The DNA Binding Preference of RAD52 and RAD59 Proteins

We examined the double-stranded DNA (dsDNA) binding preference of the Saccharomyces cerevisiae Rad52 protein and its homologue, the Rad59 protein. In nuclease protection assays both proteins protected an internal sequence and the dsDNA ends equally well. Similarly, using electrophoretic mobility shift assays, we found the affinity of both Rad52 and Rad59 proteins for DNA ends to be comparable with their affinity for internal sequences. The protein-DNA complexes were also directly visualized using atomic force microscopy. Both proteins formed discrete complexes, which were primarily found (90–94%) at internal dsDNA sites. We also measured the DNA end binding behavior of human Rad52 protein and found a slight preference for dsDNA ends. Thus, these proteins have no strong preference for dsDNA ends over internal sites, which is inconsistent with their function at a step of dsDNA break repair that precedes DNA processing. Therefore, we conclude that S. cerevisiae Rad52 and Rad59 proteins and their eukaryotic counterparts function by binding to single-stranded DNA formed as intermediates of recombination rather than by binding to the unprocessed DNA double-strand break.

The repair of DNA double-strand breaks (DSBs) 3 in cells is crucial for genomic integrity and viability. DSBs can be generated during the process of DNA replication on damaged templates or directly by exogenous DNA-damaging agents. A single unrepaired DSB can result in cell death (1). In eukaryotic cells, there are two major pathways to repair DSBs, homologous recombination and non-homologous end-joining (NHEJ). NHEJ repairs DSBs by ligating broken DNA ends, but this process is inherently error-prone (2). Homologous recombination, on the other hand, restores DSBs without the loss of genetic integrity by using a homologous sequence as template, such as a sister chromatid or homologous chromosome. In Saccharomyces cerevisiae, homologous recombination is the predominant mechanism for DSB repair but, when defective, the radiation sensitivity of NHEJ mutants is evident (3).
In S. cerevisiae, homologous recombination is mediated by proteins of the RAD52 epistasis group, including RAD50, RAD51, RAD52, RAD54, RAD55, RAD57, RAD59, MRE11, XRS2, and RFA1. A homologue of RAD54, RDH54/TID1, also plays a role in a subset of recombination-dependent DSB repair pathways (3,4). Mutation of any of these genes confers radiation sensitivity. In particular, loss of RAD52 function leads to the most severe recombination phenotype because the Rad52 protein is required for all recombination-dependent events (Ref. 5; for review, see Refs. 3 and 4). The Rad52 protein is conserved in most eukaryotic organisms ranging from yeast to human (6 -8). In vivo it accumulates at DNA damage sites, forms discrete nuclear foci, and co-localizes with the DNA strand exchange protein, Rad51 (9 -13). In vitro both yeast and human Rad52 proteins bind single-stranded DNA (ssDNA) and dsDNA, stimulate annealing of complementary ssDNA (14 -17), and facilitate Rad51 protein-ssDNA filament formation in the presence of the eukaryotic ssDNA-binding protein, replication protein A, in a species-specific fashion (18 -21). This level of functional conservation is consistent with the importance of the Rad52 protein in DNA recombination.
The Rad59 protein was identified in S. cerevisiae (22) and Kluyveromyces lactis (23) as a Rad52 protein homologue. It shares sequence similarity with the conserved N-terminal domain of Rad52 protein but lacks the C-terminal Rad51-interacting domain. Genetically, RAD59 is dispensable in RAD51dependent homologous recombination pathways but plays an important role in the RAD51-independent repair pathways (22), such as single-strand annealing (SSA) (24,25), break-induced replication (BIR) (26,27), and type II survival in telomerase-deficient cells (28,29). In vivo Rad59 interacts with Rad52 protein (30), suggesting direct participation in a subset of RAD52-dependent homologous recombination events. The Rad59 protein possesses biochemical activities that are also found in other RAD52 homologues, including the ability to bind ssDNA and dsDNA and to anneal complementary ssDNA (30 -32).
