Mismatch-repair (MMR) systems promote eukaryotic genome stability by removing errors introduced during DNA replication and by inhibiting recombination between nonidentical sequences (spellchecker and antirecombination activities, respectively). Following a common mismatch-recognition step effected by MutS-homologous Msh proteins, homologs of the bacterial MutL ATPase (predominantly the Mlh1p-Pms1p heterodimer in yeast) couple mismatch recognition to the appropriate downstream processing steps. To examine whether the processing steps in the spellchecker and antirecombination pathways might differ, we mutagenized the yeast PMS1 gene and screened for mitotic separation-of-function alleles. Two alleles affecting only the antirecombination function of Pms1p were identified, one of which changed an amino acid within the highly conserved ATPase domain. To more specifically address the role of ATP binding/hydrolysis in MMR-related processes, we examined mutations known to compromise the ATPase activity of Pms1p or Mlh1p with respect to the mitotic spellchecker and antirecombination activities and with respect to the repair of mismatches present in meiotic recombination intermediates. The results of these analyses confirm a differential requirement for the Pms1p ATPase activity in replication vs. recombination processes, while demonstrating that the Mlh1p ATPase activity is important for all examined MMR-related functions.
MISMATCH-repair (MMR) systems promote genome stability by detecting and dealing with distortions in the DNA double helix (reviewed in Harfe and Jinks-Robertson 2000a). These systems are best known for their role in removing mispaired or extrahelical nucleotides generated during DNA replication (“spellchecker” function), with defects resulting in a strong mutator phenotype. In addition to their replication-editing function, MMR systems also detect mismatches in the heteroduplex recombination intermediates that involve the pairing of single strands derived from different duplex DNA molecules. Detection of recombination-associated mismatches triggers either a repair process that restores perfect base complementarity or an antirecombination activity that prevents the recombination event from going to completion. Finally, MMR systems in some organisms are important for detecting DNA damage and for triggering appropriate cell-cycle arrest or apoptotic responses.
The MMR system of Escherichia coli contains three dedicated “Mut” proteins and has served as a paradigm for the more complicated MMR systems of eukaryotic organisms (reviewed in Modrich and Lahue 1996). MMR in E. coli is initiated when a homodimer of the MutS protein binds mismatches. A MutL homodimer then couples the MutS-dependent mismatch recognition to downstream processing steps by activating the latent endonuclease activity of the MutH protein. MutH specifically nicks the nascent strand to initiate its removal by a helicase and one or more exonucleases, and the resulting gap is filled in by DNA polymerase and sealed by ligase to complete the repair process. In eukaryotes there are multiple MutS and MutL homologs (Msh and Mlh proteins, respectively) that are involved in MMR processes, but no known MutH homologs (reviewed in Harfe and Jinks-Robertson 2000a). The active forms of the eukaryotic Msh and Mlh proteins are heterodimers instead of homodimers, with the heterodimers generally having distinct but overlapping functions in MMR. The recently solved crystal structures of bacterial MutS homodimers have revealed that they are, in fact, structural heterodimers (Lamerset al. 2000; Obmolovaet al. 2000), which can account for the existence of heterodimers rather than homodimers in eukaryotes. In the yeast Saccharomyces cerevisiae, mismatches in nuclear DNA are recognized by either an Msh2p-Msh3p or an Msh2p-Msh6p heterodimer (Johnsonet al. 1996b; Marsischkyet al. 1996), which then interacts with a MutL-like heterodimer composed of Mlh1p complexed with Pms1p, Mlh2p, or Mlh3p (Wanget al. 1999). As the Mlh1p-Mlh2p and Mlh1p-Mlh3p heterodimers play only minor roles in the repair of replication errors (Flores-Rozas and Kolodner 1998; Harfeet al. 2000) and have no reported antirecombination activity, only the Mlh1p-Pms1p heterodimer will be considered here.
Functionally important regions of the yeast Mlh1 and Pms1 proteins have been deduced by aligning MutL homologs from diverse organisms (Ban and Yang 1998; Crouse 1998) and by mutational analyses (Panget al. 1997; Tran and Liskay 2000). Protein alignments have revealed a highly conserved region at the amino terminus that contains the four domains characteristic of the GHL (gyrase b, Hsp90, and MutL) family of ATPases (Dutta and Inouye 2000). Functionally important conformational changes in the N-terminal regions of GHL family proteins are associated with ATP binding and hydrolysis, with the N-terminal ends of MutL, and with gyrase b homodimerizing upon ATP binding (Wigleyet al. 1991; Aliet al. 1993; Prodromou et al. 1997a,b; Ban and Yang 1998; Grenertet al. 1999). Studies with mutant Mlh1 and Pms1 proteins support a comparable amino-terminal heterodimerization cycle associated with ATP binding and hydrolysis (Tran and Liskay 2000). In addition, genetic studies have revealed an ATP-related asymmetry between the yeast Mlh1p and Pms1p subunits in terms of their contributions to the spellchecker function of the complex (Tran and Liskay 2000). Although Mlh1p and Pms1p share little amino acid similarity outside of the highly conserved N terminus, the C-terminal 200-300 amino acids of each protein are necessary and sufficient for ATP-independent heterodimer formation and are required for the spellchecker function of the complex (Panget al. 1997). Finally, the C-terminal 13 amino acids of yeast Mlh1p are identical to the C-terminal 13 amino acids of human MLH1, but this highly conserved motif is not present in Pms1p. This carboxy-terminal homology (CTH) motif of Mlh1p is not required for interaction with Pms1p in two-hybrid assays, but is required for spellchecker function (Panget al. 1997).
