MLH1 Mutations Differentially Affect Meiotic Functions in Saccharomyces cerevisiae

Corresponding author: Rhona H. Borts, Leicester University, University Rd., Leicester LE1 7RH, United Kingdom., rhb7{at}le.ac.uk (E-mail)

Communicating editor: M. LICHTEN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

To test whether missense mutations in the cancer susceptibility gene MLH1 adversely affect meiosis, we examined 14 yeast MLH1 mutations for effects on meiotic DNA transactions and gamete viability in the yeast Saccharomyces cerevisiae. Mutations analogous to those associated with hereditary nonpolyposis colorectal cancer (HNPCC) or those that reduce Mlh1p interactions with ATP or DNA all impair replicative mismatch repair as measured by increased mutation rates. However, their effects on meiotic heteroduplex repair, crossing over, chromosome segregation, and gametogenesis vary from complete loss of meiotic functions to no meiotic defect, and mutants defective in one meiotic process are not necessarily defective in others. DNA binding and ATP binding but not ATP hydrolysis are required for meiotic crossing over. The results reveal clear separation of different Mlh1p functions in mitosis and meiosis, and they suggest that some, but not all, MLH1 mutations may be a source of human infertility.

THE mismatch repair system plays a number of roles in maintaining genome stability. During mitosis it primarily ensures avoidance of mutations and inappropriate recombination events (reviewed in HARFE and JINKS-ROBERTSON 2000 ) while during meiosis it is involved in heteroduplex repair, crossing over, chromosome segregation, and avoidance of inappropriate recombination (reviewed in BORTS et al. 2000 ). Mismatch repair proteins function as dimers. MutS and MutL in bacteria form homodimers while their eukaryotic homologs form heterodimers. There are six MutS homologs, MSH1–6, and four MutL homologs, MLH1–3 and PMS1 (PMS2 in humans). Mutation avoidance is accomplished by mispair recognition by Msh2p/Msh6p (MutS) or Msh2p/Msh3p (MutSß) and transduction of a signal by a heterodimer of Mlh1p/Pms1p (MutL) or Mlh1p/Mlh3p (reviewed in HARFE and JINKS-ROBERTSON 2000 ) to effector molecules. The exonuclease encoded by EXO1 has been implicated in mismatch repair. However, the mutation rate is increased only moderately by deletion of the gene, indicating that other proteins are involved in mismatch removal (TISHKOFF et al. 1997 ; SOKOLSKY and ALANI 2000 ; AMIN et al. 2001 ; TRAN et al. 2001 ). In higher organisms mutation accumulation due to deficiency in mismatch repair is associated with carcinogenesis. Specifically, defects in hMLH1 and hMSH2 are found in sporadic tumors and a familial cancer syndrome, hereditary nonpolyposis colorectal cancer (HNPCC; reviewed in PELTOMAKI 2001 ). Germline mutations in hEXO1 have also been reported to be associated with HNPCC (WU et al. 2001 ). In addition to the role of mismatch repair genes in mutation avoidance MutS and MutL are responsible for the majority of repair of mismatches in heteroduplex DNA formed during meiotic recombination (WILLIAMSON et al. 1985 ; REENAN and KOLODNER 1992 ; ALANI et al. 1994 ; PROLLA et al. 1994 ; HUNTER and BORTS 1997 ; WANG et al. 1999 ). Exo1p plays little or no role in the removal of this type of mismatch (KHAZANEHDARI and BORTS 2000 ; KIRKPATRICK et al. 2000 ).

The importance of the role(s) that mismatch repair proteins play in meiosis is illustrated by the infertility found in model organisms deficient in some mismatch repair genes (reviewed in BORTS et al. 2000 ; COHEN and POLLARD 2001 ). In yeast, loss of Mlh1p, Mlh3p, Exo1p, and the meiosis-specific Msh4p and Msh5p causes defects in reciprocal recombination and chromosome segregation (ROSS-MACDONALD and ROEDER 1994 ; HOLLINGSWORTH et al. 1995 ; HUNTER and BORTS 1997 ; WANG et al. 1999 ; BORTS et al. 2000 ; KHAZANEHDARI and BORTS 2000 ; KIRKPATRICK et al. 2000 ; NOVAK et al. 2001 ; ABDULLAH 2002 ; ARGUESO et al. 2002 ). Although the phenotypes of the individual mutants are not identical, in none of the cases studied does the double mutant display a more extreme crossover defect than that of the most severe of the single mutants, msh4, suggesting that they all operate in the same crossover pathway (HOLLINGSWORTH et al. 1995 ; HUNTER and BORTS 1997 ; BORTS et al. 2000 ; KHAZANEHDARI and BORTS 2000 ; ABDULLAH 2002 ). Mice that are mutant in MLH1, MLH3, MSH4, and MSH5 have chromosome segregation or chromosome pairing abnormalities and are both male and female sterile (BAKER et al. 1996 ; EDELMANN et al. 1996 , EDELMANN et al. 1999 ; KNEITZ et al. 2000 ; COHEN and POLLARD 2001 ; LIPKIN et al. 2002 ). Cytological studies have indicated that the timing, number, and distribution of MLH1 foci in both humans and mice correlate well with that of late recombination nodules and of chiasmata, the cytological manifestations of crossing over (BARLOW and HULTEN 1998 ; ANDERSON et al. 1999 ). The Mlh3^-/- mouse has been shown to be deficient in late recombination nodules and fails to form MLH1 foci, suggesting that Mlh3p may recruit Mlh1p (LIPKIN et al. 2002 ). Neither the Mlh1^-/- nor the Mlh3^-/- mouse has functional chiasmata at diplonema (BAKER et al. 1996 ; LIPKIN et al. 2002 ). Cytological studies have also indicated that MSH4 foci appear first and are then followed by MLH1 foci (SANTUCCI-DARMANIN et al. 2000 ). Physical studies have suggested that mammalian MSH4 protein interacts with both the MLH1 and the MLH3 proteins (SANTUCCI-DARMANIN et al. 2000 , SANTUCCI-DARMANIN et al. 2002 ). The cytological data combined with the genetic data from yeast suggest a late role for Mlh1p/Mlh3p in ensuring crossover outcome that is separable from that of the Msh4p/Msh5p complex.

