Physical Interaction between Recombinational Proteins Rhp51 and Rad22 in Schizosaccharomyces pombe *

In eukaryotes, Rad51 and Rad52 are two key components of homologous recombination and recombinational repair. These two proteins interact with each other. Here we investigated the role of interaction between Rhp51 and Rad22, the fission yeast homologs of Rad51 and Rad52, respectively, on the function of each protein. We identified a direct association between the two proteins and their self-interactions both in vivo and in vitro. We also determined the binding domains of each protein that mediate these interactions. To characterize the role of Rhp51-Rad22 interaction, we used random mutagenesis to identify the mutants Rhp51 and Rad22, which cannot interact each other. Interestingly, we found that mutant Rhp51 protein, which cannot interact with either Rad22 or Rti1 (G282D), lost its DNA repair ability. In contrast, mutant Rad22 proteins, which cannot specifically bind to Rhp51 (S379L and P381L), maintained their DNA repair ability. These results suggest that the interaction between Rhp51 and Rad22 is crucial for the recombinational repair function of Rhp51. However, the significance of this interaction on the function of Rad22 remains to be characterized further.

Double strand breaks (DSBs) 1 in chromosome are very harmful, and the failure in repair of DSB results in severe genomic instabilities that can lead to cell death or, in higher eukaryotes, to cancer (1). In eukaryotes, two major pathways have been known to deal with DSBs (2). The nonhomologous end-joining pathway joins adjacent broken DNA ends, resulting in errors in the junction region. In contrast, the homologous recombination (HR) pathway takes advantage of the undamaged homologous DNA strands, resulting in accurate repair of DSBs.
RAD52 is the only gene among the RAD52 epistasis group that is required for virtually all homologous recombination events (3). Purified S. cerevisiae and human Rad52 proteins (ScRad52 and HsRad52, respectively) have an annealing activity of two complementary single-stranded DNAs (10), and they promote the strand exchange activity of Rad51 in the presence of RPA (11). In addition, HsRad52 and S. pombe Rad52 (Rad22) appear to bind to DSBs (12,13). Rad51 is an eukaryotic structural and functional counterpart of Escherichia coli RecA (14). Purified ScRad51 and HsRad51 have the homologous pairing/ strand exchange activity that is the core catalytic activity of HR (15). Rti1/Rad22B (hereafter referred to simply as Rti1) is another Rad52 homolog found in S. pombe (6,16). Deletion of both rad22 ϩ and rti1 ϩ leads to more severe defects compared with each single mutant, suggesting that the role of ScRad52 would be diverged to Rad22 and Rti1.
It has been reported that there are many protein-protein interactions in which ScRad51 and ScRad52 are involved. In accordance with its enzymatic role, ScRad51 is the center of the interactions between HR proteins in that it interacts with itself, ScRad52, ScRad54, ScRad55, and the large subunit of RPA (17)(18)(19)(20)(21). ScRad52 also interacts with itself and with all three subunits of RPA (22).
In this report we explored the interactions between Rad51 and Rad52, the two key molecules of HR, using their fission yeast homologs Rhp51 and Rad22, respectively (23,24). We determined their binding domains and investigated the effect of Rhp51-Rad22 interaction on the DNA repair function of each protein.
Plasmids and Enzymes-For yeast two-hybrid analysis, derivatives of pGBT9 (Clontech), named pGBT9 -1 and pGBT9 -2, which have different reading frames, were generated by inserting or deleting nucleotides into the EcoRI site. Various restriction or PCR fragments of the rhp51 ϩ or rad22 ϩ genes were fused in-frame with GAL4 DNAbinding domain (GAL4BD; pGBT9 derivatives) or GAL4 activation domain (GAL4AD; pDP4, -7, -12) ( (28), rad57 ϩ (29), rti1 ϩ (6), ssb1 ϩ , and ssb2 ϩ (30) genes were amplified from their respective cDNA or genomic DNA by PCR and cloned into the BamHI site of pGBT9 derivatives or pDP plasmids. For the complementation assay, SpeI-PstI fragment of p51.551 (31) were replaced by the cognate fragment from mutagenized rhp51, resulting in the full-length rhp51 genes containing each mutation. The BamHI-SmaI fragments of rad22 open reading frame containing each mutation were cloned into the equivalent sites of pREP81 (32) for the complementation assay. Restriction enzymes and modifying enzymes were purchased from New England Biolabs, Boehringer Mannheim, Promega, or Takara.
