Interactions of Exo1p with components of MutLα in Saccharomyces cerevisiae

Abstract

Previously, we reported evidence suggesting that Saccharomyces cerevisiae MutLα, composed of Mlh1p and Pms1p, was a functional member of the gyrase b/Hsp90/MutL (GHL) dimeric ATPase superfamily characterized by highly conserved ATPase domains. Similar to other GHL ATPases, these putative ATPase domains of MutLα may be important for the recruitment and/or activation of downstream effectors. One downstream effector candidate is Exo1p, a 5′-3′ double stranded DNA exonuclease that has previously been implicated in DNA mismatch repair (MMR). Here we report yeast two-hybrid results suggesting that Exo1p can interact physically with MutLα through the Mlh1p subunit. We also report epistasis analysis involving MutLα ATPase mutations combined with exo1Δ. One interpretation of our genetic results is that MutLα ATPase domains function to direct Exo1p and other functionally redundant exonucleases during MMR. Finally, our results show that much of the increase in spontaneous mutation observed in an exo1Δ strain is REV3-dependent, in turn suggesting that Exo1p is also involved in one or more MMR-independent mutation avoidance pathways.

DNA mismatch repair (MMR) is a highly conserved genome fidelity process. Phenotypes of MMR deficiency are diverse, ranging from increased spontaneous mutation rates to cancer predisposition (1, 2). Mutation avoidance is a major function of MMR and can be dissected by using facile model systems such as Escherichia coli and the yeast Saccharomyces cerevisiae (2–5). For E. coli MMR, all essential genes have been identified, and their gene products have been used to reconstitute a MMR reaction in vitro (5). The three central components of this pathway are MutS, MutL, and MutH. A MutS dimer binds to a mispair, followed by ATP-dependent complex formation with a MutL dimer. The MutS/MutL/mispair ternary complex is thought to direct downstream events, including methylation-dependent nascent strand nicking by MutH, excision of the nascent strand, repair synthesis, and ligation. Recent studies suggest that the coordination of multiple downstream events, including nicking and excision, are facilitated by the ATPase activities of the MutL dimer (6–8).

MMR-mediated mutation avoidance in S. cerevisiae involves multiple MutS homologues (MSH) and MutL homologues (MLH) (2–4). For mutation avoidance, yeast use two partially redundant MutS-like activities, MutSα (Msh2p-Msh6p heterodimer) (9–16) and MutSβ (Msh2p-Msh3p heterodimer) (10, 11). Similarly, yeast use two MutL-like activities, MutLα (Mlh1p-Pms1p heterodimer) (17–20) and MutLβ (Mlh1p-Mlh3p heterodimer) (21, 22), although, based on genetic analysis, MutLα is the major MutL-like activity. Yeast have no known sequence or structural MutH homologue, partly exemplifying the lack of insight into the mechanism of strand discrimination.

In a previous study, we reported studies suggesting that S. cerevisiae MutLα is a member of the gyrase b/Hsp90/MutL (GHL) dimeric ATPase superfamily, which is characterized by highly conserved ATPase motifs (23). Although direct evidence for ATP-binding and hydrolysis activity has not been reported, our genetic and biochemical results suggest that, similar to other GHL ATPases, yeast MutLα undergoes ATP-dependent conformational changes, highlighted by dimerization of the NH₂-terminal ATPase domains (23, 24). These ATP-dependent conformational changes in MutLα and resultant NH₂-terminal dimerization between Mlh1p and Pms1p protomers appear to be critical for MMR because mutations affecting these activities compromise yeast MMR in vivo (23). Analogous to other GHL ATPases (7, 25–32), the apparent ATP-dependent conformational changes and the NH₂-terminal dimerization of MutLα may help to direct downstream effectors in the MMR process. One such downstream effector candidate is the 5′-3′ exonuclease Exo1p, originally identified as a mutator gene in Schizosaccharomyces pombe (33, 34), and later reported for S. cerevisiae as a “two-hybrid” interactor with Mlh1p (35) and Msh2p (36). Moreover, previous genetic studies showed essentially identical phenotypic effects between exo1Δ and a missense mutation in a residue predicted to be critical for exonuclease activity but not for structural integrity of the protein. These findings were consistent with Exo1p performing a catalytic role during MutSα-dependent MMR rather than being limited to a structural role (37).

Here, we report results suggesting both physical and genetic interactions between EXO1 and the components of MutLα, Mlh1p and Pms1p. Specifically, we characterized further our initial two-hybrid interaction between Mlh1p and Exo1p. More interestingly, we report genetic interactions between mutations in MLH1, PMS1, and EXO1 that suggest that one function of the MutLα ATPase domains is to direct Exo1p and possibly other exonucleases during MMR-mediated mutation avoidance. Finally, the results suggest that Exo1p is also involved in one or more MMR-independent mutation avoidance pathways.

Materials and Methods

Strains and Media.

E. coli strain DH-10B was used for plasmid construction and amplification. Bacterial and yeast strains were grown under conditions described (17). Yeast transformations were performed by the polyethylene glycol/lithium acetate method (38).

Disruptions of EXO1 were generated as described previously (36). An exo1∷HIS3 disruption cassette was generated by PCR using strain EAY618 (E. Alani, Cornell Univ., Ithaca, NY), and transformed into yeast.

