Regulation of Replication Protein A Functions in DNA Mismatch Repair by Phosphorylation*

Replication protein A (RPA) is involved in multiple stages of DNA mismatch repair (MMR); however, the modulation of its functions between different stages is unknown. We show here that phosphorylation likely modulates RPA functions during MMR. Unphosphorylated RPA initially binds to nicked heteroduplex DNA to facilitate assembly of the MMR initiation complex. The unphosphorylated protein preferentially stimulates mismatch-provoked excision, possibly by cooperatively binding to the resultant single-stranded DNA gap. The DNA-bound RPA begins to be phosphorylated after extensive excision, resulting in severalfold reduction in the DNA binding affinity of RPA. Thus, during the phase of repair DNA synthesis, the phosphorylated RPA readily disassociates from DNA, making the DNA template available for DNA polymerase δ-catalyzed resynthesis. These observations support a model of how phosphorylation alters the DNA binding affinity of RPA to fulfill its differential requirement at the various stages of MMR.

Defects in DNA mismatch repair (MMR) 4 lead to a hypermutable phenotype and a predisposition to cancer (1)(2)(3), demonstrating the importance in maintaining genome stability. The genome maintenance functions of MMR include repair of DNA replication errors (1,3,4), suppression of DNA recombination between two divergent sequences (5)(6)(7), and participation in DNA damage signaling to trigger cell cycle arrests and/or apoptosis (8 -10). Among these functions, the correction of biosynthetic errors is best characterized.
The molecular mechanism by which MMR corrects biosynthetic errors is conserved from bacteria to human cells. Using the methyl-directed MMR system in Escherichia coli as a model, essential components required for eukaryotic MMR have been identified to include MutS␣ (a heterodimer of MSH2-MSH6), MutS␤ (a heterodimer of MSH2-MSH3), MutL␣ (a heterodimer of MLH1-PMS2 in mammalian or MLH1-PMS1 in yeast), proliferating cellular nuclear antigen (PCNA), replication protein A (RPA), EXO1, HMGB1, replication factor C (RFC), DNA polymerase ␦, and DNA ligase I (Refs. 4 and 11, and references therein). Recently, the human MMR reaction has been reconstituted using purified proteins (11,12). For 5Ј nick-directed MMR, formation of a complex between mismatched DNA and MMR proteins MutS␣ or MutS␤, MutL␣, EXO1, RPA, and HMGB1 initiates mismatch-provoked excision at the strand break (11). The excision reaction is terminated immediately after the removal of the mismatch in a manner dependent on MutL␣ and RPA (11). The resulting ssDNA gap is filled by DNA polymerase ␦, and the repair is completed by DNA ligase I (11). For 3Ј nick-directed MMR, the reaction additionally requires RFC and PCNA and is strongly enhanced by the addition of EXO1 (12).
In this study, we monitored the time-dependent association of several key MMR proteins in HeLa nuclear extracts with a biotin-streptavitin-bound mismatched DNA substrate, and demonstrate here that RPA, MSH2 (a subunit of human MutS heterodimers), MLH1 (a subunit of human MutL heterodimers), PCNA, and DNA polymerase ␦ bind to the heteroduplex in a sequential manner. Surprisingly, RPA binds to the DNA substrate at a time earlier than the known MMR initiation factors MutS␣, MutL␣, and PCNA and remains bound throughout the repair reaction. Additionally, we show that the functions of RPA in MMR are regulated by phosphorylation. Unphosphorylated RPA possesses a high DNA binding affinity and preferentially stimulates mismatchprovoked excision; phosphorylated RPA preferentially facilitates DNA resynthesis via reducing its DNA binding ability, allowing its displacement by DNA polymerases to make DNA template available for nucleotide polymerization.

EXPERIMENTAL PROCEDURES
Preparations of Nuclear Extracts and Proteins-HeLa S 3 cells were purchased from the National Cell Culture Center (Minneapolis, MN), and nuclear extracts were prepared as described (32). Three different forms of human RPA were used in this study, the wild-type RPA and its hyperphosphorylated (RPA 2D ) and non-phosphorylated (RPA 2A ) isoforms (33). These RPA isoforms were expressed in E. coli and purified essentially as described (34). Recombinant MutS␣, MutL␣, PCNA, EXO1, RFC, and pol ␦ were expressed and purified as described (11). DNA-PK was purified from HeLa nuclear extracts essentially as described (35).