How the Mlh1p/Mlh3p heterodimer exerts its function(s) is not clear. However, by analogy with Escherichia coli MutL, it is thought to act by coordinating downstream "effector" molecules such as helicases (HALL et al. 1998 ) and nucleases (BAN and YANG 1998B ; SPAMPINATO and MODRICH 2000 ). Among the possible effector proteins known to interact with Mlh1p are ReqQ helicases (yeast Sgs1p and human BLM protein; LANGLAND et al. 2001 ; PEDRAZZI et al. 2001 ) and Exo1p (TRAN et al. 2001 ). Interestingly neither Sgs1p nor the Bloom's syndrome protein has been implicated in mismatch repair, suggesting a role other than mismatch correction for the interaction of these proteins with Mlh1p. That this role is in the resolution of recombination structures has been suggested by the isolation of a complex containing Top3p, Sgs1p, Mlh1p, and Mlh3p from extracts of meiotic cells (WANG and KUNG 2002 ).

To better understand the role of MLH1 in meiosis we have assessed meiotic phenotypes conferred by a number of missense mutations that all result in defective mismatch repair (PANG et al. 1997 ; SHCHERBAKOVA and KUNKEL 1999 ; HALL et al. 2002 ; M. HALL, P. SHCHERBAKOVA and T. KUNKEL, unpublished data). Many of the known mutations map to the highly conserved amino-terminal domain of Mlh1p (Fig 1), which has been shown to have ATPase and DNA-binding activities that are essential for repair of replication errors (TRAN and LISKAY 2000 ; HALL et al. 2002 ; M. HALL, P. SCHERBAKOVA and T. KUNKEL, unpublished data). Seven mutations (yP25L/hP28L, yM32R/hM35R, yA41F/hS44F, yG64R/hG67R, yI65N/hI68N, yT114M/hT117M, and yG243D/hG244D) are analogues of human HNPCC mutations. Six of these (yP25L/hP28L, yM32R/hM35R, yG64R/hG67R, yI65N/hI68N, yA41F/hS44F, and yT114M/hT117M) are inferred to reduce the ATPase activity of Mlh1p (BAN and YANG 1998A ; BAN et al. 1999 ). Four changes (F96A, R97A, G98A, and G98V) reside in the highly conserved "GFRGEAL" box that composes the "lid" of the ATP-binding pocket (BAN and YANG 1998A ; BAN et al. 1999 ; GUARNE et al. 2001 ) and are also inferred to interfere with ATP binding or hydrolysis. Each has individually been shown to confer reduced mismatch repair (PANG et al. 1997 ). Replacement of Asn35 with alanine (N35A) results in an N-terminal domain with no ATP-binding or hydrolysis capacity, and replacement of Glu31 with alanine (E31A) results in an N-terminal domain that binds ATP but very inefficiently hydrolyzes it (HALL et al. 2002 ) and is partially repair defective (TRAN and LISKAY 2000 ; HALL et al. 2002 ). A double replacement, R273E-R274E, reduces DNA binding by the Mlh1p/Pms1p heterodimer and also confers a mismatch repair defect (M. HALL, P. SHCHERBAKOVA, J. FORTUNE and T. KUNKEL, unpublished data). The final substitution studied, G243D, maps to the interface of two domains identified in the crystal structure (BAN and YANG 1998A ). The observation that the bacterial protein with this substitution is insoluble suggests that this amino acid change causes the protein to misfold (BAN et al. 1999 ). Fourteen strains, each bearing one of these mutations, were analyzed by tetrad dissection for their effects on meiotic heteroduplex repair, crossing over, chromosome segregation, and gamete viability.

View larger version (37K): In this window In a new window Download PPT slide   Figure 1. (A) Alignment of the N termini of E. coli MutL (Z11831), S. cerevisiae, and hMLH1. Blue dots represent the HNPCC mutations, green bars highlight the ATPase domain (motifs I–IV), and magenta and orange dots identify the functionally defined mutations and GFRGEAL box mutations, respectively. (B) Crystal structure of MutL, with first the human mutations and then the equivalent yeast residue indicated. The -carbon of the residue is represented by a black ball. Green indicates the ATP-binding site, ATP is shown in red, and the gray ball is Mg2+.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Plasmids, strains, and sporulation:
MLH1 point mutations were constructed using site-directed mutagenesis (ERDENIZ et al. 1997 ; SHCHERBAKOVA and KUNKEL 1999 ) and were then introduced into Saccharomyces cerevisiae Y55 haploid strains with the following genotypes: Y55-2834 (MAT HIS4 LEU2 ADE1 trp5-1 cyh2 met13-2 lys2-c ura3-1) and Y55-2835 (MATa his4-r leu2-r ade1-1 TRP5 CYH2 MET13 lys2::InsE-A₁₄ ura3-1). The presence of the mutations was confirmed by DNA sequencing. his4-r is a 4-bp insertion mutation (BORTS and HABER 1989 ). met13-2 has a stop codon at position 278 (C A; ABDULLAH 2002 ). The lys2::InsE-A₁₄ allele contains a homopolymeric A insertion in LYS2 (TRAN et al. 1997 ; SHCHERBAKOVA and KUNKEL 1999 ). MLH1 deleted strains (mlh1) were generated using a PCR-based gene disruption method (WACH et al. 1994 ). The diploid strains used are listed in Table 1. Mating, sporulation, and tetrad dissection have been described previously (HUNTER and BORTS 1997 ; ABDULLAH and BORTS 2001 ).

