PCNA is efficiently loaded on the DNA recombination intermediate to modulate polymerase δ, η, and ζ activities

Abstract

Proliferating cell nuclear antigen (PCNA) is required for DNA homologous recombination (HR), but its exact role is unclear. Here, we investigated the loading of PCNA onto a synthetic D-loop (DL) intermediate of HR and the functional interactions of PCNA with Rad51 recombinase and DNA polymerase (Pol) δ, Pol η, and Pol ζ. PCNA was loaded onto the synthetic DL as efficiently as it was loaded onto a primed DNA substrate. Efficient PCNA loading requires Replication Protein A, which is associated with the displaced ssDNA loop and provides a binding site for the clamp-loader Replication Factor C. Loaded PCNA greatly stimulates DNA synthesis by Pol δ within the DL but does not affect primer recognition by Pol δ. This suggests that the essential role of PCNA in HR is not recruitment of Pol δ to the DL but stimulation of Pol δ to displace a DNA strand during DL extension. Both Pol η and Pol ζ extended the DL more efficiently than Pol δ in the absence of PCNA, but little or no stimulation was observed in the presence of PCNA. Finally, Rad51 inhibited both the loading of PCNA onto the DL and the extension of the DL by Pol δ and Pol η. However, preloaded PCNA on the DL counteracts the Rad51-mediated inhibition of the DL extension. This suggests that the inhibition of postinvasion DNA synthesis by Rad51 occurs mostly at the step of PCNA loading.

DNA double-strand breaks (DSBs) are introduced into the genome by several factors, including ionizing radiation, mutagenic chemicals, reactive oxygen species, and stalled DNA replication (1). Without appropriate repair, DSBs may lead to cell lethality or cancer (2–4). Homologous recombination (HR) is a widely conserved essential mechanism for high-fidelity repair of DSBs (5–8). HR is a highly coordinated multistep biochemical process, which is most elegantly demonstrated in the yeast Saccharomyces cerevisiae. Briefly, DSB ends are first processed by specialized exonucleases to generate 3′ overhangs (9–11), which are then coated with the ssDNA binding protein Replication Protein A (RPA). The Rad52 recombination mediator interacts with RPA and recruits Rad51 recombinase onto the ssDNA to form a helical nucleoprotein filament (12–15). This Rad51-ssDNA filament mediates strand invasion into a homologous dsDNA (16) to produce a DNA structure that is referred to as the D-loop (DL; see Fig. 6B). The 3′ end of the invading strand is used as the primer by DNA polymerases for DNA synthesis (postinvasion DNA synthesis), extending the DL (see Fig. 6F). After postinvasion DNA synthesis, repair may be completed by a break-induced replication, DSB repair, or synthesis-dependent strand-annealing pathway (5).

Fig. 6.

Model of PCNA functions in HR. (A) A DSB is processed and complexed with Rad51. (B) DNA strand invasion by Rad51 requires RPA, which occupies the ssDNA loop. (C) After mediating the strand invasion, Rad51 needs to be removed by Rad54. (D) RPA remaining on the ssDNA recruits RFC, which mediates the loading of PCNA onto the invading DNA strand. (F) Loaded PCNA stimulates the activity of Pol δ and Pol η to carry out the postinvasion DNA synthesis. (E) In the absence of PCNA, Pol η may extend the DNA at lower efficiency.

Postinvasion DNA synthesis is crucial to high-fidelity repair of DSBs because it recovers the genetic information that might be lost during breakage events. Several DNA polymerases, including polymerase (Pol) δ, Pol η, and Pol ζ, are possibly involved in this process. Pol δ is known as a replicative polymerase that constitutes the replisome (17, 18). Genetic studies in yeast indicate that Pol δ is involved in mitotic gene conversion (19), meiotic recombination (20), repair of γ-ray–induced DNA damage (21), and homothallic switching (HO) endonuclease-induced gene conversion (22). Studies of HR in vertebrate systems have shown that translesion polymerases Pol η and Pol ζ also contribute to HR. Pol η-deficient chicken DT40 cells have defects in gene conversion at the IgG locus (23), and purified human Pol η can catalyze DNA synthesis within a synthetic DL (24). Loss of Pol ζ function (REV3^−/−) resulted in increased sensitivity to DNA damaging reagents in mouse (25) and chicken cells (26, 27). Although Pol ζ is not essential for HR in yeast, the deletion of the yeast REV3 gene greatly decreases the mutation rate near a DSB because of its low fidelity (28), indicating a role of Pol ζ in HR.

