The present study suggests that Rec8 promotes sister bias, likely via its cohesin function, thereby inhibiting establishment of homolog bias. The role of Red1/Mek1kinase is to counteract this effect (). Despite this interplay, when Red1 and Red1/Mek1kinase are both absent, homolog bias is still established efficiently. Thus, these structural components satisfy “preconditions” for homolog bias, which is then directly implemented by other components (). During CO recombination, but not NCO recombination, bias also must be actively maintained, at the SEI-to-dHJ transition. Rec8 is required positively for this effect (). Red1/Mek1kinase might be similarly involved. All roles of Rec8 and Red1 for partner choice mirror the competing dictates of meiosis for maintenance of cohesion globally versus disruption locally at sites of recombination. Taken together with other results, our findings have additional implications.
Roles of structural components for meiotic recombination
Interplay of Rec8-mediated cohesion and Red1/Mek1kinase for establishment of homolog bias
Mcd1 substitutes efficiently for Rec8 in promoting sister bias; further, Red1/Mek1 kinase can overcome this effect as effectively as it does that of Rec8. Mcd1 also substitutes effectively for Rec8 for sister chromatid arm cohesion. Thus, Rec8− mediated sister bias is likely promoted by cohesion per se. This meiotic role of Rec8 is analogous to recently-described Mcd1 roles in promoting sister bias for recombinational repair of DSBs in non-meiotic cells (Introduction).
Meiosis requires that cohesion be robust globally, to ensure regular homolog pairing during prophase and homolog segregation at MI (Introduction). We infer that meiotic components Red1/Mek1kinase are required to counteract this cohesion locally, in the vicinity of recombinational interactions, thereby opening up the possibility for actual implementation of homolog bias via other meiosis-specific features. In this role, Red1/Mek1 likely works together with Hop1, the third yeast meiotic axis component. Hop1 interacts closely with Red1/Mek1, physically, cytologically, and functionally with respect to several activities, including homolog bias: in a hop1
Δ mutant, at HIS4LEU2
, only IS-dHJs are observed, to the exclusion of IH-dHJs (Schwacha and Kleckner, 1994
) exactly as in red1
Δ (above). This role of Hop1/Red1/Mek1kinase is the only role for these proteins in homolog bias establishment because corresponding mutations have no effect on establishment if Rec8/cohesion is absent.
The effect of Red1/Mek1kinase on Rec8-mediated cohesion could occur prior to, concomitant with, or after DSB formation, by any of several possible mechanisms. An early effect is supported by our finding that Rec8 and Red1/Mek1 play multiple roles, sometimes interactively, prior to/concomitant with DSB formation, i.e. for sister cohesion, for the levels and timing of DSBs, and in early formation of distinct spatial domains.
Homolog bias is likely implemented by components of pre/post-DSB recombinosomes, including Dmc1 (Sheridan and Bishop, 2006
). Thus: precondition effects () likely reflect a layer of structural control that is superimposed upon recombinosome-mediated events.
Our findings exclude several previous models for establishment of homolog bias. (i) With respect to Introduction Model 1: cohesion-mediated sister cohesion does not promote bias; rather it inhibits bias; also, Red1/Mek1kinase does not promote sister cohesion; rather it counteracts cohesion (see also Terentyev et al., 2010
). (ii) It was proposed that Mek1-mediated phosphorylation of Rad54 plays a role in homolog bias (Niu et al., 2009
). The present study suggests that the only role of Red1/Mek1kinase is to counteract Rec8-mediated cohesion. Mek1 phosphorylation of Rad54 may be important primarily for DNA damage “checkpoint” responses, e.g. in dmc1
Δ where Mek1/Rad54 interactions were examined; indeed, a non-phosphorylatable rad54
mutant has no phenotype in WT meiosis (Niu et al., 2009
). (iii) A recent report asserts that Mek1 mediates homolog bias independent of Rec8 (Callender and Hollingsworth, 2010
). However, that study examined only progression of DSBs (which we show here is not correlated with partner choice), and did not examine whether progressing DSBs ended up in IH or IS interactions.
