Roles for mismatch repair family proteins in promoting meiotic crossing over

Abstract

The mismatch repair (MMR) family complexes Msh4-Msh5 and Mlh1-Mlh3 act with Exo1 and Sgs1-Top3-Rmi1 in a meiotic double strand break repair pathway that results in the asymmetric cleavage of double Holliday junctions (dHJ) to form crossovers. This review discusses how meiotic roles for Msh4-Msh5 and Mlh1-Mlh3 do not fit paradigms established for post-replicative MMR. We also outline models used to explain how these factors promote the formation of meiotic crossovers required for the accurate segregation of chromosome homologs during the Meiosis I division.

1. Post-replicative DNA mismatch repair in prokaryotes

DNA mismatch repair (MMR) is a highly conserved mechanism that acts during DNA replication to remove DNA misincorporation errors that result in base-base and loop mismatches. Mutations in MMR genes confer increased mutation rates and have been linked in humans to Lynch syndrome (reviewed in [1,2] and in [3,4], this issue). In Escherichia coli, mismatches resulting from DNA polymerase misincorporation events are recognized by MutS. In an ATP-dependent reaction, MutS recruits MutL to activate MutH, a latent methylation-sensitive endonuclease that nicks the newly replicated and transiently un-methylated strand at distances up to several kilobases from the mismatch. The resulting nick promotes excision, polymerization, and ligation, resulting in removal of the mismatch (reviewed in [1] and in [5,6], this issue).

2. Eukaryotic MMR

Multiple MutS and MutL homologs (MSH and MLH proteins, respectively) have been identified in eukaryotes, however, no MutH homolog or methylation-dependent nuclease activity has been detected, suggesting that strand discrimination occurs through a different mechanism than seen in E. coli (see below). In Saccharomyces cerevisiae, base-base and small insertion/deletion mismatches are primarily recognized by the Msh2-Msh6 complex, while large insertion/deletion loops are recognized by Msh2-Msh3 heterodimers [2,7–10] (reviewed in [11] and in [12], this issue). Though not an absolute requirement for MMR, MSH complexes can interact with the replication fork via the replicative processivity clamp (proliferating cell nuclear antigen—PCNA) [13] (and reviewed in [14,15], this issue). In yeast, the Mlh1-Pms1 heterodimer (Mlh1-Pms2 in humans) is the primary MLH component in MMR and is recruited by MSH complexes to mismatched DNA. This interaction initiates downstream excision and re-synthesis steps through either Exo1-dependent or Exo1-independent pathways (reviewed in [16]). There is also a minor MMR pathway that involves mismatch recognition by Msh2-Msh3, followed by interaction with Mlh1-Mlh3, a heterodimer that participates minimally in MMR, but plays a major role in meiotic recombination [17].

MSH proteins form ring-like structures that encircle mismatched and homoduplex DNA [10,18–23] (and reviewed in [6], this issue). These proteins contain highly conserved ATP binding domains. Biochemical, structural, and biophysical studies have shown that MSH conformation and DNA binding are modulated by ATP binding and hydrolysis [24–34]. Differences in nucleotide affinities between MSH subunits have led to a model proposing that in the absence of a mismatch, ADP and ATP are bound to MSH subunits at high and low affinity sites, respectively, with the MSH complex showing a weak affinity for DNA due to the opening and closing of the MSH ring triggered by asymmetric ATP hydrolysis (Figure 1A) [24–34] (and reviewed in [12], this issue). When the MSH complex encounters a mismatch, ADP rapidly dissociates from one subunit. Both subunits are then occupied by ATP, causing the dimer to change conformation and enter a sliding clamp mode (Figure 1B). The ATP-bound sliding clamp complex is then able to recruit downstream repair factors, such as MLH complexes (Figure 1B). Upon interaction with downstream repair signals that presumably initiate excision steps, the MSH-MLH sliding clamp undergoes ATP hydrolysis followed by recycling of MSH complexes.

Figure 1. MSH sliding clamps recruit downstream factors required for MMR during DNA replication.

The model presented is based on work described in: [24–34]. It should be noted that the associations and disassociations of the factors, as well as stoichiometries are not well established in MMR. (A) MSH dimers (green and blue) have asymmetric nucleotide binding sites (high and low affinity). ATP is indicated by red T and ADP by black D. (1) In the absence of a mismatch, asymmetric ATP binding and hydrolysis results in the opening and closing of the dimeric ring. When ADP is bound to both subunits, the dimer is in the open state and has weak affinity for DNA. Exchange of ADP for ATP (2) is thought to close the ring (3). (B) Recognition of a mismatch causes the MSH complex to exchange ADP for ATP at both sites (1). (2) The MSH dimer changes conformation and enters a sliding clamp mode, moving away from the mismatch (3) to recruit downstream factors, such as MLH proteins (orange and yellow), PCNA (brown), and Exo1 (red), and a search for strand discrimination signals that are hypothesized to involve pre-existing strand discontinuities on either leading or lagging strands. (4) Excision of the mismatch occurs, leaving gaps to be filled in by Polδ.