Proliferating cell nuclear antigen (PCNA) is a DNA sliding clamp, which is conserved in all three domains of life (29). PCNA is loaded onto DNA by the “clamp-loader” Replication Factor C (RFC; complex of Rfc1, Rfc2, Rfc3, Rfc4, and Rfc5), which opens and reseals the PCNA on dsDNA in an ATP-dependent manner (30–33). Loaded PCNA encircles and slides freely along the DNA, anchoring a DNA polymerase to the DNA for processive DNA synthesis (34–36). In vitro, RFC can load PCNA onto DNA with 5′ junctions (i.e., junctions between the ssDNA template and 5′ end of the primer) and nicked dsDNA in the absence of RPA (37–40). However, in the presence of RPA, PCNA is uniquely loaded to 3′ junctions (i.e., junction between the ssDNA template and the 3′ end of the primer). RPA directly interacts with RFC to facilitate the specific binding of RFC to the 3′ junction, thereby directing RFC to load PCNA onto the specific DNA structure (31, 35, 41, 42).

A genetic study in yeast indicated that PCNA is essential for the postinvasion DNA synthesis during HR (22). Consistently, an in vitro study indicated that PCNA is required for DNA synthesis following Rad51-mediated strand invasion (43). However, several questions remain unanswered regarding the molecular functions of PCNA in HR. First, is PCNA loaded efficiently onto a DL that lacks an ssDNA region beyond the 3′ end of the invading strand, which is believed to be crucial for PCNA loading? Loading efficiency of PCNA on a DL has not been quantified because the DL is structurally different from most DNA substrates that have been analyzed for PCNA loading. If it is loaded, what is the role of RPA in the PCNA loading? Second, is the loaded PCNA oriented in the effective direction? Even if PCNA was loaded on the DL, there are two possible directions of loaded PCNA relative to the invading DNA strand. Only one direction is “productive” in stimulating polymerases (35). The ratio of the functional PCNA in loaded PCNA on the DL has not been quantified. Third, what role does Rad51 recombinase play in the PCNA loading process? To answer these questions, we investigated the loading of PCNA on a synthetic DL in vitro and the ability of the loaded PCNA to stimulate DNA synthesis by Pol δ, Pol η, and Pol ζ.

Results

PCNA Is Loaded on the DL and Primed ssDNA with Similar Efficiency.

We adopted the method published by Podust et al. (44) for quantitative analyses of PCNA loading onto DNA substrates with various structures. Essentially, we constructed ³²P-labeled PCNA (³²P-PCNA) and tested its loading onto unlabeled DNA molecules in vitro (details are provided in SI Materials and Methods). In a standard reaction, the preannealed DNA substrate was first incubated with RPA, followed by RFC and ³²P-PCNA. Three minutes after the addition of PCNA, the reaction mixture was treated with glutaraldehyde to fix the PCNA ring structure in its closed conformation on the DNA substrate and subjected to agarose gel electrophoresis (Fig. 1A). We first confirmed the loading of ³²P-PCNA on a primed Bluescript ssDNA substrate (Fig. S1). As expected, a low-mobility PCNA–DNA complex was produced in the presence of RFC and ATP. The reaction was less efficient with unprimed ssDNA or dsDNA (Fig. S1C), as previously reported (45, 46). We then analyzed the loading of the ³²P-PCNA on a synthetic substrate that mimicked the DL and on a primed ssDNA substrate (C2) (Fig. 1B). When we used the C2 substrate (Fig. 1C, lanes 1–4), ³²P-PCNA mobility was shifted in the gel in the presence of RPA, RFC, and ATP (Fig. 1C, lane 4), indicating that it was loaded on C2. Although C2 is a linear DNA molecule, loaded PCNA did not slide off from the substrate because both ends of the DNA were complexed with RPA, as previously reported (42, 47). Likewise, when the DL substrate was used (Fig. 1C, lanes 5–8), a mobility shift was observed in the complete reaction (Fig. 1C, lane 8) but not in the absence of RFC, ATP, or RPA (Fig. 1C, lanes 5–7) indicating that PCNA was loaded on the DL substrate. PCNA loading in the complete reactions on the C2 and DL substrates was comparable (Fig. 1F). The kinetics of loading on DL and C2 were similar (Fig. 1G), and loading was dependent on the amount of RFC supplied (Fig. S2). To identify the region of the DL where PCNA was loaded, we treated the product of the complete loading reaction with KpnI to cleave the dsDNA region in the DL (Fig. 1B). Five minutes of KpnI digestion did not release the loaded PCNA from the substrate; however, after 30 min of incubation, most of the ³²P-PCNA was released from the complex (Fig. 1D), indicating that KpnI digestion opens the DL and allows the loaded PCNA to slide off. These data show that the PCNA is efficiently loaded onto the dsDNA region inside the DL.

