The Logic and Mechanism of Homologous Recombination Partner Choice

SUMMARY

Recombinational repair of spontaneous double-strand breaks (DSBs) exhibits sister bias. DSB-initiated meiotic recombination exhibits homolog bias. Physical analysis in yeast reveals that, in both cases, recombination intrinsically gives homolog bias. From this baseline default, cohesin intervenes to confer sister bias, likely independent of cohesion. In meiosis, cohesin’s sister-biasing effect is counteracted by RecA-homolog Rad51 and its mediators, plus meiotic RecA- homolog Dmc1, which thereby restore intrinsic homolog bias. Meiotic axis complex Red1/Mek1/Hop1 participates by cleanly switching recombination from mitotic mode to meiotic mode, concomitantly activating Dmc1. We propose that a Rad51/DNA filament at one DSB end captures the intact sister, creating an “anchor pad”. This filament extends across the DSB site on the intact partner, precluding inter-sister strand exchange, thus forcing use of the homolog. Cohesin and Dmc1 interactively modulate this extension, giving program-appropriate effects. In accord with this model, Rad51-mediated recombination in vivo requires the presence of a sister.

INTRODUCTION

Homologous recombination is essential for repair of double-strand breaks (DSBs) in all cells. Recombination plays special roles for meiosis, creating genetic diversity and promoting pairing and segregation of homologous chromosomes (Kleckner et al., 2011). Due to their different biological imperatives, DSB repair and meiosis involve qualitatively different partner choices. For DSB repair, the sister chromatid is the preferred partner (Kadyk and Hartwell, 1992; Johnson and Jasin, 2001; Bzymek et al., 2010; Rong and Golic, 2003). Sister bias minimizes the possibility that repair will alter the state of the genome by interactions between non-allelic pseudo-homologous sequences. Moreover, crossovers between non-sister chromosomes creates inter-chromosomal connections that can disrupt regular mitotic sister segregation (Beumer et al., 1998). For meiosis, all important roles of recombination require that interactions occur between homologs.

How these two alternative partner choices are differentially specified in the two programs remains mysterious. However, an implicit or explicit cornerstone of most considerations is the idea that the biochemical process of strand exchange is neutral with respect to partner selection and that the “default option” for partner choice is use of the sister simply because it is nearby (e.g. Kadyk and Hartwell, 1992; Johnson and Jasin, 2001). In contrast, homolog bias for meiosis requires special, program-specific features (e.g. Sheridan and Bishop, 2006). The results presented below suggest that this formulation is not correct.

The biochemical steps of meiotic recombination, the nature of homolog bias, and the players involved have been defined largely by physical analysis of DNA events in budding yeast (e.g. Schwacha and Kleckner, 1997; Allers and Lichten, 2001; Hunter and Kleckner, 2001; Oh et al., 2007; Hunter, 2006; Cloud et al., 2012; Kim et al., 2010). A powerful feature of such analysis is that the inter-homolog and inter-sister versions of post-DSB intermediates can be specifically identified by their diagnostic gel mobilities, thus permitting evaluation of partner choice at those stages (Schwacha and Kleckner, 1997; Kim et al., 2010; below). The picture that emerges from such studies is as follows (Figure 1A).

Figure 1. Meiotic Recombination - Pathway and Physical Analysis.

(A) Key steps in meiotic recombination between one sister of each homolog (Hunter, 2006; Storlazzi et al, 2010; Kim et al., 2010).

(B) Physical map of HIS4LEU2 double-strand break (DSB) hot spot. Parental homologs, Mom and Dad, are distinguished via XhoI restriction polymorphisms (”X”). For all data shown, DNA is digested with XhoI, species separated on 1D or 2D gels and detected by Southern blot hybridization with “probe A” (Kim et al., 2010).

(C) Structures of SEI and dHJ intermediates. Mom- and Dad-derived DNA duplexes in red and blue, respectively. SEI, single end invasion; dHJ, double Holliday Junction. SEI species shown for invasion of “right” DSB end; analogous species occur for invasion of the “left” DSB end (see also (Kim et al., 2010)).

(D and E) 2D gel analysis of SEIs and dHJs in WT meiosis. Blue indicates IH species; Red indicates IS species. (D) Representative 2D gel. IH/IS SEI and IH/IS dHJ species indicated with arrows. IH:IS dHJ ratio indicated. (E) Quantification of SEIs and dHJs levels, as percentage of total DNA, over time during meiosis. Top panel: semi-quantitative assessment of the relative levels of IH-SEIs versus IS-SEIs is indicated in the inset. Bottom panel: IH-dHJs, IS-dHJs and the sum of the two species are plotted (blue, red and dashed black lines, respectively). IH:IS dHJ ratio is given below the graph for the time of maximal abundance. (F and G) 2D gel analysis SEIs and dHJs in red1Δ and rec8Δ strains (data from Kim et al., 2010).

Meiotic recombination initiates via programmed DSBs. One DSB end then searches for a partner and engages a homolog chromatid duplex via a nascent D-loop. The other DSB end remains associated with its sister chromatid, perhaps also in a nascent D-loop. All of these DNA events are integrated with structural features of the developing chromosome structural axes, which are concomitantly drawn together in space. As the culmination of these events, recombinational interactions comprise ~400nm bridges that link the homolog axes, with one DSB end and its associated recombinosome components associated with each axis (Storlazzi et al., 2010; Figure 1A). Formation of these bridges along the lengths of the chromosomes comprises the process of homolog pairing. The resulting configuration is “presynaptic alignment”. In some organisms that events at the two DSB ends may be controlled by two different RecA homologs, with events at the homolog-associated end dominated by meiotic RecA homolog Dmc1 and events at the sister-associated end dominated by mitotic RecA homolog Rad51 (Shinohara et al., 2000; Hunter, 2006; Kurzbauer et al., 2012). An important implication of this “bridge” stage for the recombination process is that the two DSB ends are in direct physical, and thus presumptively functional, linkage.

As recombination progresses, this ends-apart bridge ensemble undergoes a differentiation step: a subset of these intermediates are designated to become inter-homolog (IH) crossover (CO) recombination products while the remainder mature in another way, primarily as IH “noncrossovers” (NCOs). Thereafter, along the CO branch of the pathway, DNA synthesis is initiated at one of the two ends. In the majority of cases, this extension occurs at the homolog-associated end (Figure 1A, right top). As extension progresses, the sister-associated end is released and anneals to the developing ensemble, thereby being drawn into the IH recombination complex. Further events then lead to IH-double Holliday junctions (dHJs), which in turn mature specifically to IH-COs. Importantly: it is extension at the homolog-associated DSB end which commits the reaction to making a dHJ (and then a CO) between homologs, rather than between sisters. In a minority of cases, extension occurs at the sister-associated end, leading to a dHJ between sisters (and then presumably to an IS-CO) (Figure 1A, right bottom). Thus, at the particular well-characterized yeast HIS4LEU2 recombination hot spot, 85% of dHJs are IH-dHJs, giving an IH:IS dHJ ratio of ~5:1.

CO recombination also involves a second long-lived intermediate: as a prelude to the initiation of DNA synthesis during the extension phase, the nascent D-loop becomes longer, giving a species known as a single-end invasion (SEI; Hunter and Kleckner, 2001; below).