DNA Substrate Preparation and MMR Assays-A 6.4-kb circular substrate containing either a G-T mismatch and a strand break 128 bp 5Ј to the mismatch or a 171-nt gap was constructed as described (36). The circular DNA substrates were digested with BspHI to produce linear duplexes with 5Ј overhang sequences of 5Ј-CATG, which served as a template for 3Ј biotin labeling (see Fig. 1) in the presence of Klenow DNA polymerase, dCTP, and biotin-dATP as described (37). An otherwise identical linear homoduplex was similarly prepared. The biotinylated DNA substrates were incubated with streptavidin-Sepharose at 4°C for 2 h. The resulting complex was used for MMR assay essentially as described (32) with minor modifications. Briefly, the repair assay was performed in a 45-l reaction containing 140 fmol of biotin-streptavidin-attached DNA heteroduplex and 250 g of HeLa nuclear extracts. The repair reactions were incubated at 37°C for an indicated time, on a rotating rack, and the DNA samples were recovered and digested with BspDI and HindIII to score for repair on agarose gel. To analyze mismatch-provoked excision intermediates, reactions were assembled identically to the MMR reaction but in the absence of dNTPs. After digestion with SspI, DNA excision products were fractionated through a denaturing 6% polyacrylamide gel. DNA samples were subjected to Southern blot analysis using a 32 P-labeled probe (5Ј-ATTGTTCTGGATAT-TACC-3Ј) as described (38).
Protein Pulldown Assay-To pull down proteins participating in MMR, 250 g of HeLa nuclear extracts were incubated with 500 or 140 fmol of biotinylated heteroduplex or homoduplex DNA substrates attached to streptavidin at 37°C for an indicated time under the repair conditions (32). Reactions were terminated by the addition of 800 l of a low salt washing buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% Triton X-100, and protease inhibitors (0.3 mg/ml benzamidine hydrochloride, 0.5 g/ml of pepstatin A, 0.5 g/ml of leupeptin, 0.5 g/ml of antipain)). The beads were recovered by centrifugation and washed once each with a high salt washing buffer (same as the low salt washing buffer but with 500 mM NaCl) and the low salt washing buffer to remove nonspecifically bound proteins. Proteins that remained on the beads were eluted with the SDS gel loading buffer and separated by 10% SDS-polyacrylamide gels, followed by Western blot analysis.
Mismatch-excision and DNA Gap-filling Assays-Excision assays were performed in 20-l reactions containing 24 fmol of the 5Ј G-T heteroduplex (see Fig. 1A, substrate I), 5 fmol of EXO1, 400 fmol of MutS␣, 260 fmol of MutL␣, 190 fmol of RFC, 290 fmol of homotrimer of PCNA, and 800 fmol of the indicated forms of RPA. The reactions were incubated at 37°C for 10 min as described previously (11). DNA samples were recovered by phenol extraction and ethanol precipitation. Excision was scored by NheI digestion. DNA gap-filling assay was performed by incubating purified proteins (pol ␦ and one of RPA isoforms) with a circular DNA duplex containing a 171-nt gap (see Fig. 1A, substrate III). After incubation for the indicated times, DNA samples were digested with SspI, fractionated through a 6% polyacrylamide gel, and subjected to Southern blot analysis using a 32 P-labeled probe (5Ј-AAAATTTA-ACGCGAATTTT-3Ј) as described (38).
