Rad51p is a eukaryotic homolog of RecA, the central homologous pairing and strand exchange protein in Escherichia coli. Rad54p belongs to the Swi2p/Snf2p family of DNA‐stimulated ATPases. Both proteins are also important members of the RAD52 group which controls recombinational DNA damage repair of double‐strand breaks and other DNA lesions in Saccharomyces cerevisiae. Here we demonstrate by genetic, molecular and biochemical criteria that Rad51 and Rad54 proteins interact. Strikingly, overexpression of Rad54p can functionally suppress the UV and methyl methanesulfonate sensitivity caused by a deletion of the RAD51 gene. However, no suppression was observed for the defects of rad51 cells in the repair of γ‐ray‐induced DNA damage, mating type switching or spontaneous hetero‐allelic recombination. This suppression is genetically dependent on the presence of two other members of the recombinational repair group, RAD55 and RAD57. Our data provide compelling evidence that Rad51 and Rad54 proteins interact in vivo and that this interaction is functionally important for recombinational DNA damage repair. As both proteins are conserved throughout evolution from yeasts to humans, a similar protein–protein interaction may be expected in other organisms.
DNA repair is an essential part of the cellular response to DNA damage. Such damage occurs as a consequence of the cellular metabolism or by environmental exposure to chemical or physical agents. Accurate repair of DNA lesions avoids mutations and thereby ensures survival. Genetic analysis in Saccharomyces cerevisiae has identified mutations in nearly 100 genes, that affect DNA damage repair (reviewed in Friedberg et al., 1995). Some of these genes have been assigned to epistasis groups that reflect an organization into DNA damage repair pathways (reviewed in Haynes and Kunz, 1981; Game, 1983). They include the nucleotide excision repair pathway (NER, RAD3 group), the error prone/post‐replication repair pathway (RAD6 group) and the recombinational repair pathway (RAD52 group) (see Friedberg et al., 1995).
DNA lesions like double‐strand breaks (DSBs) that affect both strands of a double helix are particularly damaging. DSBs are the major genotoxic lesions induced by ionizing radiation, and the widespread use of this radiation in medical diagnosis and therapy requires a better understanding of the repair mechanisms for such lesions. In S.cerevisiae, such damage is repaired primarily by homologous recombination using the DNA sequence residing on the sister chromatid or homolog to restore the DSB in a way that is intrinsically error free (Resnick, 1976; Szostak et al., 1983). Other pathways of DSB repair include single‐strand annealing and end joining mechanisms that are intrinsically error prone (reviewed in Roth and Wilson, 1988; Haber, 1992). These pathways can be detected in S.cerevisiae (Fishman‐Lobell et al., 1992; Kramer et al., 1994), but may be of greater relative importance in higher eukaryotes (for review, see Weaver, 1995; Wood, 1996). Mutations in the RAD52 group genes result in pleiotropic defects in DNA damage repair including extreme sensitivity to ionizing radiation or other DSB‐inducing agents and alkylating agents like methyl methanesulfonate (MMS) as well as sensitivity to UV irradiation (reviewed in Game, 1993; Friedberg et al., 1995; Shinohara and Ogawa, 1995). Moreover, additional defects in various aspects of chromosome metabolism during vegetative growth and during meiosis have been described for some mutants, although not all genes have been analyzed systematically for all phenotypes (reviewed in Kleckner et al., 1991; Petes et al., 1991; Roeder, 1995).
The characterization of the RAD52 group genes identified RAD51, RAD52 and RAD54 as the most important members based on the severity of the mutant phenotypes (Game, 1993). RAD51 encodes a protein with sequence similarity to the central bacterial recombination protein RecA (Aboussekhra et al., 1992; Basile et al., 1992; Shinohara et al., 1992). The protein forms protein–DNA filaments indistinguishable from RecA protein–DNA filaments (Ogawa et al., 1993) and promotes homologous pairing and strand exchange in vitro (Sung, 1994; Sung and Robberson, 1995). Rad51 and Rad52 proteins interact as shown by genetic (Milne and Weaver, 1993; Schild, 1995), biochemical (Shinohara et al., 1992) and molecular (Donovan et al., 1994) experiments. Whereas the RecA paradigm (reviewed by West, 1992; Kowalczykowski and Eggleston, 1994) provides a framework for understanding the role of Rad51p, the function of Rad52p remains more mysterious. However, the interaction of Rad52p with Rad51p and with Rpa (Firmenich et al., 1995), the eukaryotic single‐stranded DNA‐binding protein, suggests that Rad52 protein also acts directly on the DNA during recombinational repair.