Further insight into control of recombination is provided by a recent study showing that, in a number of different yeast mutant conditions, the IH:IS dHJ ratio at HIS4LEU2 is reduced from ~5:1 to exactly 1:1 (Kim et al., 2010). Occurrence of this precise 1:1 ratio, in many different situations (see also below) is explained as follows. The ends-apart bridge intermediate forms as usual. Then, with the two DSB ends now in communication, one end or the other is selected to undergo synthesis-mediated extension, just as in WT. However, in the mutant cases, the two ends are functionally symmetrized such that either end can initiate synthesis, with equal probability (50% each, rather than 85% and 15% as in WT). As a result, dHJs form with equal probability between homologs or between sisters, irrespective of the fact that one end has formed an SEI (Kim et al., 2010; Extended Experimental Procedures). Formation of the bridge intermediate implies that a DSB has already chosen a homolog partner, rather than its sister, i.e. that homolog bias has been established. Thus, existence of a mutant phenotype involving a 1:1 IH:IS dHJ ratio implies the existence of some subsequent process which acts to maintain homolog bias after it is established by bridge formation (Figure 1, purple asterisks), with mutants exhibiting a 1:1 IH:IS dHJ ratio being defective in that process. Put another way: a mutant exhibiting a 1:1 IH:IS dHJ phenotype is “Establishment-Plus and Maintenance-Minus”.

Maintenance of homolog bias has been proposed to involve the action of a “quiescence complex” that keeps the sister-associated end inactive until it is time for its incorporation into the evolving inter-homolog reaction (Figure 1, yellow box; Kim et al., 2010). In the absence of such a complex, the sister-associated end and the homolog-associated end would have equivalent potential for undergoing extension, resulting in occurrence of IH- and IS-dHJs at equal frequency. Previous work has shown that a 1:1 IH:IS dHJ ratio is conferred by mutations that eliminate the meiotic kleisin subunit of cohesin (Rec8); compromise the strand annealing activity of Rad52; or eliminate the catalytic activity of Blooms syndrome helicase Sgs1. The quiescence complex is proposed to also contain the mitotic RecA homolog Rad51 and its associated modulator molecules plus meiotic axis components Red1/Hop1/Mek1 (below).

Budding yeast studies have also identified mutants in which recombination occurs efficiently but only between sister chromatids (discussion in Kim et al., 2010). Mutants with this phenotype are said to be defective in “establishment” of homolog bias, i.e. are “Establishment-Minus”. This defect is most straightforwardly attributed to a failure to form a normal ends-apart bridge complex, with ensuing events then occurring only between sisters (Figure 1, green asterisks; Discussion).

By this criterion, two types of molecules have been implicated in establishment of homolog bias. First are the interacting complex of meiotic axis-associated proteins, Red1, Hop1 and Mek1 kinase (hereafter RMH) (Kim et al., 2010). Red1 and Hop1 are physically-interacting abundant structural components; Mek1 is a Rad53-related kinase that associates with Red1/Hop1 along chromosome axes. Second are the RecA homologs. During wild-type (WT) meiosis, Dmc1 is responsible for strand exchange while the general RecA homolog Rad51 is recruited to specialized roles, including a central role in homolog bias establishment (Cloud et al., 2012). Dmc1 is also implicated in homolog bias, via an “interhomolog interaction function” of unknown nature (Schwacha and Kleckner, 1997; Sheridan and Bishop, 2006).

Rad51 activity is modulated by additional factors. For mitotic recombination, Rad51 activity is promoted by so-called “mediator proteins” (Krejci et al., 2012) which include essential Rad51 loading factors (Rad55/57) and the PCSS complex (Psy1, Csm2, Shu1, and Shu2), which is important but not essential for Rad51 activity (Ball et al., 2009; Krejci et al., 2012; Shor et al., 2005; Tao et al., 2012; Qing et al., 2011; Sasanuma et al., 2013). In meiosis, Rad51 strand exchange is directly and specifically inhibited by meiosis-specific factor Hed1 (Busygina et al., 2012; Tsubouchi and Roeder, 2006). Rad51 modulator roles for partner choice are addressed below.

Here we further examine the logic and mechanism of recombination partner choice by physical analysis of recombination at the HIS4LEU2 hot spot during meiosis, in WT and diverse mutant conditions. First, we analyzed diverse situations in which homolog bias establishment is defective, thus revealing the nature of the establishment process in RMH+ conditions and the nature of recombination in the particular establishment-defective condition where RMH− functions are absent. In all of these situations, Rec8 (cohesin) is present. Second, we show that, in every Establishment-Minus mutant condition analyzed, if Rec8 is then also eliminated, a 1:1 IH:IS dHJ phenotype is observed. Since this phenotype implies an Establishment-Plus and Maintenance-Minus condition (above), these observations show that elimination of Rec8 converts all “Establishment-Minus” conditions to the “Establishment-Plus Maintenance-Minus” condition. Thus, this finding implies that elimination of Rec8 has rendered irrelevant all of the factors involved in establishment of bias in RMH+ conditions. Elimination of Rec8 even permits establishment of homolog bias in RMH− conditions where (as we show) meiotic components are not functionally involved. These patterns imply that (i) the default option for recombination is homolog bias, irrespective of whether strand exchange is promoted by Dmc1 or Rad51, regardless of whether other meiotic recombination components are participating or not; (ii) Rec8 (cohesin) actively enforces sister bias; and (iii) the role of meiotic components is to counteract this active effect of cohesin. Finally, we explore interplay among Rad51, Rec8 and the sister chromatid for dHJ formation. All findings are integrated into a coherent model.

RESULTS

Experimental System

Two-dimensional (2D) gel analysis of physical DNA recombination intermediates at the HIS4LEU2 locus was carried out as described (Figure 1B-E; Hunter and Kleckner, 2001; Kim et al., 2010; Oh et al., 2007). Inter-homolog and inter-sister SEIs and dHJs have diagnostic mobilities due to their distinguishable molecular weights and shapes (Figure 1CD). In WT meiosis, the dominance of interhomolog (IH) species over inter-sister (IS) species characteristic of meiotic recombination is readily quantifiable at the dHJ stage as an IH:IS dHJ ratio of 5:1 (Figure 1DE, and below; Kim et al., 2010). IH-SEIs also predominate over IS-SEIs (Figure 1DE). Precise quantification of relative levels is difficult due to partial overlap of different signals (Extended Experimental Procedures).

Mutants that are defective in establishment of homolog bias (Introduction), are characterized by a very low IH:IS ratio and strong prominence of IS-SEIs over IH-SEIs (e.g. IH:IS dHJ = 1:9 for red1Δ; Figure 1F) as well as reductions in all IH products, COs and NCOs, as defined by one-dimensional (1D) gels (Figure S1; Kim et al., 2010).

Mutants defective in maintenance of homolog bias are defined by their diagnostic 1:1 IH:IS dHJ ratio (Introduction) (e.g. for rec8Δ; Figure 1G). In this particular mutant, IH-SEIs predominate over IS-SEIs as in WT meiosis; interestingly, however, the opposite bias can also be observed, implying that SEI status is not relevant to ultimate symmetrization of the two DSB ends for dHJ formation (Extended Experimental Procedures). We also note that the observation of a precise 1:1 IH:IS dHJ ratio, reproducibly and in diverse situations (Kim et al., 2010; text below and Table S1), suggests that IS and IH dHJs are equally long-lived, contrary to other suggestions (Goldfarb and Lichten, 2010).