Genetic analysis and statistical methods:
Genetic markers were analyzed by direct replication of dissected spore colonies to omission media as described previously (HUNTER and BORTS 1997 ; ABDULLAH and BORTS 2001 ). Non-Mendelian segregation (NMS; 6:2/2:6 conversions and 5:3/3:5 postmeiotic segregation) and reciprocal crossing over were scored only in tetrads containing four viable spores. Map distance in centimorgans was calculated according to the formula cM = 1/2 (TT + 6NPD)/(NPD + PD + TT) (PERKINS 1949 ), where PD, NPD, and TT refer to parental ditype, nonparental ditype, and tetratype segregation patterns. Statistical comparisons were carried out as follows. All of the data were compared to the wild-type and mlh1 strains. The distribution of tetrad classes with respect to the crossover and viability data were compared using a G-test of heterogeneity (SOKAL and ROHLF 1969 ). To compare NMS and the proportion of meiotic repair events, we employed Fisher's exact test, using the one-tailed distribution (http://faculty.vassar.edu/lowry/VassarStats.html). For comparisons of data sets containing >100 tetrads for which the Fisher's exact test cannot be used, we employed a two-sample z-test (http://faculty.vassar.edu/lowry/VassarStats.html). In all of the statistical comparisons, we used the Dunn-Sidak correction (SOKAL and ROHLF 1969 ) for significance testing, which is required when multiple comparisons using the same data sets are made. For example, < 0.05 is normally set as the basis for rejection of the null hypothesis when a single pairwise comparison is made. However, statistical theory necessitates that be adjusted to reflect multiple comparisons. Thus when a missense mutation was compared to both the wild-type and the mlh1 strains (e.g., crossover data and meiotic repair data) P < 0.025 was considered significant. P values <0.017 were considered significant when a given data set was compared to those of the wild-type, mlh1, and msh2 strains. The NPD ratio was calculated using the equation of PAPAZIAN 1952 , where an NPD ratio significantly lower than one indicates interference. The method of Stahl and Lande (http://www.groik.com/stahl/) was also used calculate "m" where a value of m significantly greater than zero is indicative of interference (STAHL and LANDE 1995 ).

Physical analysis of disomy:
Tetrads with two or three viable spores were analyzed for chromosomal aneuploidy using clamped homogeneous electric field (CHEF) gel analysis (KHAZANEHDARI and BORTS 2000 ). Rates of disomy were calculated by dividing the observed number of two-viable-spore asci containing disomes by the number of tetrads that it took to obtain the number of two-viable-spored tetrads that were analyzed. Of the 16 yeast chromosomes, only 10 can be assayed by intensity of the chromosome band. Thus the value obtained is an underestimate of the frequency of aneuploidy.

Alignment and protein modeling:
The E. coli MutL, S. cerevisiae MLH1, and human MLH1 were aligned using MegAlign (DNA Star) by the Jotun Hein method. Molecular representation of the MutL crystal structure (accession no. 1B63.pdb in the Brookhaven protein database) was made using Swiss-PdbViewer.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

MLH1 is dominant and haplosufficient:
All of the HNPCC and the mlh1p-N35A, E31A, and R273E-R274E mutations were studied as homozygotes (e.g., mlh1-E31A/ mlh1-E31A). However, the GFRGEAL box mutations were studied in heterozygous diploid strains (e.g., mlh1-F96A/mlh1). To confirm that this would not interfere with comparisons between strains we analyzed MLH1/mlh1. The MLH1/mlh1 strain was indistinguishable from wild type with respect to all meiotic phenotypes (Table 2, Table 3, Table 4, and Table 5), indicating that a single wild-type gene is sufficient to ensure normal levels of crossing over, gene conversion, nondisjunction, and chromosome segregation. We also analyzed mlh1-N35A/MLH1 because mlh1-N35A has been suggested to be dominant negative with respect to mitotic mismatch repair (HALL et al. 2002 ). This does not appear to be the case for meiotic functions as the heterozygous diploid is indistinguishable from both of the wild-type diploids analyzed (Table 2 Table 3 Table 4 Table 5).

Crossing over is affected only in a subset of mutants:
The MLH1 missense mutations fell into two groups when meiotic crossing over in four genetic intervals was determined (Table 2 and Table 6). Strains bearing group I mutations (mlh1-P25L, mlh1-E31A, mlh1-I65N, mlh1-T114M, mlh1-F96A, mlh1-R97A, and mlh1-G98A) had normal levels of crossing over and had crossover frequencies significantly greater than those of mlh1 (P < 0.05, G-test of homogeneity). In contrast, the group II strains (mlh1-M32R, mlh1-N35A, mlh1-A41F, mlh1-G64R, mlh1-G98V, mlh1-G243D, and mlh1-R273E-R274E) exhibited reduced crossing over in all four intervals relative to the wild type (P < 0.05). Crossing over was reduced to a level that was indistinguishable from that observed in the mlh1 strain. The observation that the mlh1-R273E-R274E protein, which displays reduced binding of DNA, is deficient for crossing over suggests that DNA binding may be important for crossing over during meiosis. Group II also includes mlh1p-N35A, whose N-terminal domain does not bind ATP, suggesting that ATP binding may also be important for meiotic crossing over. In contrast, ATP hydrolysis may be less critical, since the mlh1-E31A mutant strain has normal crossing over yet it encodes an N-terminal domain that binds but does not efficiently hydrolyze ATP.

Meiotic mismatch repair efficiencies:
The effect of each mutation on mismatch repair efficiency during meiosis was determined by assessing the frequency of postmeiotic segregation events (phenotypic sectoring of the genetic marker) that result from failure to repair heteroduplex DNA (WILLIAMSON et al. 1985 ). Repair of a 4-bp insertion at HIS4 and a mispair at MET13 were measured. The missense mutations yielded three general phenotypes with respect to efficiency of repair of meiotic heteroduplex (Table 3 and Table 6). Nine mutations (mlh1-M32R, mlh1-N35A, mlh1-G64R, mlh1-F96A, mlh1-G98A, mlh1-G98V, mlh1-T114M, mlh1-G243D, and mlh1-R273E-R274E) resulted in loss of all Mlh1p-dependent meiotic heteroduplex repair (Fisher's exact test, P < 0.025 with respect to MLH1 and P > 0.025 with respect to mlh1). The mlh1-A41F strain was clearly defective for meiotic mismatch repair of the his4-r allele whereas the data for the met13-2 allele were ambiguous. Strains with the mlh1-E31A, mlh1-P25L, and mlh1-R97A mutations displayed wild-type or near wild-type levels of repair at both loci tested (P > 0.025). Consistent with this, these three mutations have the lowest published mitotic mutation rates of the mutations analyzed (PANG et al. 1997 ; SHCHERBAKOVA and KUNKEL 1999 ; HALL et al. 2002 ). In contrast, the mlh1-I65N strain displayed allele-specific levels of repair. The mlh1-I65N strain had wild-type levels of repair at his4-r but was significantly different from both wild type (P < 0.025) and mlh1 (P < 0.025) for repair at met13-2. The effect of the missense mutations on total frequency of non-Mendelian segregation varied with no obvious pattern (Table 3).