The repair of mismatches in heteroduplex recombination intermediates can result in the replacement of one allele with the sequence of another allele (“gene conversion”), which is manifested in meiosis as the non-Mendelian segregation of allelic sequences. If mismatches in meiotic recombination intermediates are not repaired, segregation of the corresponding alleles at the next round of DNA replication will result in genetically different daughter cells (postmeiotic segregation, or PMS). In yeast, gene conversion is much more common than PMS, indicating that most mismatches are efficiently recognized and repaired by the MMR machinery (Peteset al. 1991). Although it is not known what triggers the repair vs. antirecombination activity of MMR systems, the antirecombination activity effectively limits recombination between nonidentical (“homeologous”) sequences, thereby reducing genome rearrangements and enforcing species barriers (reviewed in Harfe and Jinks-Robertson 2000a,b). In S. cerevisiae, mitotic recombination is exquisitely sensitive to potential mismatches, with a single nonidentity between 350-bp recombination substrates being sufficient to reduce the rate of recombination in a MMR-dependent manner (Dattaet al. 1997). Although elimination of yeast Msh2p, Mlh1p, or Pms1p results in identical mutator phenotypes, the antirecombination activity of Msh2p is consistently greater than that of Pms1p or Mlh1p (Chen and Jinks-Robertson 1999; Nicholsonet al. 2000). In addition, the Sgs1p helicase appears to be redundant with MMR-associated antirecombination (Myunget al. 2001), but has no known role in the repair of DNA mismatches.
The genetic differences between the MMR-associated spellchecker and antirecombination activities in yeast suggest that the Mlh1p-Pms1p-dependent steps downstream of mismatch recognition may be different during DNA replication vs. recombination. In addition, the ATPase-related functional asymmetry observed in the spellchecker functions of Mlh1p and Pms1p may extend to the recombination-related activities of the proteins as well. These issues are addressed in the current study by (1) identifying “separation-of-function” alleles of PMS1 that partially uncouple the mitotic spellchecker and antirecombination functions, (2) examining the mitotic antirecombination effects of known mutations in MLH1 or PMS1 that compromise ATP binding or hydrolysis, and (3) examining the effects of eliminating Mlh1p or Pms1p ATP hydrolysis activity on the repair of mismatches in meiotic recombination intermediates.
MATERIALS AND METHODS
Media and growth conditions: Strains were grown vegetatively at 30° and sporulated at 18°; a complete list of yeast strains is given in Table 1. Standard media and genetic techniques were used for mitotic growth, sporulation, and tetrad dissection (Sherman 1991), except as noted below. Strains were grown nonselectively in YEP medium containing either 2% glycerol and 2% ethanol (YEPGE) or 2% dextrose (YEPD). Each liter of YEPGE and YEPD was supplemented with 500 mg adenine hemi-sulfate (Sigma, St. Louis) to avoid adenine limitation during nonselective growth. For selection of yeast transformants that had incorporated the kan marker, each liter of YEPD was supplemented with 200 mg of Geneticin (Sigma).
Synthetic dextrose (SD) medium was supplemented with all but the one amino acid or base needed for selective growth (e.g., SD-His is deficient in histidine). Additional tryptophan (30 μg/ml) was added to the SD media as well as to the YEP media for growth of strains containing the trp5Δ allele (i.e., SJR1392 and its derivatives). Canavanine-resistant (Can-R) mutants were selected on SD-Arg medium supplemented with l-canavanine sulfate to a concentration of 60 μg/ml (SD-Arg + Can). Ura- segregants were selected on SD plates supplemented with required amino acids and containing 0.1% 5-fluoroorotic acid (5-FOA; Boekeet al. 1984). For the selection of His+ mitotic recombinants, dextrose in the SD medium was replaced with 2% glycerol, 2% ethanol, and 2% galactose (SGGE-His).
Sporulation of diploid cells and tetrad dissection were performed as described by Fan et al. (1995). For meiotic analyses, purified diploids were not used, since diploids homozygous for pms1 or mlh1 rapidly accumulate recessive lethal mutations. Instead, haploid parents were mated overnight on YEPD plates and then were immediately transferred to sporulation plates. Tetrads were dissected on YEPD medium and the resulting spore clones were directly replica plated to appropriate selective media. Sectored His+/His- colonies were scored by light microscopy.
Yeast strains used for mitotic studies: Strain SJR1294 was used as a host to identify plasmid-encoded pms1 alleles conferring mutator and/or hyperrecombination phenotypes. The mutator phenotype was assessed by forward mutation to canavanine resistance, while the recombination phenotype was assessed using 94%-identical HIS3::intron::cβ2 inverted-repeat (IR) substrates (see Dattaet al. 1997). In preliminary experiments, we found that the high chromosomal mutation rate of a pms1Δ host strain interfered with the efficient detection of plasmid-encoded pms1 alleles. To circumvent this problem, the endogenous PMS1 promoter was replaced with the GAL1 promoter by transforming cells with a PCR fragment generated using plasmid pFA6a-kanMX6-PGAL1 (Longtimeet al. 1998) as a template. The presence of the resulting galactose-regulated PMS1 allele (pGAL-PMS1) resulted in strong mutator and hyper-recombination phenotypes only when cells were grown in the absence of galactose.
Strain SJR1392 contains both homeologous (92% identical) and homologous (100% identical) IR recombination substrates. This strain was constructed by targeting plasmids containing homeologous HIS3::intron::cβ2/cβ7 substrates (pSR303) and homologous LYS2 substrates (pRS304) to the URA3 and LEU2 loci, respectively. Transformants containing a single copy of each plasmid were identified by Southern analysis. Ura- segregants were selected on 5-FOA medium, and retention of the homeologous recombination substrates was confirmed by the ability to produce His+ recombinants.
An mlh1Δ::URA3 allele was introduced into SJR1392 by transformation with SacI/BamHI-digested ymlh1::URA3 (Prollaet al. 1994). BstXI-digested pJH523 (Krameret al. 1989) was used in a two-step allele replacement procedure to introduce a pms1Δ allele. All mlh1Δ and pms1Δ strains were verified by PCR or Southern analysis. Derivatives containing point mutations in PMS1 or MLH1 were constructed by two-step allele replacement, and the presence of the mutation of interest was confirmed by genomic DNA sequencing. Plasmids pYI-mlh1-31, pYI-mlh1-98, pYI-pms1-61 TV II, and pYI-pms1-128 TV II were used to introduce the mlh1-E31A, mlh1-G98A, pms1-E61A, and pms1-G128A alleles, respectively (for details see Tran and Liskay 2000). The pms1-L124S, pms1-I854M, and mlh1-757stop alleles were introduced using plasmids pSR759, pSR760, and pSR746, respectively (see below).