The primary phenotype used to define functional relationships is the IH:IS dHJ ratio. IH:IS dHJ phenotypes of all analyzed strains are summarized in Table S1. Relative abundance of IH- versus IS-SEIs is of secondary importance but has been evaluated also (Extended Experimental Procedures).

Part I. Partner Choice in RMH+ Meiosis

A. The Entire Rad51 Mediator Ensemble Is Required for Establishment of Homolog Bias

The distinct roles of Dmc1 and Rad51 in promoting strand exchange and mediating homolog bias (Cloud et al., 2012; Introduction) are seen in 2D gel phenotypes. When Dmc1 is absent, no SEIs or dHJs are present and IH-COs and IH-NCOs are reduced to low levels (Figure 2B versus 2A and Figure S1E). When Rad51 is absent, high levels of SEIs and dHJs are seen as in WT but now occur almost entirely between sisters (IH:IS dHJ = 1:7 in rad51Δ versus 5:1 in WT meiosis (Figure 2C; Table S1). rad51Δ also exhibits sister bias at the SEI stage (Figure 2C) and reduced levels of IH-NCOs and IHCOs (Figure S1E and S1GH; Schwacha and Kleckner, 1997).

Figure 2. Roles of Dmc1, Rad51 and Rad51 Mediators in Meiotic Recombination.

Analysis of SEIs/dHJs (as in Figure 1D and 1E) in RMH+ meiosis (all strains were mek1as (-IN); Experimental Procedures). (A-G) WT and mutants lacking Dmc1, Rad51 or a Rad51 mediator(s). (A-C) WT, dmc1Δ, and rad51Δ. Dmc1 and Rad51 promote strand exchange and homolog partner choice respectively. (D-G) shu1Δ, psy3Δ, shu1Δ psy3Δ, and shu1Δ rad51Δ. Shu1 and Psy3 collaborate to promote homolog partner choice as mediated by Rad51. (H and I) rad55Δ and rad57Δ. Rad51-mediated homolog bias is strongly dependent on Rad51 loading factors Rad55 and Rad57. (J-M) hed1Δ strains, without or with dmc1Δ, rec8Δ, or shu1Δ. Dmc1 is required for establishment of bias while Hed1 is required for maintenance of homolog bias (text).

All factors that assist Rad51 for mitotic DSB repair also assist Rad51 in promoting meiotic homolog bias. The PCSS complex is involved: shu1Δ and psy3Δ single and the shu1Δ psy3Δ double mutant exhibit an IH:IS dHJ ratio intermediate between that of WT and rad51Δ (IH:IS = 1:2.5-1:3 in all three mutants, versus 5:1 in WT and 1:7 in rad51Δ; Figure 2D-F; Table S1) and an intermediate reduction in IH-COs and IH-NCOs (Figure S1J-L). Further, a rad51Δ shu1Δ double mutant exhibits the same, more severe, defect of a rad51Δ mutant (IH:IS dHJ = 1:7); thus, Shu1/Psy3/Shu1/Psy3 acts in the Rad51 pathway (Figure 2G and Figure S1I; Table S1). The same epistasis relationships are seen for meiotic progression timing and spore viability (Figure S2). Similar results have recently been reported by Sasanuma et al., 2013. The Rad51 paralog mediators Rad55 and Rad57 are also required, as strongly as Rad51: (IH:IS dHJs = 1:7 in rad55Δ and rad57Δ; Figure 2HI; Table S1).

Relative roles and epistasis relationships among Rad51 and its mediators for meiotic homolog bias are the same as are the same as for Rad51- mediated mitotic DSB repair: Rad55/57 is essential as Rad51; Shu1/Psy3 are important, but less so, and act downstream of Rad51(Figure S2; below; Mankouri et al., 2007). Thus, the entire mitotic RecA homolog ensemble has been coordinately utilized, as a single functional unit, for this meiosis-specific role.

B. Dmc1 Is Required for Establishment of Homolog Bias; Hed1 Is Required for Maintenance of Bias; and Dmc1 and Hed1 Play Sequential Roles for Suppression of Rad51-Mediated Strand Exchange

Dmc1 is required for establishment of homolog bias

Dmc1 has been implicated as a potential player in homolog bias (Schwacha and Kleckner, 1994; Sheridan and Bishop, 2006), but documentation of a definitive role has been precluded by the fact that a dmc1Δ mutant does not carry out strand exchange. Evaluation of this possibility has been made possible by use of a hed1Δ mutation, which permits Rad51 to carry out strand exchange when Dmc1 is absent (Cloud et al, 2012; Tsubouchi and Roeder, 2006). We find that a hed1Δ dmc1Δ mutant exhibits a strong defect in homolog bias. IS-SEIs and IS-dHJs strongly predominate over their IH counterparts (IH:IS dHJ = 1:7; Figure 2J). Thus, Dmc1 plays an important role for bias. The same observation has been made independently by Lao et al (2013). Restoration of strand exchange in this condition could result from the absence of Hed1; however, Hed1 is not essential for establishment of homolog bias (next section). We thus infer that activation of Rad51 in a dmc1D hed1D background results, in whole or in part, from the absence of Dmc1, which is thereby implicated as an inhibitor of Rad51.

Role(s) of Rad51-inhibitor Hed1

We find that a hed1Δ mutant exhibits high levels of SEIs and dHJs with an IH:IS dHJ ratio of 2:1 (Figure 2K; Table S1). This phenotype, also made independently by Lao et al (2013), could reflect a moderate defect in establishment and/or maintenance of homolog bias. Further, a shu1Δ mutation specifically affects establishment of bias (above) and the hed1Δ shu1Δ double mutant exhibits a greater defect in bias (IH:IS dHJ = 1:4; Figure 2M) than either single mutant (IH:IS = 2:1 and 1:2.5, respectively; Figures 2D and 2K above). Thus: Shu1 and Hed1 have different (apparently additive) roles. If Shu1 acts during establishment, Hed1 might only act during maintenance. Since Hed1 acts specifically and directly to inhibit Rad51-mediated strand exchange, a role for Hed1 in maintenance of bias would implicate Rad51 in that process as well. Alternatively, or in addition, Hed1 and Shu1 might affect different aspects of bias establishment, in which case Hed1 might also collaborate with Dmc1 for repression of Rad51 strand exchange.

Dmc1 suppresses Rad51 strand exchange activity to promote establishment of homolog bias

In meiosis, Rad51 mediates homolog bias, rather than carrying out strand exchange (Introduction), implying that these are two mutually exclusive states of Rad51. In a hed1Δ dmc1Δ mutant, oppositely, establishment of homolog bias is defective and Rad51 is mediating strand exchange (above), further supporting the notion of two alternative Rad51 states. What molecule is responsible for this switch? Hed1 might have been the most obvious candidate. However, evidence above suggests that Dmc1 is involved in both suppression of Rad51 strand exchange and establishment of homolog bias (above). Another study provides further evidence for Dmc1-mediated inhibition of Rad51 activity (Lao et al., 2013). Thus: Dmc1 may be the critical component of switch. Moreover, it should be required for establishment of homolog bias, at least in part because it is required to inhibit Rad51strand exchange activity, thereby enabling Rad51 to play its role in the establishment process (Discussion).

