TI - Discussion . AB - These studies demonstrate that deletion of Mus81 in mice has severe consequences for meiotic progression . Male Mus81-/- mice exhibit reduced testis size and epididymal sperm numbers , coupled with increased seminiferous tubule apoptosis that is not confined to a single stage of spermatogenesis . Meiotic DSB repair appears normal in the early stages of prophase I , as assessed by the accumulation of RAD51 and MSH4 on meiotic chromosomes , but by pachynema there is a significant increase in the numbers of MLH1 and MLH3 foci on the chromosome core of both spermatocytes and oocytes from Mus81 nullizygous animals . Since recombination rate ( and hence MLH1 focus frequency ) is tightly associated with synaptonemal complex length [40] , it is important to note that these increases in MLH1-MLH3 foci are not associated with changes in synaptonemal complex length . Moreover , we see reduced interference amongst MLH1 foci in spermatocytes from Mus81-/- males , indicating that the additional MLH1-MLH3 events are not subject to the normal strict regulation of crossover placement that is essential to ensure appropriate segregation of chromosomes at the first meiotic division . Late in pachynema , spermatocytes from Mus81-/- males show persistent and upregulated localization of BLM helicase indicating a failure to repair DSBs appropriately and/or the presence of aberrant DNA structures in late prophase I . Interestingly , however , the persistence of BLM is associated with normal chiasmata numbers at diakinesis in Mus81-/- males , despite the increase in MLH1-MLH3 foci observed at pachynema . The data herein represent the first comprehensive analysis of the effects of Mus81 mutation on meiotic progression and DSB repair in mice and demonstrate the possible existence of a second CO pathway in mammalian meiosis . Moreover , these data indicate important cross-regulatory mechanisms between the two CO pathways in mammals . In S.cerevisiae and S.pombe , the synthetic lethality of mus81.sgs1 ( or mus81rqh1 for pombe ) double mutants [12] , [41] , [42] , [43] can be rescued by deletion of rad51[44] , [45] , suggesting that the removal of homologous recombination events can prevent the accumulation of toxic recombination events during vegetative growth . Extensive studies in S.cerevisiae have also demonstrated the requirement in lagging strand replication for additional interactions between Mus81 and Sgs1 that are independent of homologous recombination [44] , and the role of Mus81 in this scenario is , in fact , primarily independent of Rad51 [46] . Li and Brill [46] have proposed a model in which Mus81 cleaves 3'-flaps present on the lagging strand to result in polymerase-directed repair in most cases . Alternatively , the 3' end may become recombinogenic , resulting in Rad51 -directed double Holliday junction ( HJ ) formation and subsequent activation of the major repair pathway involving Sgs1 . Additionally , a minor pathway may be initiated by Rad51 -mediated strand displacement and subsequent Mus81 activity [46] . Thus , in this model , Mus81 functions to remove 3' flaps prior to Rad51 -induced double HJ formation . Similar models have been proposed during meiosis in S.cerevisiae , supported by the observations that double HJs do not accumulate in mms4 mutant strains [8] , [47] , that the reduction in crossing over in this strain is extremely modest ( HJ resolvase fails to suppress the mus81 meiotic phenotype [8] . Despite the lack of evidence supporting a role for Mus81 in double HJ resolution during meiosis in yeast , mutants for mus81 or mms4 show delayed repair of DSBs and appearance/disappearance of recombination intermediates . This involvement in later stages of DSB repair may be explained by the model proposed by De Los Santos et al , in which Mus81/Mms4 are required for a subset of recombination intermediates in which over-replication of the displaced invading strand following D-loop formation ( for details see [8] ) . Taken together , these models all suggest a mechanism by which Mus81 functions upstream ( and independently of ) Rad51 , presumably through cleavage of 3' flaps , to effect DNA repair but , in addition , can act downstream of Rad51 to mediate repair involving Holliday junction intermediates ( but again through its action on 3' flaps ) . Our data on MUS81 function in mice is congruent with both possibilities presented above . An early role for MUS81 in prophase I in mice is indicated by the loss of meiotic cells from leptonema onwards , and by the observation of increased RAD51 staining associated with regions of asynapsis and synaptic disruption in zygonema ( Figure 2C ) , although it must be noted that homolog association and synapsis are unaffected in