Structural and biophysical studies on bacterial MutL, yeast Mlh1-Pms1, and human Mlh1-Pms2 have suggested that like the MSH proteins, MLH dimers also encircle DNA [35–41]. Each MutL or MLH monomer consists of a globular N-terminal and C-terminal domain (NTD and CTD, respectively) connected by a flexible linker arm. The NTD of MutL and MLH proteins contains a conserved nucleotide binding site, while the CTD is essential for MLH protein dimerization. In E. coli, MutL functions as a matchmaker to facilitate interaction between MutS and the MutH endonuclease. Interestingly, certain bacterial MutL, primarily those that lack MutH strand discrimination mechanisms, and a subset of the eukaryotic MLH proteins display endonuclease activities and contain highly conserved metal-binding motifs (DQHA(X)₂E(E)₄E) in their CTDs that are required for this activity. Two of the four yeast MLH proteins, Pms1 and Mlh3, contain this conserved motif, and both Mlh1-Mlh3 and Mlh1-Pms1, as well as their human homologs, display endonuclease activity on double stranded DNA [42–46].

Why do a subset of MLH proteins contain an endonuclease activity? Biochemical studies by Modrich and colleagues [43,44] (and reviewed in [47], this issue) have shown that this activity introduces nicks into newly replicated DNA that are used as entry points for exonucleases to promote 5’ or 3’ directed MMR. This activity may also partially facilitate strand discrimination for downstream repair. The mechanisms that direct the endonucleases to preferentially nick the daughter strand over the template in regions near a mismatch are not fully understood, but interactions within and between MLH proteins and DNA replication components are likely to be important. This is supported by the following: (1) Full length but not the CTD of B. subtilis MutL displays in vitro endonuclease activity [48]; (2) Structural studies involving B. subtilis MutL and yeast Mlh1-Pms1 suggest that MLH proteins undergo large conformational changes that involve nucleotide cofactors, the NTD, the CTD, unstructured linker arm domains, and DNA [40,48,49]; (3) Interactions have been identified between MutL/MLH complexes, the processivity clamp/clamp loader, MSH complexes, and pre-existing discontinuities in the daughter strand [43–45,50,51]; and (4) human Mlh1-Pms2 endonuclease activity is strand-specifically activated by PCNA when properly oriented by the clamp loading complex [43,51].

3. MMR roles for MSH and MLH complexes in meiosis

In addition to the roles outlined in post-replicative MMR, eukaryotic MMR proteins play critical roles in regulating genetic recombination, maintaining genome stability, and expanding trinucleotide repeat sequences (reviewed in [11,52–54]). This review focuses on the role of MMR proteins in facilitating crossing over between homologs to ensure proper chromosome segregation during the first meiotic division. Failures in this process can result in aneuploidy syndromes (e.g. Down, Turner syndromes) resulting from chromosomal nondisjunction events occurring primarily at the first meiotic division (Figure 2) [55].

Figure 2. Crossing over facilitates proper chromosomal segregation in Meiosis I.

Figure is based on models presented in [56,165,166]. During meiosis, homologs (blue and red) are duplicated, forming sister chromatids which are held together by cohesin (1). Recombination between the homologs forms chiasma (2, top, and magnified inset). Normal segregation proceeds when the crossover and cohesion between sister chromosomes allows for proper alignment and attachment to the spindle poles (3-4, top). Crossovers that do not occur (2, bottom) or do not occur in the correct location can cause segregation to the same spindle pole, resulting in nondisjunction (3-4, bottom).

3.1 Double strand break repair mechanisms

Crossing over between the homolog pairs in meiotic prophase results in the formation of chiasmata that allow for proper chromosome alignment and segregation during the first meiotic division (Figure 2; reviewed in [56]). Failure to form at least one crossover per chromosome pair, referred to as the obligate crossover, can lead to high levels of nondisjunction. As outlined below, extensive studies have been performed to understand how crossing over is controlled and distributed.