Overall, Dmc1 suppresses Rad51 strand exchange transiently during establishment of bias as part of its role in that process, perhaps with help from Hed1. Then, during maintenance of bias, Hed1 alone suppresses Rad51 strand exchange (e.g. as a component of the donor-associated quiescence complex; Figure 1A; Kim et al., 2010). Mutant phenotypes are thus explained as follows. In hed1Δ, Dmc1 inhibits Rad51 strand exchange transiently during bias establishment as it promotes that process; then, during maintenance of bias, Dmc1 no longer inhibits Rad51 strand exchange and Hed1 is absent, with resultant defects. In dmc1Δ, no Rad51-mediated strand exchange occurs (Figure 2B). Since Dmc1 is absent, Hed1 carries out strong suppression, prematurely, thus fully and permanently suppressing Rad51 strand exchange. In hed1Δ dmc1Δ (Figure 2J), Rad51-mediated strand exchange occurs because both meiotic inhibitors are absent and strong sister bias is seen because Dmc1 is not there to promote bias establishment.

C. Homolog Bias Establishment is Independent of Rad51/Dmc1 Interplay If Cohesin Is Absent

In WT meiosis, Rad51 and Dmc1 are both required for establishment of homolog bias, with IH:IS dHJ ratios of 1:7 in rad51Δ ; 1:2.5 in shu1Δ and 1:7 in hed1Δ dmc1Δ (above; Figure 2). In contrast, in all three backgrounds, when Rec8 is also eliminated, a 1:1 IH:IS dHJ ratio is observed: rad51Δ/shu1Δ rec8Δ and hed1Δ dmc1Δ rec8Δ mutants both exhibit the same 1:1 IH:IS dHJ ratio seen in a rec8Δ single mutant (Figure 3A-C and 3E). These results have two implications:

Figure 3. Rad51/Shu1 Counteract(s) an Inhibitory Effect of Cohesin.

Analysis of SEIs and dHJs in RMH+ Rec8− meiosis. All strains were mek1as (-IN) and carry a rec8Δ mutation, either without (A-E) or with (F-H) meiotically-induced expression of mitotic kleisin Mcd1 from a pRec8−MCD1 fusion construct (Kim et al., 2010). (A-C, E) All strains that are not expressing Mcd1 and have discernible levels of dHJs exhibit a 1:1 IH:IS dHJ ratio implying normal establishment of homolog bias and defective maintenance of that bias as previously defined for rec8Δ (Kim et al., 2010; Panel (A)). (B, C and E) Rad51/Shu1 is not required for establishment of homolog bias in the absence of Rec8. (D) dmc1Δ shows no SEIs/dHJs, implying that strand exchange is promoted by Dmc1 in the absence of Rec8, just as in the presence of Rec8 (Figure 2B). (F-H) Expression of Mcd1 largely substitutes for Rec8 with respect to both its roles in both establishment and maintenance of bias. (F) Mcd1 ameliorates the homolog bias maintenance defect of rec8Δ, seen as an increase in the IH:IS dHJ ratio as compared to when Mcd1 is not expressed (Panel A). (G, H) Expression of Mcd1 largely restores the dependence of homolog bias establishment on Rad51/Shu1, seen as a decrease in the IH:IS dHJ ratio compared to when Mcd1 is not expressed (Panels B, C).

- When Rec8 is absent, homolog bias can be efficiently established (with the subsequent maintenance defect resulting in a 1:1 IH:IS dHJ ratio) regardless of which RecA homolog is promoting strand exchange (Dmc1 or Rad51, respectively).

- Most importantly, even though Rad51 and Dmc1 are both required for establishment of homolog bias in WT meiosis (above), neither is required when Rec8 is absent and the other molecule is promoting strand exchange (because both mutants get to the bridge stage required for a 1:1 IH:IS dHJ ratio). Thus: the role of Rad51-Dmc1 interplay for establishment of homolog bias is to counteract a role of Rec8 that promotes sister bias. (i) When Rec8 is present and both RecA homologs are present, homolog bias is observed (the WT case). (ii) When Rec8 is present and only one of the two RecA homologs is present, sister bias is observed (because Rad51/Dmc1 interplay is required for establishment of bias). (iii) When Rec8 is absent and only one of the two RecA homologs is present, homolog bias is again observed (because, with cohesin absent, Rad51/Dmc1 interplay is no longer required to counteract cohesin’s effect).

These findings further suggest the following general logic for meiotic partner choice: (i) homolog bias is the “default option” for meiotic recombination; (ii) cohesin intervenes to confer sister bias; and (iii) Rad51/Dmc1 further intervenes to oppose the intervening effect of cohesin, thus restoring the homolog bias default option (Discussion).

The roles of Rec8 in meiotic partner choice can be substantially fulfilled by a cohesin complex containing the mitotic kleisin subunit Mcd1/Scc1 (hereafter Mcd1). In yeast, Mcd1 is poorly expressed and much less abundant during meiosis than in mitotically dividing cells, although it does have discernible roles (Kateneva et al, 2005). However: when Mcd1 is expressed at a high level (via a pRec8−MCD1 fusion) in a rad51Δ/shu1Δ rec8Δ mutant, the IH:IS dHJ ratio decreases, from 1:1 to 1:3.5/1:2 (Figure 3GH), approaching the 1:7/1:1.25 observed in rad51Δ/shu1Δ REC8 (Figure 2CD). Thus: meiotic expression of Mcd1 substantially restores the requirement for Rad51/Shu1 in promoting establishment of homolog bias. Furthermore, when Mcd1 is expressed at a high level in a rec8Δ mutant, the IH:IS ratio increases from 1:1 (in rec8Δ; Figure 3A) to 3.5:1 (Figure 3F), approaching the 5:1 ratio seen in WT (Figures 1DE and 2A). Thus: Mcd1 can also substantially substitute for Rec8 in promoting maintenance of homolog bias.

Part II. Partner Choice in RMH− Meiosis

A. Absence of the RMH Activity Toggles Recombination from Meiotic Mode to Mitotic-like DSB Repair Mode

The RMH complex is a central player in meiotic recombination partner choice (Introduction). Diverse previous studies show that complete deletion mutations of axis components Hop1 or Red1, or chemically-mediated elimination of Mek1 kinase activity all result in very strong sister bias (discussion in Kim et al., 2010). Thus: in the HIS4LEU2 locus, a red1Δ or a mek1 strain with inactive kinase (mek1as +IN; Experimental Procedures) exhibits a IH:IS dHJ ratio of ~1:9 (Kim et al., 2010; Figure 1F;Figures 4A and 4N), as also seen in a hop1Δ mutant (Schwacha and Kleckner, 1994).

Figure 4. Absence of Mek1 Kinase Activity or Red1 Protein Switches Recombination to a Mitotic-like Mode.

Analysis of SEIs/dHJs in RMH− Rec8+ meiosis (A-P) and red1::LEU2 Rec8+ meiosis (Q-S). (A-M) Mek1 kinase− strains (mek1as +IN). All strains were analyzed in parallel with the Mek1 kinase+ counterparts described in Figure 2 (Kim et al., 2010; Experimental Procedures). (N-P) strains carrying red1Δ; total SEI/dHJ levels adjusted for decreased DSB levels in red1Δ as in Kim and colleagues (2010). (Q-S) red1::LEU2. Data are from Schwacha and Kleckner (1997); see also Figure S4).