The RAD54 gene has been cloned (Calderon et al., 1983) and sequenced (Emery et al., 1991), and the predicted 898 amino acid Rad54 protein exhibits significant sequence homology to Swi2p/Snf2p. This homology includes seven sequence motifs that have been proposed to identify members of a family of putative DNA helicases (Gorbalenya and Koonin, 1993). The Snf2p/Swi2p family includes 28 eukaryotic and one prokaryotic member (Eisen et al., 1995). The proteins function in transcriptional regulation (Snf2p/Swi2p, Mot1p, lodestar, brahma) and various DNA damage repair pathways including NER (Rad16p), post‐replication repair (Rad5p), strand‐specific repair (Rad26p, ERCC6) and recombinational repair (Rad54p). Biochemical analysis of Snf2p/Swi2p (reviewed in Peterson and Tamkun, 1995) failed to demonstrate helicase activity for the protein but identified a DNA‐stimulated ATPase activity. Therefore, it is likely that Rad54p also acts directly on DNA, constituting, together with Rad51 and Rad52 proteins, a central component, critical for recombinational repair.
Recently, an astounding evolutionary conservation of the NER pathway between yeasts and humans has been established, making S.cerevisiae an attractive model system for this repair system (reviewed in Hoeijmakers, 1993). A similar situation might be true for the recombinational repair RAD52 group, although the studies in higher eukaryotes are far from being complete. Already it appears that RAD51 is conserved in most, if not all eukaryotes (Shinohara et al., 1993; reviewed in Heyer, 1994), and homologs in Archaea have been identified recently (Sandler et al., 1996). Moreover, genes with limited sequence homology to RAD52 have been identified in higher eukaryotes (Bezzubova et al., 1993; Bendixen et al., 1994; Muris et al., 1994). The third component of the putative core machinery in recombinational repair, Rad54p, is also conserved in structure and function, as the human homolog, hHR54, has been demonstrated as the first mammalian recombinational repair gene to complement a deletion in the corresponding S.cerevisiae gene (Kanaar et al., 1996).
Here we show that the S.cerevisiae Rad51 and Rad54 proteins interact in the two‐hybrid system and that an in vivo complex containing both proteins can be demonstrated biochemically by co‐immunoprecipitation. Genetic experiments provide compelling evidence that this protein–protein interaction is significant in recombinational DNA damage repair. Most importantly, overexpression of Rad54p suppressed DNA damage repair‐related phenotypes caused by a deletion in the RAD51 gene. It appears, therefore, that this protein–protein interaction is important for the molecular mechanism of recombinational repair.
Rad51p and Rad54p interact in the two‐hybrid system
Recombinational repair is likely to be mediated by a multi‐protein complex. Identifying the subunits of such a complex and their mutual interactions will be of great importance in understanding the molecular mechanism of this process. Using the two‐hybrid system (Fields and Song, 1989; Gyuris et al., 1993), we have identified an interaction between the Rad54 and Rad51 proteins of S.cerevisiae (Figure 1). The Rad51–activator construct activates transcription in combination with LexA–Rad54(1–898) but not with LexA fused to a negative control protein (Trypanosoma brucei brucei profilin) or to the C‐terminal 607 amino acids of Rad54p [Rad54(291–898)]. Some activation was detected using a LexA fusion to the first 356 amino acids of Rad54p [Rad54(1–356)], with a value significantly different from the others as the standard deviations did not overlap. No evidence for Rad54p–Rad54p interaction was found in this system. Both fusion constructs that showed interaction, pEG202‐RAD54(1–898) and pJG4‐5‐RAD51(1–400), are biologically active as they complemented the MMS‐sensitive phenotype of the respective deletion mutations in RAD54 and RAD51.
Protein complexes containing Rad51 and Rad54 protein exist in vivo
To corroborate this interaction, we demonstrated that complexes containing Rad51 and Rad54 proteins exist in S.cerevisiae cells. For this purpose, we used anti‐Rad54p antibodies in immunoprecipitation experiments and showed that the two proteins can be co‐precipitated (Figure 2A, lanes 4). In a control strain lacking Rad54 protein, no Rad51p was precipitated (Rad51p was tagged with the HA epitope, lanes 5), showing that the HA‐Rad51p is not simply precipitating under the experimental conditions. In strains lacking HA‐Rad51 and Rad51 proteins, Rad54p was precipitated (lane 6 left), but no signal was detected with the anti‐HA antibody (lane 6 right). Omitting the anti‐Rad54p antibody in the precipitation (lanes 1–3) and analyzing extracts of the strains used (lanes 7–9) by immunoblotting demonstrated the specificity of the antibodies used. The biological significance of the occasional appearance of lower molecular weight forms of Rad54p and HA‐Rad51p visible in lanes 4 and 6–9 is unclear. They probably constitute proteolytic degradation products.