Strains used for meiotic recombination studies: Diploid strains used for meiotic recombination experiments were constructed by mating isogenic derivatives of the HIS4 strain AS4 (Stapleton and Petes 1991) and the his4-AAG strain PD73 (Detloffet al. 1991). All diploids thus are heterozygous for the his4-AAG mutant allele, which has a single-base-pair change at the second position of the HIS4 start codon. Haploid derivatives containing an mlh1Δ::URA, pms1Δ, mlh1-E31A, or pms1-E61A allele were constructed by transformation as described above for SJR1392. The mlh1Δ::kanMX4 and pms1Δ::kanMX4 alleles were introduced by transformation with PCR deletion cassettes generated using pFA6-kanMX4 (Wachet al. 1994) as a template. To determine forward mutation rates at the CAN1 locus, ARG4 derivatives of haploid strains with the AS4 genetic background were constructed by transformation with AgeI-digested pMW52 (Whiteet al. 1993).
Plasmids: Plasmid pSR303 contains the HIS3::intron::cβ2/cβ7 homeologous recombination substrates and was constructed by combining 5′ cβ2 and 3′ cβ7 recombination cassettes as inverted repeats (Figure 1A). Plasmid pSR266 contains a full-length HIS3::intron gene, with a unique BamHI site within the intron, and was used to generate both the 5′ and 3′ cassettes (see Dattaet al. 1996). Plasmids pSR273 and pSR301 were constructed by inserting an 800-bp BamHI/BglII cβ2 and a 783-bp BamHI/BglII cβ7 fragment, respectively, into the BamHI site of plasmid pSR266. The cβ7 3′ cassette plasmid pSR302 was derived from pSR301 by deleting the SalI fragment upstream of the cβ7 sequences (i.e., the 5′ portion of HIS3 and the 5′ part of the intron). A SmaI fragment containing the 5′ cβ2 cassette (from pSR273) was then inserted into the filled-in SpeI site of pSR302 in reverse orientation relative to the 3′ cβ7 cassette.
Plasmid pSR304 contains the lys2Δ5′ and lys2Δ3′ homologous recombination substrates oriented as IRs (Figure 1B) and was constructed using LYS2 sequences derived from pDP6 (Fleiget al. 1986). First, a 2.7-kb HincII/HindIII fragment containing the 3′ end of LYS2 (lys2Δ5′ allele) was directionally cloned into SmaI/HindIII-digested pRS305 (Sikorski and Hieter 1989), yielding plasmid pSR300. A 3-kb XbaI/StuI fragment containing the 5′ end of LYS2 (lys2Δ3′ allele) was then inserted into XbaI/SstI(blunt)-digested pSR300, with the resulting plasmid (pSR304) containing the lys2Δ5′ allele downstream of and in inverted orientation relative to the lys2Δ3′ allele. The region of overlap between the lys2Δ5′ and lys2Δ3′ alleles is ∼900 bp.
Plasmid pSR758 contains the 2715-bp PMS1 open reading frame and was constructed by cloning a 4-kb chromosomal BglII/SalI fragment (from YIp5-PMS1; obtained from D. Maloney) into BamHI/SalI-digested pRS315 (LEU2-CEN vector; Sikorski and Hieter 1989). pSR758 was the source for the PMS1 fragments that comprise the three deletion plasmids used in gap-repair experiments. pRS315-PMS1Δ1 (pSR764) has a deletion of the first 590 bp of the PMS1 coding sequence between the MluI and HinP1I sites at -33 and +591, respectively, relative to the start codon; pRS315-PMS1Δ2 (pSR765) has a centrally located 1090-bp deletion extending from the Eco0109I site at +387 to the FokI site at +1477; and pRS315-PMS1Δ3 (pSR766) has a 927-bp deletion encompassing the C-terminal region of PMS1, extending from the BspHI site at +1713 to the NcoI site at +2640. Each of the three deletion plasmids contains a unique BamHI site between the PMS1 fragments that flank the deleted segment.
pSR761 contains the MLH1 locus and was derived by inserting a 7-kb chromosomal SacI fragment (from YEp24-MLH1; Prollaet al. 1994) into pRS306 (Sikorski and Hieter 1989). The mlh1-757stop allele (pSR746) was constructed by replacing codon 757 of the MLH1 coding sequence with a TAG stop codon using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The resulting mutation was confirmed by DNA sequencing.
Random mutagenesis of the PMS1 coding sequence and incorporation of mutations by gap repair: Appropriate PMS1 fragments were generated by mutagenic PCR and were then recombined in vivo into plasmids pRS315-PMS1Δ1, pRS315-PMS1Δ2, or pRS315-PMS1Δ3 using a standard yeast gap-repair procedure (Muhlradet al. 1992). Taq DNA polymerase was used for the PCR mutagenesis; the error frequency of the enzyme was increased by doubling the concentration of dATP or dGTP relative to the other dNTP’s and by increasing the MgCl2 concentration to 3 mm. Plasmids for the in vivo gap repair were prepared by digestion with BamHI, followed by treatment with shrimp alkaline phosphatase. Gap repair of pRS315-PMS1Δ1 was accomplished using a 687-bp PCR fragment extending from -61 to +626 of the PMS1 sequence; gap repair of plasmid pRS315-PMS1Δ2 was effected using an 1167-bp PCR fragment extending from +349 to +1516 of the PMS1 sequence; and gap repair of pRS315-PMS1Δ3 was done using a 1029-bp PCR fragment extending from +1668 to +2697 of the PMS1 sequence.
Strain SJR1294 was cotransformed with 1 μg of purified PCR fragment and 0.1 μg of gapped vector, and transformants were selected on SD-Leu medium. Control experiments with gapped vector only indicated a gap-repair efficiency of >95%. Approximately 1000 transformants derived from each of the three gap-repair reactions were selectively purified. Transformants were patched onto SD-Arg + Can or SGGE-His medium to score mutation or homeologous recombination frequency, respectively. Approximately 30% of the transformants exhibited phenotypes characteristic of a pms1Δ strain and were assumed to contain plasmid-encoded null alleles. Plasmid DNA was isolated from those transformants that consistently exhibited a separation-of-function phenotype (either a mutator or a hyper-rec phenotype, but not both phenotypes) and was used to retransform SJR1294. Following the confirmation of a separation-of-function phenotype, the relevant portion of the mutagenized pms1 allele was sequenced. The separation-of-function alleles pms1-L124S and pms1-I854M were identified in this manner. For integration into the yeast genome, the pms1-L124S and pms1-I854M alleles were transferred to the integrating vector pRS306 (Sikorski and Hieter 1989) as SacII/KpnI fragments, yielding plasmids pSR759 and pSR760, respectively.