Analysis of RMH− conditions. (A, N) In the absence of other mutations (”WT”), strong inter-sister bias is observed. (B, H, M, O) Absence of Dmc1 (dmc1ΔΔ) has no effect on SEI/dHJ patterns when compared with isogenic DMC1 strains (A, E, K, N). (CG, I, J, P) Rad51 is responsible for strand exchange. SEI/dHJ levels are eliminated or reduced by absence of Rad51/55/57 or Shu1/Psy3 respectively with the same hierarchy of effects seen for mitotic DSB repair (text; Figure S2); sister bias is not altered for residual SEIs/dHJs seen in the absence of Shu1/Psy3. (K-M) Hed1 has no discernible role. hed1Δ exhibits strong inter-sister bias, alone or in the presence of shu1Δ (but with reduced SEI/dHJ levels) or dmc1Δ. (Q) IH-dHJs are observed in red1::LEU2 but not in red1Δ (see Figure S4 for isogenic comparison). (R) The IH-dHJs that arise in red1:LEU2 are dependent on Dmc1, implying that a residual Red1 function activates Dmc1 to promote homolog bias (text). (S) Absence of Rad51 confers a slight reduction in the residual homolog bias in red1::LEU2, implying a Rad51 role.

Further examination of these mutant conditions reveals a simple general explanation for their phenotype: absence of RMH toggles the recombination process, as an entire discrete unit, from the normal meiotic mode to a mode that, by diverse criteria, seems very similar to mitotic DSB repair.

1. The strong sister bias in RMH− meiosis (Figure 1F; Figures 4A and 4N) corresponds to the strong sister bias in mitotic DSB repair (Kadyk and Hartwell, 1992; Johnson and Jasin, 2001).

2. RMH− meiosis is characterized by dramatic hyper-resection of DSBs to give very long 3′ single-strand (ss) DNA “tails” (Kim et al., 2010) as compared to the carefully controlled resection of meiotic DSBs in WT meiosis (Hunter and Kleckner, 2001). Hyper-resection also is a prominent characteristic of mitotic DSB repair (Chung et al., 2010).

3. In mitotic DSB repair, Rad51 promotes strand exchange. We now show that, in RMH− conditions, elimination of Rad51 (in the presence or absence of Shu1), Rad55 or Rad57 abolishes SEIs/dHJs (Figure 4CD and 4IJ versus 4A; Figure 4P versus 4N). Further, elimination of Shu1, Psy3, or both, reduces SEI/dHJ levels (in the presence or absence of Hed1) (Figure 4E-G versus 4A; Figure 4L versus 4K). Furthermore, the rad51II-3A mutation, which diminishes but does not eliminate Rad51’s strand exchange activity (Cloud et al., 2012), diminishes but does not eliminate SEIs/dHJs (Figure S3). In contrast, elimination of Dmc1 has little or no effect (in the presence or absence of Shu1 or Hed1) (Figure 4B versus 4A; Figure 4H versus 4E; 4M versus 4K; Figure 4O versus 4N). Thus, in RMH− conditions, Rad51 is promoting strand exchange, as during mitotic DSB repair, while Dmc1 has no discernible influence. This feature of the RMH condition was not previously appreciated.

4. In mitotic DSB repair, meiosis-specific factors are not expressed and thus are not participating. In RMH− meiosis, meiosis-specific factors are presumably expressed but are, nonetheless, again not participating. Elimination of Dmc1, Hed1 or Dmc1 and Hed1 has no effect (above and Figure 4K versus 4A). Since most other meiotic factors work in specific concert with Dmc1 (e.g. Hop2/Mnd1; Mei5/Sae3; Rdh54-Tid1; Cloud et al., 2012; Nimonkar et al., 2012; Pezza et al., 2007), those factors also should not be participating. Since all of these molecules are presumably expressed in RMH− meiosis, they appear to be present but unable to act in the absence of RMH function.

5. During mitotic DSB repair, recombination tends to proceed to the NCO fate via synthesis-dependent strand annealing (SDSA) rather than to the CO fate via SEIs and dHJs (Bzymek et al, 2010). The same is true in RMH− meiosis, where the majority of inter-homolog recombination events appear to be resolved as NCOs rather than COs: CO/NCO = 1.3 in WT meiosis versus 0.5 in RMH− meiosis (Figure S1F).

B. RMH Function Is Required for Meiotic Homolog Bias Because It Activates Dmc1

Given that the RMH complex activates the entire meiotic recombination process, the next question is: how this activation might occur, particularly regarding the ability of the RMH complex to promote homolog bias. Insight is provided by strains carrying a red1::LEU2 insertion/disruption allele. In a previous analysis, we detected a significant level of IH-dHJs in this mutant (Figure 4Q; Schwacha and Kleckner, 1997). Moreover, occurrence of these IH-dHJs was specifically dependent on Dmc1: when Dmc1 is eliminated, Rad51 promotes dHJ formation but only IS events occur (Figure 4R versus 4Q; Schwacha and Kleckner, 1997). Put another way: in the red1::LEU2 background, Dmc1 is required specifically for IH events, not for strand exchange in general, which is promoted by Rad51. It was thus inferred that Dmc1 has a specific “interhomolog interaction function” (Schwacha and Kleckner, 1997). We now show above that Dmc1 is important for establishment of homolog bias. However, in contrast to the situation in red1::LEU2, no IH-dHJs are observed in RMH− (red1Δ) meiosis, even when Dmc1 is present (Figure 4N). Importantly, our previous study used a more complex version HIS4LEU2 locus. However, we have confirmed that this same intermediate defect occurs in a red1::LEU2 strain carrying the HIS4LEU2 allele used above for analysis of red1Δ and mek1as+IN null mutants (Figure S4).

Together these results suggest that red1::LEU2 is not a null allele and that it retains a sub-function which allows Dmc1 to now exert its influence on partner choice, i.e. that the Dmc1’s “inter-homolog interaction function” is conferred (directly or indirectly) by Red1. Thus: one role of the RMH complex for meiotic homolog bias is to activate Dmc1 as a factor for promotion of homolog bias establishment. This activation could be exerted directly between Red1 and Dmc1 or indirectly via Red1/Dmc1-dependent components. Since all other meiosis-specific recombination components act via Dmc1, activation of Dmc1 would be a convenient key lynchpin for the switch from mitotic to meiotic mode.

Interestingly, the IH:IS dHJ ratio in red1::LEU2 RAD51 is 1:2 while the ratio in red1::LEU2 rad51Δ is lower, 1:3 (Figure 4S versus 4Q). Thus, Rad51 has a role for homolog bias even in this compromised situation.

C. Mitotic-like RMH− Meiosis Also Exhibits Homolog Bias if Cohesin Is Absent

We showed previously that, in RMH− conditions, strong sister bias for recombination is converted to normal homolog bias if cohesin is removed, as shown by a 1:1 IH:IS dHJ in RMH− rec8Δ strains (Kim et al., 2010; Figure 5A). Moreover, in these RMH− Rec8− conditions, if mitotic cohesin is then expressed at a high level, strong sister bias is restored (IH:IS dHJ = 1:5 in rec8Δ pRec8−MCD1 versus 1:1 in RMH− rec8Δ (Kim et al., 2010; Figure 5G versus 5A).