Yeast Two-hybrid Analysis-A nonlethal ␤-galactosidase plate assay was performed as described by Duttweiler (33).
GST Pull-down Assay-Lysates of E. coli cells harboring pET22 (12) and pGST51 (31) were incubated with a 10-l bed volume of glutathione S-transferase (GST)-Sepharose beads in an 0.5-ml reaction containing 50 mM Tris-HCl (pH 7.4), 0.5 mM EDTA, 0.5 mM dithiothreitol, and 100 mM NaCl at 4°C for 2 h with rotation. The beads were precipitated and washed six times with the same buffer. The protein complexes were eluted and analyzed by SDS-polyacrylamide gel electrophoresis followed by immunoblotting.
Co-immunoprecipitation-Methods and conditions for coimmunoprecipitation of Rhp51 and Rad22 in S. pombe wild type cells are described by Kim et al. (31).
Random Mutagenesis and Screening of Binding Mutants-The hydroxylamine mutagenesis method was employed (34,35) to introduce random point mutations into rhp51 ϩ or rad22 ϩ genes. Plasmid DNA pBD51-(102-365) and pAD22-(310 -469) harboring Rhp51 residues 102-365 in pGBT9 -2 and Rad22 residues 310 -469 in pDP12, respectively, were subjected to mutagenesis. Each plasmid was incubated in 0.5 ml of reaction buffer (50 mM sodium pyrophosphate (pH 7.0), 1 M NaCl, 2 mM EDTA, and 0.5 M hydroxylamine) at 75°C for 30 min and purified by spun-column. After amplification through E. coli, the mutagenized pBD51-(102-365) or pAD22-(310 -469) plasmids were transformed into S. cerevisiae Y190 harboring pAD22 or pBD51, respectively, for yeast two-hybrid screening. Each transformant was replica plated onto solid media lacking histidine. The colonies showing a His Ϫ phenotype were subjected to a ␤-galactosidase color assay. Plasmid DNAs were recovered from the white colonies and recloned into the same intact plasmid to eliminate mutations in plasmid backbone. The changes of nucleotide sequence in insert DNAs were determined using Sequenase version 2.0 kit (Amersham Pharmacia Biotech). In case multiple mutations were found in a single clone, they were separated into each mutation by restriction enzyme digestion or site-directed mutagenesis.
Survival Test-Complementation of methylmethane sulfonate (MMS) and UV sensitivity of rhp51⌬ and rad22⌬ strains by mutant rhp51 and rad22 genes has been described previously (31).

Homo-and Heterotypic Interactions of Rhp51 and Rad22-
We examined the association of Rhp51 and Rad22 in the wild type cell by co-immunoprecipitation. Fig. 1A shows that Rad22 protein can be co-precipitated with Rhp51, indicating that these two proteins can form a protein complex not only when they are overproduced (31) but also at normal levels. We also confirmed their interaction by GST pull-down assay. Recombinant Rad22 protein was co-eluted with GST-Rhp51 fusion protein but not with GST protein (Fig. 1B, upper panel, lane 4), indicating that these two proteins can associate directly without the help of any other proteins in S. pombe. The interaction between Rhp51 and Rad22 was also visualized by yeast twohybrid ␤-galactosidase plate assay (Fig. 1C). As expected, cells harboring both reciprocal sets of Rhp51 and Rad22 proteins fused with GAL4AD and GAL4BD turned blue. In addition, both proteins were self-interacting. Taken together, these results indicate that Rhp51 and Rad22 proteins directly associate both in vivo and in vitro and that each of them also self-associates.
Mapping Domain of Rhp51 Confers Interaction with Rad22 or with Itself-To determine the region of Rhp51 required for association with Rad22 or with itself, a yeast two-hybrid assay was employed using various truncations of Rhp51 fused with GAL4BD (Fig. 2). For interaction with Rad22, the carboxylterminal two-third of Rhp51 (residues 102-365, fragment 2) was required, which encompasses the central RecA homology region and the carboxyl-terminal region. On the other hand, for self-association of Rhp51, a shorter fragment (residues 102-312, fragment 12) was sufficient. Therefore the carboxyl-terminal 53 amino acids appeared to have the residues critical for Rad22 binding. However, the amino-terminal region (residues 1-102), which corresponds to the region of ScRad51 responsible for the interaction with ScRad52 or with itself (18), was not required for the Rad22 binding to Rhp51.