Fig. 1.

PCNA is efficiently loaded on the DL. (A) Illustration of the PCNA loading assay. A detailed description is provided in SI Materials and Methods. (B) Structures of synthetic DNA substrates (DL and C2) used for the PCNA loading analysis. Numbers are lengths of DNA segments (in nucleotides). (C) Loading reactions of ³²P-PCNA on C2 and DL were performed in the presence or absence of ATP, RFC, and RPA as indicated. (D) Complete PCNA loading reaction with DL substrate was followed by incubation with (lanes 2 and 4) or without (lanes 1 and 3) KpnI for indicated times. (E) PCNA loading reactions were carried out using DNA substrates with different structures (DL, DL-L, DL-R, DL2-L, and DL2-R), which are illustrated in H. (F and H) Percentages of loaded PCNA were calculated from C and E, respectively, with repeated experiments (n = 3). (G) Same reactions as shown in lane 4 (C2) and lane 8 (DL) of C were stopped at the indicated time points. Error bars represent the SDs.

RPA on the ssDNA Loop Stimulates PCNA Loading.

Previous studies using primed ssDNA substrates indicated that efficient PCNA loading required an ssDNA region beyond the 3′ junction of the primer and template (37, 38). This ssDNA region is believed to be complexed with RPA, which provides a binding site for RFC for subsequent PCNA loading. However, our DL substrate had only a one-nucleotide ssDNA region at the 3′ end of the invading DNA strand, which is too small for RPA binding (48, 49) yet required RPA for PCNA loading. Therefore, we performed a series of experiments to identify the location of RPA in PCNA loading using different DNA structures (Fig. 1 E and H). Interestingly, the loading of PCNA on the DL substrates containing 3′ junctions [DL-left (L) and DL2-L] was not better than that on the DL substrates containing 5′ junctions [DL-right (R) and DL2-R]. That was also confirmed by analyzing the PCNA-dependent extension of the DLs by Pol δ (Fig. S3), where DL, DL2-L, and DL2-R gave comparable amounts of full-length products, indicating that PCNA was loaded onto these substrates with similar efficiency. All these data indicate that PCNA loading on the DL was not dependent on a 3′ junction, in contrast to the previous studies using primed ssDNA substrates (30–33). Therefore, RPA most likely binds to the displaced ssDNA loop to stimulate the PCNA loading reaction. To investigate this hypothesis, we analyzed PCNA loading on a modified DL substrate, where the ssDNA loop was annealed to another oligonucleotide to make it double-stranded (Fig. 2A, DL-CAP). The loading of ³²P-PCNA on the DL-CAP was greatly reduced, supporting the notion that RPA binding to the displaced ssDNA region of the DL is necessary for PCNA loading.

Fig. 2.