Two-hybrid assays: Plasmids used in two-hybrid assays were constructed by inserting the coding sequences of wild-type Pms1p and Mlh1p into vectors pGAD424 (Bartelet al. 1993) and pBTM116 (Vojteket al. 1993), respectively. pGAD424 and pBTM116 contain the Gal4p activation and the LexA DNA-binding domains, respectively. The pGAD-PMS1 construct was mutagenized to yield pGAD-pms1-I854M using the QuikChange site-directed mutagenesis kit (Stratagene). Interactions of proteins were assessed by cotransforming pGAD and pBT derivatives into yeast strain L40, which contains both lacZ and HIS3 reporter constructs (Vojteket al. 1993). β-Galactosidase activity was measured in liquid assays as described previously (Panget al. 1997).
Rate measurements and statistical analyses: The method of the median (Lea and Coulson 1949) was used to calculate mutation and recombination rates. Data from at least 16 independent cultures (typically 8 cultures from each of two independent isolates) were used for each rate determination. For the experimentally derived medians, 95% confidence intervals (CIs) were determined (Dixon and Massey 1969) and these were then used to calculate 95% CIs for the corresponding rates. For rate determinations using SJR1392 and its derivatives, individual colonies were inoculated into 5 ml YEPGE medium and grown for 3 days on a roller drum. Cells were harvested by centrifugation, washed with H2O, and resuspended in 1 ml H2O. Aliquots of appropriate dilutions were plated on SD-Arg + Can to select Can-R mutants, on SD-His to select His+ (homeologous) recombinants, on SD-Lys to select Lys+ (homologous) recombinants, and on YEPD to determine the number of viable cells. Plates were incubated for 2 days (YEPD, SD-Arg + Can, SD-Lys) or 4 days (SGGE-His) before counting colonies. Mutation rates to Can-R in AS4 and PD73 and in their derivatives were similarly determined, except that cultures were grown overnight in YEPD before selective plating, total viable cells were determined by plating on SD-Arg, and colonies arising on SD-Arg + Can were counted after 3 days. Comparisons of the distributions of meiotic spore classes derived from different diploids were made by Fisher’s exact test with two-tailed P values. Results were considered significant if P < 0.05.
Mutagenesis of PMS1 and identification of mitotic separation-of-function mutations: A PMS1 gene contained on a CEN vector was randomly mutagenized and the resulting alleles were screened for a mitotic separation-of-function phenotype in a strain devoid of the wild-type Pms1 protein (see ‘materials and methods for details of the mutagenesis). Specifically, transformants containing the mutagenized plasmids were screened for an associated increase in either the spontaneous mutation or the homeologous recombination frequency, but not both. The mutator phenotype was assessed by replica plating transformants to canavanine medium, which selectively identifies forward mutations at the CAN1 locus (Can-R mutants). The level of homeologous recombination was assessed by replica plating transformants to histidine-deficient medium, which selects for inversion events that reconstitute a full-length HIS3::intron gene (Figure 1A). All candidate separation-of-function plasmids identified in the screen conferred little or no mutator phenotype, but resulted in a clearly elevated level of homeologous recombination. No candidates with the opposite phenotype were identified.
To confirm that the elevated recombination conferred by the putative pms1 separation-of-function alleles was specific for homeologous substrates, the plasmid-encoded alleles were introduced into the PMS1 locus of strain SJR1392, which contains identical (“homologous”) LYS2 recombination substrates as well as the homeologous HIS3::intron substrates. The LYS2-based homologous system (Figure 1B) is composed of inverted repeats and thus is comparable in structure to the homeologous HIS3::intron system. As with the HIS3::intron system, replication between the lys2 inverted repeats reorients the region between them, resulting in a full-length LYS2 gene whose presence can be identified on lysine-deficient medium.
As shown in Table 2, elimination of Pms1p in the SJR1392 strain background resulted in a 60-fold increase in the rate of Can-R mutants. When normalized to the homologous recombination rate, the increase in the homeologous recombination rate was 11-fold in the pms1Δ mutant relative to the PMS1 strain; similarly, normalized rates are used when describing homeologous recombination in pms1 (or mlh1) missense mutants. Quantitation of recombination and mutation rates in the pms1 mutants identified in the screen confirmed only two separation-of-function pms1 alleles. Each allele resulted in a significant (3- to 4-fold) increase in the homeologous recombination rate, but no significant increase in the forward mutation rate at the CAN1 locus (Table 2). DNA sequence analysis of the separation-of-function alleles revealed a mutation resulting in a leucine-to-serine change at amino acid 124 in one mutant (pms1-L124S allele) and a mutation causing an isoleucine-to-methionine change at amino acid 854 in the other mutant (pms1-I854M allele).
Pms1p-I854M interacts normally with Mlh1p in two-hybrid assays: The pms1-I854M allele alters a single amino acid in the C-terminal region of Pms1p, a region that is essential for interaction with Mlh1p in two-hybrid assays (Figure 2; Panget al. 1997). One possible explanation for the separation-of-function phenotype conferred by the pms1-I854M allele is that more of the Mlh1p-Pms1p complex is needed to regulate mitotic recombination than is needed to correct DNA replication errors. A decrease in the amount/stability of the complex would thus be expected to elevate homeologous recombination rates to a greater extent than mutation rates. To determine whether the I854M change significantly affects the level of the Pms1p-Mlh1p complex in vivo, we used two-hybrid assays to compare the interaction of Mlh1p with the wild-type vs. the I854M mutant Pms1 protein. As shown in Figure 3, the interactions were indistinguishable in a qualitative phenotypic assay as well as in a quantitative β-galactosidase assay, suggesting that the I854M change affects neither the stability of Pms1p nor its interaction with Mlh1p.