The crossover defect does not predict the degree of aneuploidy:
The crossover defect of mlh1 has previously been shown to be associated with a moderate amount of nondisjunction (HUNTER and BORTS 1997 ). To determine what the contribution of nondisjunction was to meiotic inviability in the strains with missense mutations we measured disomy rates by CHEF gel analysis (Table 4 and Table 6). Because the sample sizes for the missense mutation strains are individually too small to allow statistical analysis, we pooled the data from all of the mutant strains exhibiting crossover frequencies indistinguishable from those of the mlh1 strain. These strains have a disomy rate of 1.1% (10/908). This is significantly lower (P < 0.05, z-test) than that found in mlh1 (5/131, 3.8%). In contrast, the crossover-proficient strains were indistinguishable from the wild-type strain (0/1066 vs. 1/970).

Nondisjunction contributes to gamete death:
Gamete death in mlh1 strains is due to at least two factors whose relative contributions are unknown, aneuploidy and the accumulation of haplolethal mutations (including synthetic lethal mutations) that are uncovered by meiosis (HUNTER and BORTS 1997 ). To assess the relative contributions of the mitotic mutator phenotype and nondisjunction to gamete viability, we compared the spore viability of strains with the crossover-defective missense mutations to that of msh2 and mlh1 (Table 5 and Table 6). Since the msh2 and mlh1 strains have equivalent mitotic mutation rates, but msh2 strains have no crossover or segregation defects (HUNTER and BORTS 1997 ), gamete viability in the msh2 strain provides an estimate of the contribution of mitotically acquired haplolethals and meiotic repair deficiency to gamete death. Consistent with the previous report, the msh2 mutant strain had viability intermediate between wild-type and mlh1 strains. As might be predicted, the three mutant strains, mlh1-R97A, -E31A, and -P25L, reported to have moderate mutation rates (PANG et al. 1997 ; SHCHERBAKOVA and KUNKEL 1999 ; HALL et al. 2002 ) and without a crossover defect had wild-type or intermediate levels of spore viability. In addition, all of the missense mutant strains with repair defects had significantly poorer viability than that of the wild type. Furthermore, the mismatch repair-defective, crossover-proficient missense mutations had the same pattern of spore viability as msh2. However, the crossover-deficient strains fell into two classes. They were either msh2-like (mlh1-M32R, mlh1-N35A, and mlh1-R273E-R274E) or intermediate between mlh1 and msh2. None were mlh1-like except perhaps mlh1-G98V, which could not be distinguished from either msh2 (P = 0.09) or mlh1 (P = 0.06). This indicates that apparently equivalent crossover and repair defects do not translate directly into an equivalent defect in viability and suggests that Mlh1p may be playing a role in meiotic viability separable from its role in crossing over.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Meiotic mismatch repair generally reflects mitotic repair efficiency:
The efficiency of repair of meiotic heteroduplex DNA by the strains with missense substitutions is for the most part consistent with the published mutation rates. However, mlh1-I65N displays wild-type repair at one of the alleles studied despite high mitotic mutation rates. This allele-specific effect could reflect different functional requirements for repair of a single mispair compared to a four-base insertion or could reflect the position of the marker relative to the double-strand break, which initiates meiotic recombination. Current models for the repair of mismatches in meiotic heteroduplex envisage distinctly different fates for alleles close to the double-strand break and those far away (reviewed in BORTS et al. 2000 ). Thus mutations in MLH1 might differentially affect the processing of mismatched heteroduplex in a context-specific manner, i.e., if it is coupled to strand invasion vs. being directed by Holliday junction resolution (ALANI et al. 1994 ; GILBERTSON and STAHL 1996 ). The observation that total levels of non-Mendelian segregation vary with no apparent pattern may be an indication of the complexity of the interrelationship between the repair of meiotic heteroduplex and crossing over. Another possibility is that these differential repair defects reflect different levels of the various mutant proteins that are partially or even fully active but are limiting in different contexts. This explanation has been proposed to account for the phenotype of the temperature-sensitive MLH1 mutants found by ARGUESO et al. 2002 . In a systematic site-directed mutagenesis of MLH1, a mutation (mlh1-2) that is partially defective for both meiotic repair and meiotic crossing over and abolishes the gene conversion gradient at ARG4 was identified (ARGUESO et al. 2003 ). Such a mutant might reflect an absence of crossover resolution-directed repair.

Structure function relationships revealed by the missense mutations:
The results with strains bearing mutant proteins that have known biochemical defects (mlh1p-E31A, mlh1p-N35A, and mlh1p-R273E-R274E) begin to offer some insights into the importance of ATP binding, ATP hydrolysis, and DNA binding by Mlh1p for different meiotic functions. The observations that the N-terminal domains of the mlh1p-N35A and mlh1p-R273E-R274E substituted proteins have reduced binding of ATP and DNA, respectively, and the corresponding mutants are defective for crossing over and heteroduplex repair suggests that both substrate-binding properties of Mlh1p are important for these meiotic functions. From a comparison of the meiotic phenotypes of mlh1-E31A and mlh1-N35A and the observation that mlh1p-E31A is capable of binding but not hydrolyzing ATP while mlh1p-N35A does neither, we conclude that ATP binding is sufficient for executing the crossover functions of Mlh1p. This conclusion is supported by data from a similar study where it was shown that a mutation of E31 to lysine is recombination defective (ARGUESO et al. 2003 ). In E. coli, a change in a nearby conserved glutamic acid (E32, E34 in yeast) to lysine reduces ATP binding and the interaction of MutL with MutH (SPAMPINATO and MODRICH 2000 ), suggesting that a lysine substitution at E31 also abolishes ATP binding. However, the relationship between ATP interactions and crossing over is complex. This is indicated by the fact that both groups I and II contain substitutions for highly conserved residues that, on the basis of the crystal structures of E. coli MutL (BAN and YANG 1998A ; BAN et al. 1999 ) and human PMS2 (GUARNE et al. 2001 ), should alter ATP interactions. Perhaps this distinction in phenotype can be used to infer how some of the amino acid substitutions influence protein function. For example, G64R and I65N substitutions both result in completely defective mismatch repair in mitotic cells and are predicted to interfere with ATP binding and/or hydrolysis. However, only G64R affects crossing over, suggesting that perhaps only the G64R substitution interferes with ATP binding whereas the I65N substitution affects only hydrolysis. These predictions can be supported only by biochemical studies.