Correspondingly, Rad54p–Rad51p‐containing complexes were also identified after precipitation of the Rad51 protein partner (Figure 2B, lanes 4). The control experiments demonstrated again the specificity of the antibodies used and the specificity of the interaction. Omitting the anti‐HA antibody in the precipitation (lanes 1–3) and analyzing extracts of the strains used (lanes 7–9) demonstrated the specificity of the antibodies in precipitation and immunoblotting. The specificity of the Rad51p–Rad54p association (Figure 2, lanes 4) is demonstrated by the control experiments in lanes 5 and 6, showing that Rad54p is not simply precipitating under the experimental conditions. The amount of HA‐Rad51 protein was consistently lower in the absence of Rad54p (compare Figure 2A and B, lanes 7 and 8 right, and in Figure 2B, lanes 4 and 5 right), suggesting a possible stabilization of HA‐Rad51p by the complex formation with Rad54p.
In the co‐immunoprecipitation experiments, the proteins were overexpressed resulting for Rad54p in a 50‐ to 100‐fold increased protein level compared with the level originating from the chromosomal RAD54 gene (B.Clever and W.‐D.Heyer, unpublished result). The use of the tag precluded the determination of how much HA‐Rad51p was overproduced in comparison with native Rad51p. However, co‐precipitation of HA‐Rad51p with Rad54p did not require such elevated levels of Rad54p, as HA‐Rad51p could also be co‐precipitated from extracts of a strain containing only the normal chromosomal copy of RAD54 (data not shown).
To assess the stability of the Rad51p–Rad54p interaction, we performed co‐immunoprecipitations using increasingly specific washing conditions. Immunoprecipitations were done exactly as for Figure 2, and precipitates were washed subsequently four times in buffer with increasing concentrations of NaCl and Triton X‐100 before immunoblot analysis. As shown in Figure 3, under the most stringent conditions tested (200 mM NaCl/1% Triton X‐100), essentially no loss of signal for the co‐precipitating HA‐Rad51 protein was detected after immunoblot analysis. These conditions are more stringent than in vivo. Therefore, the stability of the protein–protein interaction is consistent with the in vivo existence of a Rad51p–Rad54p complex.
Co‐overexpression of Rad51 and Rad54 proteins causes negative effects in S.cerevisiae
Genetic experiments gave evidence for the biological relevance of the Rad51p–Rad54p interaction. Overexpression of neither Rad54 nor Rad51 proteins alone resulted in negative effects in S.cerevisiae (Table I), whereas co‐overexpression of both proteins led to a synthetic phenotype of 3‐fold reduced survival in the presence of MMS. Mutations in all RAD52 group genes led to sensitivity against MMS which induces DNA damage repaired by the recombinational repair pathway (reviewed in Friedberg et al., 1995; see also Figure 6).
Furthermore, in the absence of MMS, i.e. under normal growth conditions, co‐overexpression of both proteins significantly reduced the cell survival in wild‐type diploid cells (see Table I for strains). This can be demonstrated by determining the ratio of the colony‐forming units (c.f.u.) on galactose‐containing medium (overexpression condition) to the c.f.u. on glucose (non‐overexpressing condition). Whereas in double vector control (pGal/YEp13) and single overexpressors (pGalRad54/YEp13, pGal/YEp13RAD51) no negative effect on galactose was observed, in cells co‐overexpressing both Rad51 and Rad54 proteins survival on galactose was reduced significantly to 0.50 ± 0.15.
Overexpression of Rad54p suppresses some, but not all DNA damage repair defects caused by a deletion in the RAD51 gene
Overexpression of the Swi2p/Snf2p homolog Rad54p in cells lacking the RecA homolog Rad51p surprisingly led to the suppression of DNA damage repair phenotypes caused by the rad51Δ mutation (Figures 4,5,6). rad51Δ causes hypersensitivity against UV irradiation (reviewed by Friedberg et al., 1995; see Figure 4). This phenotype was partially suppressed by Rad54 protein overexpression. At 100 J/m2, survival of a rad51Δ strain is reduced ∼1400‐fold compared with wild‐type. Overexpression of Rad54p in a rad51Δ cell restored survival 90‐fold. This survival is below that of wild‐type cells (16‐fold), which is a statistically significant difference.