Role of Pms1p ATP binding/hydrolysis in mitotic MMR functions: The L124S change is immediately adjacent to conserved motif III of the GHL family of ATPases, which is important in ATP binding and/or associated conformational changes (Banet al. 1999; see Figure 2). The identification of the pms1-L124S allele in the separation-of-function screen suggested that ATP binding/hydrolysis by Pms1p might be more important for its antirecombination activity than for its spellchecker function. To pursue this further, we introduced the pms1-G128A and pms1-E61A alleles, which are predicted to compromise ATP binding (and/or associated conformational changes) and hydrolysis, respectively, into the SJR1392 strain background. These alleles were previously reported to have little, if any, effect on the spellchecker functions of the corresponding proteins in the CAN1 mutation assay (Tran and Liskay 2000). The results obtained with the pms1-E61A and pms1-G128A alleles were indistinguishable from those obtained with the psm1-L124S allele. As shown in Table 2, there was no significant increase in the rate of Can-R colonies, but there was a significant (three- to fourfold) increase in the rate of homeologous recombination in the pms1-E61A and pms1-G128A strains.
The separation-of-function phenotype conferred by mutations in the N-terminal ATP binding/hydrolysis domains of Pms1p was very similar to that associated with the C-terminal pms1-I854M allele. To determine whether the C- and N-terminal mutations affect Pms1p in fundamentally different ways, we constructed strains containing the double-mutant pms1-E61A,I854M or pms1-G128A,I854M allele. Both double-mutant strains exhibited significantly higher mutation and homeologous recombination rates than those observed with the corresponding single-mutant strains (Table 2), suggesting that the individual mutations have functionally distinct consequences. The mutator phenotype of the double mutants was very weak, however, with the mutation rates being 10-fold lower than that of an isogenic pms1Δ strain. In contrast, the ratio of homeologous to homologous recombination in the double mutants was similar to that in a pms1Δ strain. The double-mutant proteins thus retain most of their spellchecker activity, but appear to be completely defective for the mitotic antirecombination activity.
Role of Mlh1p ATP binding/hydrolysis in mitotic MMR functions: A functional asymmetry in the ATPase activities of Pms1p and Mlh1p has been demonstrated previously, with disruption of Mlh1p ATP binding/hydrolysis resulting in stronger mutator phenotypes than those resulting from comparable changes in Pms1p (Tran and Liskay 2000). We therefore examined the effects of the mlh1-E31A and mlh1-G98A alleles (direct counterparts of the pms1-E61A and pms1-G128A alleles, respectively; see Figure 2) on the mitotic spellchecker and antirecombination functions of the encoded mutant proteins (Table 3). In the SJR1392 strain background, deletion of MLH1 resulted in a 53-fold elevation in the rate of forward mutation at CAN1 and in a 9.7-fold increase in the rate of homeologous (relative to homologous) recombination. These effects are statistically the same as those observed in the pms1Δ mutant. Both the mlh1-E31A and mlh1-G98A alleles were indistinguishable from the mlh1Δ allele in terms of the rate of His+ recombinants, indicating that both ATP binding and hydrolysis are essential for the antirecombination activity of Mlh1p. In terms of the spellchecker function, the mlh1-G98A allele resulted in a 48-fold increase in the rate of Can-R colonies while the mlh1-E31A allele resulted in a lesser, 20-fold increase. In agreement with an earlier study (Tran and Liskay 2000), it thus appears that Mlh1p retains residual spellchecker activity when ATP hydrolysis, but not binding, is compromised. Finally, we constructed an mlh1-G98A pms1-G128A double-mutant strain. The spellchecker and antirecombination phenotypes of the double mutant were indistinguishable from those of an mlh1Δ or pms1Δ mutant.
Role of the Mlh1p CTH domain in mitotic MMR functions: The final 13 amino acids of yeast Mlh1p and human MLH1 are identical and constitute the CTH domain. This domain is not required for interaction between yeast Mlh1p and Pms1p in two-hybrid assays, but is required for the spellchecker function of the complex (Panget al. 1997). To examine the role of the CTH domain in mitotic antirecombination, we replaced the codon specifying the first amino acid of the CTH domain with a stop codon (mlh1-757stop allele; see Figure 2). As shown in Table 3, the mlh1-757stop allele produced mutator and recombination phenotypes that were indistinguishable from those of a null (mlh1Δ) mutant.
Roles of Pms1p and Mlh1p ATP hydrolysis in the repair of mismatched meiotic recombination intermediates: In addition to the mitotic spellchecker and antirecombination activities, the yeast MMR system also detects and repairs the mismatch formed when a heterozygous marker (e.g., alleles A and a) is included in a heteroduplex recombination intermediate. Efficient MMR is associated with high levels of gene conversion and low levels of PMS for heterozygous markers, and inefficient MMR results in low levels of conversion and high levels of PMS. PMS tetrads with two A spore colonies, one a spore colony, and one sectored A/a spore colony are called “5A:3a” tetrads whereas those with one A spore colony, two a spore colonies, and one sectored A/a colony are called “3A:5a” tetrads. Using this nomenclature (derived from eight-spored fungi), we define Mendelian segregation as 4A:4a and gene conversion events as 6A:2a or 2A:6a.
Diploid strains heterozygous for the his4-AAG mutation in the HIS4 start codon were used to analyze the effects of defects in Pms1p- or Mlh1p-associated ATP hydrolysis (pms1-E61A and mlh1-E31A alleles, respectively) on the repair of mismatches in heteroduplex recombination intermediates. The PD83 strain background was used in these experiments because of the very high level of meiotic recombination at HIS4 (Naget al. 1989), which occurs as a consequence of a high frequency of meiosis-specific double-strand breaks near the 5′ end of the gene (Fanet al. 1995). Depending on which DNA strand is transferred during heteroduplex formation, a heteroduplex composed of a wild-type strand and a strand with the his4-AAG substitution will contain either an A-A or a T-T mismatch (Detloffet al. 1991). As shown in Table 4, a strain with normal MMR (PD83) repaired both mismatches efficiently (Detloffet al. 1991), resulting in high levels of gene conversion and low levels of PMS.