In S. cerevisiae, the recombination process leading to crossover formation begins in early meiotic prophase with the formation of ~150-200 programmed double strand breaks (DSBs) distributed throughout the genome (Figure 3) [57–65]. DSBs are resected, creating 3’ single-stranded tails that are coated with the single-strand binding protein RPA. These tails can invade the double-stranded homologous template in a reaction catalyzed by Rad51, Dmc1, Rad52, and Rad54, creating a displacement-loop (D-loop). Stable heteroduplex DNA intermediates created in this pathway are known as single-end invasion intermediates (SEIs) [66]. Models predict that SEIs, which in baker's yeast are biased to form between homologs, rather than sister chromosomes [67], can be processed by a number of mechanisms outlined in Figure 3 . SEIs can be disassembled by the Sgs1-Top3-Rmi1 helicase/topoisomerase complex (see section 5.2) [68–70], producing non-crossover products by the synthesis-dependent strand-annealing pathway (SDSA), or can be further processed by a second-end capture mechanism facilitated by RPA, Rad52, and Rad54 [71–78], ultimately maturing into a dHJ intermediate. In vivo studies suggest that dHJs stabilized by a family of proteins termed the ZMMs (described in section 4.1) are resolved in a biased orientation producing crossover products (Figure 3B) [79–82]. In yeast, a small percentage of crossovers form independently of the ZMM pathway (see section 5 below; Figure 3C). It is important to note that evidence for the recombination intermediates described above has been provided by a combination of genetic, molecular (two-dimensional gel electrophoresis), and biochemical studies [66,67,83].

Figure 3. Multiple pathways for strand invasion intermediates.

In early prophase I, Spo11 catalyzes the formation of ~150-200 DSBs that undergo 5’ to 3’ end resection. The 3’ single-stranded tails can then invade the homolog to form a D-loop in a reaction catalyzed in part by RPA, Rad52, Rad51, Dmc1, and Rad54. The resulting D-loop intermediate can be: (A) unwound by Sgs1-Top3-Rmi1, followed by synthesis and re-annealing to form non-crossovers (SDSA pathway); (B) stabilized by ZMM proteins (Zip3 and Msh4-Msh5) to enable second-end capture (catalyzed in part by RPA, Rad52, and Rad54), DNA synthesis, branch migration, and ligation, forming a dHJ that is resolved asymmetrically by Mlh1-Mlh3 to form only crossover products; (C) form dHJ independent of ZMM stabilization resolved by Mus81-Mms4 (or Slx1-Slx4 or Yen1) to give a mixture of crossover and non-crossover products.

3.2 MMR in double strand break repair: meiotic gene conversion and heteroduplex rejection

DNA mismatches are formed in heteroduplex DNA during double-strand break repair (DSBR) when the donor and acceptor molecules differ in sequence. The repair of these mismatches by the MMR machinery has been well studied and results in gene conversion when the MMR machinery repairs mismatches on one chromosome to sequences present on the other (reviewed in [79,84,85]). Yeast tetrad analysis and high-throughput DNA sequencing of meiotic progeny have provided valuable insights into how this process occurs [79,85–90]. Numerous studies in yeast have monitored the segregation of genetic markers as a function of their position relative to an initiating DSB site (gene conversion gradient). This relationship is linear for several loci, where there are high levels of aberrant segregation of markers near the DSB site and low levels further away. Strains deficient in MMR (null mutations in MSH2, PMS1, or MLH1) display gene conversion gradients with a more gradual slope; as the distance from the DSB increases, the frequency of aberrant segregations decreases by a smaller amount compared to MMR-proficient strains [85,90–92]. This observation has given rise to a model where discontinuities in DNA direct the orientation of repair analogous to strand-discrimination steps proposed for post-replicative repair. At sites near the initial DSB, the break directs repair to form gene conversions, but mismatches more distant from the DSB are repaired either randomly or are directed by nicks formed through the activity of Holliday junction resolvases to yield a higher frequency of restoration events. It is also likely that the frequency of heteroduplex formation contributes to gradient formation [91]. In addition to their roles in meiotic pathways that result in gene conversion, MMR proteins prevent recombination between divergent DNA sequences (reviewed in [85,93]). This meiotic role is similar to the anti-recombination role seen for MMR proteins during DSBR in non-meiotic cells [94,95] (and reviewed in [96], this issue).

4. A majority of meiotic crossovers depend on a subset of MSH and MLH complexes

As outlined above, eukaryotic MSH and MLH complexes play essential roles in recognizing mismatches in heteroduplex DNA and the subsequent promotion of downstream repair events, including gene conversion and heteroduplex rejection. In addition to these canonical roles, in some organisms, e.g. S. cerevisiae and mammals, a subset of MSH and MLH proteins participate directly in crossover-promoting processes in roles distinct from their MMR functions. These are the Msh4-Msh5 heterodimer, which has no role in the recognition or repair of mismatches, but is vital in meiosis [97–100], and Mlh1-Mlh3, which has a minor role in MMR, but is critical to maintain wild-type levels of crossing over [82,83,101–103].