Although both the homologous recombination and NHEJ pathways can repair DSBs, each functions differently in cells. NHEJ function is enhanced in haploid cells and in the G 1 phase of the cell cycle (33), whereas recombination functions in the S-G 2 phase (11,34) or in diploid cells (35). It appears that pathway choice is determined by availability of a homologous sequence. However, it is unclear how and by what molecular mechanism pathway choice is regulated.
Human Rad52 (hRad52) protein was reported to localize preferentially at dsDNA ends and to protect dsDNA ends from nuclease degradation (36). These observations led to the hypothesis that Rad52 protein serves as a "molecular switch" or a "gatekeeper" to channel DSBs into the homologous recombination repair pathway rather than into the NHEJ pathway (36,37). Soon afterward, a Rad52 homologue from the fission yeast Schizosaccharomyces pombe, the Rad22 protein, was reported to have the same ability to protect dsDNA ends from nuclease digestion and to localize to sites of DSBs in vivo (38). However, a subsequent report used atomic force microscopy (AFM) (39) to show that hRad52 has no preference for dsDNA ends; rather, it was shown to preferentially bind to ssDNA regions in dsDNA, regions even as small as 2 or 4 nucleotides. Because the functions of Rad52 protein are conserved among different organisms, we tested whether the S. cerevisiae Rad52 protein bound preferentially to dsDNA ends or to ssDNA. Answering this question could help address the issue of whether Rad52 regulates the channeling of DSB repair into different repair pathways in budding yeast. As an alternative, we also entertained the idea that Rad59 protein in S. cerevisiae might instead have assumed the role of gatekeeper, so we examined the dsDNA binding preference of Rad59 protein as well. We found no significant differences in dsDNA end binding versus internal binding for the yeast Rad52 and Rad59 proteins. Furthermore, we confirmed that the hRad52 had only a weak preference for dsDNA ends (39). Therefore, for these reasons and because these proteins bind ssDNA preferentially over dsDNA, we conclude that Rad52 and Rad59 proteins act only at later stages of recombination, after the DSBs are processed into ssDNA.

EXPERIMENTAL PROCEDURES
Proteins-Exonuclease III (ExoIII) was purchased from Promega. Restriction endonucleases and T4 polynucleotide kinase were purchased from New England Biolabs. Yeast Rad52 protein was purified as described (18), except that the Superose-12 column was replaced with a Sephacryl-300 column. Rad59 protein was overexpressed in BLR(DE3) pLysS cells and purified as described (32). Human Rad52 protein was a generous gift from Dr. P. Sung of Yale University. In reactions where a protein was omitted, an equal amount of corresponding protein storage buffer was added instead.
DNA Substrates-All DNA concentrations are expressed in nucleotides. The complementary oligonucleotides PB77 and PB78, 100 nucleotides in length, were purchased, purified, and annealed as described (40). Concentrations of PB77 and PB78 were determined using nucleotide extinction coefficients of 8891 and 9737 M Ϫ1 cm Ϫ1 at 260 nm, respectively. PB77 and PB78 were first annealed to each other, and the resulting PB77-78 dsDNA was labeled by T4 polynucleotide kinase at the 5Ј-end on both strands. Unincorporated [␥-32 P]ATP was removed using a MicroSpin G25 desalting column (GE Healthcare). Plasmid pBR322 DNA was purified by conventional alkaline lysis followed by equilibrium ultracentrifugation in a CsCl-ethidium bromide gradient. Purified pBR322 was linearized with the indicated restriction endonucleases and purified by phenol extraction followed by ethanol precipitation. In all cases agarose gel electrophoresis was used to verify that the DNA was completely digested by the restriction endonuclease (data not shown). DNA concentration was determined using an extinction coefficient of 6500 M Ϫ1 cm Ϫ1 at 260 nm.