Stopped-flow Assays-The 21-mer (5Ј-GCTGAAGCAGAA-GGCTTGCAA-3Ј) and its 5Ј-hexachlorofluorescein-labeled analog were synthesized by Gene Link (Hawthorne, NY) and purified as described (39). Concentrations were determined spectrophotometrically using extinction coefficients calculated according to Cantor et al. (40). Real-time fluorescence changes were measured using a KinTek SF 2001 stopped-flow spectrophotometer (KinTek Instruments, State College, PA) with upgraded Hamamatsu light source/monochromator fitted with a 150-watt xenon arc lamp. Reactions containing the indicated RPA and oligonucleotides were maintained at a constant 37°C with a Neslab RTE-111 refrigerated bath. Intrinsic protein fluorescence was excited at ex ϭ 290 nm while monitoring total emission at wavelengths Ͼ325 nm using an Oriel 51960 filter. Hexachlorofluorescein fluorescence was excited at ex ϭ 535 nm while monitoring total emission at wavelengths Ͼ560 nm using a Corion LG-560-F filter. Time courses were fitted by nonlinear least-square regression to a single of exponential function, y ϭ A 1 (1 Ϫ e Ϫktt ) ϩ C.

RESULTS
Dynamic Protein-DNA Association during the MMR Process-A biotinylated linear DNA heteroduplex containing a G-T mismatch and a nicked 128 base pairs 5Ј to the mismatch immobilized on streptavidin beads (Fig. 1A, substrate II ) was constructed to use in pulldown assays during MMR. In HeLa nuclear extracts, the linear heteroduplex was efficiently repaired ( Fig. 2A) with appearance of repair products after 6 min incubation at 37°C. Mismatch-provoked excision assay under the condition of limited DNA synthesis (38) showed little excision before and at 1 min (Fig. 2B), as almost all DNA molecules remained as nicked substrate (second band from top) and a small amount of directly ligated side product (top band). Onset of bands corresponding to smaller excision products indicated the initiation of mismatch-provoked excision at 2 min. By 4 min, extensive excision was observed. However, as little repair products were detected at 6 min in the presence of dNTPs ( Fig. 2A), DNA resynthesis past the mismatch required at least 6 min for the majority of molecules.
To correlate the repair progress with the repair proteins assembled on the heteroduplex substrate in real time, the resin-bound substrate was used to pull down proteins involved in MMR (Fig. 2F). To ensure recovery of sufficient proteins for analysis, 500 fmol of the streptavidin-bound heteroduplex (3.5fold more compared with the repair reaction) was incubated with 250 g of HeLa nuclear extract under the repair conditions. Proteins pulled down at various times were analyzed by Western blotting to profile proteins bound during the course of the MMR reaction (Fig. 2C). As expected, MMR initiation components, MSH2, MLH1, and PCNA, were detected 0.5 min after incubation at 37°C, the amounts of these proteins apparently increased, particularly after the mismatch had been removed (at 6 min). However, this increase in the level of MSH2 and MLH1 may not necessarily reflect their requirement for the late steps of the reaction. Instead, it is likely due to their association with unrepaired heteroduplexes, where these proteins could spread along the nick-ligated and ends-blocked molecules after loading to the mismatch. Due to the poor specificity of the commercial antibodies, EXO1 was not directly detectable; however, the appearance of pol ␦ binding at ϳ2 min ( Fig. 2C) implied that excision must have occurred at or prior to 2 min, consistent with the excision profile shown in Fig. 2B.
When antibodies against the 70-(RPA1) and 34-kDa (RPA2) subunits of RPA were used to monitor the status of RPA during MMR, two surprising phenomena were observed: (i) RPA bound to the DNA substrate earlier than MSH2, MLH1, and PCNA ( Fig. 2C), suggesting involvement of RPA in the initiation step of MMR; (ii) the RPA2 antibody detected two species (a fastmigrating and a slow-migrating) of RPA2, with the fast migrating species appearing at the earlier steps of the reaction and the slow migrating one covering the later steps of the reaction (Fig. 2C). The two RPA2 species are reminiscent of phosphorylated (slowmigrating) and unphosphorylated (fast-migrating) RPA2 observed during cellular response to DNA damage in published reports (41,42). This was confirmed in experiments supplemented with 4 M wortmannin, a potent inhibitor of PI 3-kinases, where the fast migrating species was no longer detectable (data not shown), suggesting strongly that the slow migrating RPA2 was likely formed from the fast migrating species via phosphorylation by PI 3-kinases (43)(44)(45). Comparison with the excision profile ( Fig. 2B) revealed that unphosphorylated RPA, appearing from 0 to 6 min ( Fig. 2C), was associated with excision initiation and complete removal of the mismatch (from 0 to 6 min, see Fig. 2B), whereas phosphorylated RPA appeared just before the binding of pol ␦ to the DNA substrate and accumulated throughout the resynthesis phase of the reaction (Fig. 2C). These results, therefore, suggest that unphosphorylated RPA is required for the excision step, whereas phosphorylated RPA is necessary for the resynthesis step of MMR.