Other structural or functional inferences can be drawn from the phenotypic data. A comparison of our observation that the mlh1-R273E-R274E strain is crossover deficient with the observation that when the adjacent arginines are replaced with alanines the resultant strain is crossover proficient (ARGUESO et al. 2003 ) leads us to predict that the alanine substitutions do not impair DNA binding. By analogy with the E. coli data on MutL-G238D, which indicate that the protein is insoluble (BAN et al. 1999 ), one might predict that mlh1-G243D would be phenotypically identical to mlh1. This is not the case, as it falls into the class of mutants that have better viability and better disjunction than the deletion. As discussed below we interpret the improved disjunction and viability with respect to the deletion to mean that Mlh1p has a structural role in segregation. This inferred structural role seems to be fulfilled by the mutant protein encoded by mlh1-G243D.

The strains bearing mutations in the GFRGEAL box also fall into both groups. Two different substitutions for the same amino acid (e.g., G98A vs. G98V) result in proteins with differential effects on crossing over vs. meiotic and mitotic mismatch repair. Gly98 is in the GFRGEAL box that not only contacts the nucleotide but also is implicated in dimerization of the N-terminal domain upon ATP binding (BAN and YANG 1998A ; BAN et al. 1999 ; TRAN and LISKAY 2000 ). The valine substitution alters the interaction of Mlh1p with Pms1p (TRAN and LISKAY 2000 ) while the alanine substitution does not. Thus the role Gly98 plays in crossing over can be accomplished when it is replaced by alanine but not when it is replaced by valine, suggesting that the lid interaction with the nucleotide may not be as important for meiotic recombination as it is for mitotic mismatch repair. It has been proposed previously (BAN et al. 1999 ; TRAN and LISKAY 2000 ; HALL et al. 2002 ) that ATP binding induces the conformational changes leading to changes in partner binding while the hydrolysis restores the previous conformation. In this context, we suggest that ATP binding is sufficient to ensure that the downstream effector molecules for crossing over are capable of interacting functionally. If, as suggested, the dimerization of Mlh1p with Mlh3p is similar to its dimerization with Pms1p, then the crossover defect in mlh1-G98V strains may be attributable to an effect on dimerization with Mlh3p. Due to the difficulty demonstrating the known interaction between Mlh1p and Mlh3p with wild-type proteins (ARGUESO et al. 2002 ) we have been unable to test this hypothesis.

As discussed above, the conformational change associated with ATP binding is also thought to signal the effector molecules (BAN et al. 1999 ). Among the proteins known to interact with Mlh1p and possible effectors of its meiotic functions are Mlh3p (WANG et al. 1999 ; BORTS et al. 2000 ), Msh4p (SANTUCCI-DARMANIN et al. 2000 ), Exo1p (AMIN et al. 2001 ; TRAN et al. 2001 ), and Sgs1p (LANGLAND et al. 2001 ; PEDRAZZI et al. 2001 ; WANG and KUNG 2002 ). Three of the severely crossover-defective mutations are known to be (A41F and G98V; PANG et al. 1997 ) or presumed to be (N35A) defective in their N-terminal interaction with Pms1p. If Mlh3p interacts with Mlh1p in a manner similar to that of Pms1p, as suggested by studies of the human proteins (KONDO et al. 2001 ) and MutL (BAN and YANG 1998A , BAN and YANG 1998B ), these mutations can be predicted to interfere with the Mlh1p-Mlh3p interaction and this may account for their crossover defect. The role of the interactions between Exo1p and Mlh1p in crossing over is unclear. We have shown previously that exo1 has a defect similar to that of mlh1 in crossing over and segregation but has no defect in repair of mismatched heteroduplex, although total non-Mendelian segregations are reduced at some loci (KHAZANEHDARI and BORTS 2000 ). The single amino acid change T117M in human MLH1 is reported to disrupt the interaction with hEXO1 (JAGER et al. 2001 ). However, strains with the corresponding T114M mutation in yeast Mlh1p do not display a defect in crossing over as might have been expected if an interaction between Mlh1p and Exo1p were functionally important for crossing over. Perhaps Mlh1p and Exo1p do not interact via this residue in yeast to exert their crossover function or their respective roles in crossing over do not require them to interact. Alternatively, they may be involved in different types of crossovers, as has been suggested (KHAZANEHDARI and BORTS 2000 ). It has recently been hypothesized that the role of the Mlh1p/Mlh3p heterodimer is to recruit Sgs1p/Top3p to the sites of late recombination intermediates to aid in their resolution as crossovers (WANG and KUNG 2002 ). It will be interesting to determine if any of the crossover-defective mutants interfere with a meiotic Sgs1p/Mlh1p interaction.