Figure 5. In the Absence of Cohesin, Rad51-promoted Strand Exchange Exhibits Homolog Bias.

(A-I) Physical analysis in the indicated strains in a Mek1 kinase− Rec8− background: without (A-F) or with (G-J) meiotically-induced expression of mitotic kleisin Mcd1 from a pRec8−MCD1 fusion construct. All strains were mek1as (+IN) rec8ΔΔ. In all cases where SEIs/dHJs form, a 1:1 IH:IS dHJ ratio is observed. This ratio indicates normal establishment of homolog bias (and defective homolog bias maintenance) (Kim et al., 2010; text). Observation of this phenotype in the absence of Dmc1 (Panel B) demonstrates that normal establishment of homolog bias occurs when Rad51 is promoting strand exchange. Absence/reduction of SEIs/dHJs in rad51Δ/shu1Δ, with 1:1 IH:IS dHJ ratio in the latter case (C and D) confirms that, in this cohesin-minus background, Rad51-promotes strand exchange with efficient establishment of homolog bias. (G-J) Mcd1 expression can largely substitute for Rec8 in conferring sister bias as seen in both the presence and absence of Dmc1 and for the reduced levels of strand exchange seen in the absence of Shu1. Note: IH-SEIs are not indicated on the corresponding gel panels due to the difficulty of unambiguous identification but may well be present in many/all of these cases (Extended Experimental Procedures).

The results presented above now further reveal that, in RMH− Rec8+ conditions, recombination exhibits strong similarities with mitotic DSB repair, including the fact that Rad51 is responsible for strand exchange (Figure 4A-Q).

Taken together, these two sets of results suggest that mitotic-like recombination, promoted by Rad51, and in the apparent absence of functional contributions from meiotic components, intrinsically tends to exhibit homolog bias when cohesin is absent; then, when cohesin is present, this intrinsic tendency for homolog bias is converted into sister bias.

However: this important conclusion depends critically on the assumption that the dHJ phenotypes seen in RMH− Rec8− conditions reflect recombination promoted by Rad51, just as in RMH− Rec8+ conditions. We have confirmed this critical assumption. In RMH− Rec8− meiosis, absence of Dmc1 has no effect on partner choice, either in Hed1+ (Figure 5B versus 5A) or in Hed1− (Figure 5F versus 5E): a 1:1 IH:IS dHJ ratio is again observed, as in the presence of Dmc1. Thus: in RMH− Rec8− conditions, Rad51 promotes strand exchange with normal homolog bias.

Elimination of Dmc1 does reduce steady state dHJ levels somewhat. This likely reflects increased dHJ turnover rather than direct participation of Dmc1 in strand exchange. Accordingly, absence of Rad51 essentially completely eliminates SEIs and dHJs, and absence of Shu1 significantly reduces SEIs and dHJs (Figure 5CD versus 5A). Strand exchange is also strongly reduced by a rad51 non-null allele that strongly reduces Rad51-mediated strand exchange which retaining Rad51-mediated functions for homolog bias (Figure S3).

Furthermore, Mcd1 can substantially substitute for Rec8 in RMH− conditions, just as in RMH+ conditions (above). Mcd1 expression in RMH− Rec8− restores strong sister bias regardless of whether Dmc1 is present or absent (Figure 5G versus 5A; 5H versus 5B) and even when strand exchange is reduced by absence of Shu1 (Figure 5J versus 5D), with strand exchange still fully dependent on Rad51 (Figure 5I).

These results show that, in conditions where meiotic factors are not participating and where recombination has strong similarities to mitotic DSB repair (i.e. RMH− meiosis), Rad51-mediated strand exchange recombination exhibits homolog bias if cohesin is absent and sister bias if cohesin is present. The suggestion from these results is that the intrinsic tendency for Rad51-mediated mitotic-like recombination, i.e. the “default option” for this process, is homolog bias, not sister bias as usually assumed; moreover, cohesin actively intervenes in this default to promote sister bias (Discussion).

Synthesis Thus Far

The observations presented above suggest the following general scenario for RecA homolog-promoted partner choice in the mitotic and meiotic programs (Discussion). (i) The basic mechanistic default option for all RecA homolog-promoted recombination is selection of a homolog partner. (ii) Cohesin interferes with this basic tendency to actively confer sister bias. (iii) In meiosis, Dmc1/Rad51 collaboration, dependent on RMH activation of Dmc1, counteracts this cohesin-mediated sister bias activity, thereby restoring the intrinsic homolog bias default.

Part III. Requirements for Execution of dHJ formation

A. Interplay among Rad51, Cohesin and the Sister Chromatid Is Required for Stimulation of Dmc1-Mediated dHJ formation in RMH+ Meiosis

Rad51 and Cohesin Have Overlapping Roles

The above results reveal that the levels of SEIs and dHJs are severely reduced when both Rec8 and Rad51 are absent (rec8Δ rad51Δ/shu1Δ; Figure 3BC) but not when either component is absent alone (in rec8Δ or rad51Δ/shu1Δ (Figures 3A and Figure 2C-G)). This pattern suggests that Rec8 and Rad51 have overlapping stimulatory roles in promoting dHJ formation for meiosis. Rad51 has been implicated previously as a positive activator of Dmc1 (Cloud et al., 2012). The role of Rec8 for this stimulatory effect can be carried out reasonably successfully by the mitotic kleisin Mcd1 as shown by increased levels of SEIs/dHJs in rad51Δ/shu1Δ rec8Δ pRec8−MCD1 as compared to isogenic rec8Δ strains (Figure 3GH versus Figure 3BC).

The Sister and Cohesin Have Overlapping Roles

Meiotic recombination can occur with reasonable efficiency under conditions where prophase occurs without a preceding round of DNA replication (pMCD1-CDC6; Hochwagen et al., 2005). Since recombination in this condition occurs in the absence of a sister chromatid, examination of this condition permits evaluation of possible roles of the sister per se.

Recombination in the absence of a sister exhibits significant levels of SEIs, dHJs and COs, confirming and extending earlier studies (Figure 6A; Hochwagen et al., 2005). Furthermore, this recombination is absolutely dependent on Dmc1, as in WT meiosis (Figure 6B), with DSBs accumulating dramatically in the absence of Dmc1 (Figure 6F versus 6E).

Figure 6. Analysis of Recombination in the Absence of a Sister.

(A-H) Recombination in the absence of DNA replication (pMCD1-CDC6; Hochwagen et al., 2005). (A) Virtually all dHJs form between homologs, in accord with a tight block to replication and resultant absence of sister chromatids. (B) No dHJs are observed in dmc1Δ implying strand exchange is promoted by Dmc1. (C) No dHJs are observed in rec8Δ implying strong essential role for Rec8 in Dmc1-promoted recombination despite the absence of a sister and thus via a role for cohesin independent of its role in cohesion per se. (D) No dHJs are observed in Mek1-, implying that in mitotic mode, when Rad51 is promoting strand exchange, presence of a sister is essential for SEI/dHJ formation. In (A), SEI/dHJ levels are multiplied by two to give per-chromatid levels comparable to WT meiosis. (E-H) Full 2D gels for strains in (A-D) showing DSB levels; quantification at right.