In control experiments performed with an otherwise identical, nicked DNA substrate without the mismatch, little MSH2 and MLH1 binding to the homoduplex substrate was detected initially as expected (MLH1 was detected only at a much later reaction time, Fig. 2D). The binding profile of other protein components to the nicked homoduplex also appeared similar. However, RPA2 phosphorylation in the homoduplex reaction appeared much faster as complete phosphorylation was observed at 4 min as compared with the 10 min required for the heteroduplex reaction (Fig. 2, C and D). These results are repeatable and likely reflect the known differences in the length of the ssDNA gap generated in these two reactions (see "Discussion" for details). Significantly, in either case, RPA2 phosphorylation consistently coincided with or slightly preceded pol ␦ binding, suggesting again that RPA2 phosphorylation may play a critical role in triggering DNA synthesis, even in the homoduplex reaction where a significant amount of DNA synthesis unrelated to MMR has been reported previously (46,47). Additionally, although RPA (RPA2) was readily pulled down by both the nicked hetero-and homo-duplexes, identical constructs lacking the nick were poor substrates for RPA binding (Fig. 2E). The requirement for a nick in the substrate implies that RPA likely binds to the DNA substrate at or via the strand break.
Unphosphorylated RPA Promotes the Excision Reaction, but Phosphorylated RPA Facilitates the Resynthesis in MMR-The human MMR system has recently been reconstituted using purified proteins, and it can be subdivided into the excision and resynthesis reactions (11). To determine the role of RPA phosphorylation in MMR, two RPA isoforms that functionally mimic phosphorylated and unphosphorylated RPA (48) were used in the reconstituted excision and resynthesis reactions. These RPA isoforms were obtained by substituting the phosphorylation sites of wild-type RPA2 with aspartate (RPA2 D ) and alanine (RPA2 A ), respectively (33), and the individual RPA trimers containing RPA2 D and RPA2 A were referred to as RPA 2D or RPA 2A here, respectively. The mobility of RPA2 A and RPA2 D in SDS-PAGE correlated well with that of unphos-phorylated and hyperphosphorylated RPA2, respectively (Fig. 1B). Fig.  3A shows the excision assay in the reconstituted system in the presence of these different RPA isoforms. Relative to the reaction with the phosphorylated isoform, RPA 2D , 2.2-and 1.7-fold higher excision activities were observed in reactions using unphosphorylated wild-type RPA (RPA WT ) and the unphosphorylated isoform, RPA 2A . These results suggest a preferential role for the unphosphorylated form of RPA in stimulating the mismatch-provoked excision phase of the reaction.
To explore the role of RPA phosphorylation in the resynthesis step of MMR, gap-filling activity was assayed in a defined system containing purified pol ␦ and either phosphorylated or unphosphorylated RPA using a circular DNA substrate with a 171-nt single-stranded gap (Fig. 1A, III ) to mimic the MMR excision product. Resynthesis products were detected by Southern blot analysis (Fig. 3B) and the percentage of resynthesis products are plotted as a function of time in Fig. 3C. As previously observed, pol ␦ by itself can efficiently fill the 171-nt gap (11). The addition of RPA WT to the reaction resulted in a severe, greater than 4-fold reduction in the rate of gap-filling (Fig. 3C). Gap filling activity was also similarly inhibited in reactions supplemented with RPA 2A . However, a greater than 2-fold recovery of gap filling activity was observed when RPA 2D was substituted for RPA WT or RPA 2A , which partially mitigated but did not completely abolish the overall inhibitory effect of adding RPA (Fig. 3, B and C). These results suggest a preferential role for phosphorylated RPA during DNA resynthesis.