In our previous studies, the efficiency of mismatch repair was determined by dividing the number of tetrads in which one or more spores exhibit PMS by the total number of tetrads with non-Mendelian (aberrant) seg-regation. This approach has two inherent problems. First, tetrads with multiple PMS or multiple conversion events are counted as equivalent to tetrads with a single PMS or conversion event. Second, it is not clear how to count a tetrad that contains both a conversion event and a PMS event. Consequently, here we used a different method to measure the efficiency of meiotic hetero-duplex repair, which is based on counting the number of individual PMS and gene conversion spore colonies rather than tetrads. Using the definitions of tetrad classes given in Detloff et al. (1991), the number of PMS spore colonies (indicated in parentheses) counted in each class is the following: normal 4:4 (0), 6:2 (0), 2:6 (0), 5:3 (1), 3:5 (1), aberrant 4:4 (2), aberrant 6:2 (2), aberrant 2:6 (2), deviant 5:3 (3), deviant 3:5 (3), deviant 4:4 (4), 7:1 (1), 1:7 (1), 8:0 (0), and 0:8 (0). The number of gene conversion spore colonies counted for each class of tetrad is the following: normal 4:4 (0), 6:2 (1), 2:6 (1), 5:3 (0), 3:5 (0), aberrant 4:4 (0), aberrant 6:2 (0), aberrant 2:6 (0), deviant 5:3 (0), deviant 3:5 (0), deviant 4:4 (0), 7:1 (1), 1:7 (1), 8:0 (2), and 0:8 (2).
The tetrad/spore data for PD83 and mutant derivatives are presented in Table 4. The levels of aberrant segregation tetrads in all strains were similar, varying between 57 and 60%. As expected from previous studies (Williamsonet al. 1985; Prollaet al. 1994), diploids homozygous for mlh1Δ or pms1Δ exhibited an increase in the relative frequencies of PMS spore colonies from 18% in wild type to 89 or 78%, respectively, indicating inefficient meiotic mismatch repair. Statistical comparison of the relative number of PMS vs. gene conversion spore colonies indicates that mlh1Δ strains had significantly less mismatch repair than the pms1Δ strains (P < 0.0001). Although the mlh1-E31A strain had significantly less MMR than the wild-type strain (P < 0.0001), it had significantly more repair than the mlh1Δ strain (P < 0.0001). Similarly, the pms1-E61A strain has less mismatch repair than the wild-type strain (P < 0.002), but more repair than the pms1Δ strain (P < 0.0001). Finally, relative to the corresponding null allele, the pms1-E61A allele did not appear to confer as severe a defect in meiotic MMR as the mlh1-E31A allele.
The frequency of PMS events at HIS4 in the wild-type strain PD83 was higher than that observed in most studies involving different mutant alleles in other genetic backgrounds. Although one interpretation of this finding is that PD83 has a less efficient MMR system than that of other wild-type strains, we prefer a different explanation: that the efficiency of MMR is context dependent. One argument in support of this conclusion is based on an analysis of aberrant segregation of the heterozygous arg4-17 allele in PD83. In 482 tetrads, we found 37 conversion events and no PMS events, a significant difference (P < 0.002) in the relative number of conversion and PMS tetrads compared to that observed for the his4-AAG marker. Since a heteroduplex formed between arg4-17 and ARG4 would contain either an A/A or a T/T mismatch (the same type of mismatch as expected for the his4-AAG marker), these results argue that the efficiency of meiotic MMR is affected by the context of the mismatch. In a previous study, Fogel et al. (1979) found that A/G or C/T mismatches failed to get repaired in a wild-type strain at frequencies (expressed as the percentage of PMS tetrads divided by total aberrant tetrads) of 12, 4, and 0% for mismatches located at the HIS4, ARG4, and TRP1 loci, respectively.
Meiotic crossovers and spore viability in pms1 and mlh1 mutants: It has been observed previously that deletion of MLH1, but not PMS1, reduces crossovers in a variety of intervals (Hunter and Borts 1997; Wanget al. 1999). We calculated map distances for several genetic intervals on chromosome III in wild-type and mutant strains using only tetrads in which both markers for each interval underwent Mendelian segregation (Table 5). This analysis confirmed the previous observation that the mlh1Δ mutation significantly reduces crossovers in most genetic intervals examined (Hunter and Borts 1997; Wanget al. 1999). In contrast, the mlh1-E31A mutation significantly reduced crossovers only in the HIS4-CEN3 interval. The difference in tetrad classes for the mlh1Δ compared to the mlh1-E31A strain was significant only for the MAT-CEN3 interval. Neither the pms1Δ nor the pms1-E61A mutation significantly affected crossovers in any of the intervals examined.
Diploid strains with mutations in MLH1 or PMS1 have reduced spore viability compared to wild-type strains, with mlh1 alleles having a stronger effect than pms1 alleles (Prollaet al. 1994; Hunter and Borts 1997; Wanget al. 1999). Compared to the wild-type strain (84% spore viability), we found significant (P < 0.0001) decreases in total spore viability with the mlh1Δ, pms1Δ, and mlh1-E31A strains (69, 77, and 79% spore viabilities, respectively), but not with the pms1-E61A strain (84% spore viability; P = 0.2). There also were significant (P < 0.01) elevations in the proportion of tetrads with two viable and two inviable spores for the mlh1Δ, pms1Δ, and mlh1-E31A strains but not for the pms1-E61A strain; none of the MMR-deficient strains had a significant elevation in the proportion of tetrads with three viable spores (Figure 4). The mlh1Δ and mlh1-E31A strains were significantly different from each other in all spore viability classes except the class with three viable spores.
