|Home | About | Journals | Submit | Contact Us | Français|
During DNA double-strand-break (DSB) repair by recombination, the broken chromosome uses a homologous chromosome as a repair template. Early steps of recombination are well characterized: DSB-ends assemble filaments of RecA-family proteins that catalyze homologous pairing and strand-invasion reactions. By contrast, the post-invasion steps of recombination are poorly characterized. Rad52 plays an essential role during early steps of recombination by mediating assembly of RecA-homolog, Rad51, into nucleoprotein filaments. The meiosis-specific RecA-homolog, Dmc1, does not show this dependence, however. By exploiting the Rad52-independence of Dmc1, we reveal that Rad52 promotes post-invasion steps of both crossover and noncrossover pathways of meiotic recombination in Saccharomyces cerevisiae. This activity resides in the N-terminal region of Rad52, which can anneal complementary DNA strands, and is independent of its Rad51-assembly function. Our findings show that Rad52 functions in temporally and biochemically distinct reactions, and suggest a general annealing mechanism for reuniting DSB-ends during recombination.
The template-directed repair of chromosomes by homologous recombination plays essential roles in DNA replication, DNA repair and homolog segregation during meiosis (Heyer, 2006; Krogh and Symington, 2004; Hunter, 2006). The in vivo events of meiotic recombination are particularly well characterized. Spo11-catalyzed DNA double-strand-breaks (DSBs) are processed to produce long single-stranded 3’-tails. Resected DSB-ends then assemble filaments of RecA-like proteins, which catalyze homologous pairing and strand-invasion reactions. Most eukaryotes possess two RecA homologs: Rad51, which functions in both mitotic and meiotic cells, and the meiosis-specific Dmc1 (Neale and Keeney, 2006; Shinohara and Shinohara, 2004). Assembly of Rad51 and Dmc1 nucleoprotein filaments requires accessory factors to overcome the inhibitory effect of the single-stranded binding protein, RPA, bound to DSB-ends. The archetypal mediator is Rad52, which catalyzes assembly of Rad51 filaments onto RPA-coated single-stranded DNA (Gasior et al., 2001; Gasior et al., 1998; Krejci et al., 2002; New et al., 1998; Shinohara and Ogawa, 1998; Sugawara et al., 2003; Sung, 1997). During meiosis, assembly of Rad51 into chromatin-associated immunostaining foci shows a strict dependence on Rad52 (Gasior et al., 1998). Dmc1 does not share this dependence, however, and its assembly may be mediated by dedicated factors such as the Mei5-Sae3 complex (Bishop, 1994; Hayase et al., 2004).
Products of strand-invasion can be detected in vivo as Single-End-Invasion (SEI) intermediates, in which one DSB-end has undergone strand-exchange with a homologous chromosome (Hunter and Kleckner, 2001) (Figure 1C). Interaction of the second DSB-end and recombination-associated DNA synthesis leads to formation of the double-Holliday junction (dHJ), which is subsequently resolved into crossover products (Schwacha and Kleckner, 1995) (Figure 1C). The majority of meiotic DSBs are repaired not as crossovers, however, but as noncrossovers without reciprocal exchange of chromosome arms. Noncrossovers are thought to form primarily via synthesis-dependent strand-annealing (SDSA) in which an invading DSB-end is first extended by DNA synthesis, displaced from the template and then annealed to the other DSB-end (Nassif et al., 1994; Paques and Haber, 1999).
In contrast to the early steps of meiotic recombination, proteins that catalyze post-invasion steps are ill defined. Along the crossover pathway, the SEI-to-dHJ transition requires the second DSB-end to be “captured” by the SEI. Theoretically, this could occur via a second Dmc1/Rad51-catalyzed strand-invasion reaction. Alternatively, the canonical DSB-repair model (Szostak et al., 1983) posits that the second DSB-end anneals to the displaced-loop of the SEI, which is enlarged as DNA polymerase extends the invading DSB-end. In the context of this model, the complementary strand-annealing activity of Rad52 is proposed to catalyze second-end-capture because, in addition to its mediator function, Rad52 uniquely catalyzes the annealing of complementary DNA strands coated with RPA (Mortensen et al., 1996; Shinohara et al., 1998; Sugiyama et al., 1998; Sugiyama et al., 2006; Wu et al., 2006).