4.1 Meiotic roles for Msh4-Msh5

Msh4-Msh5 is a member of the ZMM (Zip1-4, Mer3, and Msh4, 5) group of meiotic proteins needed to form crossovers. Early in meiosis, pairing between homologous chromosomes is initiated by strand invasion events facilitated by the synaptonemal complex, composed of proteins that form a zipper-like structure along the lengths of homologous chromosomes (reviewed in [104]). zmm mutants are defective in synaptonemal complex formation, suggesting a role for ZMM proteins in their assembly. Additionally, zmm mutants have reduced numbers of crossovers, and localization studies suggest that ZMM proteins localize to sites of recombination, and specifically mark sites for crossing over [83,105–109]. Also in baker's yeast, expression of ZMM proteins, including Msh4-Msh5, is specifically induced in meiosis [97,98,110]. These observations suggest roles for ZMM proteins in chromosome pairing and crossover formation.

Msh4 and Msh5 share homology with four of the five domains present in other MSH proteins including the structural domain containing the ATP-binding site. These proteins, however, lack domain I, one of two domains critical for the binding of MSH proteins to mismatch DNA [20,97,98]. Genetic studies have shown that Msh4 and Msh5 are essential for maintaining wild-type levels of crossing over [97–99,102], and Msh4 and Msh5 have early and late meiotic roles. Work in Sordaria suggested that Msh4 can act upstream of the other ZMM proteins during the leptotene stage of meiotic prophase to facilitate chromosome pairing [111]. In budding yeast, Msh4-Msh5 has been shown to stabilize strand invasion intermediates in the later zygotene stage [105]. In both mouse spermatocytes and in human oocytes Msh4 foci are seen at all DSB sites during leptotene, but by late pachytene Msh4-Msh5 foci are only found at sites for crossover formation [112].

4.2 Msh4-Msh5 and Mlh1-Mlh3 are involved in a meiotic crossover resolution pathway

Previous studies have indicated that in addition to roles in MMR, Mlh1 is important to form crossovers in meiosis, and Mlh1 and Mlh3 interact in the same meiotic pathway responsible for crossover formation [103,113,114]. Genetic analysis in baker's yeast has defined a pathway involving Msh4-Msh5 and a dHJ resolvase activity suggested to be Mlh1-Mlh3; strains deficient in both Mlh1 and either Msh4 or Msh5 show the same decrease in crossover formation as mlh1 single mutants [102,115,116]. Furthering the evidence for a Msh4-Msh5/Mlh1-Mlh3 pathway are mouse spermatogenesis data showing that Mlh1-Mlh3 foci form following the appearance of Msh4-Msh5 foci. Msh4-Msh5 foci localize to DSB sites early in meiosis and are reduced in frequency as meiosis progresses. By mid-pachytene Mlh1 foci appear and mostly co-localize with Msh4 foci, and Mlh1 and Mlh3 foci are only present at sites for crossover resolution [117–120].

5. Endonucleases act to resolve Holliday junction intermediates that form prior to crossover resolution in meiosis

As described in section 2, Mlh3 contains a highly conserved metal-binding motif found in MLH proteins that display endonuclease activity. This observation, coupled with localization and genetic analysis of Mlh1-Mlh3, suggest that it acts as a dHJ resolvase. Curiously, Mlh1-Mlh3 does not share conservation with the known structure-selective endonuclease superfamilies: XPF, URI-YIG, and Rad2/XPG.

The conserved metal-binding motif found in S. cerevisiae Mlh1-Mlh3 and Mlh1-Pms1 and human Mlh1-Pms2 is required for the endonuclease and MMR functions of these complexes [43,44]. A genetic analysis of this domain in baker's yeast showed that a mlh3-D523N mutation, where the highly conserved aspartic acid residue within the DQHA(X)₂E(E)₄E motif was changed to asparagine, conferred an mlh3-null-like phenotype in meiosis, illustrating the importance of this putative endonuclease function in the meiotic Mlh1-Mlh3 pathway [101]. Yeast and human Mlh1-Mlh3 display in vitro endonuclease activity that is not seen in the purified Mlh1-mlh3-D523N complex. To date cleavage by Mlh1-Mlh3 of synthetic Holliday junctions or related small substrates has not been observed [45,46].