The specific increase in the proportion of tetrads with two live:two dead spores is consistent with either meiosis I nondisjunction resulting from reduced crossing over or segregation of a heterozygous recessive lethal mutation. Meiosis I nondisjunction involving chromosome III can be readily assessed, with the two surviving spores predicted to be nonmaters because of heterozygosity at MAT. Wang et al. (1999) reported that 20 of 1632 tetrads in an mlh1Δ deletion strain had this segregation pattern, whereas no such tetrads were observed in wild-type or pms1Δ strains. Among the 1936 tetrads derived from the mlh1Δ and mlh1-E31A homozygous strains examined here, we found only 3 with the pattern of two nonmating spores and two dead spores. PCR analysis with MATa- and MATα-specific primers indicated that only 1 of the tetrads contained spores disomic for chromosome III (data not shown). In samples of 1074, 975, and 696 tetrads from wild-type, pms1Δ, and pms1-E61A strains, respectively, none had the segregation pattern characteristic of meiosis I nondisjunction of chromosome III. These results suggest that, although the mlh1Δ mutation clearly reduces the frequency of crossing over on chromosome III, this reduction is not sufficient to result in elevated meiosis I nondisjunction. The difference between our results and those of Wang et al. (1999) may reflect the different genetic backgrounds used in the two studies; in our genetic background, the HIS4 locus (on chromosome III) is an extraordinarily strong recombination hotspot (Whiteet al. 1993). In addition, in our experiments, the strains were sporulated at 18° rather than at room temperature.
Different efficiencies of mismatch repair in different strain backgrounds: As described above, the pms1-E61A allele had no significant effect on mutation rates at the CAN1 locus in the haploid strain SJR1392 (Table 2), while the mlh1-E31A allele resulted in a mutator phenotype intermediate between those of wild-type and mlh1Δ strains (Table 3). Because these mutator assays were done in a strain background unrelated to the strains used in the meiotic experiments, we repeated the CAN1 mutator assay in derivatives of the haploid parental strains used to construct the diploids (Table 6). In agreement with earlier studies (Tran and Liskay 2000) and the results reported here with the SJR1392 derivatives, the effect of the mlh1-E31A allele is about one-half that observed in the mlh1Δ mutant, whereas the effect of pms1-E61A on mutation rate is very subtle. Although the mutator phenotypes reported here are qualitatively similar in different strain backgrounds, the absolute effects of null mutations in MLH1 and PMS1 on the forward mutation rate at the CAN1 locus vary considerably between strains. Relative to the isogenic wild-type strain, deletion of MLH1 or PMS1 elevates the CAN1 mutation rate ∼60-fold in the SJR1392 background (Tables 2 and 3), 30-fold in the AS4 background, and only 15-fold in the PD73 background (Table 6). Although rate differences of this sort generally are attributed to minor variations in the mutator assay as performed in different labs at different times, the mutator phenotypes for AS4 and PD73 were determined simultaneously. In addition, we performed side-by-side CAN1 mutation rate measurements for SJR1392 and PD73 and confirmed that the mutation rates are elevated to different extents in mlh1Δ mutants (data not shown). The strain-to-strain differences in mutation rates documented here likely reflect strain-dependent differences in the fidelity of DNA polymerase and/or differences in the efficiency of MMR.
Although the precise roles of MutL homologs are not known, it is generally assumed that heterodimers of these proteins serve as “matchmakers” to couple MutS-dependent mismatch recognition to the appropriate processing steps (see Harfe and Jinks-Robertson 2000a). The nature of the downstream processing steps is not clear, however, nor is it known whether these steps are identical in all MMR-related processes. For example, the repair of mismatch-containing replication intermediates must incorporate a strand discrimination step, which may not be relevant to the recognition/repair of mismatches in recombination intermediates. To address possible dissimilarities between MMR mechanisms in replication vs. recombination, we randomly mutagenized the PMS1 gene and screened for alleles that differentially affected the mitotic spellchecker and antirecombination activities of the corresponding proteins. Two such separation-of-function alleles were identified in the screen (pms1-L124S and pms1-I854M), each of which resulted in a reduction in the antirecombination activity of Pms1p, but had no significant effect on the spellchecker function. When the pms1-I854M mutation was combined in the same gene with mutations similar to the pms1-L124S mutation (pms1-E61A,I854M and pms1-G128A,I854M double-mutant alleles; Table 2), the mutations were not epistatic and, therefore, likely affect the function of the Mlh1p-Pms1p complex in fundamentally different ways.
The alteration conferred by the pms1-I854M allele is within the C-terminal 200 amino acids of Pms1p, a region that is highly conserved with the human PMS2 protein, but is not represented in the yeast and human Mlh1 proteins or in the bacterial MutL protein (Crouse 1998). Because the conserved C-terminal region of Pms1p is required for ATP-independent heterodimer formation with Mlh1p (Panget al. 1997), the separation-of-function phenotype associated with the pms1-I854M allele might simply reflect a reduction in the amount of the Mlh1p-Pms1p complex, with full antirecombination activity requiring more of the complex than the spell-checker function. Our two-hybrid analysis (Figure 3) indicates, however, that the I854M amino acid change does not significantly affect the level of the Mlh1p-Pms1p interaction. An alternative to a stability-related explanation for the pms1-I854M separation-of-function phenotype derives from studies of the C-terminal region of the bacterial MutL protein, which not only is important for analogous homodimer formation (Drotschmannet al. 1998), but also is required for interaction of MutL with the MutH endonuclease and with the UvrD helicase in two-hybrid assays (Hallet al. 1998; Hall and Matson 1999). It thus is intriguing to speculate that the pms1-I854M mutation alters an interaction with a protein that is important in mitotic antirecombination, but that plays little, if any, role in mutation avoidance. Candidates for such a protein might include members of the RAD52 epistasis group of recombination proteins (Sunget al. 2000) if mismatches are sensed during the strand invasion process or might include proteins involved in disrupting or destroying a recombination intermediate (e.g., helicases or nucleases) if mismatches are detected after formation of stable hetero-duplex DNA.
