a way that promotes e xtensi v e origin unwinding.Although RPA modulates the activity of other helicases through direct interactions ( 62 , 68 ), this has not been observed for the CMG helicase.Indeed, the ability of EcSSB and RPA OB-F WH to substitute for RPA indica tes tha t direct pr otein-pr otein interactions between RPA and the helicase are not r equir ed.Instead, we consider that RPA stimulates origin unwind-ing by interacting with the excluded strand during CMG translocation (Supplemental Figure S5D).Structural studies of the CMG complex found that the excluded strand interacts with the helicase central pore (69)(70)(71).Recent evidence suggests this interaction leads to CMG stalling, and RPA binding to the excluded strand stimulates the rate of the CMG helicase ( 19 ).Similarly, studies of the E. coli replicati v e helicase, DnaB, that showed a ppl ying tension to the excluded strand, but not the encircled strand, stimulates helicase activity ( 72 ).In contrast, a similar study using the T4 bacteriophage replicati v e helicase, Gp41, demonstrated that tension on the excluded strand inhibited DNA unwinding ( 73 ).If this type of tension on the excluded strand is important for CMG stimulation, these data would explain the ability of EcSSB, but not T4 Gp32 to function in the unwinding assay.

Replication r equir es the RPA OB-F and WH domains
Specific RPA interactions are required at the eukaryotic replica tion fork.W hen substituted for RPA, both EcSSB and RPA OB-F WH display ed dr amaticall y decreased DN A synthesis in reconstituted replication assays, suggesting defects in nascent strand initiation or elongation (Figure 5 B, C and Supplemental Figure S3D).At higher EcSSB or RPA OB-F WH concentrations, DNA synthesis increased, but the products displayed altered distributions of leadingand / or lagging-strand products.Thus, in addition to overall DNA synthesis defects, our results show that specific interactions between RPA and the eukaryotic replication machinery are required for appropriate replication fork function.This is consistent with previous observations in SV40 r eplication, wher e neither EcSSB nor yeast RPA can efficiently substitute for human RPA ( 42 , 44 ).Similarly, the bacterial r eplisome r equir es specific interactions with the EcSSB C-terminal tails ( 74 ).New structural data of the eukaryotic replisome shows Pol-␣/ primase directly contacts the CMG helicase ( 75 ).These replisome structures were assembled without RPA, raising the possibility that RPA facilita tes Pol-␣/ primase incorpora tion into the replisome.Such a function could explain the reduced DNA synthesis in replication reactions with EcSSB or the RPA proteininteraction domain mutants.
A pre vious study observ ed that EcSSB supports only Pol ε -dependent rolling-circle replication products in reconstituted yeast replication assays on circular DNA templates ( 76 ).Thus, Pol ε synthesis may explain the longer replication products and absence of short characteristic Okazaki fragments in reactions with EcSSB or RPA OB-F, WH on our linear templates.These findings suggest that the OB-F and WH domains likely regulate Pol-␣/ primase and / or Pol ␦ activity at the replication for k.An alternati v e e xplanation is that the long products observed consist of long Okazaki fragments, possibly due to decreased priming on the lagging strand.
Consistent with the Rfa2 WH domain r equir ement in nascent strand initiation or Pol-␣/ primase recruitment to the r eplisome, incr easing Pol-␣/ primase concentrations partially r estor ed the synthesis of short replication products (Figure 6 A).This result suggests that the Rfa2 WH domain facilitates Pol-␣/ primase binding or activity on ssDNA.Consistent with this hypothesis, previous studies found that deletion of the Rfa2-WH domain reduced RPA interaction with Pol-␣/ primase ( 77 ).We can also gain insight from structures of the conserved RPA-like CST complex with Pol-␣/ primase in its acti v e conformation.These structures re v eal direct contacts between Pol-␣/ primase and the Stn1 winged-helix domains, which are analogous to but di v ergent from the Rfa2-WH domain, and these contacts are proposed to promote Pol-␣/ primase activity and processivity ( 78 , 79 ).
Our RPA OB-F results implicate an important role for the Rfa1 OB-F domain in regulation of lagging-strand synthesis.We observed tha t replica tion assays with this mutant had weak leading-strand synthesis and accumulated Okazaki fragments that were shorter than those in control r eactions (Figur e 5 D and Supplemental Figure S3D).Inter estingly, incr easing Pol-␣/ primase concentrations in reactions with WT RPA display ed a similar ly skewed leading:la gging product distrib ution with shorter Okazaki fra gments (Figure 6 A).This suggests that excess Pol-␣/ primase acti vity could e xplain the results we hav e observ ed.In support of this hypothesis, we observed that RPA OB-F lacked the ability to limit Pol-␣/ primase ssDNA binding that we observed with full-length RPA (Figure 6 B).Such a negati v e regulatory function for RPA could play a critical role in determining the frequency of lagging-strand priming gi v en the recent finding suggesting that Pol-␣/ primase associates continuously with the replisome ( 75 ).
It is also possible that this negati v e regulation of Pol-␣/ primase is necessary for polymerase switching.Failure of this activity could prevent or delay handoff from Pol-␣/ primase to the more processi v e Pol ␦ and / or Pol ε , affecting both leading and lagging synthesis.Indeed, replication reactions with RPA OB-F do not recover leading-strand synthesis, e v en in the presence of excess Pol-␣/ primase (Figure 6 A).Similarly, SV40 replication assays performed without the Pol ␦ processivity clamp PCNA display only short replication products ( 59 ), suggesting that accumulation of short products may be due to failure of polymerase switching.Indeed, genetic data supports a role of the OB-F domain in regulating not only Pol-␣/ primase but also Pol ␦.Mutations in the Rfa1 OB-F domain are synthetic lethal with mutant alleles of P ol-␣/ primase, P ol ␦ and RFC ( 80 , 81 ).
Overall, our results re v eal that the Rfa1 OB-F and Rfa2 WH domains each regulate a distinct step of nascent strand initiation or elongation.A function of the WH domain in stimulating priming activity and OB-F domain in negati v ely regulating this e v ent fits with our analysis of the double and single mutants.We found that the WH deletion mimics the double m utant, w hereas the OB-F m utant shows a distinct phenotype, consistent with the WH domain stimulating an activity that is r equir ed befor e the activity regulated by the OB-F domain.Future studies in volving assa ys f or specific steps of replication initiation will be r equir ed to determine the precise mechanism by which the OB-F and WH domains regulate replication fork function.