A structural role for Mlh1p in segregation?
Some of the missense mutations are as defective as the deletion strain for both mismatch repair and crossing over, yet have significantly better viability and less nondisjunction than the deletion strain. There are a number of possible explanations for the poor correlation between crossover defectiveness, nondisjunction, and viability. One possibility is that the intervals studied are not an accurate reflection of the crossing over in the genome as a whole. Possibly, the deletion of MLH1 is affecting another interval to a greater extent than the missense mutations and that crossing over in this interval is more relevant to segregation. Given recent suggestions that there are at least two types of crossovers in yeast, this is not an unreasonable hypothesis (ROSS-MACDONALD and ROEDER 1994 ; ZALEVSKY et al. 1999 ; KHAZANEHDARI and BORTS 2000 ; ABDULLAH 2002 ). However, one class of these crossovers, those known to be dependent on Msh4p, display a nonrandom distribution of exchanges indicative of a phenomenon termed interference (ROSS-MACDONALD and ROEDER 1994 ; NOVAK et al. 2001 ). If Mlh1p acted in the same complex as Msh4p, then one would predict that its absence should lead to loss of interference. This is not the case as indicated by strong interference detected (NPD ratio of 0.38, P < 0.05, 1 < m < 2 in the TRP5-CYH2 interval) in the mlh1 strain and in a previous study (ARGUESO et al. 2002 ). These data further support separable roles for the MutS and MutL homologs during meiosis. Another possibility for the poor correlation between nondisjunction and viability is that the greater nondisjunction defect in the mlh1 strain as compared to some of the missense mutations is caused by the loss of the protein that impairs formation of a complex important for chromosome segregation but not exchange at the DNA levels. One possibility is that it is a component of the proteinaceous structure associated with chiasmata such as a "chiasma binder" suggested by CARPENTER 1994 .

Implications for human fertility:
Our results indicate that Mlh1p has at least three meiotic functions, heteroduplex repair, crossing over, and chromosome segregation, that are separable from each other and from mismatch repair of replication errors in mitotic cells. To date, no infertility has been linked to MLH1 HNPCC patients, perhaps due to the absence of homozygous individuals or the rarity of loss of heterozygosity in the germline. The results presented here suggest that MLH1-dependent aneuploidy leading to reduced fertility would be specific to certain mutations. Hence not all HNPCC carriers would be at equal risk for fertility problems, which may be a reason why it has not been previously noted. Perhaps even polymorphisms in the general population may result in reduced fertility due to impaired crossing over. For example, three single-nucleotide polymorphisms have recently been shown to reduce the interaction between hMLH1 and hPMS2 (YUAN et al. 2002 ) in vitro. Such "polymorphisms" may confer defects in crossing over and may be a possible source of infertility.

	ACKNOWLEDGMENTS

We thank M. Liskay for the GFRGEAL mutations, A. Aziz for preparing media, and E. Alani for sharing unpublished data. We thank E. Louis, C. Griffin, and the anonymous reviewers for helpful comments on the manuscript. We thank V. Cotton, B. Herbert, and R. Watson for technical assistance. This work was supported by the Wellcome Trust; E.R.H. was supported by a Prize Studentship from the Wellcome Trust.

Manuscript received September 24, 2002; Accepted for publication November 8, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ABDULLAH, M. F. F., 2002 Transcription factors and mismatch repair proteins in meiotic recombination. Ph.D. Thesis, University of Oxford, Oxford.

ABDULLAH, M. F. F. and R. H. BORTS, 2001  Meiotic recombination frequencies are affected by nutritional states in Saccharomyces cerevisiae.. Proc. Natl. Acad. Sci. USA 98:14524-14529.[Abstract/Free Full Text]

ALANI, A., R. A. REENAN, and R. D. KOLODNER, 1994  Interaction between mismatch repair and genetic recombination in Saccharomyces cerevisiae.. Genetics 137:19-39.[Abstract]

AMIN, N. S., M. N. NGUYEN, S. OH, and R. D. KOLODNER, 2001  exo1-Dependent mutator mutations: model system for studying functional interactions in mismatch repair. Mol. Cell. Biol. 21:5142-5155.[Abstract/Free Full Text]

ANDERSON, L. K., A. REEVES, L. M. WEBB, and T. ASHLEY, 1999  Distribution of crossing over on mouse synaptonemal complexes using immunofluorescent localization of MLH1 protein. Genetics 151:1569-1579.[Abstract/Free Full Text]

ARGUESO, J. L., S. SARIN, A. KIJAS, and E. ALANI, 2002  Analysis of conditional mutations in the Saccharomyces cerevisiae MLH1 gene in mismatch repair and crossing over. Genetics 160:909-921.[Abstract/Free Full Text]

ARGUESO, J. L., A. W. KIJAS, S. SARIN, J. HECK, and M. WAASE et al., 2003  Systematic mutagenesis of the Saccharomyces cerevisiae MLH1 gene reveals distinct roles for Mlh1p in meiotic crossing over and in vegetative and meiotic mismatch repair. Mol. Cell. Biol. 23:873-886.[Abstract/Free Full Text]

BAKER, S. M., A. W. PLUG, T. A. PROLLA, C. E. BRONNER, and A. C. HARRIS et al., 1996  Involvement of mouse Mlh1 in DNA mismatch repair and meiotic crossing over. Nat. Genet. 13:336-342.[Medline]

BAN, C. and W. YANG, 1998a  Crystal structure and ATPase activity of MutL: implications for DNA repair and mutagenesis. Cell 95:541-552.[Medline]

BAN, C. and W. YANG, 1998b  Structural basis for MutH activation in E. coli mismatch repair and relationship of MutH to restriction endonucleases. EMBO J. 17:1526-1534.[Medline]

BAN, C., M. JUNOP, and W. YANG, 1999  Transformation of MutL by ATP binding and hydrolysis: a switch in DNA mismatch repair. Cell 97:85-97.[Medline]

BARLOW, A. L. and M. A. HULTEN, 1998  Crossing over analysis at pachytene in man. Eur. J. Hum. Genet. 6:350-358.[Medline]

BORTS, R. H. and J. E. HABER, 1989  Length and distribution of meiotic gene conversion tracts and crossovers in Saccharomyces cerevisiae.. Genetics 123:69-80.[Abstract/Free Full Text]

BORTS, R. H., S. R. CHAMBERS, and M. F. F. ABDULLAH, 2000  The many faces of mismatch repair in meiosis. Mutat. Res. 451:129-150.[Medline]

CARPENTER, A. T. C., 1994  Chiasma function. Cell 77:959-962.