Investigating putative post-invasion functions of Rad52 during Rad51-dependent DSB-repair is thwarted by its epistatic function in assembling Rad51 filaments. By exploiting the Rad52-independence of Dmc1, however, we show that Rad52 promotes both the SEI-to-dHJ transition and the completion of noncrossover recombination. This activity of Rad52 resides in the conserved N-terminal region, which possesses strand-annealing activity. These data suggest a model in which crossover and noncrossover pathways share a common late step – Rad52-mediated annealing of DSB-ends – and differ only with respect to the fate of the original invading DSB-end.
Cultures of wild-type and rad52Δ null mutant cells were induced to undergo meiosis and intermediate steps of recombination were monitored using the HIS4LEU2 physical assay system (Figure 1) (Hunter and Kleckner, 2001). The HIS4LEU2 locus contains a hotspot for formation of meiosis-specific DSBs and DNA events at this site are monitored using a series of gel electrophoresis and Southern hybridization assays. Cell samples (12–15 per time-course) are first treated with psoralen to produce DNA interstrand crosslinks, which stabilize the joint molecule (JM) intermediates, SEIs and dHJs (Figure 1C). XhoI restriction site polymorphisms between parental “Mom” and “Dad” homologs produce fragments diagnostic for parental and recombinant chromosomes, DSBs and JMs. Each hybridizing signal is quantified using a phosphorimager. DSBs and crossovers are quantified from one-dimensional gels (Figure 1B), whereas SEIs and dHJs are analyzed using native/native two-dimensional (2D) gels, which reveal the branched structure of these intermediates (Figure 1D) (Bell and Byers, 1983; Hunter and Kleckner, 2001).
Spore formation in rad52Δ cells is inefficient and only dead spores are formed (Game et al., 1980) (data not shown). Using physical assays, significant levels of crossover products are detected in rad52Δ cells, however (~40% of wild-type levels; Figure 2A and 2B), consistent with the study of Borts et al. (Borts et al., 1986). Moreover, crossover formation in rad52Δ cells shows the same dependence on the pro-crossover factor, Msh5 (Borner et al., 2004; Hollingsworth et al., 1995), as seen in wild-type cells suggesting that the normal differentiated pathway is being utilized (Supplemental Figure S1).
DSBs in rad52Δ cells form with wild-type timing and the majority of breaks appear to turnover, albeit with a delay of ~3 hrs (Figure 2A and 2B); this is mirrored by a similar delay of the first meiotic division (Figure 2B). At late times (>5 hrs), persistent hybridizing species arise that appear to be ~0.3 kb longer than initial DSB molecules (Figure 2C, central panel). Further analysis shows that these molecules are likely to be hairpins resulting from stem-loop formation within the 3’- strand of the DSB-end followed by intra-strand priming of DNA synthesis (see Supplemental Figure S2). We suggest that these aberrant structures form at persistent DSB-ends that failed to be captured in rad52Δ mutant cells (see below). In sharp contrast to rad52Δ cells, rad52Δ dmc1Δ double mutants accumulate high levels of DSBs that persist for the duration of the experiment, indicating a severe block at the DSB stage (Figure 2B). Thus, meiotic recombination in rad52Δ cells progresses beyond the DSB stage in a Dmc1-dependent fashion.
2D gel analysis reveals that rad52Δ cells are defective in the transition from SEIs to dHJs (Figure 2D and 2E). While SEIs reach wild-type levels, peak steady-state levels of interhomolog-dHJs (IH-dHJs) are reduced by 8-fold, from 1.2% of hybridizing DNA in wild type to 0.15% in rad52Δ. By contrast, formation of dHJs between sister-chromatids (IS-dHJs) does not appear to be significantly altered by the rad52Δ mutation. The difference between IH-dHJ and IS-dHJ levels in rad52Δ cells is at least partially explained by the observation that a larger fraction of DSBs is repaired between sister-chromatids in the absence of Rad51 filaments (Schwacha and Kleckner, 1997) (J.P.L and N.H., unpublished). In addition, dependence of dHJ formation on Rad52 may differ between IH and IS pathways. Overall, peak steady-state dHJ levels (IH-dHJs + IS-dHJs) are reduced ~3.5-fold in rad52Δ cells (“total dHJs” in Figure 2E).