DA T A A V AILABILITY
The data underlying this article will be shared on reasonable request to the corresponding author.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.

Figure 2 .
Figure 2. EcSSB, but not Gp32, can support Mcm2-7 unwinding of origin DNA.( A ) Schematic of the origin DNA unwinding assay.( B ) Origin DNA unwinding assays with a 2-fold titration series of Gp32 in the absence of RPA. ( C ) Origin DNA unwinding assay substituting the indicated concentrations of EcSSB for RPA.In both origin unwinding experiments, 'no protein' controls (lanes 1-2) contain only plasmid + / -Topoisomerase I (Topo I) to show migration of supercoiled (sc) and relaxed (r) plasmids.

Figure 3 .
Figure 3. RPA subcomplexes containing two DNA-binding domains do not support origin DNA unwinding.( A ) Structures of RPA AB and RPA TriC subdomains (1JMC, 6I52, respecti v ely).( B ) Coomassie-stained SDS-PAGE gel of purified RPA, RPA AB and RPA Tri-C .( C ). Fluorescence polarization results of 2.5 nM 6-FAM oligo-dT30 incubated with indicated concentrations of RPA AB (cyan; K d,app ∼ 18.35 nM) or RPA Tri-C (green; K d,app ∼ 2.97 nM).The wild-type RPA results ( K d,app ∼ 1.7 nM) from Figur e 1 ar e included for comparison.( D ) Origin DNA unwinding assays with two-fold titration of RPA AB or RPA TriC (see also Figure 2 legend).

Figure 4 .
Figure 4.A high-affinity four DNA-binding domain RPA mutant is not sufficient for origin unwinding.( A ) Diagrams of RPA ABAB and RPA OB WH mutants.( B ) Coomassie-stained SDS-PAGE gel of purified wild-type and mutant RPAs.( C ) Fluorescence anisotropy results of 2.5 nM 6FAM-oligo-dT30 incubated with indicated concentrations of RPA ABAB ( K d,app ∼ 1.45 nM) or RPA OB-F WH ( K d,app ∼ 2.04 nM).The wild-type RPA results ( K d,app ∼ 1.7 nM) from Figure 1 are included for comparison.( D ) Origin DNA unwinding assa y with two-f old titration of RP A ABAB or RP A OB-F WH .

Figure 5 .
Figure 5. RPA OB-F and WH domains control replica tion elonga tion.( A ) Schema tic of the reconstituted DNA replication assay.(B-E) In vitr o replica tion with a two-fold titration series of ( B ) EcSSB, ( C ) RPA OB-F WH , ( D ) RPA OB-F or ( E ) RPA WH substituting for RPA.Locations of normal leading-and la gging-strand product distrib utions are labeled in (B).As with the origin DNA unwinding assay, replication is dependent on RPA (compare lanes 1 and 2) and DDK (compare lanes 3 and 4).

FFigure 6 .
Figure 6.Pol-␣/ primase mediates the defects of the OB-F and WH deletions.( A ) Titration of Pol-␣/ primase in in vitro replication assays containing 150 nM of the respecti v e RPA pr otein.( B ) Fluorescence anisotr opy of Pol-␣/ primase titration with a 2.5 nM 6FAM-oligo-dT60 pre-incubated with an equimolar amount of the indicated RPA. ( C ) Diagram of Pol-␣/ primase (purple) with CMG (brown) and RPA at a replication fork with OB-F and WH interactions.