5.1 Characterization of resolvases that act on model Holliday junctions

Cleavage of synthetic Holliday junctions and other branched intermediates constructed from oligonucleotides is considered an important biochemical signature of Holliday junction resolvases. In S. cerevisiae (and humans), three eukaryotic nucleases that are capable of nicking such junctions in vitro have been identified and characterized using model four-way junctions: Mus81-Mms4 (MUS81-EME1) (XPF family), Slx1-Slx4 (SLX1-SLX4) (URI-YIG family), and Yen1 (GEN1) (Rad2/XPG family) (Figure 4) (reviewed in [121]). In vitro, Mus81-Mms4 nicks flapped DNA substrates and four-way Holliday junctions, among other branch structures [122,123]. This complex has been implicated as the primary resolvase in a crossover pathway that does not require ZMM proteins to stabilize crossover intermediates (Figure 3C), and has been suggested to prevent chromosome entanglements in both crossover and non-crossover pathways [102,115,124]. In both in vitro and in in vivo assays, Mus81-Mms4 is activated by phosphorylation by Cdc5 kinase, suggesting that its meiotic resolvase activity is carefully regulated through post-translational modification [125,126]. Yeast and human Slx1-Slx4 are URI-YIG family endonucleases that are also structure-specific, capable of cleaving forked junctions, 5’-flapped substrates, and four-way Holliday junctions in vitro [127–129]. Both Mus81-Mms4 and Slx1-Slx4 are essential for the formation of meiotic crossovers in the absence of Sgs1 [82]. Yen1 is a member of the Rad2/XPG superfamily and has been shown to cleave Holliday junction structures in vitro [130,131] and can at least partially compensate for loss of Mus81-Mms4 function in vivo [82].

Figure 4. Substrate specificities of S. cerevisiae dHJ resolvases.

Figure is adapted and based on data reviewed in [121] and determined for Mlh1-Mlh3 in [45,46]. As described in section 5.1, Mus81-Mms4, Slx1-Slx4, and Yen1 are structure-specific endonucleases that cleave branched structures at discrete locations (green triangle). Mus81-Mms4 displays weak endonuclease activity on continuous four-way junctions (grey) relative to pre-nicked junctions and other indicated branched structures. Slx1-Slx4's in vitro nicking positions are currently not determined (ND) on model replication forks. Nicking of model four-way junctions and other branched structures by Mlh1-Mlh3 has not been observed, and activity on supercoiled or open circle DNA substrate has been observed to be non-specific (NS).

Curiously, the combined activities of the above nucleases account for the formation of only a minority of meiotic crossovers; the majority are dependent on a pathway requiring the MMR complex Mlh1-Mlh3 [82]. In addition to appearing unrelated in homology to the established resolvases, Mlh1-Mlh3 is also distinct in its actions as a putative resolvase. Resolution of dHJ intermediates by Mus81-Mms4, Slx1-Slx4, and/or Yen1 results in a mixture of crossover and non-crossover products, whereas resolution by Mlh1-Mlh3 results in only crossover products [82].

5.2 Evidence for a Msh4-Msh5, Mlh1-Mlh3, Sgs1-Top3-Rmi1, and Exo1 meiotic crossover pathway

The failure of Mlh1-Mlh3 to conform to paradigms set by previously characterized resolvases suggests that it may be directed and positioned at dHJs in a way that is unique and involves the cooperation of other protein components. Candidates for directing Mlh1-Mlh3 functions are Msh4-Msh5, Sgs1-Top3-Rmi1, and Exo1.

The RecQ family helicase Sgs1 in S. cerevisiae (BLM in humans) has been established as a key crossover regulator and is involved in early crossover regulation steps. In meiosis, Sgs1 functions in a complex with type-I topoisomerase (Top3) and Rmi1 [68–70,132–134]. Initially, Sgs1 was assigned as an anti-crossover factor because it facilitates unwinding and decatenation events [73,135–138]. More recently it has been labeled a pro-crossover factor based on the following observations [69,70,82,139]. (1) In S. cerevisiae, in the absence of Sgs1, Top3, or Rmi1, SEIs and dHJs accumulate and are primarily resolved by Mus81-Mms4 [69,70,82,139]. (2) In budding yeast Zakharyevich, et al. [82] found that in an mms4 slx4 yen1 triple mutant, crossing over at a meiotic hotspot was reduced to ~70% of wild-type levels. (3) In an mms4 slx4 yen1 sgs1 quadruple mutant, the number of crossovers at this hotspot was reduced to ~10% of wild-type levels [82]. Similar results were found in strains deficient in Mms4, Slx4, Yen1 and either Top3 or Rmi1 [69,70]. These data suggest that the Sgs1-Top3-Rmi1 complex plays an important role in promoting crossover formation and does so in conjunction with a resolvase that is not Mus81-Mms4, Slx1-Slx4, or Yen1. Consistent with this, when an mms4 slx4 yen1 mlh3 mutant was made, crossovers were reduced to ~10% of wild-type levels, the same extent seen with the mms4 slx4 yen1 sgs1 quadruple mutant. These data suggest a pathway where Sgs1-dependent crossovers require Mlh1-Mlh3. Furthermore, non-crossovers were reduced ~four-fold, compared to wild-type, in an mms4 slx4 yen1 sgs1 quadruple mutant, but in the mms4 slx4 yen1 mlh3 mutant, non-crossovers form at the same levels as in wild-type strains. These data suggest a role for Sgs1, but not Mlh3, in the formation of non-crossovers. Collectively, these studies indicate that the majority of crossovers form through the actions of either Mlh1-Mlh3 or Mus81-Mms4, with Sgs1-Top3-Rmi1 regulating both [69,70,82,139]. This was consistent with earlier data showing that mms4 mlh1 double mutants display significant decreases in crossing over compared to mlh1 or mms4 single mutants, suggesting that Mlh1-Mlh3 and Mus81-Mms4 are responsible for the majority of meiotic crossovers [102].