This suppression effect is dependent on active Rad54 protein, as the rad54‐K341R mutation abolishes the suppression activity of Rad54p (Figure 4). This mutation affects the central lysine residue in the putative ATP‐binding fold (Walker A‐box) of the Rad54 protein and abolishes function. The rad54‐K341R allele did not complement the MMS phenotype of a rad54Δ mutant as a centromere‐plasmid borne copy or transplaced in the genome (J.Schmuckli‐Maurer and W.‐D.Heyer, unpublished data).
Moreover, Rad54p overexpression also suppressed the MMS hypersensitivity in rad51Δ cells (Figure 5). At 0.01% MMS, survival of the rad51Δ strain is reduced 84‐fold compared with wild‐type. High Rad54p levels improves this 70‐fold, to a survival only 1.2‐fold below wild‐type, which is not statistically significant. As for UV survival, the suppression was dependent on active Rad54p, as the rad54‐K341R allele did not exhibit this suppression effect (Figure 5).
The defects of rad51Δ cells in γ‐ray survival, spontaneous and UV‐induced mitotic recombination (intragenic recombination his7‐1×his7‐2), and repair of the HO endonuclease‐mediated DSB during mating type switching were not suppressed by overexpression of Rad54p (data not shown).
The expression levels of wild‐type and mutant Rad54 proteins in cells grown in glucose‐ and galactose‐containing medium were determined by immunoblot analysis (data not shown). The pGalRAD54 and the pGalrad54‐K341R vectors resulted in ∼10‐fold overexpression of proteins in glucose and 50‐ to 100‐fold overexpression in galactose. This analysis therefore established that the loss of suppression activity with rad54‐K341R was due to non‐functional protein, as stable rad54‐K341Rp was detectable.
The suppression of rad51Δ by high Rad54p levels is genetically dependent on both RAD55 and RAD57
To characterize further the suppression of the rad51Δ MMS‐hypersensitive phenotype by Rad54p overexpression, we analyzed the genetic requirements for this effect. The suppression was dependent on both the RAD55 and RAD57 genes, which encode proteins with sequence similarity to Rad51 and RecA proteins (reviewed in Heyer, 1994).
Mutations in RAD55 and/or RAD57 exhibit overlapping phenotypes which are less severe than those of rad51, rad52 or rad54 mutants (compare in Figure 6 rad51Δ with rad55Δ rad57Δ). This is especially the case at 30°C, the temperature used here, as deletions in RAD55 or RAD57 exhibit an unusual, cold‐sensitive phenotype (Lovett and Mortimer, 1987). The triple mutant (rad51Δ rad55Δ rad57Δ) exhibits exactly the same MMS sensitivity as the rad51Δ single mutant (see Figure 6). Overexpression of Rad54p in rad51Δ cells again showed the suppression (see Figure 6 and legend) already seen in Figure 5. Overexpression of Rad54p in the triple mutant (rad51Δ rad55Δ rad57Δ; Figure 6) or in either double mutant (rad51Δ rad57Δ, rad51Δ rad55Δ; data not shown) did not have any effect on the MMS sensitivity of the cells. Based on this quantitative difference, we conclude that the Rad54p suppression effect depends on RAD55 and RAD57. If this effect were independent of Rad55 and Rad57 proteins, the survival in the triple mutant strain overexpressing Rad54p should have been equal to the survival of the rad55Δ rad57Δ double mutant strain, which is clearly not the case (see Figure 6). To exclude a possible negative effect of Rad54p overexpression in a rad55Δ rad57Δ background, we tested this possibility and no negative effect was detected (see Figure 6).
The rad55Δ rad57Δ mutations could be complemented to wild‐type level by a plasmid expressing both proteins (Figure 6). Overexpression of Rad54p did not suppress the MMS sensitivity of rad55, rad57 or rad55 rad57 mutants under the conditions analyzed (25 or 30°C, glucose or galactose; data not shown).