Although it is unclear how the pms1-I854M mutation impacts protein function, the pms1-L124S allele changes an amino acid that is immediately adjacent to conserved motif III of the GLH superfamily of ATPases (Figure 2). Motif III is part of the “ATP lid,” which undergoes dramatic conformational change when the N-terminal fragment of MutL (LN40) binds ATP (Banet al. 1999). This ATP-dependent conformational change is accompanied by dimerization of LN40, which is required for subsequent ATP hydrolysis and is speculated to be critical for the interaction of MutL with proteins that participate in the downstream steps of MMR. The weak ATPase activity of MutL presumably returns the protein to its starting conformation so that it can participate in another round of MMR. The situation is more complex in eukaryotes than in bacteria, with the active forms of the comparable MutL-like complexes being heterodimers. A functional asymmetry in the yeast Mlh1p-Pms1p heterodimer has been observed in genetic studies in which targeted mutations were introduced into the ATP binding/hydrolysis domains of the individual subunits (Tran and Liskay 2000). In these studies, mutations that impacted ATP binding or the associated conformational changes (mlh1-G98A and pms1-G128) resulted in a stronger mutator phenotype than that of the ATP hydrolysis mutations (mlh1-E31A and pms1-E61A), and the mutations in Mlh1p resulted in a stronger mutator phenotype than that of the comparable mutations in Pms1p (Tran and Liskay 2000). On the basis of these results, it was suggested that ATP binding by the individual subunits in the Mlh1p-Pms1p complex may be sequential, with ATP binding by Mlh1p preceding and perhaps facilitating ATP binding/hydrolysis by Pms1p. In support of this model, recent biochemical studies have shown that an N-terminal fragment of yeast Mlh1p binds ATP with 10-fold higher affinity than does a comparable N-terminal fragment of Pms1p (Hallet al. 2002; see also Tomeret al. 2002). The asymmetry between Mlh1p and Pms1p may not be limited only to ATP binding, but may also extend to ATP hydrolysis, as each N-terminal fragment has inherent ATPase activity (Guarneet al. 2001; Hallet al. 2002). This is in contrast to an N-terminal fragment of the bacterial MutL protein, where only the homodimer exhibits ATPase activity (Banet al. 1999).
According to the model described above, the Mlh1p-Pms1p complex would be expected to retain some function if the ATPase activity of Pms1p is compromised, but would be expected to retain little, if any, function if the ATPase activity of Mlh1p is eliminated. This prediction not only is consistent with spellchecker phenotypes reported previously (Tran and Liskay 2000), but also is supported by the recombination analyses reported here, with mutations in the ATPase domain of Mlh1p resulting in a more severe in vivo phenotype than those in the ATPase domain of Pms1p. The functional asymmetry observed for the ATPase domains of Mlh1p and Pms1p in correcting DNA replication errors (Tran and Liskay 2000) thus extends to the mitotic antirecombination function of the MMR system as well as to the repair of mismatches in meiotic recombination intermediates.
In addition to the asymmetry between Mlh1p and Pms1p observed in both the spellchecker and recombination assays, the results reported here indicate that disruption of the ATPase activity of Pms1p impacts the recombination-related functions of the Mlh1p-Pms1p complex more than the replication-related spellchecker function. The differential effect was evident when examining mitotic recombination between homeologous substrates (Table 2) and when assessing the repair of mismatches in meiotic recombination intermediates (Table 4). These results suggest that the cycles of conformational changes induced by ATP binding/hydrolysis by Pms1p are more important in the recognition or processing of DNA mismatches in recombination intermediates than in the recognition or processing of DNA mismatches resulting from DNA replication errors. Although the purpose of these conformational changes is not clear, it is likely that they are important in interactions of the Mlh1p-Pms1p heterodimer with other proteins involved in MMR-related functions. The proteins required for processing mismatch-containing recombination vs. replication intermediates may be different or the proteins simply could be present at different levels in recombination vs. replication intermediates. For example, proliferating cell nuclear antigen (PCNA), which is known to interact with MMR proteins (Johnsonet al. 1996a; Umaret al. 1996; Flores-Rozaset al. 2000), would be expected to be found at high concentrations at DNA replication forks, but may not necessarily be present at high levels in a heteroduplex recombination intermediate. Thus, the Pms1p ATPase motifs might be necessary for formation and/or stability of a functional mismatch-repair complex in the absence of PCNA. Alternatively, the ATPase activity of Pms1p may be required to direct the activity of downstream factors in the absence of a replication-associated signal that determines which strand is to be repaired.
Although we favor the explanation that the pms1-E61A, pms1-L124S, and pms1-G128A mutations partially separate the functions of the yeast Mlh1p-Pms1p complex in replication and recombination, there are several caveats to this conclusion. First, since the assays used to monitor recombination-related functions of the MMR system were quite different from those used to assess the spellchecker function, we cannot rule out the possibility of DNA sequence-specific or chromosome context-specific effects on MMR activity. In addition, there may be competing systems of repair that operate differently on recombination vs. replication intermediates. Finally, recombination-related processes may be more sensitive to the concentration of the Mlh1p-Pms1p complex than are replication-related processes. It is formally possible that the separation-of-function mutations in PMS1 decrease the overall stability of the protein and thereby reduce the concentration of the Mlh1p-Pms1p complex. Such a stability explanation has been invoked to explain MLH1 separation-of-function alleles that affect the repair of mismatches in meiotic recombination intermediates more than meiotic crossing over (Arguesoet al. 2002). Although we have not directly examined the stabilities of the mutant Pms1p proteins, as reported here for the pms1-I854M allele, introduction of the PMS1 or MLH1 ATP binding/hydrolysis mutations does not affect the stability of two-hybrid fusion proteins (Tran and Liskay 2000).
In summary, the results presented here demonstrate that it is possible to mutationally separate the replication vs. recombination roles of the yeast Pms1p protein. The identification of pms1 separation-of-function alleles is consistent with the notion that the Mlh1p-Pms1p complex couples mismatch recognition to the appropriate downstream processing steps and suggests that the downstream steps may differ, depending on the context of the mismatch. A major goal of future MMR studies in yeast will be to define the relevant downstream steps in replication vs. recombination processes.
We thank D. Brenner for help with the genetic analysis. This work was supported by National Institutes of Health (NIH) grants GM-24110 to T.D.P., GM-38464 to S.J.-R., and GM-45413 to R.M.L. P.T.T. was supported by NIH training grant HL-07781; J.S. and H.M.K. were supported by NIH training grant GM-07092.
Communicating editor: A. Nicolas
- Received March 27, 2002.
- Accepted August 30, 2002.
- Copyright © 2002 by the Genetics Society of America