The MMR component Exo1 was also shown to be important in the Mlh1-Mlh3 resolution pathway. Exo1, a Rad2/XPG family nuclease, displays 5’ to 3’ exonuclease activity and an endonuclease activity capable of nicking 5’-single strand flaps. Exo1 has been implicated in a wide variety of cellular processes, including MMR, replication, and recombination. Meiotic crossing over is reduced by as much as two-fold in S. cerevisiae exo1 strains, suggesting an important role for the protein in at least one meiotic crossover pathway [140–142]. In S. cerevisiae Zakharyevich, et al. [143] showed that mlh3 and exo1 strains displayed similar levels of SEI and dHJ intermediates; however, crossover formation in both mutants (as well as in an mlh3 exo1 double mutant) was similarly reduced. These observations support a late resolution step involving both Exo1 and Mlh1-Mlh3. Surprisingly, they observed that Exo1's catalytic activity was not required to maintain wild-type levels of meiotic crossing over, suggesting an exonuclease-independent and perhaps structural role for Exo1 in this pathway. Zakharyevich, et al. [82] presents data supporting a structural role for Exo1 in the Mlh1-Mlh3 resolution pathway. These observations are supported by work performed by Borts and colleagues [144] who found Exo1 activity to be important in both early and late stages of meiosis, and that these roles could be genetically separated in the exo1-D173A exonuclease mutant which was functional for its late role in meiosis. Together these data suggest that Exo1's exonuclease activity is not required for its late role in promoting meiotic crossovers.

Work in baker's yeast and in mammals have led to the proposal of a pathway where a majority of crossovers occur through the functions of Msh4-Msh5, Sgs1-Top3-Rmi1, Exo1, and the Mlh1-Mlh3 endonuclease. This pathway is subject to crossover interference, a mechanism that ensures that crossovers are spaced non-randomly and that formation of a crossover in a specific region of the chromosome reduces the probability that another crossover will form in a nearby region [83,97,145]. Interestingly, crossovers formed by Mus81-Mms4, Slx1-Slx4, or Yen1 are insensitive to interference mechanisms. The exact processes that confer Mlh1-Mlh3-dependent resolution of dHJs into interference-dependent crossovers are unknown. The formation of crossovers via a dHJ in this Mlh1-Mlh3 pathway requires the actions of a nuclease activity that nicks with a bias between the two junctions; one junction of the dHJ is resolved in one orientation, while the other junction is resolved in an orthogonal orientation (Figure 3B). This finding suggests an asymmetric loading of meiotic protein complexes onto the two Holliday junctions (see below).

5.3 Model for resolution of dHJs into crossover products

Is a Msh4-Msh5/Mlh1-Mlh3 supercomplex critical for dHJ resolution? In MMR, interactions between MSH, MLH proteins, and other repair factors are thought to be important for activating and directing endonuclease activity. Therefore it is reasonable to hypothesize that physical interactions between Msh4-Msh5 and other meiotic protein factors orient Mlh1-Mlh3 to permit asymmetric cleavage of dHJs.

Many studies have been performed to elucidate physical interactions that occur among the proteins implicated in the ZMM-directed dHJ resolution pathway. For example, yeast two-hybrid screens and co-immunoprecipitation assays have suggested that human Sgs1 (BLM) and Mlh1 physically interact [146]. These data were supported by immunoprecipitation studies in meiotic yeast cells where Mlh3 was shown to interact with Sgs1 [147]. Mlh1 has also been shown to physically interact with Exo1, an interaction likely also used in MMR, but perhaps exploited in meiosis as a structural bridge [113,148]. Exo1 is known to interact with MSH proteins during MMR, so conceivably interactions with Msh4 or Msh5 occur during meiosis [149,150]. Physical interactions have also been detected between human Msh4 and human Mlh1 and Mlh3 during meiosis [120,151]. Despite the many interactions that occur, a model for how such interactions convey asymmetry is not immediately clear.