Maintenance of bias during CO recombination
For homolog bias maintenance, Rec8 is required and Mcd1 does not effectively substitute. Thus, meiosis-specific Rec8 functions are involved. Such roles might still be cohesion-related, or not. Intriguingly, Red1/Mek1kinase may work together with Rec8 for maintenance of bias (despite working in opposition to Rec8 during bias establishment). Similarly, Red1/Mek1kinase is implicated in promoting sister cohesion (despite also counteracting its inhibitory effects). Perhaps Red1/Mek1 and Rec8 roles for bias maintenance both reflect meiotic cohesion-favoring effects.
Maintenance of homolog bias is required specifically during CO recombination. Perhaps this is because CO recombination, but not NCO recombination, involves accompanying local exchange of individual chromatid axes (Introduction), and thus is more dependent on sister stabilization factors to maintain overall chromosome integrity during disruptive recombinational transitions (Storlazzi et al., 2008
Establishment and maintenance of homolog bias via programmed quiescence and release of first- and second-DSB ends
During CO recombination, the two ends of each DSB interact with a partner duplex in ordered sequence (Introduction; ). A “first” DSB end engages the partner in stable strand invasion (SEI-formation), then primes DNA extension synthesis and resultant formation of pre-dHJs. After pre-dHJ formation, this end is captured into the developing recombination complex by single-strand annealing. Apparently, during the intervening period, the second end remains associated with its sister, at both the DNA and axis levels (Introduction). This “ends-apart” scenario has further implications. (i) At the time of DSB formation, both DSB ends would be sister-associated. (ii) The first DSB end would be released from this association to permit interaction with a homolog chromatid. (iii) The “second” DSB end must remain biochemically quiescent while the first DSB end progresses. (iv) The second DSB end must also eventually be released from its sister to permit its capture into the recombination complex, during the SEI-to-dHJ transition, which occurs at early/mid-pachytene when SC is fully formed (Hunter and Kleckner, 2001
). Since early/mid pachytene is an important global transition point for meiosis (Kleckner et al., 2004
), release of quiescence could be a regulated event, which in turn would imply that quiescence itself is specifically programmed.
In correspondence to these implications (): (i) Sister-association of DSB ends is supported by our finding that cohesin Rec8 is relevant to events prior to and during DSB formation as well as immediately ensuing homolog bias.
(ii) Rec8/cohesion concomitantly promotes sister bias and inhibits use of the homolog. Perhaps it inhibits release of the first DSB end from its sister. Red1/Mek1kinase would then counteract this inhibition, making first end release possible, thereby satisfying preconditions for meiotic homolog bias. Recombinosome components would then ensure that the released end selects a homolog partner rather than its sister.
(iii) Rec8 could mediate maintenance of bias at the SEI-to-dHJ transition by mediating second-end quiescence. The events that normally give rise to in IH-dHJ are initiated at the first/homolog-associated DSB end (above). If these same events initiated, instead, at the second, sister-associated DSB end, the consequence would be formation of an IS-dHJ rather than an IH-dHJ (). The rec8Δ phenotype of loss of bias at the SEI stage can be explained, and in such a way as to give a 1:1 IH:IS dHJ ratio, if Rec8-mediated second-end quiescence would be defective such that pre-dHJ formation can be initiated with equal probability on either end. Red1/Mek1kinase might also contribute to second-end quiescence (above).
Initiation of pre-dHJ formation at both ends of the same DSB seems to be quite rare. Such events would yield MCJMs (Oh et al., 2007
). While somewhat elevated in Rec8− strains, MCJMs are not dramatically prominent (K.K. unpublished). To explain this and other features of the data we suggest that communication between the two DSB ends, via a recombination intermediate that spans the SC (Storlazzi et al., 2010
), may ensure that initiation of pre-dHJ formation (i.e. initiation 3’ extension synthesis) can initiate on only one of the two ends of any given DSB. In WT, Rec8 acts to favor initiation at the homolog-associated end; in Rec8−, this bias is lost. Also, the Rec8− phenotype is probably not explained by a failure to resolve MCJMs, because resolution-defective mutants still exhibit reasonable homolog bias (IH:IS dHJ = 3:1; e.g. Oh et al., 2007
(iv) Modulation of Rec8-mediated sister association would be required for second-end release ().