Rad51 and Rad54 proteins form a complex in vivo
Recombinational repair of DSBs and other lesions to DNA in S.cerevisiae is controlled by the evolutionarily conserved RAD52 group of genes. Rad51 and Rad54 proteins perform crucial functions in this process (reviewed by Game, 1993). Here we provide strong evidence that Rad51 and Rad54 proteins interact in vivo (Figures 1,2,3). The interaction measured in the two‐hybrid system is specific and requires a distinct region of Rad54p. Rad54 protein is organized grossly in two regions, an N‐terminal region (amino acids 1–326) called Rad54A and a C‐terminal region (amino acids 327–898) called Rad54B. Rad54B includes the region of sequence homology to the Swi2p/Snf2p family and the putative DNA helicase motifs (Eisen et al., 1995), whereas Rad54A is specific for Rad54p and shows only significant sequence homology to Rad54 proteins from other organisms (Kanaar et al., 1996; Muris et al., 1996). The data from the two‐hybrid system suggest that the Rad54A part may be involved in the Rad51p interaction, whereas no indication for an involvement of the Rad54B part has been obtained in this system. The B‐part fusion was found to be as abundant as the full‐length fusion protein in the cell (data not shown). Unrelated two‐hybrid experiments also suggested that the Rad54B fusion is stable in vivo (H.Interthal and W.‐D.Heyer, unpublished results). As no other protein in S.cerevisiae shares significant homology with the Rad54A part, it is unlikely that the interaction with Rad51p is fortuitous. This might have been argued had the interaction involved the Rad54B part, as several S.cerevisiae proteins share significant homology with Rad54p in this region. Moreover, using a Rad54p fragment encompassing the first 114 amino acids, in vitro interaction with purified Rad51p was observed (A.Shinohara, personal communication). Therefore, the Rad51p‐specific interaction appears to be mediated by the N‐terminal region of Rad54p.
Functional interaction between Rad51 and Rad54 proteins
The genetic experiments reported here (Figures 4,5,6 Table I) provide compelling evidence that the Rad51p–Rad54p interaction is functionally significant for the mechanism of recombinational repair. Rad51p, like its prokaryotic counterpart RecA protein, is probably active in the central step of recombinational repair, i.e. search for homology and homologous pairing. Rad51p is a DNA‐dependent ATPase that forms nucleoprotein filaments active in homology search in vitro (Ogawa et al., 1993; Sung, 1994; Sung and Robberson, 1995). Rad54p has no prokaryotic counterpart to serve as a paradigm. However, information is available about two other S.cerevisiae members of the Swi2p/Snf2p family. Swi2p/Snf2p itself is a DNA‐stimulated ATPase that acts in a high molecular weight complex involved in overcoming the general repressive effect of chromatin on transcription (reviewed in Peterson and Tamkun, 1995). Mot1p, another family member, is also a DNA‐stimulated ATPase that is capable of removing TATA‐binding protein (TBP) from DNA, thereby resulting in global repression of RNA polymerase II transcription (Auble et al., 1994). Despite some superficial similarities between Rad51p, Swi2p/Snf2p and Mot1p being DNA‐dependent/stimulated ATPases, Rad51p is different from the other proteins (see Kowalczykowski and Eggleston, 1994; Shinohara and Ogawa, 1995). Therefore, it was surprising to find that high Rad54p levels suppress certain DNA damage repair defects of a deletion in the RAD51 gene. Suppression was found for the MMS and UV sensitivity but not for the γ‐ray sensitivity, the spontaneous and UV‐induced mitotic recombination defects, and the mating type switching defect caused by a deletion of the RAD51 gene. These results point to a heterogeneity in the genetic requirements within the RAD52 pathway for different types of DNA damage repair and different genetic endpoints.
The Rad54p‐mediated suppression of certain DNA damage repair‐related phenotypes caused by rad51Δ could either be direct, by partial functional replacement of Rad51p by Rad54p, or indirect, involving other proteins. Recently, the crystal structure of the PcrA protein from Bacillus stearothermophilus has been solved (Subramanya et al., 1996). PcrA protein is a DNA helicase with seven sequence domains shared by many DNA helicases and other DNA‐dependent or stimulated ATPases including Rad54p (see Gorbalenya and Koonin, 1993). Surprisingly, the ATP‐binding domain of this DNA helicase was structurally highly similar to that of the Escherichia coli RecA protein (Story and Steitz, 1992; Story et al., 1992), as pointed out by Subramanya et al. (1996). As Rad51 protein forms nucleoprotein filaments equivalent to RecA–DNA filaments (Ogawa et al., 1993) and has an amino acid sequence that can be modeled into the RecA crystal structure (Story et al., 1993), partial functional replacement of Rad51p by Rad54p cannot be dismissed. Such an interpretation would predict that this effect is autonomous, i.e. independent of other genes. Testing this hypothesis, we found that the suppression by high Rad54p levels was dependent on Rad55 and Rad57 proteins. Therefore, the suppression is not achieved by functional replacement but rather by an indirect route involving other proteins. The inability of Rad51p overexpression to suppress a deletion in the RAD54 gene (data not shown) is consistent with this interpretation. Interestingly, RAD54 has been isolated as a high copy suppressor of the meiotic arrest and spore inviability phenotypes of a dmc1 mutant (D.Bishop, personal communication). Dmc1p is another S.cerevisiae protein with homology to E.coli RecA that specifically functions during meiosis (Bishop et al., 1992).