Nuclease Protection Assay-The ability of Rad52 and Rad59 proteins to protect dsDNA from degradation by nucleases was assayed in nuclease protection buffer (20 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol) with the concentration of magnesium acetate indicated. Duplex DNA (PB77-78 that was 5Ј-end-labeled at both ends, 10 M) was preincubated with the indicated amount of protein at 37°C for 10 min before the addition of nuclease to start DNA degradation. ExoIII and HaeIII were titrated to give ϳ80% degradation of the free DNA substrate for the time specified. After incubation at 37°C for 15 min, the reactions were deproteinized with 0.7% SDS and 1 mg/ml proteinase K (Roche Applied Science) and subjected to native polyacrylamide gel (8%) electrophoresis in 1ϫ TBE buffer (89 mM Tris borate, pH 8.3, 2 mM EDTA). DNA substrate with the electrophoretic mobility of the remaining intact dsDNA was quantified using a Storm 820 system (GE Healthcare). The average band intensity for the dsDNA from control reactions lacking nuclease in each gel (Figs. 1-3, panels A and B, lanes 1 and 10) was defined as 100% protection. A box was drawn around the position of the intact dsDNA in the same gel (lanes 1 and 10); boxes of the same size and electrophoretic position were used in the sample lanes to determine the amount of intact dsDNA remaining. The band intensity in nuclease-only control (lane 2) was defined as 0% protection. The degree of dsDNA protection was determined using the formula dsDNA protected (%) ϭ (band intensity Ϫ 0% control)/(100% control Ϫ 0% control) ϫ100%.
Electrophoretic Mobility Shift Assay-The reactions contained 10 M 5Ј-end-labeled dsDNA (100 bp, PB77-78) and various amounts of unlabeled competitor dsDNA in electrophoretic mobility shift buffer (30 mM K ϩ -MOPS, pH 7.3, 1 mM dithiothreitol). Rad52 (1 M), Rad59 (2.2 M), or hRad52 (1.2 M) protein was added to the dsDNA, and the reactions were incubated at 37°C for 5 min to allow protein-DNA complex formation. The concentration of NaCl contributed by the protein storage buffer was 40, 50, and 25 mM in the Rad52-, Rad59-, and hRad52-containing reactions, respectively. Under these conditions, in the absence of unlabeled competitor DNA, the mobility of more than 90% of the DNA substrate was shifted. The reactions were analyzed by electrophoresis using 8% polyacrylamide gels in 1ϫ TBE buffer and quantified using a Storm 820 system. AFM-EcoRV-or PstI-linearized pBR322 DNA (10 M) was incubated with various amount of yeast Rad52 protein (0.4 -350 nM) in binding buffer (30 mM K-MOPS, pH 7.3, 3 mM magnesium acetate, 20 mM NaCl) in a total reaction volume of 20 l at 37°C for 5 min. The sample was applied onto freshly cleaved mica; after 5 min the surface was washed with 1 ml of distilled water and dried with compressed nitrogen gas. Protein-DNA complexes were examined using a Nanoscope IIIa AFM (Digital Instruments) operated in tapping mode. Images were captured at scan sizes of 0.6 -1.7 m and processed by first-order flatten-ing to remove sample tilt. For Rad59 protein, magnesium acetate and protein concentrations were optimized to 5 mM and 100 nM, respectively, to facilitate discrete complex formation and to minimize aggregation.