Mapping Domains of Rad22 Conferring Interaction with Itself or with Rhp51-Yeast two-hybrid analysis was also applied to verify the region of Rad22 responsible for the interaction with Rhp51 or with itself (Fig. 3). Fusion proteins of GAL4AD with truncations of Rad22 and GAL4BD with Rhp51 or with Rad22 were subjected to a ␤-galactosidase plate assay. As shown in Fig. 3, the regions for Rhp51 and self-association were completely separated in the Rad22 protein. For self-interaction, the amino-terminal region (residues 1-162, fragment 6) was sufficient, whereas the carboxyl-terminal region (residues 310 -469, fragment 11) dealt with the interaction with Rhp51. However, as the fragment became shorter, the signals for Rhp51 binding progressively decreased, suggesting that the entire structure may also be required for the interaction with Rhp51 or for stabilization of the protein expression.
Screening of Point Mutations of Rhp51 or Rad22 That Interrupt Protein-Protein Interactions of Both Proteins-To identify the significance of Rhp51-Rad22 interaction in recombinational repair, we designed a screening method to identify mutations in Rhp51 that disrupt the interaction with Rad22 and vice versa. Based on the binding domain mapping, we introduced random point mutations into plasmids pBD51-(102-363) and pAD22-(310 -469) by the hydroxylamine mutagenesis method. In yeast two-hybrid screening using these mutant libraries as prey and pAD22 or pBD51 as bait, we isolated five Rhp51 and four Rad22 mutant clones that did not show a blue color. We determined the causal mutations by nucleotide sequence anal-ysis, separated multiple mutations in a single clone into single mutations by site-directed mutagenesis, and thereby acquired five rhp51 and four rad22 mutant genes carrying single point mutation.
All five mutations of the rhp51 gene were localized in the central RecA-homologous core region (Fig. 4A). None of the mutations was found in the conserved ATP-binding motifs (36) or the corresponding residues of putative RecA DNA binding sites (37). Except for one mutation that resulted in the stop codon (Q228NS), all four mutations caused single amino acid substitution. As shown in Fig. 4A, G177S, C179F, Q228NS, and G282D completely impaired the Rad22 binding of Rhp51, whereas the L274P mutation reduced its binding ability. Western blot analysis revealed that all mutant proteins except Q228NS were expressed well (data not shown), indicating that the failure in interaction was not caused by the lack of protein expression.
Each of the rad22 Mutations Caused a Single Amino Acid Substitution-Interestingly, all four mutations (S377F, S379L, P381S, and P381L) were found in very close proximity, and two were different substitutions of the same amino acid (Fig. 4B), suggesting that this region could be a binding epitope for Rhp51.
Interaction between Rhp51 Mutants and Recombination Factors-We performed two-hybrid analysis employing Rhp51, Rad22, Rhp54, Rhp55, Rhp57, Rti1, and the large and middle subunits of RPA. These results revealed that Rhp51 interacts with itself, Rad22, Rhp54, Rhp57, and Rti1, whereas Rad22 interacts with itself, Rhp51, and Rti1. However, we could not find the interactions between Rad22 and the large or middle subunit of RPA that have been reported in S. cerevisiae (data not shown). Our results were similar to those of Tsutsui et al. (38), except that the self-interactions of Rhp51 and Rhp57 and the interaction between Rhp51 and Rti1 were newly found. Based upon these results, we examined whether the mutant Rhp51 could interact with its binding partners other than Rad22. All five mutations had different effects on the interaction of Rhp51 with its binding partners (Fig. 5A). The Rhp51 C179F could not bind with any of the proteins examined, indicating that this cysteine residue could be crucial for the entire structure of Rhp51, for example, by disulfide bridge formation. Rhp51 G177S and G282D, which had failed to interact with Rad22, also did not interact with Rti1, a Rad22 homolog, suggesting that Rad22 and Rti1 may bind a similar epitope in Rhp51. Rhp51 G177S also did not interact with Rhp54. Rhp51 Q228NS, the nonsense mutation, did not interact with itself or Rti1 and showed reduced interaction with Rad22 and Rhp54. Rhp51 L274P interacted with Rhp57, Rad22, and Rti1 but not with Rhp51 or Rhp54. These results showed that none of the five mutants was restricted to a single interaction. However, given that Rad22 and Rti1 are homologs, Rhp51 G282D could be considered a Rad22-specific mutant.