Programmed quiescence and release of the second DSB end also explains other findings (). (i) Yeast encodes both Dmc1, a meiosis-specific RecA homolog implicated specifically in IH interactions, and Rad51, the general RecA homolog; meiosis also specifies a direct inhibitor of Rad51, Hed1; and it is proposed that Dmc1 binds to the first DSB end while Rad51 binds to the second DSB end (Hunter, 2006
; Sheridan and Bishop, 2006
). Thus: a key role of Rad51/Hed1 could be to promote second end quiescence. Accordingly, a rad52
allele specifically defective in abundant loading of Rad51 confers the same 1:1 IH:IS dHJ ratio as a Rec8− mutant (Lao et al., 2008
(ii) Components of preDSB recombinosomes, e.g. Rec102 in yeast and Spo11 transesterase in several organisms, remain on the chromosomes after DSB formation, into pachytene; further Rec102 is released abruptly, specifically at early/mid-pachytene, i.e. at the time of second-end release (Kee et al, 2004
; Romanienko and Camerini-Otero, 2000
). preDSB recombinosome components may remain bound (at the second DSB end) in order to mediate second-end quiescence.
(iii) Retention of a Rad51-mediated second end/sister interaction leaves open the possibility for return to a mitotic-like inter-sister DSB repair reaction if meiotic IH recombination goes awry, with IS events triggered by activation of second-end release. Accordingly: (i) in mouse, DSBs that lack an homologous partner sequence remain unresolved until early-mid pachytene; and (ii) in allohexaploid wheat, recombinational interactions between homeologous sequences are specifically lost, presumptively to IS repair, at this same stage (Mahadevaiah, 2004; Zickler and Kleckner, 1999
Establishment of DSB/homolog connections via a nucleus-scaled homology-searching tentacle
Tethered-loop axis complexes are clearly present shortly after DSB formation by both molecular and cytological criteria (Blat et al., 2002
; Zickler and Kleckner, 1999
). It is less clear whether this association is created prior to DSB formation, concomitant with development of axial structure, or after DSB formation, with post-DSB complexes associating with already-developed structure. One prior finding points to pre-DSB recombinosome/axis association: DSBs and DSB-associated Dmc1 complexes occur, preferentially, half way between flanking axis association sites, rather than randomly with respect to those sites (Blat et al., 2002
; Kuguo et al, 2009; F.Klein, personal communication). Thus: developing recombination complexes and axis-association sites must communicate prior to DSB formation. Here we provide additional evidence to this effect. (1) All known meiotic axis components are required for maximal levels of DSBs including Rec8 as shown here and elsewhere. (2) Red1/Rec8 interplay is important for the timing of DSB formation. (3) Red1 and Rec8 localize in abundant domains that exhibit longitudinal linearity before DSBs form.
Together, these results support a picture in which DSBs occur in tethered-loop axis complexes that contain both sisters, with DSBs occurring preferentially mid-way between flanking axis-association sites (, ). If so, release of a first DSB end (above) will release a tentacle whose length is approximately half the length of a chromatin loop (). Budding yeast loops are 10–15kb in length (Blat et al., 2002
). A released tentacle would thus be ~7kb, i.e. ~0.3 or ~2µm of nucleosomal filament or naked DNA respectively. These lengths are similar to the diameter of the meiotic yeast nucleus, ~2 µm. Release of a tentacle would thus permit a DSB to search for a homologous partner without the dramatic stirring forces that would otherwise be required to bring DSB ends in contact with homologous partners. Recent findings support long-distance homology recognition (Storlazzi et al., 2010
). Importantly, chromatin loop size scales with genome size (Zickler and Kleckner, 1999
; Kleckner, 2006
), which in turn scales with nucleus size. Thus, DSB formation should universally release a nucleus-scaled homology-searching tentacle ().
Structure-mediated control of recombinational progression
Previous considerations suggest that meiotic chromosome structure plays a central role in controlling the timing of recombination progression in WT meiosis (e.g. Börner et al., 2008
). Our results suggest that Red1/Mek1 and Rec8 are involved in "putting the brakes" on recombination progression and that they act via distinct effects. As a result, when both types of components are absent, biochemical events proceed extremely rapidly.