To demonstrate that the faster gap filling observed with RPA 2D was due to RPA phosphorylation, gap filling assay was conducted using RPA WT in the presence or absence of purified DNA-PK, which has been shown to be capable of phosphorylating RPA in vitro (17,26). Consistent with previous observations, we found that DNA-PK purified from HeLa cells (see Fig.  1B) could efficiently phosphorylate recombinant RPA (data not shown), although we do not know whether or not RPA was hyperphosphorylated in this manner. As shown in Fig. 3D, addition of DNA-PK to the reaction containing pol ␦ and RPA WT stimulated the gap filling activity by more than 1.7-fold, and the stimulation by DNA-PK was abolished in the presence of wortmannin, a potent inhibitor of PI 3-protein kinases Excision assays were performed essentially the same as the repair assay except no dNTPs. After digestion with SspI, the DNA samples were fractionated through a denaturing 6% polyacrylamide gel, and subjected to Southern blot analysis. Schematic representation of the SspI fragment is shown on the right side of the gel. The position of the mismatch is indicated by an asterisk, and the solid bar indicates where the 32 P-labeled probe anneals. C and D, detection of proteins bound to the nick-containing G-T and A-T substrates, respectively. 250 g of HeLa nuclear extracts and 500 fmol of DNA substrates attached to streptavidin beads were incubated at 37°C for the indicated times under the repair conditions. Proteins bound to DNA substrates were fractionated through SDS-PAGE, followed by Western blot analysis using antibodies against the indicated proteins. E, RPA2 pulldown using nicked and non-nicked duplexes. The non-nicked substrates were prepared by ligation of the corresponding nicked substrates as described (38). Ku80 served as a pulldown control. F, diagram of the protein pulldown experiments.
including DNA-PK (49). These observations provide definitive evidence that RPA phosphorylation facilitates repair DNA synthesis. Taken together, our results demonstrate that while unphosphorylated RPA is preferred for the excision phase of MMR, its subsequent phosphorylation facilitates DNA gap filling by pol ␦.
Surprisingly, RPA, regardless of its phosphorylation status, was observed to inhibit pol ␦-catalyzed DNA gap-filling, a result that may seem refractory to the observed requirement of RPA for DNA resynthesis in a cell-free extract MMR system (16). However, we must also keep in mind the indispensable function of RPA in protecting the ssDNA template from degradation by nucleases that are present in vivo and in extracts but are absent in the purified system. Therefore, the slightly inhibitory effect of RPA on DNA synthesis may be a small but requisite price when balanced against its in vivo value for protecting the integrity of the template genetic information for resynthesis.
Phosphorylation Reduces the ssDNA Binding Ability of RPA-RPA is required to protect the ssDNA during excision (16), but for DNA re-synthesis it must disassociate from the singlestranded template. To determine whether RPA phosphorylation fulfills this transition, competition experiments were per-formed to examine the DNA binding capability of RPA 2D , RPA 2A , and RPA WT . First, equimolar mixtures of RPA 2D with RPA 2A and RPA 2D with RPA WT were incubated with increasing amounts of the biotinylated streptavidin-immobilized ssDNA beads, and the bound proteins were recovered and subjected to Western blot analysis. As shown in Fig. 4A, whereas significant amounts of RPA 2A and RPA WT were detected regardless of DNA concentrations, RPA 2D was only found in reactions containing a higher DNA concentration, indicating that RPA 2D possesses a lower DNA binding affinity than RPA 2A and RPA WT .
We then performed competition assays using a 171-nt gapped circular DNA substrate (Fig. 1A, substrate IV), which mimicked the excision product generated during mismatchprovoked excision. In this assay, one form of RPA was initially allowed to interact with the gapped DNA substrate, and the preformed complex was challenged with varying concentrations of a different form of RPA. The proteins bound to the DNA substrate were analyzed by Western blot. As expected, although RPA 2D alone could bind to the DNA substrate, the bound RPA 2D was readily displaced by RPA WT (Fig. 4B, top  panel) or RPA 2A (Fig. 4B, middle panel) at concentrations higher than the initial RPA 2D concentration. In contrast, RPA 2A prebound to the DNA substrate could not be displaced by RPA 2D regardless of the concentration used (Fig. 4B, bottom  panel). Interestingly, RPA 2D appeared to be binding to the DNA in addition to the prebound RPA 2A , and an increase in binding was associated with increasing RPA 2D concentrations. Because there was no detectable decrease in RPA 2A binding, the majority of which presumably stayed with the ssDNA area, the observed RPA 2D binding is likely in the double-stranded areas that were not occupied by RPA 2A . Similar results were observed for prebound RPA WT -DNA (data not shown). These results suggest that RPA 2D possesses a lower DNA binding affinity than RPA 2A and RPA WT for the single-stranded gap, implying that phosphorylation reduces the DNA binding affinity of RPA. The stopped-flow results are consistent with the single-step binding mechanism previously reported on wild-type RPA (50).