Yeast Rad52 and Rad59 Proteins and hRad52 Protein Protect DNA Ends and Internal Sites Equally from Nucleolytic
Degradation-We wanted to know whether the Rad52 and the Rad59 protein from budding yeast display a preference for binding to DNA ends as was reported for Rad52 homologues from human and fission yeast. Previously, this preference was demonstrated as an ability to protect DNA more effectively from exonucleolytic degradation than from endonucleolytic cleavage (36,38). We, therefore, conducted such nuclease assays in the presence of Rad52 or Rad59 protein. The DNA substrate used in the nuclease assays was a blunt-ended, 5Ј-end labeled-100-bp DNA duplex (labeled on both strands) that contained one HaeIII site. This substrate was chosen to ensure that limited exonuclease degradation would be detected as a change in electrophoretic mobility. The DNA substrate was preincubated with various amounts of Rad52 or Rad59 protein and was then incubated with either ExoIII or HaeIII enzyme, each of which was present at an amount that degraded ϳ80% of the naked dsDNA substrate under the given reaction conditions. For both of the ExoIII or HaeIII protection assays, the percentage of label remaining at the position of the intact dsDNA was quantified because it is directly related to protection. We used two magnesium ion concentrations, 0.66 mM (low) and 10 mM (high), to permit comparison to previously published work (38). At these two Mg 2ϩ concentrations, the activities of ExoIII and HaeIII were the same (data not shown); therefore, the same amount of each nuclease was used at the two different assay conditions. We found that Rad52 protein protected dsDNA from degradation in a protein concentration-dependent manner in both the ExoIII-and HaeIII-containing reactions (Fig. 1,  A and B). In the presence of ExoIII, the dsDNA migrated as a slightly faster and broader band due to removal of nucleotides by the exonuclease (Fig. 1A, lane 2). The addition of increasing amounts of Rad52 protein increased the fraction of dsDNA substrate that remained intact (Fig. 1A, lanes 3-9). Consistent with a previous report showing that DNA binding activity is sensitive to increased Mg 2ϩ concentration (16), Rad52 protein protected the dsDNA better at 0.66 mM Mg 2ϩ (Fig. 1C) than it did at 10 mM Mg 2ϩ (Fig. 1D). Likewise, in the HaeIII nuclease protection assay, Rad52 protein prevented dsDNA cleavage (Fig. 1B), and it did so better at 0.66 mM Mg 2ϩ (Fig. 1C) than it did at 10 mM Mg 2ϩ (Fig. 1D). Most importantly, when the protection patterns for ExoIII and HaeIII were compared at the same magnesium ion concentrations (Fig. 1, C and D), protection was essentially the same, within experimental error. Thus, yeast Rad52 protein shows no preference for dsDNA end binding as measured by this nuclease protection assay.
The behavior of Rad59 protein was also evaluated in these assays. Similar to Rad52 protein, Rad59 protein protected the dsDNA from both exonuclease and endonuclease degradation in a protein concentration-dependent manner (Fig. 2, A and B). In agreement with the previously reported sensitivity of DNA binding to increased Mg 2ϩ concentration (31), Rad59 protein protected the dsDNA to a greater extent at the lower Mg 2ϩ concentration (Fig. 2C) than at the higher Mg 2ϩ concentration (Fig. 2D). More importantly, Rad59 protein showed almost the same level of protection against ExoIII-and HaeIII-mediated degradation at the low Mg 2ϩ concentration and, actually, better protection against HaeIII digestion at the high Mg 2ϩ concentration at the highest protein concentration used. Thus, the results obtained from the nuclease assays do not demonstrate preferential DNA end binding for either Rad52 or Rad59 proteins.
Finally, we tested the DNA end binding preference of human Rad52 protein using the same nuclease protection assays. Like the yeast Rad52 and Rad59 proteins, hRad52 protein protected dsDNA in a protein concentration-dependent manner (Fig. 3, A  and B). The protection from HaeIII degradation was better than from ExoIII degradation at the low Mg 2ϩ concentration (Fig.  3C) and marginally better at the high Mg 2ϩ concentration (Fig.  3D). Unlike the yeast Rad52 and Rad59 proteins, however, the The DNA Binding Preference of RAD52 and RAD59 Proteins DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 protection by hRad52 protein was only slightly reduced at the higher Mg 2ϩ concentration (Fig. 3D), indicating the DNA binding activity of hRad52 protein is less sensitive to an increased concentration of Mg 2ϩ than that of yeast Rad52 and Rad59 proteins.