RPA on the ssDNA loop stimulates PCNA loading on the DL. (A) PCNA loading was performed on the DL (lanes 3 and 4) and its derivative that had an additional oligonucleotide covering the ssDNA loop (DL-CAP; lanes 1 and 2). (B) PCNA loading on the C2 (lanes 1–3) and the DL (lanes 4–6) was performed under standard conditions in the presence or absence of RFC (0.08 μM), RPA (0.8 μM), and E. coli SSB protein (0.8 μM) as indicated. (C) Percentages of loaded PCNA were calculated from B with repeated experiments (n = 3–4). Error bars represent the SDs.

Because studies have shown that RPA/RFC protein–protein interaction is involved in the PCNA loading (31, 41, 42) on simple primed substrates, we investigated whether this is also the case in the PCNA loading on the DL. To do this, we replaced RPA with Escherichia coli single strand binding (SSB) protein in our loading protocol (Fig. 2 B and C). SSB only poorly stimulated the loading of PCNA on the DL (Fig. 2B, lane 5) and C2 (Fig. 2B, lane 2). Additionally, RPA titration indicated that the amount of RPA required for PCNA loading roughly corresponded to that required for full occupancy of the displaced ssDNA loop (Fig. S4 C and D). Based on these results, we conclude that RPA on the ssDNA loop provides the binding site for the RFC, leading to PCNA loading onto the dsDNA region within the DL.

Rad51 Inhibits PCNA Loading on the DL.

Although Rad51-dependent DNA strand invasion is crucial for HR, an increasing number of studies have indicated that Rad51 inhibits the process after DNA strand invasion (50–52). Therefore, it is of special interest to investigate the effect of Rad51 in PCNA loading on the DL (Fig. 3). When the RPA–DL complex was incubated with an increasing amount of Rad51 and then subjected to the PCNA loading reaction, the loading of ³²P-PCNA was inhibited by Rad51 in a concentration-dependent manner (Fig. 3 A and D, red open circle). At the concentration of Rad51 that was sufficient to saturate all dsDNA regions of the DL substrate (Rad51/DNA = 32), almost no loading was observed. In contrast, when Rad51 was added to the DL after the PCNA loading reaction, a supershift of the signal was observed (Fig. 3B), indicating the formation of a Rad51–PCNA–DNA complex. No significant effect on the percentage of the loaded PCNA was observed (Fig. 3D, PCNA→Rad51). These results indicate that Rad51 inhibits the loading of PCNA but does not affect the stability of the loaded PCNA. Because the reaction contained RPA that should occupy the ssDNA loop, Rad51 should preferentially bind to the dsDNA region of the DL (Fig. 3A). This was confirmed by measuring the ATPase activity of Rad51 under the same conditions (Fig. S5). Therefore, the Rad51-mediated inhibition of PCNA loading was not due to the removal of RPA from the DL or the melting of the DL structure into ssDNA but to the Rad51–dsDNA complex in the DL. Most likely, Rad51 bound to the invading DNA strand acts as a physical obstruction preventing the loading of PCNA.

Fig. 3.

Rad51 inhibits PCNA loading. (A) Rad51 (0, 0.4, 0.8, 1.6, 2.4, and 3.2 μM from lanes 2–7) was added to 0.1 μM preannealed DL preincubated with 0.4 μM RPA. ³²P-PCNA and RFC were then added to start the PCNA loading reaction. (B) Same experiment as in A was performed, except that Rad51 was added after PCNA loading on the DL. (C) Rad51–ssDNA complex was first produced by incubating Rad51 (0, 0.4, 0.8, 1.6, 2.4, and 3.2 μM from lanes 2–7) with 0.1 μM 29-mer oligo (TSO319), which mimicked invading strand for 5 min at 37 °C. The reaction mixture was further incubated with preannealed DL (0.1 μM) without the 29-mer for 15 min and then with 0.4 μM RPA for another 3 min at 37 °C. ³²P-PCNA and RFC were then added to start the PCNA loading reaction. (D) Experiments shown in A (red open circle), B (blue square), and C (black triangle) were repeated, and percentages of PCNA that were in the complexes were calculated and plotted against the Rad51/DL molecular ratio. Error bars represent the SDs (n = 3).