Under the conditions of excess 21-mer, the binding reaction can be modeled as a pseudo-first ordered reaction, where R, D, and RD denote RPA, 21-mer, and RPA-21-mer complex, respectively, where the observed rate constant, k obs , can be calculated as k obs ϭ k on [D] ϩ k off (50 -52). Therefore, the slopes of the line in Fig. 5B defined the second order rate constants, k on value, for initial binding of DNA by the different forms of RPA.
These association experiments, however, failed to yield accurate values for k off as indicated by the near zero values of the y intercepts; therefore, separate experiments were conducted to directly measure the dissociation of DNA from preformed protein-DNA complexes. Using a 5Ј-hexachlorofluorescein-labeled 21-mer (40 nM), fluorescent complexes were preformed with RPA (40 nM) and chased at time 0 with 600 nM nonfluorescent 21-mer. Fig. 5C shows single-exponential loss of fluorescence for each of the RPA-DNA complexes following challenge with the unlabeled 21-mer with apparent dissociation rate constants of 0.107 Ϯ 0.002, 0.072 Ϯ 0.001, and 0.377 Ϯ 0.007 s Ϫ1 for RPA WT , RPA 2A , and RPA 2D , respectively. Control experiments (data not shown) using 1000 nM nonfluorescent 21-mer as trap showed identical results, indicating that the trap used in these experiments were both sufficient and non-interfering. Additionally, binding experiments using the fluorescently labeled 21-mer showed identical rates of association to RPA as the unlabeled 21-mer (data not shown), indicating that the presence of the fluorophore did not alter the binding characteristics of the DNA. Table 1 summarizes the quantitative binding parameters for RPA WT , RPA 2A , and RPA 2D with single-stranded DNA. Whereas RPA 2A showed binding kinetics similar to wild-type, RPA 2D , the phosphorylation isoform, showed a modest 1.6-and 1.8-fold decrease in k on from RPA wt and RPA 2A , respectively. More strikingly, RPA 2D showed a 3.5-and 5.2-fold increase in k off relative to RPA WT and RPA 2A , respectively. These k on and k off values couple to produce a significant 6.2-8.1-fold increase in K D .

DISCUSSION
Role of RPA in Nick Sensing-RPA has been widely implicated in all DNA metabolic pathways including replication, recombination, repair, and DNA damage response (17,18,22,53). However, exactly how RPA can differentially function in so many different reactions remains poorly understood. Because of the high affinity and specificity of RPA for binding to ssDNA, it is generally assumed that ssDNA species produced during DNA metabolism provide the major target for RPA binding and participation in these reactions. In this study, we show that RPA binds to DNA substrates earlier than the known MMR initiation components that are required to generate ssDNA regions (Fig. 2C), indicating that the binding of RPA to DNA in this case occurs prior to the formation of ssDNA. Instead, we find that RPA binding required only the presence of a nick in the DNA substrate (Fig. 2E), suggesting that RPA may act as a sensor for DNA strand breaks. Consistent with this notion, RPA has been regarded as an initiator for DNA damage response and doublestrand break repair (22). For example, cellular response to UV irradiation requires ATR to activate Chk1, but RPA is essential for localizing ATR to the damage site through an interaction with ATRIP, the interacting partner of ATR (54). Recent studies have shown that RPA co-localizes with the double-strand break binding factor ␥-H2AX (55) and ATR (56) at the sites of DNA damage immediately following ionizing radiation and camptothecin treatment, respectively, further supporting the role of RPA as a nick-sensor. This proposed role for RPA would provide a novel model for explaining how RPA may gain entry into such a wide variety of different DNA metabolic transactions, as strand breaks are common early intermediates in DNA replication, recombination, and individual repair pathways.