In addition, and in contrast to WT meiosis, recombination in this condition is absolutely dependent on Rec8 (Figure 6C). Since no sister is present, this result shows that Rec8 stimulates Dmc1-mediated recombination in the absence of a sister and thus via an effect that does not involve sister cohesion. A cohesin-independent role of Rec8 was also suggested by analysis of rec8 separation-of-function alleles (Brar et al., 2009). In this situation, DSBs do not accumulate, implying that they are proceeding, but to some non-SEI/dHJ fate.

This result also permits a further conclusion. In the presence of a sister, absence of Rec8 does not dramatically reduce the levels of SEIs/dHJs (Figure 3A). In the presence of Rec8, absence of a sister does not dramatically reduce the levels of SEIs/dHJs (Figure 6A). However, in the absence of both Rec8 and a sister chromatid, only low levels of SEIs/dHJs occur (Figure 6C). Thus: the presence of a sister has an overlapping stimulatory role with Rec8 for Dmc1-mediated dHJ formation.

Rad51-mediated Stimulation of Dmc1 in RMH+ Meiosis Requires the Presence of the Sister

Rad51 and Rec8 have overlapping roles for stimulation of Dmc1-mediated dHJ formation; the sister chromatid and Rec8 have overlapping roles for stimulation of Dmc1-mediated dHJ formation (above). The obvious suggestion that emerges from these two results is that Rad51 and the presence of a sister collaborate for the same role in stimulation of Dmc1-mediated dHJ formation, which role is overlapping with a sister-independent role of Rec8. Put another way: Rad51-mediated stimulation of Dmc1-promoted dHJ formation depends upon the presence of a sister chromatid (as seen in the absence of Rec8).

A prediction of this possibility is that (in the absence of Rec8), elimination of either Rad51 or the sister chromatid or both will all confer the same strong reduction of Dmc1-mediated dHJ formation. This prediction is fulfilled for the “single” defect cases: SEIs/dHJs are virtually undetectable in both rec8Δ rad51Δ and in the absence of a cohesin and a sister (rec8Δ pMCD1-CDC6) (Figures 3B and 6C). The “double” defect case is difficult to test because pMCD1-CDC6 rad51Δ rec8Δ exhibits mitotic growth defects.

B. Rad51-mediated dHJ formation in RMH− Meiosis Requires the Presence of the Sister

The above results suggest that collaboration between Rad51 and the sister chromatid is important in promoting Dmc1-mediated dHJ formation in RMH+ meiosis. We were interested to know whether such a collaboration might also be important for Rad51-mediated dHJ formation under conditions where meiotic recombination components have little or no effect. We therefore examined recombination in RMH− conditions, where Rad51 is responsible for strand exchange and Dmc1 is not playing a prominent role (above). If the sister is required for Rad51 strand exchange activity, no SEIs/dHJs will occur in cdc6 RMH− conditions. Remarkably, this prediction is fulfilled: no SEIs or dHJs occur in pMCD1-CDC6 mek1as+IN (Figure 6D). DSBs occur in all pMCD1-CDC6 strains (Figure 6E-H); thus absence of SEIs/dHJs does not reflect a failure to enter meiosis or initiate recombination. Taken together these findings provide strong support for the idea that the presence of a sister chromatid is essential for both Rad51-mediated dHJ formation in mitotic-like mode and for Rad51-mediated stimulation of Dmc1-mediated dHJ formation in meiotic mode.

DISCUSSION

The presented results permit a synthetic view of recombination partner choice for both general (mitotic) DSB repair and for meiosis.

Homolog Bias in Wild-type Meiosis

Analysis of recombination in RMH+ conditions further defines the nature of meiotic homolog bias and roles in this process of meiotic and mitotic RecA homologs and several associated modulators. The relationships defined above are summarized in Figure 7A. In brief: the default option for recombination is homolog bias; cohesin intervenes in the process to confer sister bias; and meiotic components counteract the effect of cohesin to restore homolog bias. For the meiotic process, the RMH ensemble activates Dmc1 which acts (at least in part) to inhibit Rad51 strand exchange activity, perhaps with help from Hed1. As a result that Rad51 and its entire ensemble of mediator proteins now functions to antagonize cohesin. It is most straightforward to think that all of these effects occur on the DSB donor chromosome, prior to release of a DSB end to search for homolog. However, alternative or additional effects are not excluded, including features required to ensure that DSB extension by synthesis occurs preferentially at the homolog-associated DSB end (Figure 1A).

Figure 7. Logic and Proposed Anchor Pad Mechanism of RecA Homolog Recombination Partner Choice.

(A) New partner choice paradigm. Recombination exhibits an intrinsic mechanistically-specified default towards selection of the homolog (homolog bias). Cohesin intervenes in this mechanism to channel recombination into use of the sister. This is the appropriate situation for mitotic DSB repair. However, in meiosis, homolog bias is required. This outcome is achieved by the collaborative action of Rad51 and Dmc1, with the RMH complex (and Red1 specifically; text) required to activate Dmc1. These factors also intervene to counteract the sister-channeling effect of cohesin. (B) Proposed anchor pad model for mechanistically-specified homolog bias. Rad51 filament on the ssDNA of one DSB end captures the intact sister, creating the anchor pad. Extension of the filament along the intact sister, across the site corresponding to the DSB, precludes use of the sister, thus forcing use of the homolog. This effect can also promote release of the other DSB end from association with the sister, thus permitting loading of Rad51 on that end to promote inter-homolog strand exchange. Effects of cohesin and meiotic counteracting components come into play after anchor pad formation to modulate extension of the filament along the intact sister. (C) Cohesin modulates the homolog bias default to give sister bias by altering the parameters of the Rad51 filament so that it promotes strand exchange rather than polymerizing along the intact sister. In meiosis, the same step is further modulated by RMH-activated Dmc1, which counteracts the sister-channeling effect of cohesin.

Direct activation of Dmc1 by RMH could be the lynchpin for bringing other meiosis-specific molecules into play, since all of those molecules work via Dmc1 (e.g. Tid1/Rdh54, Mei5-Sae3, Hop2-Mnd1). However, the RMH complex likely also plays other roles, e.g. to provide a locally cohesin-depleted zone (Kim et al., 2010) such that effects of cohesin on partner choice are more manageable.

Rad51 inhibitor Hed1 is likely required for maintenance of bias. Hed1 would exert its role as part of the quiescence complex that keeps the sister-associated end from undergoing synthesis-mediated extension and ensuing IS-dHJ formation, in collaboration with Rad51 (Kim et al., 2010; Figure 1A). A role for Hed1 at a later stage in recombination has also been suggested by Tsubouchi and colleagues (Busygina et al., 2012). Hed1 may also act earlier, in concert with Dmc1 (above).

RMH Activity Toggles Recombination between Mitotic-Like and Meiotic Modes

The presented results show that the RMH cleanly switches recombination from a mode that has strong similarities to general DSB repair to meiotic mode, with its additional features. It is well known that RMH− meiosis has mitotic-like qualities; however the existence of such a clean switch was not previously appreciated. This feature is biologically interesting. First, it suggests that the RMH complex played a critical role in the evolution of meiotic recombination from mitotic DSB repair. Second, the existence of such a switch underlies the fact that yeast cells can interrupt meiosis and return smoothly to the mitotic program in a process that involves (a) rapid loss of meiotic chromosome axis components; and (b) diversion of recombination from inter-homolog to inter-sister mode (Schwacha and Kleckner, 1997; Zenvrith et al., 1997; Goldfarb and Lichten, 2010). This “return to growth” regime thus appears to involve toggling of the RMH switch from meiotic mode back a mitotic mode analogous RMH− mutant conditions.