S.cerevisiae mms4 mutants [8] . However , the fact that RAD51 focus numbers are largely unaffected in Mus81 nullizygous mice , together with the observation that MSH4 focus frequency is normal in these animals , would argue that the early stages of DSB repair are unaffected by the absence of MUS81 . It is possible that the subset of DSB events that are destined to become subSTRates for RAD51 is unaffected by the loss of MUS81 and that only those MUS81 -dependent DSBs are then left unrepaired , perhaps becoming the SUBstrate for DNA mismatch repair processes . However , this is inconsistent with our observation that the increase in MLH1/MLH3 focus numbers does not occur until mid-to-late pachynema , much later than the appearance of these aberrant DNA structures . Moreover , that the additional foci of MLH1 are equally represented by additional foci of MLH3 implies that these extraneous MutL heterodimers are of the MLH1/MLH3 variety ( involved in recombination events ) and not of the MLH1/PMS2 variety ( involved in canonical mismatch repair ) , although analysis of PMS2 localization is prevented by the absence of a functional antibody . Nonetheless , we cannot exclude the possible involvement of a non-canonical mismatch repair complex in these events , nor can we dismiss the possible importance of other repair pathways in these processes . Alternatively , such aberrant structures , which fail to be processed by MUS81 in early leptonema , could proceed through prophase I to pachynema , whereupon they become SUBstrates for MLH1/MLH3 accumulation ( giving rise to the increased focus numbers for these MutL homologs ) . The lack of an appreciable increase in MSH4 foci would argue against this possibility , although the predicted increased may be too small to be evident above the normal level of MSH4 accumulation . Previous studies from other organisms have shown Mus81 to be important in the processing of interference -independent COs [8] , [9] We believe non-interfering COs might be generated in the same way in mice , as Mus81-/- males show irregularities in processing of late recombination intermediates , characterized by a significant increase in interference -independent MLH1 foci . Intriguingly , this increase in MLH1 does not correspond to an increase in chiasmata in the Mus81-/- mice , when compared with WT ( discussed below ) . Thus , Mus81 deletion represents the first single null mutant in which an increase in MLH1 foci is not correlated with increased SC length and , more importantly , in which the increase in MLH1 focus numbers does not result in an increase in the final tally of chiasmata . That these increased MLH1 foci are associated with similar increases in MLH3 suggests that the MLH1 function at these sites is one of recombination rather than of DNA mismatch repair ( which would utilize PMS2 ) . Unfortunately , the lack of a MUS81 antibody that detects the protein on chromosome spreads precludes detailed analysis of MUS81 localization in spermatocytes ( Holloway and Cohen , unpublished data ) . These observations point to alterations in crossover control at the level of atleast two distinct , but converging , recombination repair pathways and not to regulation of crossover frequency by the SC per se . Given the late prophase I increase in MLH1-MLH3 foci in Mus81 nullizygous animals , it is possible that MUS81 may play both an early and a late role in recombination events in the mouse , the two possibly being delineated by RAD51-independence versus RAD51-dependency as suggested by Li and Brill [46] . For the later , RAD51 -dependent function , there are two possible models for the interplay between MSH4-MSH5 , MUS81 and MLH1-MLH3 in generating COs ( Figure 7 ) . The first posits that MSH4-MSH5 will bind model dHJs [48] , recruit MLH1-MLH3 to the majority of CO sites , while recruiting MUS81 to the remaining subset of CO sites ( or the presence of MUS81 at this subset prevents MLH1-MLH3 recruitment here ) . In the absence of MUS81 , MSH4-MSH5 directs all COs to be processed by MLH1-MLH3 . There is evidence from tomato that MLH1-positive COs and MLH1-negative COs may arise from the same early precursors [39] . Moreover , data from mice and humans show that genetic maps are consistently longer than those estimated from MLH1 focus counts , suggesting that MLH1 foci do not represent all the physical COs generated during meiosis [7] , [49] , [50] . This model , however , would require physical interaction between MSH4-MSH5 and MUS81 , a prerequisite for which there is no published data . In addition , evidence from A.thaliana suggests that , if this were the case , Atmsh4 mutants and Atmsh4.mus81 double mutants would show the same number of residual chiasmata , which is not the case [20] .