Interaction between Rad22 Mutants and Recombination Factors-In our two-hybrid analysis, Rad22 interacted not only with Rhp51 but also with itself and Rti1, a Rad22 homolog (data not shown). We also investigated the interactions of mutant Rad22 with Rad22 or Rti1 (Fig. 5B). Rad22 S379L and P381L interacted with Rad22 and Rti1. On the other hand, Rad22 S377F and P381S did not interact with Rad22 and Rti1. These results suggested that S379L and P381L specifically disrupted the interaction of Rad22 with Rhp51, leaving other interactions intact. Immunoblot experiment revealed that Rad22 S379L, S377F, and P381S were not detected with our antibody (data not shown). However, because anti-Rad22 antibody recognizes the carboxyl-terminal region of Rad22, it is unclear whether the immunoblot results reflect the lack of protein expression. Rather, it is likely that the structural change of these three proteins may cause the failure in detection of protein by immunoblot.
Effects of Protein-Protein Interaction on DNA Repair Function of Rhp51 and Rad22-To investigate the significance of protein-protein interactions on the function of Rhp51, we examined whether the mutant rhp51 genes could rescue the DNA damage sensitivity of the rhp51 deletion mutant (Fig. 6A). All five mutant rhp51 genes barely rescued MMS and UV sensi-tivity of the rhp51 deletion mutant. The complementation efficiency of these mutant rhp51 genes were slightly better than an empty plasmid, suggesting that these mutations severely impaired the DNA repair ability of Rhp51. Interestingly, Rhp51 G282D, which is specifically unable to interact with Rad22 homologs, were biologically inactive, indicating that the interaction with Rad22 homologs would be indispensable for the DNA repair function of Rhp51. However, we could not evaluate the significance of the individual interaction between Rhp51 and other binding partners because of the lack of specific mutation that disrupted each individual interaction.
We also examined whether mutant rad22 genes could rescue DNA damage sensitivity of rad22 deletion mutants. As expected, Rad22 S377F and P381S were unsuccessful in complementing MMS and UV sensitivity. However, Rad22 S379L and P381L, which were specific to the interaction with Rhp51, were biologically active in their DNA repair function (Fig. 6B). These results suggested that the interaction with Rhp51 might not be essential for the DNA repair ability of Rad22.

Domains of Rhp51 for Protein-Protein Interactions-
The amino-terminal region of ScRad51, where sequence homology exists between eukaryotic Rad51 homologs but is missing in RecA, is known as a domain of self-interaction and for ScRad52 binding (18). However, recent studies of ScRad51 and HsRad51 argued that the amino-terminal domain is not required for the interaction with either Rad52 or Rad51 but is responsible for the DNA binding (39 -41). In our mapping, we also found the binding region of Rhp51 for the Rad22 or itself located in the broad regions spanning the central RecA homology region and the carboxyl terminus but not in the amino terminus. In addition, point mutations that affect these interactions were dis- tributed throughout the central RecA homology region of Rhp51. Therefore it is likely that binding domains would be constituted by the cooperation of several domains or that distantly located residues might shape a single domain.
A comparison of multiple interactions between mutant Rhp51 and its binding partners revealed that a single mutation could affect multiple interactions. Among the mutations, at least three mutations appeared to have a severe influence on the global structure of protein by substitution of a cysteine residue (C179F), by introduction of proline (L274P), or by the truncation of polypeptide (Q228NS). In contrast, G177S and G282D seem to be involved specifically in the interactions of Rhp51 with Rad22 and Rti1. Amino acid sequence alignment of Rhp51, ScRad51, and HsRad51 demonstrated that both Gly 177 and Gly 282 are well conserved in Rad51 homologs (Fig. 7A, Table I). Interestingly, the neighboring residues of Gly 177 and Gly 282 were found to be involved in the Rad51-Rad52 interaction in ScRad51 and HsRad51. Mutation in Phe 259 of HsRad51, which corresponds to Phe 281 of Rhp51, disrupted the interaction with HsRad51-HsRad52 in a GST pull-down assay (41). Three-dimensional modeling by similar mutational study in ScRad51 revealed that Gly 210 , Gly 211 , and Ala 320 , which correspond to Gly 175 , Gly 176 , and Ala 284 of Rhp51, constituted a single Rad52-binding domain (40). Therefore, it is highly likely that residues around Gly 177 and Gly 282 of Rhp51 could also comprise a single domain that is responsible for the Rad22 binding.