The molecular phenotype of the original rad52 allele, rad52-1 (Game et al., 1980) (an A90V amino-acid change) was also examined and found to be qualitatively very similar to the rad52Δ null mutation (Supplemental Figure S3). If anything, rad52-1 exaggerates the defects described in rad52Δ null cells, perhaps because Rad52-1 protein can still localize to recombination sites and block access by other enzymes. Moreover, formation of both IH-dHJs and IS-dHJs is defective in rad52-1 cells.
The dramatic reduction of IH-dHJs in rad52Δ cells could, in theory, be explained by a large reduction in the lifespan of IH-dHJs. To address this possibility, we measured dHJ levels in strains carrying an ndt80Δ mutation, which causes dHJs to accumulate (Allers and Lichten, 2001a) (Figure 3). This analysis confirms the inferences made in NDT80 strains. In fact, the effects of the rad52Δ mutation are more exaggerated in the ndt80Δ background. IH-dHJs accumulate to 20-fold lower levels in rad52Δ ndt80Δ cells than in the ndt80Δ single mutant, and total dHJs are reduced by 8-fold. In contrast, IS-dHJ levels are indistinguishable between the two strains. The difference in IH-dHJ levels in NDT80 and ndt80Δ backgrounds could be explained if rad52Δ NDT80 cells do in fact form very few IH-dHJs, but these intermediates have a longer lifespan than normal (relative to both IH-dHJs in wild-type cells and IS-dHJs in rad52Δ cells); as such, the steady-state dHJ level in rad52Δ NDT80 cells will not accurately report the absolute reduction in IH-dHJs. Consequently, the IH-dHJ/IS-dHJ ratio in rad52Δ NDT80 cells will be higher than that in rad52Δ ndt80Δ cells. We conclude that in addition to its well-characterized mediator function, Rad52 is also required for efficient progression from SEIs to dHJs.
Bishop (Bishop, 1994; Gasior et al., 1998) showed that rad52Δ mutation differentially affects assembly of Rad51 and Dmc1 into immunostaining foci. This result was confirmed by staining spread meiotic chromosomes with antibodies against Rad51 and Dmc1 (Supplemental Figure S4). Dmc1 foci form with essentially normal timing in rad52Δ cells, although the average number of foci per nucleus is reduced ~2-fold. In sharp contrast, assembly of Rad51 foci is completely defective in rad52Δ cells. Consistently, all strand-exchange in rad52Δ cells is Dmc1-dependent: no SEIs or dHJs, and only residual levels of crossovers are detected in rad52Δ dmc1Δ cells. In fact, recombination in rad52Δ dmc1Δ cells is indistinguishable from that in rad51Δ dmc1Δ cells lacking both RecA homologs (Figure 2B and 2E) (Shinohara et al., 1997). Although Dmc1 foci assemble with normal timing in rad52Δ cells, their disassembly is severely delayed, by as much as ~5 hrs (Figure 3B). These persistent Dmc1 foci mirror the delay in DSB turnover and execution of the first meiotic division (above; Figure 2B).
During the SEI-to-dHJ transition, the second DSB-end could be captured via a second strand-invasion reaction. In this case, defective dHJ formation in rad52Δ cells could be a consequence of failure to assemble Rad51 filaments. Alternatively, second-end-capture may occur via Rad52-mediated annealing of the second DSB-end to the displaced-loop of the SEI (Sugiyama et al., 1998; Szostak et al., 1983). These two possibilities were distinguished in vivo using a separation-of-function allele, rad52-327, which encodes a truncated Rad52 protein lacking the Rad51-interaction domain required for mediator function, but retains the conserved N-terminal region comprising oligomerization, DNA binding and complementary strand-annealing activities (Asleson et al., 1999; Boundy-Mills and Livingston, 1993; Krejci et al., 2002). Consistent with absence of mediator activity, rad52-327 cells fail to assemble Rad51 foci (Supplemental Figure S4). Also, like rad52Δ null mutants, rad52-327 cells assemble immunostaining foci of Dmc1 with normal timing, although again the average number of foci per nucleus is reduced, by ~2.5-fold. rad52-327 cells appear to turnover Dmc1 foci more efficiently than rad52Δ cells, however, with a delay of only ~2 hrs relative to wild type.