Models suggesting that MSH-MLH interactions are important for biased Mlh1-Mlh3 nicking follow logically from known physical interactions involving MSH and MLH factors and the observation that a majority of crossovers occur via the Msh4-Msh5 pathway. Such models, however, do not account for the fact that other organisms (e.g. Drosophila and Schizosaccharomyces pombe) do not have Msh4 and Msh5 homologs, yet achieve efficient crossover resolution using endonucleases belonging to the major superfamilies described in section 5.1 [152–155]. Additionally, there are organisms, such as Caenorhabditis elegans, that have Msh4 and Msh5 orthologs which are required for nearly all meiotic crossovers, despite the fact that roles for Mlh1-Mlh3 orthologs have not been found [156]. In C. elegans crossover resolution occurs with absolute interference (only one crossover per homolog pair) by pathways involving an Slx1-Slx4 ortholog, a Mus81 ortholog, and an additional XPF-1 family endonuclease [157–159]. Redundancy among nucleases and the fact that efficient crossover formation can occur in other organisms in the absence of Msh4-Msh5 or Mlh1-Mlh3 suggests that crossing over can be achieved with different sets and permutations of proteins.

A recent bioinformatic study further suggests that an MSH/MLH complex may not be required for imposing asymmetric nicking of dHJs [124]. Sequencing of recombination products in yeast tetrads derived from msh4 strains showed that biased nicking of dHJs was retained. However, in a zip3 mutant (a SUMO E3 ligase and member of the ZMM family), biased cutting was lost. By monitoring crossover-associated gene conversion tracts, Fung and colleagues [124] suggest that Zip3 is involved in promoting second-end capture followed by ligation to form the dHJ, possibly through intermediate stabilization. This is supported by localization data showing late meiotic roles for Zip3 in budding yeast, where late-forming Zip3 foci localize to a subset of crossover sites [106]. These data suggest a model where the asymmetry for dHJ resolution is established by Zip3 and possibly other proteins involved in second-end capture, branch migration, and ligation to form the dHJ (e.g. RPA, Rad52, and Rad54) (Figure 5) [71–78,124]. In the aforementioned bioinformatic study, the retention of biased resolution in the msh4 mutant suggests that Msh4-Msh5 may act to stabilize strand invasion intermediates and dHJs, and not to asymmetrically direct Mlh1-Mlh3 endonuclease activity [124].

Figure 5. Model for creating an asymmetric intermediate to facilitate biased cleavage.

Model is discussed in section 5.3 and is based in part on work presented by Fishel and colleagues, Fung and colleagues, and Kowalczykowski and colleagues [71,124,160], Msh4-Msh5 (purple) binds to invasion intermediates (1). Encountering open-ended duplex allows ATP to be hydrolyzed and the clamp to dissociate (2), allowing iterative loading of Msh4-Msh5 prior to second end capture (3-4). The second-end capture action (5) and its facilitating proteins (e.g. RPA, Rad52, and Rad54) (grey), likely including Zip3 (green) create an asymmetric substrate for the dHJ resolving proteins (6-7).

Genetic and molecular studies suggest that Msh4-Msh5 and the other ZMM proteins act to stabilize SEIs and dHJ intermediates [105]. Experiments performed with purified human Msh4-Msh5 suggest that the complex specifically recognizes model four-way junctions constructed from oligonucleotides [160]. Recognition of the junction appears to cause an exchange of ADP for ATP, triggering sliding clamp formation. Msh4-Msh5 can then slide onto the homologous arms of a four-way junction, and appears to encircle two duplex arms of a stacked junction (closed, antiparallel conformation) [160,161]. ATP hydrolysis is triggered upon encountering an open DNA end [161]. Prior to second-end capture, Msh4-Msh5 loaded onto invasion intermediates can encounter open-ended substrate and release. This dissociation can cause iterative loading of Msh4-Msh5 onto the Holliday junction on the invading side of the DSB (Figure 5). Zip3 and proteins that facilitate re-annealing at the opposing side of the DSB may force asymmetry by preventing Msh4-Msh5 from binding to the other junction. The result of this process is a symmetric DNA structure that is not equivalent in terms of protein occupancy and by extension loading of Mlh1-Mlh3 resolvase.