Domains of Rad22 for Protein-Protein Interactions-Unlike broad mapping of Rhp51 domains, those of Rad22 for Rhp51 binding and self-association were separately located in the carboxyl and the amino terminus, respectively. Yeast twohybrid assays demonstrated that the Rad51-binding domains of ScRad52 and HsRad52 also exist in their C-terminal regions (17,42). In addition, the existence of mutations that specifically disrupt binding with Rhp51 (S379L, P381L) also supports the possibility that the carboxyl-terminal region indeed forms Rhp51-binding domain. Because the carboxyl-terminal region lacks sequence homology between Rad52 homologs, Rad51-Rad52 interaction, although conserved from yeast to human, is likely to be species-specific. Interestingly, sequence alignment of Rad22 and Rti1 demonstrates that two short regions in the carboxyl terminus of Rad22 (residues 334 -346 and 365-384) have a relatively higher sequence similarity than other regions in the carboxyl terminus (Fig. 7B). Moreover, Ser 379 and Pro 381 residues are conserved between Rad22 and Rti1. Because Rti1 also binds with Rhp51, these two regions may serve as an Rhp51-binding domain of Rad22 and Rti1. The lack of Rhp51 binding in the short fragments of residues 310 -380 or 366 -469 (Fig. 3) suggests that both of these regions are required and that they might comprise a single Rhp51-binding domain.
Significance of Rhp51-Rad22 Interaction in Recombinational Repair-All five mutant Rhp51 proteins were unable to interact with at least two binding partners and were nonfunctional in DNA repair. Therefore, although the significance of each individual interaction is uncertain, some or all of such interactions are likely to be required for the proper functioning of Rhp51.
On the other hand, four Rad22 mutants were divided into two groups. Rad22 S379L and P381L were specifically defective in interaction with Rhp51 and biologically active, whereas Rad22 S377F and P381S had neither protein binding ability nor DNA repair activity. The latter phenotypes could be the result of a lack of protein expression because Rad22 S377F and P381S were not observed in immunoblot. However, because Rad22 S379L, which is functional, was not detected either by immunoblot and our anti-Rad22 antibody recognized the carboxyl terminus region of Rad22, these phenotypes are presumed to be caused by a severe change in protein structure rather than by lack of expression.
In our experiments, the significance of Rhp51-Rad22 inter-FIG. 6. Complementation of MMS and UV sensitivity of rhp51 or rad22 deletion mutants by mutant rhp51 or rad22 genes. A, rhp51 deletion mutants harboring each mutant rhp51 gene in multicopy plasmids were serially diluted by 10-fold and spotted onto plates containing 0.002% MMS or irradiated by 100 J/m 2 of UV after spotting. Plates were incubated for 3 days and then photographed. B, rad22 deletion mutants harboring each pREP81 mutant rad22 were grown in the absence of thiamine for induction of protein expression and spotted onto plates in the presence of MMS (0.004%) or UV (200 J/m 2 ). action appears to be different for each protein. Impairment of this interaction significantly affected Rhp51 function (Rhp51 G282D) but not that of Rad22 (Rad22 S379L and P381L). We can postulate at least two possibilities for this discrepancy. First, Rad22 S379L and P381L may retain enough binding activity to complement in vivo, although their Rhp51 binding is undetectable by yeast two-hybrid analysis; this is probably because we employed a multicopy plasmid with an attenuated nmt1 promoter for the complementation assay. Second, Rad22 S379L and P381L may interact indirectly with Rhp51 via a protein that can bind to both Rhp51 and Rad22. Rti1, a Rad22 homolog in S. pombe, fits this assumption well. In our twohybrid analysis, Rti1 was the only protein that could evidently bind to both Rhp51 and Rad22. In addition, Rhp51 G282D lost its ability to interact with Rti1, whereas Rad22 S379L and P381L could bind to Rti1. Therefore, it is plausible that Rad22 S379L and P381L might interact with Rhp51 indirectly via Rti1, resulting in no effect in Rad22 function.
Our observation that interaction between Rhp51 and Rad22 is essential to the DNA repair function of Rhp51 suggests that there would be an essential but unknown role of Rad22 other than as a co-factor of strand exchange reaction. One possible role of Rad22 in terms of interaction with Rhp51 is that of a mediator to direct Rhp51 into the site of action, as there are a few reports that Rad52 homologs bind to the end of duplex DNA. However, because there is a controversy on this property of Rad52, this possibility will require extensive verification.