TI - Two possible models for late action of MUS81 in mammalian meiosis . . AB - The second model assumes that MSH4-MSH5 and MUS81 act in separate pathways , between which there is some degree of crosstalk . This model also assumes that mammalian MUS81 acts at HJs in vivo , which has been shown in vitro using recombinant human MUS81 [16] . The presence of MUS81 may then prevent MLH1-MLH3 binding to a subset of CO sites , resulting in processing of these COs down a MUS81 -dependent pathway , which is , by inference from the yeast data , interference -independent [8] . In the absence of MUS81 in Mus81-/- spermatocytes , MLH1-MLH3 would not be repressed at these CO sites , thus increasing overall MLH1 focus number , and decreasing the amount of interference between MLH1 foci , while maintaining the eventual number of chiasmata ( see below ) . MUS81-MLH1-MLH3 crosstalk may be mediated by BLM , as BLM is known to interact with MUS81 in somatic cells [51] and with key components of the DNA mismatch repair family , including MLH1 [52] . These data predict that interference is laid down prior to MLH1-MLH3 accumulation , and remains in place regardless of the pathway by which COs are processed . This represents the first data on interference decision and timing in mouse spermatocytes and is supported by data concerning the role of MUS81-EME1 in cleaving HJs in vitro and in vivo . MUS81 readily cleaves non-HJ SUBstrates in vitro[12] , [45] , [53] , [54] , [55] , and can efficiently catalyse symmetrical and coordinated cleavage of intact HJs in vitro[14] , [16] . The potential importance of this intact HJ resolution activity is supported by indirect evidence that MUS81-EME1 cleaves HJ in vivo , atleast in fission yeast [12] , [15] , but has yet to be demonstrated definitively in other eukaryotes . A third possibility is that the extra MLH1 foci seen in Mus81-/- mice are not representative of extra COs , but instead of aberrant repair structures that require MUS81 protein to be repaired correctly , or which arise as a result of failure of MUS81-driven processes ( as discussed above ) . From our data , it is difficult to say whether this is the case . The persistent BLM and RAD51 foci seen in the mutant spermatocytes indicate repair defects , rather than crossover anomalies yet , at the same time , since we see increased MLH3 foci in addition to the increase in MLH1 foci , it is equally likely that these structures represent nascent COs rather than unrepaired breaks . In Mus81-/- mice , the number of chiasmata was not different from that of WT mice , despite the increase in MLH1 foci at pachynema . However chiasma numbers were significantly reduced with respect to MLH1 focus number in Mus81-/- pachytene cells , indicating either that some MLH1 foci are lost before they are resolved as chiasmata or that cells with higher number of MLH1 foci undergo apoptosis following pachynema . Cells with MLH1-MLH3 foci too close together may be subject to additional processing by downstream factors limiting the crossover number , such as EXO1 or BLM ( Sgs1 in yeast ) [56] , [57] which , we suggest , would normally overlook interference -independent MUS81 foci . Unfortunately , analysis of the localization of EXO1 is precluded by the absence of an antibody that is functional on chromosome spreads . However BLM appears to persist into late pachynema in Mus81 spermatocytes and may account for this additional surveillance mechanism . Oh et al.[57] have proposed that Sgs1 in yeast functions to suppress closely spaced COs by preventing the formation of complex recombination intermediates involving multiple chromatids . In this model , Mlh1-Mlh3 complexes act to promote inter-homolog strand invasion , in part by antagonizing Sgs1 . At the same time , Sgs1 acts to disassemble complex recombination intermediates that might include those resulting from closely spaced DSB processing events . Given these suggestions , the current data indicate that the absence of MUS81 in mice results in closely spaced interference -independent MLH1-MLH3 events that then become the target for directed BLM action . We suggest , therefore , that CO control in the mouse involves complex integration between MUS81 and MLH1-MLH3 , perhaps mediated by BLM . The current study provides strong evidence that COs in mammals are controlled by atleast two pathways , and that MUS81 is responsible for processing a distinct subset of these events . Whether this control is mediated early in prophase I , at the level of RAD51 acquisition , or later , upon accumulation of MLH1/MH3 , remains to be seen . Clearly , however , the function of MUS81 in meiosis differs between eukaryotic species and also possibly betweeen its function in the context of replication and meiosis . Continued cross-species analysis of MUS81 action will help to elucidate the nature of its activity in these diverses processes .