Unexpectedly, even when Rad51 and ssDNA were preincubated before reconstituting the Rad51–DL complex, maximum inhibition required Rad51 that could saturate the entire dsDNA region of the DL (Rad51/DL ratio = 32; Fig. 3 C and D). This might suggest that Rad51 molecules outside the invading DNA strand may also contribute to the inhibition. Alternatively, Rad51 on the invading DNA strand may quickly redistribute along the dsDNA region after DNA strand invasion. Consistent with the latter idea, the Rad51-ssDNA preincubation inhibited the PCNA loading more strongly during first 2 min of the loading reaction (Fig. S6, Left; compare lower two curves).

PCNA Stimulates Pol δ to Displace DNA Strand During the Postinvasion DNA Synthesis.

Genetic and biochemical studies in yeast indicated that both PCNA and Pol δ play major roles in HR (19–22). Studies in vertebrate systems showed that Pol η and Pol ζ are also involved in HR (23, 26). To characterize the roles of these polymerases in HR further, we investigated the ability of Pol δ (Pol3–Pol31–Pol32 complex), Pol η (Rad30), and Pol ζ (Rev3–Rev7 complex) to extend the DL in the presence and absence of loaded PCNA (Fig. 4). To distinguish primer recognition and strand displacement in DL extension, we used the DL substrate that had a one-nucleotide ssDNA region beyond the 3′ end of the primer (Fig. 1B). The polymerase can extend the primer by 1 nt before encountering the dsDNA region of the DL substrate. Among the three yeast polymerases we tested, Pol δ showed the poorest ability to extend the DL in the absence of the loaded PCNA [Fig. 4 C (compare lanes 2, 4, and 6) and D], consistent with the previous observations (24, 53). Interestingly, however, Pol δ extended almost all the primers for 1–2 nt (Fig. 4C, lane 4), indicating that Pol δ efficiently recognized the primer in the DL in the absence of PCNA but that the synthesis was obstructed by dsDNA. When Pol δ was added to the DL that had loaded PCNA (Fig. 4 C and D, lanes 5 and 11–13, respectively), the majority of primers were fully extended, indicating that PCNA greatly stimulated Pol δ’s ability to displace the DNA strand during DNA synthesis. Both the inability of Pol δ to extend the DL and the stimulation by PCNA were also reproduced with the plasmid-sized substrate with a 457-nt loop [Fig. 4 B (“Bubble” and “Y-shape”) and F (lanes 7, 8, and 11)], indicating that these activities are unrelated to the loop size.

Fig. 4.

Extension of the DL by Pol δ, Pol η, and Pol ζ is differently modulated by PCNA. (A) Illustration of DNA polymerase assay on the DL containing a ³²P-labeled invading DNA strand (red arrow). Nonradiolabeled PCNA was loaded on the DL, and Pol δ, Pol η, or Pol ζ was then added to start DNA synthesis. (B) Plasmid-sized substrates that were used in F are shown. A ³²P-labeled primer (40-mer) was annealed to 3.0 kb of dsDNA containing a 457-nt noncomplementary region (Bubble), which was produced from Bluescript SK(+) and SK(−) ssDNA molecules (details are provided in SI Materials and Methods). The same primer was annealed to a Y-shaped derivative (Y-shape) and linearized Bluescript SK⁺ ssDNA (Linear). (C) DL substrate was incubated with PCNA and RFC for 3 min as indicated, and then with Pol η, Pol ζ, or Pol δ for 15 min. DNA products were analyzed by denaturing PAGE. (D) Time courses of the DNA synthesis in DL substrate by indicated polymerases were analyzed in the presence or absence of PCNA/RFC. (E) Same reactions as in D were repeated, except that simple primed substrate (C2; Fig. 1B) was used instead of DL. (F) Plasmid-sized DNA substrates shown in B were used for the PCNA loading and DNA synthesis reactions by Pol δ, Pol η, Pol ζ, or Klenow fragment (K). Lanes 1, 10, and 13 contain size markers that were produced by labeling a 100-bp DNA ladder by T4-kinase followed by denaturing in formamide.