Role of RPA in MMR Initiation-RPA is an essential component for MMR (13,16). Our demonstration that RPA binds to a DNA substrate earlier than MutS␣ and MutL␣ (Fig. 2C) also suggests that RPA plays an important role in MMR initiation. As DNA binding by RPA requires a strand break, the site where in vitro MMR reaction starts, RPA likely plays a critical role in facilitating the formation of the initiation complex at the site of the strand break.
Such a model readily encompasses any of the three current models for MMR initiation. For the translocation model (57) and the sliding clamp model (58 -60), the binding of RPA at the strand break may function as a translocation termination signal by physically blocking MutS␣ migration, so that a repair initiation complex could be formed at the site. In the alternate DNAbending-mismatch verification model (61), where the MutS protein stays at the mismatch and interacts with the MutL proteins to bring together the mismatch and the strand break via DNA loop formation (62), RPA bound at the nick could serve as a beacon marking the position of the nick for loop formation. Therefore, our novel observation that RPA remains bound to the DNA substrate throughout the MMR reaction implicates involvement of RPA in all stages of the MMR reaction.
Role of RPA Phosphorylation in MMR-It is well established that RPA plays roles in each stage of the MMR reaction, including at least stimulating mismatch-provoked excision, protecting the ssDNA region produced during the excision, and facilitating repair DNA synthesis (11,15,16). However, the continued involvement of RPA in the repair reaction itself poses a mechanistic conundrum. Given the dynamic changes in DNA structures, reaction types, and protein components involved in each step of MMR, the continued binding of RPA to the DNA substrate would require at least 3 different binding modes. At the start of the reaction, a tight RPA-DNA interaction at a nick seems appropriate for the MMR machinery to recruit repair components at the nick position. During the subsequent excision phase, tight and perhaps cooperative binding to ssDNA becomes critical to the function of RPA in protecting the ssDNA gap (16) and displacing other MMR proteins, e.g. MutS␣ (11,63). However, during the resynthesis phase, an asyet-uncharacterized loose form of interaction between RPA and DNA would appear to be necessary to expose the gapped ssDNA for use as the resynthesis template. Thus, understanding the regulation and modulation of the RPA-DNA interaction is mechanistically essential for understanding MMR.
In this study, we provide direct evidence that RPA phosphorylation fulfills this critical mechanistic role in modulating the  RPA-DNA interaction affinity. At the beginning of the reaction, no phosphorylated RPA was detectable; during the period in which excision and resynthesis occur simultaneously, both the phosphorylated and unphosphorylated RPAs were observed; and at the end of the reaction, all participating RPAs were phosphorylated (see Fig. 2C). These observations indicate that unphosphorylated RPA functions during the initiation and the excision steps while phosphorylation is required for resynthesis. Indeed, our dissection experiments in a defined system show that unphosphorylated RPA preferentially facilitates the excision reaction, and that phosphorylated RPA preferentially promotes the resynthesis reaction (Fig. 3). Considering the different requirements for the RPA-DNA interaction before and during resynthesis discussed above, we expect that unphosphorylated RPA would possess a stronger ssDNA binding affinity than phosphorylated RPA.  (65) have recently reported that both phosphorylated and unphosphorylated RPA bind with equal affinity to single-stranded dT 30 , whereas hyperphosphorylated RPA binds only slightly less efficiently than unphosphorylated RPA to a 93.3% purine oligonucleotide. However, the solution conditions differ between our studies and theirs; most notably, their dT 30 data were collected at 1 M NaCl, and all of our reactions were conducted in buffers with 110 mM KCl, a condition close to physiological. Additionally, we observed the same differential binding of phosphorylated and unphosphorylated RPA to 3 different DNA sequences: the 21-mer mixed sequence oligonucleotide in the stopped-flow studies (67% purine), the circular gapped DNA substrates in the competition experiments (54.1% pyrimidine in the gap region), and the PCR-generated DNA substrate (56.7% pyrimidine). With all these DNA substrates, we show that unphosphorylated RPA possesses a moderately stronger binding affinity than its phosphorylated form. We are therefore confident that our results are representative of the physiologically relevant random sequence encountered in the single-stranded gap generated during excision. The discrepancies with Patrick et al. (65) may thus reflect either differences in buffer conditions and/or peculiarities of the specialized DNA sequences used in that study.