The role of Rad55 and Rad57 proteins
Functional suppression of a rad51Δ mutation by high levels of Rad54p involves both Rad55 and Rad57 proteins. They are also members of the recombinational repair group (reviewed in Friedberg et al., 1995; Shinohara and Ogawa, 1995) and exhibit overlapping mutant phenotypes which are less severe than those of rad51, rad52 or rad54 mutants. Rad55p (Lovett, 1994) and Rad57p (Kaus and Mortimer, 1991) share certain sequence similarity with E.coli RecA and Rad51p (reviewed in Heyer, 1994; Shinohara and Ogawa, 1995). Both proteins have been shown to interact (Hays et al., 1995; Johnson and Symington, 1995), consistent with our result of their mutual requirement for the Rad54p‐mediated suppression effect. In addition, Rad55p and Rad51p interact (Hays et al., 1995; Johnson and Symington, 1995). Overexpression of Rad51p can partially suppress DNA damage repair phenotypes of mutations in the RAD55 and RAD57 genes (Hays et al., 1995; Johnson and Symington, 1995), but Rad54p overexpression does not. Our data are the first evidence that Rad55 and Rad57 proteins correspondingly are involved in suppression of a rad51 mutation, at least under conditions of high Rad54p levels. The molecular details of these complex interactions between four recombinational repair proteins remain to be determind. Rad55 and Rad57 proteins may serve as accessory factors to the Rad51p–DNA filament during homology search, providing functions similar to, and partially redundant with, Rad51p during recombinational repair. Alternatively, these data might be interpreted to mean that deletion of RAD55 and/or RAD57 affects MMS sensitivity independently of rad51. However, such an interpretation is not consistent with the fact that the rad51 mutation is epistatic to the rad55 and/or rad57 mutation, as the single rad51 mutation has the same MMS sensitivity as the rad51 rad55 rad57 triple mutant (Figure 6) or the rad51 rad55 and rad51 rad57 double mutants (data not shown).
Protein–protein interactions in recombinational repair: possible roles for Rad54p
Specific protein–protein interactions are important in numerous DNA metabolic processes, including DNA replication (reviewed in Kornberg and Baker, 1992), mismatch repair (reviewed in Kolodner, 1996), transcription (see Struhl, 1995), NER (reviewed in Wood, 1996) and recombination in E.coli (reviewed in Kowalczykowski et al., 1994) and phage T4 (reviewed in Kowalczykowski and Eggleston, 1994). Therefore, it is not surprising to find that protein–protein interactions play a significant role during recombinational repair in eukaryotes. Identifying the protein partners interacting in a putative recombinational repair complex or complexes and understanding their mutual interactions will give important insights into the mechanism of this process. Multiple individual interactions between recombinational repair proteins (Rad51p–Rad52p, Shinohara et al., 1992; Rad51p–Rad55p, Rad55p–Rad57p, Hays et al., 1995; Johnson and Symington, 1995; Rad52p–Rpa1p, Firmenich et al., 1995; for additional references see above; Rad51p–Rad54p, this study) have been identified in S.cerevisiae. They provide a functional framework which has been interpreted as evidence for a ‘putative recombinosome’ including heterotrimeric Rpa, Rad51, Rad52, Rad54, Rad55 and Rad57 proteins (Hays et al., 1995; this study). It is unclear presently whether these proteins are assembled in one complex or in different sub‐complexes. The existence of different sub‐complexes may be indicated by the heterogeneity in the genetic requirement found in the suppression of rad51 phenotypes by Rad54p overexpression. In addition, these individual interactions might not necessarily all occur simultaneously. The existence of a large ‘repairosome’ in NER (Svejstrup et al., 1995) recently has been questioned (Guzder et al., 1996). The individual interactions discussed above, and in particular the interaction between Rad51 and Rad54 proteins, might be relevant at several steps during recombinational repair. The biochemical data and the RecA paradigm strongly suggest a role for the Rad51p–DNA filament in homology search and pairing, a central step in this process. Rad54p might be involved in the assembly of such a filament as an assembly factor or molecular matchmaker (see Sancar and Hearst, 1993). Alternatively or in addition, Rad54p may have an active role during Rad51p‐mediated homology search as an accessory protein similar to gp41 helicase in phage T4 recombination (Salinas and Kodadek, 1995). It has been suggested that Rad54p (and Rad51, Rad55, Rad57 proteins) are accessory proteins needed to gain access to otherwise inaccessible regions of the chromosome (Sugawara et al., 1995), possibly in a way similar to the role of Swi2p/Snf2p in overcoming transcriptional repression by chromatin (reviewed by Peterson and Tamkun, 1995). Finally, Rad54p could act in the turnover of the Rad51 protein–DNA filament in analogy to the action of Mot1p during RNA polymerase II transcription (see Auble et al., 1994). This would allow assembly of the crucial Rad51p filament at the site of DNA damage, which under conditions of limiting Rad51p concentration might result in severe mutant phenotypes for rad54. A similar speculation might explain the suppression of rad51Δ by high levels of Rad54p through an increased turnover of Rad55 and Rad57 proteins.