Yeast Rad52 and Rad59 Proteins and Human Rad52 Protein Bind to DNA Ends with Minimal Preference Relative to Internal Sites-To confirm the results from the nuclease protection assays, we designed a direct competition experiment to exam-ine the dsDNA binding specificity of Rad52 and Rad59 proteins. The same 5Ј-end-labeled 100-bp dsDNA substrate was incubated with a fixed concentration of protein so that ϳ90% of the dsDNA substrate had a resultant electrophoretic mobility that caused it to remain in the well of the gel. This protein-DNA complex was then titrated with increasing amounts of unlabeled, EcoRV-or HaeIII-digested pBR322 DNA. There is only one EcoRV site in pBR322 DNA, but there are 22 HaeIII sites; thus, the HaeIII-linearized pBR322 DNA yields 22 times more dsDNA ends than the EcoRV-linearized DNA. Therefore, if a protein has a high affinity for dsDNA ends, then the same concentration of HaeIII-linearized pBR322 DNA should be a 22-fold better competitor for protein binding than the EcoRVlinearized DNA.
The results obtained with Rad52 protein are shown in Fig.  4A. The retention of labeled dsDNA in the well was inversely dependent on the concentration of the unlabeled competitor DNA. Quantification of the gel (Fig. 4B) shows that the amount of the EcoRV-digested pBR322 needed to liberate the labeled dsDNA due to competitive binding of the Rad52 protein is similar to the amount of the HaeIII-digested pBR322. This indicates that the EcoRV-digested pBR322 dsDNA, although possessing 22 times fewer dsDNA ends, competes for Rad52 protein binding as efficiently as the HaeIII-digested pBR322 DNA.
Identical experiments were performed using Rad59 protein. Only the quantification is shown here (Fig. 4C) because the gels were essentially the same as those shown for Rad52 protein. Just as for Rad52 protein, the HaeIII-digested pBR322 competed equally well for Rad59 protein binding compared with the EcoRV-linearized pBR322. These results are consistent with our nuclease protection studies (Figs. 1 and 2), and they show that neither Rad52 protein nor Rad59 protein has a strong preferential affinity for dsDNA ends.
When the human Rad52 protein was examined using this assay, the results were similar, although the HaeIII-linearized pBR322 DNA was found to compete for hRad52 protein binding slightly better than the EcoRV-linearized pBR322 DNA (Fig. 4D). The concentration of competitor DNA required to dissociate the hRad52 protein-dsDNA complex to free ϳ50% of the labeled dsDNA was about 20 and 40 M for the HaeIII-and EcoRV-cut DNA, respectively. Although this difference suggests that hRad52 protein has a somewhat higher affinity for dsDNA ends than for internal sites, this 2-fold difference is much less than the 22-fold difference expected. We, therefore, estimate that the hRad52 protein has at most an ϳ100-fold greater affinity for dsDNA ends than internal dsDNA sites. However, the results from the nuclease protection experiments suggest little or no preferential binding.
Visualization of Rad52-and Rad59-DNA Complexes-We used AFM to directly visualize yeast Rad52-and Rad59-DNA complexes on linearized plasmid dsDNA. We observed that Rad52 protein often forms discrete bead-like complexes on plasmid dsDNA and that it binds to both dsDNA ends and internal sites (Fig. 5A). In AFM, tip convolution results in measured widths that are larger than the actual widths of DNA or protein-DNA complexes. The width of naked dsDNA was measured to be 11.7 Ϯ 1.8 nm (n ϭ 19), which is ϳ10 nm larger  than expected (2 nm). Similarly, the diameter of the Rad52 protein complexes was measured to be about 28.1 Ϯ 3.4 nm (n ϭ 10), implying that its actual diameter is at most 18 nm. To verify that Rad52 protein did not prefer to bind dsDNA ends, we reduced the protein concentration until only a few Rad52 protein complexes were bound to each dsDNA molecule. On 40 clearly resolved individual EcoRV-linearized pBR322 molecules, 88 bound Rad52 proteins were discerned; 79 of these proteins (90%) were bound to internal DNA sites, whereas only 9 were located close to a DNA end (10%). Notably, even though some of the complexes visualized contained as many as five Rad52 protein bound to the DNA, we did not observe any dsDNA molecules with both ends occupied. Our results with yeast Rad52 protein are consistent with the prior AFM study of human Rad52 protein-dsDNA complexes, which showed that only 6 -18% of hRad52 protein was bound at the end of bluntended dsDNA (39). In contrast, with PstI-linearized pBR322, which has 4-nucleotide 3Ј-overhangs on both ends, 17 of 55 Rad52 complexes observed (31%) were bound to ends. This result shows that the presence of an ssDNA tail as short as 4 nucleotide results in 3-fold more binding of Rad52 at the DNA ends and is the same as that reported for hRad52 protein (39).