Although phosphorylation plays an important role in modulating the function of RPA in MMR, what triggers RPA phosphorylation in the repair process is unclear. Previous studies in both in vivo and in vitro DNA replication have established that phosphorylation of RPA2 is associated with S phase and promoted by the generation of ssDNA (66,67). Consistent with previous observations in DNA replication, our data shown in Fig. 2 appear to suggest that RPA2 phosphorylation in MMR is also triggered by the generation of ssDNA, most likely following the binding of RPA to the ssDNA gap. Because both DNA replication and the MMR resynthesis involve DNA polymerases, it is also possible that interactions between DNA polymerases and the DNA-bound RPA activate a PI 3-kinase, which in turn phosphorylates the DNA-bound RPA. However, based on the fact that RPA2 phosphorylation seemed to occur slightly earlier than the recovery of pol ␦ on the DNA substrate (Fig. 2C), RPA phosphorylation in MMR is most likely triggered by the binding of RPA to the ssDNA gap just before pol ␦ is recruited to the site for resynthesis. This model is also in agreement with RPA phosphorylation at the early phase of cellular response to DNA damage, during which only ssDNA, but not DNA polymerases, is available (22,54).
Taken together, our results described here support a model for modulation of RPA functions in MMR by phosphorylation (Fig. 6). In this model, unphosphorylated RPA binds initially to heteroduplex DNA substrates at the strand break to facilitate the initiation of MMR. During the excision phase of MMR, unphosphorylated RPA binds to and "coats" the exposed ssDNA region to protect this fragile intermediate from nuclease damage. As repair proceeds toward the resynthesis phase, however, the RPA-coated single-stranded gap must become available for DNA polymerase to use as the repair template. This is accomplished via the phosphorylation of the RPA2 subunit, which likely occurs following the binding of the protein to the ssDNA region and results in reduction of RPA DNA binding affinity. As DNA resynthesis proceeds, the phosphorylated RPA becomes released from the DNA or loosely attached to the double-stranded resynthesis product. Although the molecular details for RPA FIGURE 6. Model for regulation of RPA functions in MMR. RPA1, RPA2, and RPA3 are depicted as an oval, square, and circle, respectively. Phosphorylated or unphosphorylated RPA2 is indicated by black or white squares, respectively. Excision and polymerization machineries are depicted as broad arrows. Newly synthesized DNA is depicted by a dotted line. See text for full description of the model. release remains unknown, a possible model involves the transient displacement of a single phosphorylated RPA trimer by the DNA polymerase at the immediate site of DNA synthesis. This model seems to be suitable for DNA replication, as RPA phosphorylation is required for the process (43,68). Therefore, RPA functions in DNA metabolic pathways are likely modulated via the same mechanism, i.e. phosphorylation.
Lastly, we return to the issue of RPA phosphorylation in the homoduplex reactions (Fig. 2D), which occurred much faster in the homoduplex relative to the heteroduplex reaction (see Fig. 2, C and D). This may be correlated to the different lengths of the ssDNA fragments generated in these two reactions. Because of the coupling of a highly processive excision reaction with a moderately processive resynthesis reaction in MMR (11), an extensive ssDNA region is produced, which would require extensive protection by unphosphorylated RPA. Thus, it becomes essential to maintain a balance or "trade-off " between phosphorylated and unphosphorylated forms of RPA to support both resynthesis and excision reactions, which occur simultaneously at a point for 5Ј nick-directed MMR. However, in the case of the homoduplex reaction, excision is likely to be followed immediately by DNA resynthesis, resulting in a nick translation-like overall reaction that generates a very short ssDNA gap. Because of the limited size of this gap, extensive protection afforded by unphosphorylated RPA is no longer required, making this nick translation-like resynthesis better facilitated by phosphorylated RPA. However, the exact mechanism underlying the discrepancy requires thorough investigations.