6. Conclusions and questions for future studies

As described above, not all organisms have co-opted MMR homologs to form meiotic crossovers. This suggests that there are other strategies available to the cell to achieve crossover formation, and that the mechanisms that ensure efficient crossover formation are distinct from MMR.

Biochemical reconstitution of the Mlh1-Mlh3 crossover pathway has been elusive due in part to the difficulty of purifying the many components involved. Purification strategies for yeast and human Sgs1/BLM, Top3, Rmi1, and Exo1 have been available for some time, but only recently have active yeast and human Mlh1-Mlh3 been purified [45,46]. A procedure to purify yeast Msh4-Msh5 is not yet available, despite the fact that the human complex was purified and characterized some time ago. In addition, the full complement of required proteins is not yet established nor is the exact DNA substrate. Requirements for second-end capture proteins (e.g. RPA, Rad52, and Rad54) [71–78] and additional ZMM proteins are not well established. In addition to an incomplete identification of protein components, the biochemical properties of the known crossover promoting components are not well understood. For instance, Mlh1 and Mlh3 each contain nucleotide-binding domains, but roles for ATP hydrolysis by Mlh1-Mlh3 in meiosis are unclear. Genetic analysis suggested that ATP binding, but not hydrolysis by Mlh1, is critical for wild-type levels of crossing over, but another study suggested that hydrolysis is a requirement for both subunits [114,162]. In vitro experiments also do not shed light on the role of ATP hydrolysis in Mlh1-Mlh3 function. Purified yeast Mlh1-Mlh3 exhibits a weak ATPase activity; however this activity is not enhanced in the presence of DNA substrate. Furthermore, the complex's in vitro endonuclease activity was not activated by the presence of ATP or other nucleotides [45,46]. The possibility that Mlh1-Mlh3's ATPase may need activation by external factors has not been exhaustively studied. In addition, an as yet unidentified co-factor or post-translational modification may clarify the in vivo role of Mlh1-Mlh3 ATPase activity.

Currently, Mlh1-Mlh3's in vitro endonuclease functions have only been observed on large circular DNA substrates (Figure 4). Mlh1-Mlh3 binding to Holliday junction substrates has been observed in vitro, but with preference for an open conformation over stacked junctions, suggesting that other protein components may be required in vivo to create DNA substrate in a preferred conformation for Mlh1-Mlh3 binding to occur [46]. The presence of protein aggregates in gel shift experiments with Mlh1-Mlh3 along with the evidence that endonuclease activity has only been observed on large DNA substrates suggest that MLH proteins bind DNA cooperatively [45,46]. This is consistent with filter binding experiments involving various sized DNA substrates and purified yeast Mlh1-Pms1 where cooperative binding was observed [163]. Biochemical analysis of yeast Mlh1-Mlh3, Mlh1-Pms1/2 from yeast and human, and MutH from E. coli has shown that sliding clamps, such as MSH proteins and processivity factors, can enhance the nicking activity of these endonucleases, suggesting that interactions or handoff mechanisms may direct nicking [43–45,164].

Specific questions that may elucidate how Mlh1-Mlh3 resolves dHJs into crossover are: (1) What is the role of ATP hydrolysis for Mlh1-Mlh3? (2) Do other factors interact with Mlh1-Mlh3? (3) Are protein-protein interactions between Mlh1-Mlh3 and the other subunits implicated in the pathway essential? (4) Are all of the proteins involved in the pathway identified? (5) Is the role of Exo1 purely structural? (6) There is evidence that post-translational modifications activate resolvases—what if any role does post-translational modification play in activating the pathway? (7) Is Mlh1-Mlh3 capable of making symmetric cuts across a Holliday junction? (8) If Mlh1-Mlh3 does nick a Holliday junction, what activates this activity and where does the protein nick? (9) What roles do chromatin modifications play in this process?

Further bioinformatic approaches and physical assays need to be performed with additional mutant strains to screen for other possible players in the Mlh1-Mlh3 pathway, with particular focus on proteins catalyzing the second-end capture mechanism. Additional biochemical studies further exploring Mlh1-Mlh3's in vitro endonuclease activity and searching for factors that may enhance or specify its functions will also be valuable.

Highlights.

• Many organisms have co-opted MMR homologs for meiotic crossing over.

• Mlh1-Mlh3 is an endonuclease that acts in both MMR and meiosis.

• Mlh1-Mlh3 endonuclease activity is required for meiotic crossing over.

• Msh4-Msh5, Mlh1-Mlh3 act in meiotic recombination in a pathway distinct from MMR.

• We outline models for an MLH-mediated Holliday junction resolution pathway in meiosis.