We also examined complexes of Rad59 protein and EcoRVlinearized pBR322 dsDNA. We saw that Rad59 protein also formed discrete complexes (Fig. 5B). The size and shape of the Rad59 complexes are highly variable, suggesting the presence of different oligomeric states. For these molecules, the width of the naked dsDNA was measured to be 9.5 Ϯ 0.8 nm (n ϭ 15), which is ϳ7.5 nm larger than the actual width of DNA. The observed diameter of the Rad59 complexes ranged from 13 to 30 nm (n ϭ 33). The average diameter of the smallest complexes was measured to be 14.8 Ϯ 1.2 nm (n ϭ 16) or at most 7 nm after correction, suggesting these complexes are Rad59 monomers. To determine whether Rad59 protein displays any preference for dsDNA ends, we decreased the Rad59 protein concentration to 100 nM (the resultant ratio was 1 protein/100 nucleotides). The 35 individual dsDNA molecules that could be clearly resolved contained 105 bound Rad59 proteins; 99 of the proteins (94%) were bound to internal sites, whereas only 6 were bound to a dsDNA end (6%). As for Rad52 protein, even though some of the DNA molecules visualized contained as many as 10 Rad59 protein bound to the DNA, none had both ends occupied by Rad59 protein. Thus, our observations using AFM confirm the biochemical data that neither Rad52 nor Rad59 protein have a preference for dsDNA ends.

DISCUSSION
In this report we showed that the S. cerevisiae Rad52, S. cerevisiae Rad59, and human Rad52 proteins do not have a significant preference for binding to dsDNA ends. Although our conclusions differ, our experimental results are not necessarily inconsistent with those previously reported for hRad52 and S. pombe Rad22 proteins (36,38), where an apparent end binding preference had been inferred. In fact, many of the previous interpretations can be explained by our observations. First, the protein concentration required to achieve efficient protection from exonuclease digestion was extremely high compared with that of dsDNA ends (158 proteins per dsDNA end). A reasonable explanation for this requirement of excess protein over DNA ends is that most of the Rad52 protein is sequestered by binding to the internal DNA sites. Second, in the previous reports DNA end binding was measured either by the release of label from a uniformly labeled plasmid DNA using trichloroacetic acid precipitation (36) or by the change in mobility of plasmid DNA using agarose gel electrophoresis (38). ExoIII, the nuclease used in the exonuclease protection assays, is a distributive enzyme that hydrolyzes ϳ6 nucleotides before dissociating from a dsDNA end (41), leaving a ssDNA tail. Given that  DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 Rad52 (16), hRad52 (39), and Rad59 (31) proteins have a greater binding affinity for ssDNA than for dsDNA, the previous studies cannot eliminate the alternative explanation that Rad52 protein protected the remainder of the plasmid DNA substrate by binding to the ssDNA region that was generated by a limited extent of exonucleolytic degradation. Indeed, the limited degradation (evident in the slight smearing of the mobility) of the dsDNA in the ExoIII protection assay (Fig. 1A) indicates that Rad52 protein cannot efficiently protect the dsDNA substrate until the ends are processed into ssDNA tails. Furthermore, previous AFM results (39) with hRad52 and our findings with yeast Rad52 show that the presence of just a 4-nucleotide ssDNA tail results in a 2-3-fold increase in Rad52 binding to a DNA end. Thus, the previous findings and our interpretation of those findings are consistent with our results here showing that the affinity of Rad52 protein for internal DNA sequences is comparable with its affinity for dsDNA ends.