Materials and methods
Strains, media and growth conditions
The S.cerevisiae strains used in this study are listed in Table II. Standard media and standard growth conditions were as described (Sherman et al., 1982; Bähler et al., 1994). MMS was added to solid medium cooled to 50°C before pouring, and plates were used within 1 day. The S.cerevisiae strains were transformed using the lithium acetate method modified according to Schiestl and Gietz (1989).
Plasmid constructions and DNA methods
Plasmids were constructed by standard methods. For the LexA fusions: pEG202‐RAD54(1–898) contains the NcoI (Klenow resected)–SalI RAD54 fragment of pWDH251 (Kanaar et al., 1996) in EcoRI (Klenow fill in)–SalI‐digested pEG202 (Gyuris et al., 1993); pEG202‐RAD54(1–356) derives from pEG202‐RAD54(1–898) by ApaI–SalI digestion, Klenow fill‐in and religation; pEG202‐RAD54(291–898) is the RcaI (Klenow fill‐in)–SalI RAD54 fragment of pEG202‐RAD54(1–898) in EcoRI (Klenow fill‐in)–SalI‐digested pEG202.
Activator fusions: pJG4‐5‐RAD51(1–400) contains an EcoRI (partial cleavage)–SalI RAD51 fragment of pEG202‐RAD51 in EcoRI–XhoI‐digested pJG4‐5 (Gyuris et al., 1993); pEG202‐RAD51 was constructed by inserting the StuI–DraI RAD51 fragment of YEp13‐RAD51‐23 (Calderon et al., 1983) in the SmaI site of pEG202; pJG4‐5‐RAD54(1–898) contains the RAD54 EcoRI fragment of pEG202‐RAD54 in EcoRI‐digested pJG4‐5.
pGal vector is pWDH181 described previously (Bashkirov et al., 1995), and YEp13 is described in Broach et al. (1979). pGalRAD51 was constructed by ligating the StuI–DraI 1.4 kb RAD51 fragment of YEp13‐RAD51‐23 (Calderon et al., 1983) into pWDH129 (Holler et al., 1995) cut with EcoRI and NheI after fill‐in with Klenow. pWDH251 expresses a His(6)‐tagged Rad54 protein using the pT7‐7 system (Tabor and Richardson, 1985). The XhoI–EcoRI RAD54 fragment of pGalRAD54 was cloned in the XhoI–EcoRI‐digested pTrcHisC (Invitrogen) to result in pWDH202. The NcoI (partial, Klenow fill‐in)–EcoRI fragment encoding His(6)‐Rad54 of pWDH202 was cloned in NdeI (Klenow fill‐in)–EcoRI‐digested pT7‐7 to result in pWDH251. pGalRAD55‐RAD57 also derives from pWDH181, placing RAD55 and RAD57 individually under the control of the Gal promoter in a head‐to‐head fashion (J.King and W.‐D.Heyer, unpublished; details to be published elsewhere).
The rad54‐K341R allele was generated by site‐directed mutagenesis using the oligonucleotide 5′‐G CTG GGT AGG ACA TTG C‐3′, otherwise pGalrad54‐K341R is identical to pGalRAD54. The rad51Δ::kanMX, rad55Δ::kanMX and rad57Δ::kanMX alleles were generated by PCR product‐mediated transformation as described (Wach et al., 1994), deleting the entire open reading frame.
Genetic and physiological methods
The two‐hybrid system in S.cerevisiae strain EGY48 was used as described previously (Gyuris et al., 1993). Quantitative β‐galactosidase assays were performed as described (Harshman et al., 1988). The LexA activator fusions did not appreciably induce reporter gene expression by themselves. They exhibited activity in the repression assay, demonstrating that the respective LexA fusion proteins were synthesized, transported to the nucleus and active in binding the LexA operator.