Pol δ, Pol η, and Pol ζ Respond Differently to PCNA on the DL.

When yeast Pol η was added, the DL was extended efficiently without the loaded PCNA (Fig. 4 C and D, lanes 2 and 2–4, respectively), consistent with the previous observations (24, 43, 53). This extension was not significantly affected by loaded PCNA (Fig. 4 C and D, lanes 3 and 5–7, respectively). However, on the plasmid-sized DL, extension by Pol η was moderately but clearly stimulated by the loaded PCNA (Fig. 4F, lanes 3 and 4). This may suggest that Pol η has lower efficiency to interact with PCNA but that once they interact, Pol η-PCNA can extend longer than 100 nt. In control reactions (Fig. 4E), both Pol η and Pol δ extended the primer on simple primed substrate both in the presence and absence of the loaded PCNA. Pol ζ extended the simple primed substrate less efficiently than the same amount of Pol η and Pol δ (Fig. 4E). Interestingly, Pol ζ extended the DL substrate slightly more efficiently than Pol δ in the absence of PCNA (Fig. 4 C and D, lanes 6 and 14–16, respectively), indicating that Pol ζ extension was not strongly inhibited by the DL structure. However, no stimulation was observed in the presence of PCNA (Fig. 4 C, D, and F; Pol ζ).

We detected an unexpected premature termination product in the Pol δ extension reactions on the simple primed substrate C2 (indicated by the asterisk in Fig. 4E). It was not detected on the DL that had the same sequence as C2 (Fig. 4D) or on other primed DNA substrates (Fig. S3, substrate C3). We think that this sequence-specific termination is due to the secondary structure of the C2 substrate. We also observed a small fraction of unused primers in the presence of PCNA and RFC (Fig. 4 C–E). This apparent inhibition of primer recognition was observed consistently in the reactions containing RFC and was independent of PCNA or polymerase type. Because the reaction contained fourfold more RFC than DNA substrate, the excess RFC on the DNA might inhibit primer recognition by polymerases. We do not believe that this inhibition is biologically significant.

Loaded PCNA Counteracts the Rad51-Mediated Inhibition of DNA Synthesis.

A previous study has shown that on the invading DNA strand, Rad51 inhibits DNA synthesis (54). We have shown here that Rad51 inhibits the loading of PCNA on the DL. Therefore, we expected that Rad51 would also inhibit PCNA-dependent DL extension by polymerases. However, it was unclear whether Rad51 would inhibit the DL extension if PCNA had been loaded already (Fig. 5A). To address this question, we examined the effect of Rad51 on DNA synthesis by Pol δ and Pol η in the DL in the presence or the absence of loaded PCNA (Fig. 5). As expected, in the absence of PCNA, Rad51 strongly inhibited both Pol δ and Pol η extension of the DL substrate [Fig. 5 B and C (lanes 2–8) and D and E (blue curves)]. Importantly, if PCNA was preloaded, Rad51 inhibition of DNA synthesis by Pol η [Fig. 5 B (lanes 9–15) and D] and Pol δ [Fig. 5 C (lanes 9–15) and E] was less pronounced. These results indicate that loaded PCNA lessens the inhibition by Rad51.

Fig. 5.

Preloaded PCNA counteracts Rad51 inhibitory effect on Pol δ and Pol η. (A) Rad51 was added to DL preloaded with PCNA. DNA synthesis was then started by means of the addition of Pol δ or Pol η. (B and C) Rad51 (0, 0.2, 0.4, 0.6, 0.8, 1.2, and 1.6 μM) was added to 0.02 μM DL (lanes 2–8) or DL preloaded with PCNA (lanes 9–15). Then, DNA synthesis was started by adding 0.02 μM Pol η (B) or Pol δ (C). (D and E) Quantification of the relative percentage of primer use in B and C, respectively. The values were normalized by setting the value of no Rad51 reaction to be 100%. Error bars represent SDs (n = 3).