COHEN, P. E. and J. W. POLLARD, 2001  Regulation of meiotic recombination and prophase I progression in mammals. Bioessays 23:996-1009.[Medline]

EDELMANN, W., P. E. COHEN, M. KANE, K. LAU, and B. MORROW et al., 1996  Meiotic pachytene arrest in MLH1-deficient mice. Cell 85:1125-1134.[Medline]

EDELMANN, W., P. E. COHEN, B. KNEITZ, N. WINAND, and M. LIA et al., 1999  Mammalian MutS homologue 5 is required for chromosome pairing in meiosis. Nat. Genet. 21:123-127.[Medline]

ERDENIZ, N., U. H. MORTENSEN, and R. ROTHSTEIN, 1997  Cloning-free PCR-based allele replacement methods. Genome Res. 7:1174-1183.[Abstract/Free Full Text]

GILBERTSON, L. A. and F. W. STAHL, 1996  A test of the double-strand break model for meiotic recombination in Saccharomyces cerevisiae.. Genetics 144:27-41.[Abstract]

GUARNE, A., M. S. JUNOP, and W. YANG, 2001  Structure and function of the N-terminal 40 kDa fragment of human PMS2: a monomeric GHL ATPase. EMBO J. 20:5521-5531.[Medline]

HALL, M. C., J. R. JORDAN, and S. W. MATSON, 1998  Evidence for a physical interaction between the Escherichia coli methyl-directed mismatch repair proteins MutL and UvrD. EMBO J. 17:1535-1541.[Medline]

HALL, M. C., P. V. SHCHERBAKOVA, and T. A. KUNKEL, 2002  Differential ATP binding and intrinsic ATP hydrolysis by amino terminal domains of the yeast Mlh1 and Pms1 proteins. J. Biol. Chem. 277:3673-3679.[Abstract/Free Full Text]

HARFE, B. D. and S. JINKS-ROBERTSON, 2000  DNA mismatch repair and genetic instability. Annu. Rev. Genet. 34:359-399.[Medline]

HOLLINGSWORTH, N. M., L. PONTE, and C. HALSEY, 1995  MSH5, a novel MutS homolog, facilitates meiotic reciprocal recombination between homologs in Saccharomyces cerevisiae but not mismatch repair. Genes Dev. 9:1728-1739.[Abstract/Free Full Text]

HUNTER, N. and R. H. BORTS, 1997  Mlh1p is unique among mismatch repair proteins in its ability to promote crossing-over during meiosis. Genes Dev. 11:1573-1582.[Abstract/Free Full Text]

JAGER, A. C., M. RASMUSSEN, H. C. BISGAARD, K. K. SINGH, and F. C. NIELSEN et al., 2001  HNPCC mutations in the human DNA mismatch repair gene hMLH1 influence assembly of hMutLalpha and hMLH1-hEXO1 complexes. Oncogene 20:3590-3595.[Medline]

KHAZANEHDARI, K. and R. H. BORTS, 2000  EXO1 and MSH4 differentially affect crossing-over and segregation. Chromosoma 109:94-102.[Medline]

KIRKPATRICK, D. T., R. J. FERGUSON, T. D. PETES, and L. S. SYMINGTON, 2000  Decreased meiotic intergenic recombination and increased meiosis I nondisjunction in exo1 mutants of Saccharomyces cerevisiae.. Genetics 156:1549-1557.[Abstract/Free Full Text]

KNEITZ, B., P. E. COHEN, E. AVDIEVICH, L. ZHU, and M. F. KANE et al., 2000  MutS homolog 4 localization to meiotic chromosomes is required for chromosome pairing during meiosis in male and female mice. Genes Dev. 14:1085-1097.[Abstract/Free Full Text]

KONDO, E., A. HORII, and S. FUKUSHIGE, 2001  The interacting domains of three MutL heterodimers in man: hMLH1 interacts with 36 homologous amino acid residues within hMLH3, hPMS1 and hPMS2. Nucleic Acids Res. 29:1695-1702.[Abstract/Free Full Text]

LANGLAND, G., J. KORDICH, J. CREANEY, K. H. GOSS, and K. LILLARD-WETHERELL et al., 2001  The Bloom's syndrome protein (BLM) interacts with MLH1 but is not required for DNA mismatch repair. J. Biol. Chem. 276:30031-30035.[Abstract/Free Full Text]

LIPKIN, S. M., P. B. MOENS, V. WANG, M. LENZI, and D. SHANMUGARAJAH et al., 2002  Meiotic arrest and aneuploidy in MLH3-deficient mice. Nat. Genet. 31:385-390.[Medline]

NOVAK, J. E., P. B. ROSS-MACDONALD, and G. S. ROEDER, 2001  The budding yeast Msh4 protein functions in chromosome synapsis and the regulation of crossover distribution. Genetics 158:1013-1025.[Abstract/Free Full Text]

PANG, Q. S., T. A. PROLLA, and R. M. LISKAY, 1997  Functional domains of the Saccharomyces cerevisiae Mlh1p and Pms1p DNA mismatch repair proteins and their relevance to human hereditary nonpolyposis colorectal cancer-associated mutations. Mol. Cell. Biol. 17:4465-4473.[Abstract]

PAPAZIAN, H. P., 1952  The analysis of tetrad data. Genetics 37:175-188.[Free Full Text]

PEDRAZZI, G., C. PERRERA, H. BLASER, P. KUSTER, and G. MARRA et al., 2001  Direct association of Bloom's syndrome gene product with the human mismatch repair protein MLH1. Nucleic Acids Res. 29:4378-4386.[Abstract/Free Full Text]

PELTOMAKI, P., 2001  DNA mismatch repair and cancer. Mutat. Res. 488:77-85.[Medline]

PERKINS, D. D., 1949  Biochemical mutants in the smut fungus Ustilago maydis.. Genetics 34:607-626.[Free Full Text]

PROLLA, T. A., D.-M. CHRISTIE, and R. M. LISKAY, 1994  Dual requirement in yeast DNA mismatch repair for MLH1 and PMS1, two homologs of the bacterial mutL gene. Mol. Cell. Biol. 14:407-415.[Abstract/Free Full Text]

REENAN, R. A. and R. D. KOLODNER, 1992  Characterization of insertion mutations in the Saccharomyces cerevisiae MSH1 and MSH2 genes: evidence for separate mitochondrial and nuclear functions. Genetics 132:975-985.[Abstract]