Homolog Bias Is the Mechanistically-specified Default Option for All RecA Homolog Recombination

In RMH− meiosis, strand exchange is promoted by Rad51. Known meiotic components can be explicitly deleted (dmc1Δ/hed1Δ) or can be present but apparently making no functional contribution to recombination. Nonetheless, when cohesin is absent, RMH− recombination exhibits establishment of homolog bias, as indicated by occurrence of a 1:1 IH:IS dHJ ratio; and when cohesin is present, recombination exhibits establishment of sister bias, with virtual absence of all IH species. These observations imply that homolog bias is the default option for mitotic-like Rad51-promoted strand exchange and that cohesin overrides this default to actively promote sister bias (Figure 7A bottom).

Correspondingly: in a study of bona fide mitotic DSB repair, lower and higher levels of cohesin resulted, respectively, in higher and lower levels of interhomolog versus intersister recombination (Covo et al, 2010).

The role of cohesin in actively promoting sister bias probably does not involve its role in mediating sister chromatid cohesion, for three reasons. (1) Given that homolog bias is a built-in feature of Rad51-mediated recombination, the relative spatial proximities of the sister and the homolog should be irrelevant. There is no reason to suppose that holding the sister closer to the DSB would have any effect. (2) During meiotic recombination, cohesin can play roles for recombination independent of the presence of a sister and thus independent of its cohesion role (above). (3) A non-cohesion role of cohesin is required for limitation of DSB-initiated ectopic recombination yeast, an effect that could also involve channeling of the DSB into an inter-sister interaction (D. Koshland and J. Heidinger-Pauli, personal communication).

The situation in meiosis is precisely analogous except that, in addition, Rad51 and Red1-activated Dmc1 collaborate to override the effect of cohesin thereby restoring the homolog bias default, as described above. Moreover, homolog bias is observed in meiotic (RMH+) conditions when cohesin is absent regardless of whether strand exchange is being promoted by Rad51 (with Dmc1 absent) or Dmc1 (with Rad51 absent). We conclude that homolog bias is the basic mechanistic default option for RecA homolog-mediated recombination (Figure 7A).

These findings further imply that program-appropriate outcomes in mitotic DSB repair and meiosis arise by modulation of the basic mechanistic inter-homolog default. For mitotic DSB repair, sister bias is the programmatically appropriate outcome and is ensured by cohesin. During meiosis, the appropriate outcome is homolog bias. However, cohesin cannot be globally eliminated because of its multiple additional roles. Thus, locally, at the sites of recombination, recombinosome components directly counteract the effects of cohesin, thereby providing the programmatically appropriate outcome by restoring the homolog bias default.

A Synthetic Model

How could a RecA homolog reaction exhibit intrinsic mechanistically-specified homolog bias? We propose the following model (Figure 7BC). Assembly of Rad51 on the ssDNA tail of a DSB leads immediately to capture of a nearby DNA duplex giving an “anchor pad” interaction. In G2, or during meiosis, this duplex will nearly always be the sister chromatid. Rad51 then polymerizes outward from this anchor pad, along the intact partner duplex, across the site opposite the DSB. This polymerization will preclude use of the sister as a partner, thus forcing use of the homolog. This polymerization could also have the effect of promoting release of the other DSB end from the sister, thus making it available for loading of additional Rad51 or, in meiosis Dmc1, to create a “leading end” that can search for an available homologous partner DNA that will, necessarily, be on the homolog.

Program-appropriate partner bias could be imposed on this process by intervention at the step of polymerization outward from the anchor pad. Cohesin would intervene by altering the parameters of the filament so that polymerization across the DSB site is disfavored whereas initiation of strand exchange would be favored. During meiosis, Dmc1 could also intervene at this step, eliminating the effect of cohesin while concomitantly blocking Rad51-mediated strand exchange (as observed), e.g. by direct interaction with Rad51 (Cloud et al., 2012).

This mechanism also has the attractive feature that it automatically provides for a functional and temporal sequence of events at the two DSB ends. This is a prominent feature of meiotic recombination (Figure 1A) and could also be an important feature in mitotic recombination, particularly during inter-homolog recombination. In this context, the Rad51 ensemble would be retained at the anchor pad end so that it can mediate “second end quiescence” (in collaboration with Hed1; above) and also to provide a possible “backup” for intersister recombination if interhomolog recombination goes awry as previously suggested (Hunter, 2006).

Rad51 Activity for dHJ Formation Requires the Presence of a Sister Chromatid

The proposed anchor pad model predicts that the sister chromatid would play an important role in Rad51-mediated homolog-mediated recombination. Investigation of this possibility motivated our analysis of recombination in the absence of a sister (above). Mutant phenotypes suggest that Rad51 and the sister chromatid do collaborate to promote Rad51-mediated stimulation of Dmc1-mediated dHJ formation. Furthermore, there is no Rad51-mediated dHJ formation in mitotic mode (RMH−) in a cdc6 background where no sister is present, although this result could reflect some defect resulting from absence of DNA replication other than absence of the sister per se. Also, DSBs do progress to some other fate in the Cdc6− RMH− condition, suggesting that some type of Rad51-mediated strand exchange remains possible. Nonetheless, these striking and unanticipated results, not predicted by any previous considerations, provide encouraging support for the central feature of our proposed model.

Conclusion

The presented results provide a synthetic new view of how a DSB selects a partner in both the mitotic and meiotic programs, with inputs from the basic biochemical mechanism, chromosome structure components and, during meiosis, direct interplay of the meiotic RecA homolog Dmc1 with its general counterpart Rad51.

EXPERIMENTAL PROCEDURES

Yeast Strains, Meiotic Time Courses and DNA Physical Analysis

Detailed genotypes and strain constructions are listed in Table S2. Procedures for meiotic time course and recombination physical analyses have been described (Hunter and Kleckner, 2001; Kim et al., 2010; Extended Experimental Procedures). Most analyzed strains carry the inhibitor-sensitive mek1as allele (Wan et al., 2004). In all time courses with such strains, a single pre-meiotic culture was split into two SPM cultures; fresh 1μM 1-NA-PP1 (USBiological) in DMSO was added directly to one of the two cultures (Kim et al., 2010); and the two cultures were then carried in parallel through meiosis to give directly comparable Mek1+ and Mek1 kinase-minus results (mek1as(-IN)) and mek1as(+IN)) respectively. For every genotype where both Mek1+ and Mek1 kinase-minus time courses are presented in the text, the Mek1+ data was derived from a mek1as(-IN) culture analyzed in parallel with the corresponding Mek1 kinase-minus case in this way.

Calculation of IH:IS dHJ Ratios

For each time course, the IH:IS dHJ ratio is given for the time point at which dHJs were at their maximum level. Ratios denoted as 1:1 were, for all experiments with all analyzed strains, 1:1 + 0.17 (n = 14; range 1:0.8 - 1:1.2). All other values were rounded off to the nearest increment of 0.5 (e.g. a ratio of 1:2.7 is given as 1:3 and a ratio of 3.4:1 is given as 3.5:1).