Discussion

In this study, we describe unique findings about the functional interactions of RFC, PCNA, DNA polymerases, and Rad51 recombinase on the DNA recombination intermediate. Based on our results, we propose the updated DSB repair model, focusing on the events after Rad51-mediated strand invasion (Fig. 6). First, we showed that PCNA loading on the DL structure was very efficient (Fig. 1). In addition, PCNA loading onto the DL required the binding of RPA to the displaced ssDNA loop, thereby providing a binding site for RFC, which mediated the PCNA loading (Fig. 2). Loaded PCNA greatly stimulated Pol δ extension within the DL (Fig. 4). Without PCNA, Pol δ was very poor at extending the DL. The DL structure itself did not interfere with the polymerase–primer interaction. Therefore, the major role of PCNA in the DL extension is to stimulate Pol δ to displace the DNA strand. In contrast, Pol η extended the DL much more efficiently than Pol δ in the absence of PCNA, but most of the products were shorter than 100 nt. PCNA stimulated the reaction to extend a much longer distance. Taken together, these results suggest that Pol δ and Pol η may have different roles in HR, dependent on the availability of PCNA (Fig. 6 D and E). In our experiments, Pol ζ demonstrated limited ability to extend DL that was not stimulated by PCNA. However, our Pol ζ preparation did not include the Rev1 subunit previously suggested to have a critical role in Pol ζ function in HR (55). Recently, three groups reported that Pol31 and Pol32, which were considered subunits of Pol δ, were essential subunits of functional Pol ζ (56–58). The potential roles of these proteins in DL extension by Pol ζ remain to be elucidated.

It is known that only one side of the PCNA ring can interact with polymerases (29, 35). Therefore, if the PCNA loading is oriented randomly, only half of the loaded PCNA can stimulate polymerase. As shown in Fig. 4D, the majority of DL substrate produced the full-length product in the presence of PCNA. Importantly, the reaction contained the same amounts (0.2 pmol) of DNA substrate and PCNA, indicating that the majority of PCNA in the reaction was loaded on the DL in the correct orientation (oriented to the 3′ end of the primer). Previous studies using the simple primed ssDNA substrates showed that the 3′-oriented loading of PCNA depends on RPA, which directionally binds to the ssDNA template (29, 38). However, in the DL analyzed in this study, RPA binds to the displaced ssDNA, which is running in the opposite direction to the template strand. Precise molecular architecture of the RPA–RFC–PCNA complex on the DL needs to be demonstrated to reconcile this apparent inconsistency.

A previous study indicated that Rad51 recombinase inhibited the primer extension by means of the Klenow enzyme (54). We found that Rad51 inhibited both PCNA loading on the DL (Fig. 3) and DNA synthesis (Fig. 5). Therefore, Rad51 can inhibit two essential biochemical processes of the postinvasion DNA synthesis. It is not clear which of these two inhibitions is more important in regulating HR. However, it seems more likely that PCNA loading is a key regulatory target of Rad51, because PCNA loading precedes the Pol δ–primer interaction (18). Two pieces of data in this paper are consistent with this idea. First, in Rad51 titration experiments (Figs. 3D and 5D and E), the Rad51 concentration needed to inhibit PCNA loading was lower than that required to inhibit primer extension, indicating that Rad51 is a more effective inhibitor of PCNA loading than of DNA synthesis per se. Second, when PCNA was preloaded on the DL, Rad51 allowed polymerase to extend the DL more easily (Fig. 5). Although we do not know whether PCNA can be loaded before removing Rad51 in vivo, this in vitro result suggests that the inhibition of PCNA loading is more crucial for regulation of postinvasion DNA synthesis. These Rad51-mediated inhibitions can be cleared up by Rad54, which removes Rad51 from dsDNA (54, 59) (Fig. 6C). Rad54-mediated stimulation of PCNA loading is yet to be demonstrated in vivo and in vitro. This study contributes to our understanding of the mechanisms of HR-mediated DNA repair and maintenance of genome integrity.

Materials and Methods

Details of plasmid construction, protein purification, and assays are provided in SI Materials and Methods. Sequences of synthetic oligonucleotides are shown in Table S1. Oligonucleotides used to produce each substrate are shown in Table S2.