ROSS-MACDONALD, P. and G. S. ROEDER, 1994  Mutation of a meiosis-specific MutS homolog decreases crossing-over but not mismatch correction. Cell 79:1069-1080.[Medline]

SANTUCCI-DARMANIN, S., D. WALPITA, F. LESPINASSE, C. DESNUELLE, and T. ASHLEY et al., 2000  MSH4 acts in conjunction with MLH1 during mammalian meiosis. FASEB J. 14:1539-1547.[Abstract/Free Full Text]

SANTUCCI-DARMANIN, S., S. NEYTON, F. LESPINASSE, A. SAUNIERES, and P. GAUDRAY et al., 2002  The DNA mismatch-repair MLH3 protein interacts with MSH4 in meiotic cells, supporting a role for this MutL homolog in mammalian meiotic recombination. Hum. Mol. Genet. 11:1697-1706.[Abstract/Free Full Text]

SHCHERBAKOVA, P. V. and T. A. KUNKEL, 1999  Mutator phenotypes conferred by MLH1 overexpression and by heterozygosity for mlh1 mutations. Mol. Cell. Biol. 19:3177-3183.[Abstract/Free Full Text]

SOKAL, R. R., and F. J. ROHLF, 1969 Biometry. W. H. Freeman, San Francisco.

SOKOLSKY, T. and E. ALANI, 2000  EXO1 and MSH6 are high-copy suppressors of conditional mutations in the MSH2 mismatch repair gene of Saccharomyces cerevisiae.. Genetics 155:589-599.[Abstract/Free Full Text]

SPAMPINATO, C. and P. MODRICH, 2000  The MutL ATPase is required for mismatch repair. J. Biol. Chem. 275:9863-9869.[Abstract/Free Full Text]

STAHL, F. W. and R. LANDE, 1995  Estimating interference and linkage map distance from two-factor tetrad data. Genetics 139:1449-1454.[Abstract]

TISHKOFF, D. X., A. L. BOERGER, P. BERTRAND, N. FILOSI, and G. M. GAIDA et al., 1997  Identification and characterization of Saccharomyces cerevisiae EXO1 a gene encoding an exonuclease that interacts with MSH2.. Proc. Natl. Acad. Sci. USA 94:7487-7492.[Abstract/Free Full Text]

TRAN, H. T., J. D. KEEN, M. KRICKER, M. A. RESNICK, and D. A. GORDENIN, 1997  Hypermutability of homonucleotide runs in mismatch repair and DNA polymerase proofreading yeast mutants. Mol. Cell. Biol. 17:2859-2865.[Abstract]

TRAN, P. T. and R. M. LISKAY, 2000  Functional studies on the candidate ATPase domains of Saccharomyces cerevisiae MutL. Mol. Cell. Biol. 20:6390-6398.[Abstract/Free Full Text]

TRAN, P. T., J. A. SIMON, and R. M. LISKAY, 2001  Interactions of Exo1p with components of MutL in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 98:9760-9765.[Abstract/Free Full Text]

WACH, A., A. BRACHAT, R. POHLMANN, and P. PHILIPPSEN, 1994  New heterologous modules for classical or PCR-based gene disruptions in Saccharomyces cerevisiae.. Yeast 10:1793-1808.[Medline]

WANG, T. and W. KUNG, 2002  Supercomplex formation between Mlh1-Mlh3 and Sgs1-Top3 heterocomplexes in meiotic yeast cells. Biochem. Biophys. Res. Commun. 296:949.[Medline]

WANG, T.-F., N. KLECKNER, and N. HUNTER, 1999  Functional specificity of MutL homologs in yeast: evidence for three Mlh1-based heterocomplexes with distinct roles during meiosis in recombination and mismatch correction. Proc. Natl. Acad. Sci. USA 96:13914-13919.[Abstract/Free Full Text]

WILLIAMSON, M. S., J. C. GAME, and S. FOGEL, 1985  Meiotic gene conversion mutants in Saccharomyces cerevisiae. I. Isolation and characterization of pms1-1 and pms1-2. Genetics 110:609-646.[Abstract/Free Full Text]

WU, Y., M. J. BERENDS, J. G. POST, R. G. MENSINK, and E. VERLIND et al., 2001  Germline mutations of EXO1 gene in patients with hereditary nonpolyposis colorectal cancer (HNPCC) and atypical HNPCC forms. Gastroenterology 120:1580-1587.[Medline]

YUAN, Z. Q., B. GOTTLIEB, L. K. BEITEL, N. WONG, and P. H. GORDON et al., 2002  Polymorphisms and HNPCC: PMS2-MLH1 protein interactions diminished by single nucleotide polymorphisms. Hum. Mutat. 19:108-113.[Medline]

ZALEVSKY, J., A. J. MACQUEEN, J. B. DUFFY, K. J. KEMPHUES, and A. M. VILLENEUVE, 1999  Crossing over during Caenorhabditis elegans meiosis requires a conserved MutS-based pathway that is partially dispensable in budding yeast. Genetics 153:1271-1283.[Abstract/Free Full Text]

This article has been cited by other articles:

				A. Svetlanov, F. Baudat, P. E. Cohen, and B. de Massy Distinct Functions of MLH3 at Recombination Hot Spots in the Mouse Genetics, April 1, 2008; 178(4): 1937 - 1945. [Abstract] [Full Text] [PDF]

				E. Avdievich, C. Reiss, S. J. Scherer, Y. Zhang, S. M. Maier, B. Jin, H. Hou Jr, A. Rosenwald, H. Riedmiller, R. Kucherlapati, et al. Distinct effects of the recurrent Mlh1G67R mutation on MMR functions, cancer, and meiosis PNAS, March 18, 2008; 105(11): 4247 - 4252. [Abstract] [Full Text] [PDF]

				A. E. Gammie, N. Erdeniz, J. Beaver, B. Devlin, A. Nanji, and M. D. Rose Functional Characterization of Pathogenic Human MSH2 Missense Mutations in Saccharomyces cerevisiae Genetics, October 1, 2007; 177(2): 707 - 721. [Abstract] [Full Text] [PDF]