The DNA Binding Preference of RAD52 and RAD59 Proteins
Moreover, in the work reporting the DNA end binding preference of the S. pombe Rad22 protein (38), the two nuclease protection assays were conducted in buffers recommended by the manufacturers, i.e. 0.66 mM Mg 2ϩ for ExoIII and 10 mM Mg 2ϩ with 50 mM NaCl for HaeIII. 4 Because the DNA binding affinity of Rad52 and Rad59 proteins is reduced by increasing concentrations of monovalent and divalent cations (14,16,31), the ability of these proteins to protect DNA in nuclease protein assays is concomitantly reduced (Figs. 1 and 2). If the DNA binding affinity of S. pombe Rad22 protein is similarly sensitive to Mg 2ϩ and NaCl concentration, then it is likely that the Rad22 protein also would be less able to protect dsDNA from both endo-and exonuclease digestion at high Mg 2ϩ and NaCl concentrations. Hence, the comparison of nuclease protection experiments that were performed at different reaction conditions could lead to an erroneous conclusion. If one were to compare our exonuclease protection results obtained for Rad52 protein at 0.66 mM Mg 2ϩ (Fig. 1C, ExoIII trace), where strong protection is observed, to the endonuclease protection results at 10 mM Mg 2ϩ (Fig. 1D, HaeIII trace), where weak protection is observed, then one would incorrectly conclude that Rad52 protein had a significant binding preference for dsDNA ends. It remains to be determined whether our simple explanation affords a straightforward explanation for the apparent inconsistency between our data with the S. cerevisiae proteins and the S. pombe homologue (38).
Our results suggest that the hRad52 and the S. pombe Rad22 proteins protect dsDNA from exonuclease degradation by binding to the ssDNA tail of dsDNA or by saturating the entire dsDNA molecule rather than by preferential binding to the blunt dsDNA end. All of the Rad52 homologues that we examined protected dsDNA from endonuclease degradation as well as they protected the DNA from exonuclease degradation with no strong bias. In addition, when we substituted T7 gene 6 exonuclease (a 5Ј to 3Ј exonuclease) for ExoIII in our end-protection assays (using 5Ј-end-labeled dsDNA), comparable protection was observed for Rad52, hRad52, and Rad59 to that seen with ExoIII. 5 In the electrophoretic mobility shift experiments we also observed that a dsDNA competitor with a low concentration of ends was as efficient a competitor for Rad52 protein binding as dsDNA with a high concentration of ends. Experiments using AFM to directly observe Rad52 protein-dsDNA or Rad59 protein-dsDNA complexes also revealed no strong preference for dsDNA ends by either protein. A similar conclusion was reached for hRad52 using AFM to show that the protein bound to ssDNA regardless of the presence of dsDNA ends or the dsDNA end structure (39); in competition experiments with DNA that had an internal ssDNA gap or an ssDNA tail as part of the dsDNA substrate, hRad52 protein bound exclusively to the ssDNA region. These findings are consistent with our conclusion that there is no preferential dsDNA end binding by Rad52 protein and its homologues. Rather, because they have a much greater affinity to ssDNA, these proteins are localized to DSB sites after the dsDNA end is processed into ssDNA.