Meiotic recombination is also more efficient in rad52-327 cells. Despite the absence of Rad51 foci, dHJ formation is relatively efficient (Figure 2D and 2E). A distinct difference between rad52-327 and wild-type cells is a change in choice of recombination partner. In wild-type cells, IH-dHJs form with a >3-fold preference over IS-dHJs at the HIS4LEU2 locus (Figure 2E) (Hunter and Kleckner, 2001; Schwacha and Kleckner, 1997). Inter-homolog bias makes biological sense during meiosis because intersister recombination does not contribute to chiasma formation. In rad52-327 cells, the ratio of IH-dHJs/IS-dHJs is decreased to one. This phenotype appears to be caused by the absence of Rad51 filaments in rad52-327 cells and is consistent with the loss of interhomolog bias observed in rad51Δ null mutants (Schwacha and Kleckner, 1997) (J.P.L. and N.H. unpublished). Regardless, comparison of total dHJ levels suggests that rad52-327 cells form dHJs at essentially wild-type levels.
dHJ levels were also measured in rad52-327 ndt80Δ cells and the basic patterns observed in NDT80 cells were reiterated (Figure 3). IH-dHJs in rad52-327 ndt80Δ cells accumulate to ~50% of the levels observed in the ndt80Δ single mutant, an increase of 9.5-fold relative to rad52Δ cells. Total dHJ levels accumulate to ~65% of ndt80Δ levels, representing a 5-fold increase over rad52Δ levels. Taken together, these data show that the N-terminal region of Rad52 promotes the transition from SEIs to dHJs.
Crossing-over at HIS4LEU2 in rad52-327 cells reaches ~70% of wild-type levels (Figure 2B). This level of crossing-over is close to that expected given the reduction in inter-homolog bias observed in the rad52-327 strain (~11% crossovers expected versus 13% detected). Moreover, rad52-327 cells sporulate efficiently and produce 29% viable spores (91 of 316 spores analyzed) implying that some or all of the crossovers formed create functional chiasmata that facilitate homolog disjunction.
The specific annealing mechanism we invoke for second-end-capture during dHJ formation is analogous to the annealing step of the SDSA model for noncrossover formation (see Discussion and Figure 4). This similarity prompted us to examine the requirement for Rad52 in noncrossover formation. Noncrossovers at HIS4LEU2 are detected by virtue of a BamHI/NgoMIV polymorphism situated immediately at the DSB site (Figure 1E) (Martini et al., 2006). In wild-type cells, noncrossovers plateau 8 hrs after induction of meiosis, at 7.6% of hybridizing DNA (Figure 2F and 2G). Noncrossovers are strongly Rad52 dependent as shown by the 7-fold reduction observed in rad52Δ cells. In contrast, noncrossovers reach ~50% of wild-type levels in rad52-327 cells, a >3-fold increase over rad52Δ levels. Failure to form noncrossovers at wild-type levels can be explained by the loss of interhomolog bias in rad52-327 cells, described above. Assuming that the reduction in interhomolog bias detected for dHJs reflects a general reduction in interhomolog recombination, we expect ~5% noncrossovers to form in rad52-327 cells; almost 4% noncrossovers are detected in this strain. We conclude that Rad52-catalyzed strand-annealing is also an important step in noncrossover formation.