To establish MMS survival curves, transformants were grown in minimal medium selective for the presence of the plasmids to stationary phase, diluted in H2O, and aliquots were plated on medium lacking leucine and/or uracil with glucose or galactose (2% w/v) as carbon source and the indicated concentration of MMS. UV survival curves were performed at fluencies of 0–100 J/m2 on plates containing galactose. Colonies were counted after 6 days of incubation at 30°C. γ‐Ray survival curves were performed at 0–200 Gray irradiating the cells in solution using a 137Cs source and plating appropriate dilutions on plates containing glucose and galactose.
Recombination rates between the his7‐1 and his7‐2 hetero‐alleles were determined by fluctuation tests using the method of the median (Lea and Coulson, 1949). For measurement of UV‐induced recombination, cells were irradiated with 25 J/m2. Repair of the HO endonuclease‐mediated DSB at the MAT locus was monitored as survival after inducing the HO gene controlled by the GAL1‐10 promoter and determining the number of colonies on medium containing glucose and galactose.
For co‐immunoprecipitation experiments, strain FF181079‐2 transformed with plasmid pGalRAD54 (labeled pRAD54 in Figure 2) and pJG4‐5‐RAD51 (labeled pRAD51 in Figure 2) as well as strain FF18973 transformed with either plasmid were grown in medium selecting for the presence of the plasmids containing raffinose (2% w/v). Both plasmids encoded functional Rad51 and Rad54 proteins as they complemented deletion mutations in the respective genes. Five hours after addition of galactose (2% w/v), cells were harvested, washed, resuspended in lysis buffer (20 mM HEPES pH 7.4, 100 mM KOAc, 2 mM MgOAc) and frozen in liquid N2. Fifty OD units of frozen cells were thawed on ice, resuspended in 200 μl of lysis buffer containing 2 mM phenylmethylsulfonyl fluoride (PMSF), and extracted using 0.45 μm glass beads. Three hundred μg of extract were mixed with 500 μl of IP buffer (50 mM Tris–HCl pH 7.5, 0.2 % Triton X‐100, 50 mM NaCl) containing 0.5 mg/ml bovine serum albumin (BSA), to which either 0 or 1 μg of anti‐Rad54 antibodies or 0 or 5 μg of anti‐HA antibodies (12CA5, Boehringer Mannheim) were added. After overnight rocking at 4°C, 100 μl of protein G–Sepharose FF (Pharmacia) slurry previously washed in IP buffer were added and incubated for 1 h. Immunoprecipitates were washed twice with 500 μl of IP buffer containing BSA and twice in IP buffer before the beads were resuspended in 1× Laemmli buffer. Proteins were separated by SDS–PAGE. Immunoblot detection was by ECL (Amersham). Additional co‐immunoprecipitation experiments using cells not overexpressing Rad54p were done exactly as above but using 4 mg of whole cell extract. Purification of the recombinant His(6)‐Rad54 protein from E.coli cells transformed with pWDH251, production of polyclonal antisera in rabbits and immunoaffinity purification of antibodies were done essentially as described (Santos‐Rosa et al., 1996) and will be described fully elsewhere (B.Clever, J.Schmuckli‐Maurer and W.‐D.Heyer, in preparation).
Note added in proof
After submission of the final version of this manuscript the paper by H.Jiang, Y.Xie, P.Houston, K.Stemke‐Hall, U.H.Mortensen, R.Rothstein and T.Kodadek (1996) Direct association between the yeast RadS1 and RadS4 recombination proteins. J. Biol. Chem., 271, 33181–33186, appeared describing also the interaction between RadS1 and RadS4 proteins mediated by the N‐terminus of RadS4p.
We are grateful to C.Naula for the profilin control plasmid, to R.Brent for generously providing the two‐hybrid system, to D.Bishop and A.Shinohara for communicating results prior to publication, to P.Linder for the HO plasmid, and all members of the Bern recombination groups for fruitful discussions. We very much appreciated the kind help of F.Fabre in supplying strains and for his hospitality during the stay of H.I. in Paris. We thank S.West for his kind help with the manuscript as well as for supplying pT7‐7, R.Walmsley for critically reading the manuscript, and R.K.Mortimer for his help in the initial stages. This work was supported by a career development award (START) and a research grant by the Swiss National Science Foundation to W.‐D.H. J.K. is the recipient of a Human Frontiers Science Program postdoctoral fellowship.
↵† B.Clever and H.Interthal contributed equally to this work
- Copyright © 1997 European Molecular Biology Organization