It is our interpretation that, because Rad52 protein does not have a significant specificity for dsDNA ends, it cannot control the choice of repair pathway for DSBs, homologous recombination versus NHEJ. One of the most notable differences between the NHEJ and homologous recombination pathways is the level of DNA processing at the first step of repair. Although extensive nucleolytic degradation may occur before end joining (33,42), typically only limited DNA processing is involved in NHEJ (33,43). In contrast, DSB repair via homologous recombination requires extensive DNA processing to generate 3Ј-terminated ssDNA (44). If Rad52 protein indeed competed for dsDNA ends with the NHEJ pathway, then one would expect that in the absence of RAD52 the rate of DSB processing would be greatly reduced. Instead, DSB processing is either unaffected (45) or slightly increased (46,47) in rad52 mutant cells. The increased DNA degradation seen in the rad52 mutant is likely due to the inability of the homologous recombination pathway to repair DSBs at a later stage, presumably because Rad52 protein is not available to promote the assembly of Rad51 protein onto ssDNA tails. Furthermore, in a plasmid-rejoining experiment where yeast cells were directly transformed with linear dsDNA containing various lengths of an ssDNA overhang, the efficiency of dsDNA rejoining was either solely dependent on the yKU70 and yKU80 or on the RAD52 (48); when the ssDNA overhang was less than eight nucleotides, plasmid re-ligation was solely dependent on the yeast Ku70-80 proteins but not Rad52; with longer ssDNA overhangs, plasmid re-ligation was largely dependent on RAD52, presumably via a SSA-type mechanism but not on yKU70-80. These results also imply that the NHEJ pathway is essentially not in competition with the homologous recombination pathways.
Cytological evidence also argues that Rad52 protein cannot be the first recombination factor to act on the DSB in homologous recombination pathways. During meiosis, in response to DSB formation, Rad52 protein forms nuclear foci that extensively co-localize with, but do not form earlier than replication protein A foci (12). This observation suggests that localization of Rad52 protein onto the DSB sites requires processing of the DSB into ssDNA. In contrast, in irradiated cells, ϳ50% Mre11 foci do not co-localize with replication protein A foci, and for-mation of Mre11 foci precedes that of Rad52 foci (49), indicating MRE11 acts upstream to RAD52 before DSB resection.
Taken together, Rad52 protein is unlikely to be a gatekeeper in DSB repair. The steps at which Rad52 protein acts are after the DSB has been resected to form ssDNA. Rad52 protein is important in several critical post-processing aspects of homologous recombination, such as facilitating Rad51-ssDNA complex formation (18 -21, 50) and the annealing of complementary ssDNA in the presence of replication protein A (14 -17). For all of these essential biochemical activities, Rad52 protein acts on ssDNA rather than dsDNA. Likewise, Rad59 protein has no preferential end binding affinity, but rather, it binds to ssDNA with a greater affinity than to dsDNA (31). Rad59 protein interacts with Rad52 protein in vivo (30) and enhances the annealing function of Rad52 under suboptimal conditions in vitro (32). Hence, we believe that the substrate on which Rad59 protein acts is also ssDNA.
If Rad52 protein does not regulate the choice of DSB repair pathways, then which proteins(s) channels DSBs to the appropriate repair pathways? The S. cerevisiae Mre11⅐Rad50⅐Xrs2 (MRX) complex or the mammalian Mre11⅐Rad50⅐Nbs1 (MRN) complex is likely to be such a regulator. The MRX/MRN complex is important in both repair pathways as well as in intra-S phase checkpoint control (51)(52)(53). Mre11 localizes to DSB sites independently of DSB processing (49,54). Deletion of MRE11, RAD50, and XRS2 retards DSB processing (44,45,55) in unsynchronized cells and completely abolishes it in G 2 -arrested cells (54). All of these phenotypes are expected for proteins that are important in the channeling function. In S. pombe, another pathway was also reported to act at the stage of DSB processing (56). This pathway is under the control of cellular Cdc2-cyclin B activity and specifically regulates homologous recombination events in G 2 independently of and in parallel with RAD50. Thus, it seems that there are multiple regulatory mechanisms to channel DSBs to different repair pathways and that the choice of DSB repair pathway must be made at a stage much earlier than the Rad52-dependent steps. The mechanism of this choice is yet remains to be elucidated.