We present in vivo evidence that post-invasion steps of meiotic recombination in S. cerevisiae are catalyzed by the strand-annealing activity of Rad52. Human RAD52 has recently been shown to mediate single-strand annealing in an in vitro replication-dependent D-loop expansion reaction that recapitulates the second-end-capture mechanism proposed by Szostak et al. (Szostak et al., 1983) (M. McIlwraith and S.C. West, personal communication). Additional features of our in vivo data support a different mode of second-end-capture, however. Both SEIs and dHJs appear to be crossover-specific precursors (Allers and Lichten, 2001a; Borner et al., 2004), but paradoxically, while IH-dHJs are reduced at least 8-fold in rad52Δ cells, crossovers form at ~40% of wild-type levels (18% in wild-type versus 8% in rad52Δ; Figure 2B). To reconcile this observation, we propose that SEIs formed in rad52Δ cells may be resolved to produce nonreciprocal half-crossovers (Figure 4B). This idea derives from the observation of Haber and Hearn (Haber and Hearn, 1985), that very rare crossovers formed in vegetative rad52Δ cells are always non-reciprocal. Specifically, we suggest that the SEI-to-dHJ transition often involves migration of the SEI D-loop away from the DSB-site and “end-first” displacement of the extended DSB-end (Figure 4A), as previously proposed by Allers and Lichten to explain the occurrence of DSB-distal JMs that lack intervening heteroduplex DNA (Allers and Lichten, 2001b). In wild-type cells, the displaced 3’-strand of an SEI would subsequently undergo Rad52-catalyzed annealing to the second DSB-end. In rad52Δ cells, in which second-end capture fails, we propose that the D-loop continues to migrate into the double-stranded region of the DSB-end, converting the SEI into a dHJ lacking a chromosome arm. These cryptic-dHJs could be resolved via a canonical resolution mechanism to form a half-crossover plus a broken chromatid (Figure 4B). Regular formation of cryptic dHJs may be peculiar to the rad52Δ mutant because this predicts the appearance of symmetric heteroduplex associated with crossover products, for which there is little evidence in S. cerevisiae (e.g. Nicolas and Petes, 1994). Thus, in wild-type cells, we predict that second-end capture occurs before a cryptic dHJs can form (Hunter, 2006). It remains possible, however, that significant levels of symmetric heteroduplex are not detected in S. cerevisiae because the inter-HJ distance in dHJs is only ~260 bp (Cromie et al., 2006). We note that there is good evidence for symmetric heteroduplex in other fungi (e.g. Nicolas and Petes, 1994).
It is also possible that Rad52-independent crossovers form via the non-dHJ mechanism proposed to explain crossing-over mediated by Mus81-Eme1, a structure-selective endonuclease in S. pombe (Mus81-Mms4 in S. cerevisiae) (Osman et al., 2003). An argument against this possibility is that Rad52-independent crossovers are strongly dependent on MutS homolog Msh5, which defines a Mus81-Mms4-independent pathway of crossing-over in S. cerevisiae (Supplemental Figure S1) (Argueso et al., 2004; De Los Santos et al., 2003). We note, however, that the non-dHJ mechanism also includes a second-end annealing step that could be catalyzed by Rad52.
During noncrossover formation, the two DSB-ends would also be reunited via Rad52-catalyzed annealing, and crossover and noncrossover pathways would differ only with respect to the fate of the original invading strand. Along the crossover pathway, the invading strand undergoes end-first displacement to preserve the joint molecule; whereas during SDSA, the invading strand is displaced “end last” to dissociate the joint molecule (Figure 4A and 4D).
Unregulated crossing-over can cause chromosome rearrangements, missegregation and homozygosis of deleterious mutations (Richardson et al., 2004). To minimize these risks, somatic cells actively suppress crossing-over (Ira et al., 2003; Johnson and Jasin, 2001). In contrast, meiotic crossover control mechanisms must act to ensure that each homolog pair obtains at least one crossover (Jones, 1984). Thus, regulating the formation of crossover-specific dHJs is a critical aspect of DSB-repair in all cell types. A potential target of such regulation is Rad52-mediated second-end-capture. In particular, it may be important to regulate the strand-annealing activity of Rad52 in order to prevent spontaneous dHJ formation and aberrant crossing-over.
We thank Dennis Livingston (University of Minnesota) for plasmids carrying the rad52-1 and rad52-327 alleles, Michael McIlwraith and Steve West for communicating unpublished data and Eric Kofoid and John Roth for assisting in identification of stem-loop motifs. This work was supported by NIH NIGMS grant GM074223 to N.H.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.