Red1/Mek1 impedes recombination in both WT and Rec8− strains. Further, Mek1 is Rad53-related, and Rad53 is the primary downstream target of ATR, the replication and DSB repair regulatory surveillance kinase. Thus, Red1/Mek1 might monitor local developments within individual recombinational interactions, ensuring that each biochemical step is completed and new components properly loaded before the next biochemical step can occur (Schwacha and Kleckner, 1997
). These effects likely also involve Pch2 (Börner et al., 2008
). How might Rec8 participate in progression timing? Perhaps Rec8 responds to global regulatory signals derived from the cell cycle, licensing major transitions nucleus-wide. Such effects would link recombination progression to overall cell status and periodically reinforce nucleus-wide synchrony. Together, Red1/Hop1/Mek1 and Rec8 would integrate local "surveillance" signals and global cell cycle-related signals to control progression at both levels.
Domainal differentiation and evolution of the meiotic inter-homolog interaction program
Red1 and Rec8 play functionally distinct roles in every process examined here: sister association and several aspects of recombination including (i) opposing effects for homolog bias establishment; (ii) cooperative roles for maintenance of homolog bias; and (iii) distinct roles for regulation of recombination progression. However, in a mutant lacking both Rec8 and Red1, recombination is still executed normally: initiation, establishment of homolog bias and CO/NCO differentiation occur; CO recombination proceeds via SEIs and dHJs; and CO and NCO products are both formed efficiently. Thus: these structural components only modulate basic biochemical events, which are directly executed by other (i.e. recombinosome) components.
Red1 and Rec8 tend to be enriched in spatially distinct domains along chromosomes on a per-cell basis. We propose that Red1 and Rec8 carry out their distinct but coordinated roles (for cohesion, homolog bias and recombinational progression) via corresponding spatially distinct domains. We proposed previously that meiotic chromosomes might comprise two functionally and structurally different types of regions: interaction domains and stabilization domains, which would occur alternately along chromosomes (Zickler and Kleckner, 1999
). Interaction domains would encourage structural destabilizations needed for pairing and recombination; stabilization domains would provide structural snaps that counteract such destabilization, thereby maintaining chromosome integrity. Red1−rich regions (which are also Hop1-rich regions; Börner et al., 2008
) and Rec8-rich regions could be these two types of domains. In support of this idea: (i) CO sites are associated primarily with Red1/Hop1 domains (Joshi et al., 2009
); and (ii) Red1 is more strongly required for DSB formation and, separately, to ensure that a DSB gives an IH product (i.e. homolog bias) in domains where it is more abundant than in domains where it is less abundant (Blat et al., 2002
). Domainal recombinosome/axis organization could arise easily if each emerging pre-DSB recombination complex tends to nucleate development of a surrounding Red1 domain, concomitantly constraining positions of Rec8 domains.
In the context of domainal control, a specific idea regarding homolog bias emerges. Red1 domains might comprise zones in which, because of they way they developed, Rec8-mediated cohesion is relatively depleted and where, additionally, Red1/Mek1 mediates another type of sister association. This alternative mode would compensate for the deficit of Rec8 but, unlike cohesin-mediated cohesion, would be susceptible to recombination-directed destabilization. Rec8 domains, in contrast, would comprise zones of cohesin-mediated cohesion that is robust and insensitive to recombinosome-directed effects. This model can explain how Red1 could act both positively and negatively for sister cohesion. Further, when Red1 is absent, recombinosome-nucleated formation of Red1 domains would not occur, and unconstrained loading of Rec8 would confer sister bias.
We previously proposed that meiosis evolved by integration of elements from mitotic DSB repair and elements of late-stage mitotic (G2-anaphase) chromosome morphogenesis, with functional linkage achieved via tethering of recombinosomes to structural axes (Kleckner et al., 2004
; Kleckner, 1996
). These two sets of evolutionary inputs could be implemented via spatial and functional domainal organization along the chromosomes.
Red1/Hop1/Mek1kinase domains would mediate effects evolved from mitotic DSB repair, modulating execution of recombination and controlling local progression (above), while Rec8 domains would mediate effects evolved from modulation of cohesion status that normally occur during the latter stages of the mitotic cell cycle.