Disruptions of REV3 were created by transforming the pYPG101 construct (D. Hinkle, Rochester University, Rochester, NY) after KpnI digestion and selecting for Ura⁺ prototrophs. Genotyping of strains was performed by PCR or Southern blot analysis (specifics are available on request). The generation of the other strains used in this study has been described (23).

Plasmid Construction.

All DNA manipulations were performed by using standard molecular biology procedures (39). All automated sequencing was done with an Applied Biosystems automated sequencer.

pBTM-MSH2 and pBTM-MSH6 were constructed by cloning the coding sequences for yeast MSH2 and MSH6 into the two-hybrid “bait” vector pBTM116. The other constructs used in this study were described (17).

Two-Hybrid Screening and Mating and β-Galactosidase Assays.

The two-hybrid screening was performed as described (40) with a yeast cDNA expression library (S. Elledge, Baylor College of Medicine, Waco, TX). Candidates were retested directly “one-on-one” by mating as described (23). Growth at 30°C for 2–3 days on −uracil −tryptophan −leucine (−UTL) plates indicated efficiency of mating, whereas growth on −tryptophan −histidine −uracil −leucine −lysine (−THULL) plates indicated “bait–prey” interaction.

Diploid L40/AMR70 is homozygous for a second chromosomal lexA-GAL4A reporter system, URA3∷(lexAop) ₈-lacZ. β-Galactosidase liquid assays were performed as described (17). One β-galactosidase unit = [(OD₄₂₀/OD₆₀₀)60]/ min, where OD₄₂₀ and OD₆₀₀ are the optical densities at 420 and 600 nm, respectively.

Measurement of Mutation Rates and CAN1 Mutational Spectra Analysis.

Strains were streak purified, individual colonies were grown to saturation in YPD medium, then various dilutions were plated onto complete synthetic medium, −threonine, and +canavanine (+CAN) [at 60 μg/ml] plates, and colonies were counted after 2–3 days growth at 30°C. Rates were determined as described (17). The 95% confidence intervals (CI) of relative mutation rates were determined by using PRISM 2.0a software (GraphPad, San Diego).

Canavanine-resistance (CAN^R) mutations were identified from genomic preparations by using the glass bead lysis method, followed by PCR of the CAN1 gene as described (23), and direct sequencing of the QIAquick (Qiagen, Chatsworth, CA) purified PCR amplicon. χ² analysis was used to determine whether changes in mutational spectra were statistically significant (P < 0.05). Rates of frameshifts (FS) and base substitutions (BS) at CAN1 were calculated by using absolute mutation rates determined at CAN1 multiplied by the frequency with which FS or BS mutations occurred in the particular strain. As the calculated rates of FS and BS at CAN1 in Table 3 possess the product of two different forms of error, we were unable to perform statistical analysis on these values.

Table 3.

Relative estimated rates of FS and BS mutations at CAN1

Strain	Relevant genotype	Fold mutator rate
		FS*	BS†
GCY35	Wild type	1	1
PTY100	mlh1Δ	73	12
PTY200	mlh1-E31A	15	6
PTY300	mlh1-G98A	53	6
PTY101	pms1Δ	64	10
PTY104	pms1Δmlh3Δ	63	10
PTY202	pms1-E61A mlh3Δ	1‡	1‡
PTY302	pms1-G128A mlh3Δ	9	2
PTY105	exo1Δ	12	9
PTY106	mlh1Δexo1Δ	72	22
PTY207	mlh1-E31A exo1Δ	104	18
PTY307	mlh1-G98A exo1Δ	46	36
PTY107	pms1Δexo1Δ	68	21
PTY205	pms1-E61A exo1Δmlh3Δ	31	6
PTY305	pms1-G128A exo1Δmlh3Δ	38	13

Relative rates were calculated from Table 1 and frequency of FS and BS mutations from Table 2 as described in Materials and Methods. 

^*Relative to the wild-type rate of 1.05 × 10⁻⁷ for FS at CAN1. 

^† Relative to the wild-type rate of 1.66 × 10⁻⁷ for BS at CAN1. 

^‡ Assumed that PTY202 mimics the wild type for ratio of FS/BS. 

Results

Experimental Rationale.

In a previous report we referred for convenience to the mutations mlh1-E31A and pms1-E61A, which are predicted to affect ATP hydrolysis with little or no effect on ATP binding, as “ATP-hydrolysis” mutations (23). Similarly, we referred to a second pair of mutations, mlh1-G98A and pms1-G128A, which are predicted to cause a deficiency in ATP-binding and/or ATP-binding-dependent conformational changes, as “ATP-binding” mutations. Interestingly, we found that alterations in both of these putative ATPase motifs of Mlh1p produced more severe effects on mutation avoidance than did the corresponding “ATPase” mutations in Pms1p. These differential effects of the mlh1 ATPase mutations versus the pms1 ATPase mutations were not because of partially redundant functions of MLH3 and PMS1, and therefore suggested a functional asymmetry within the MutLα heterodimer (23).

As discussed above, one possible function of the ATPase domains of MutLα is the recruitment and/or activation of additional proteins. One candidate is Exo1p, which we identified previously from a two-hybrid screen as an Mlh1p interactor (35) (see below). In the present study we have analyzed further initial two-hybrid results and have performed epistasis analysis with the MutLα “ATPase” mutations combined with exo1Δ. The epistasis analyses included both mutation-rate measurements at hom3-10 and CAN1, and mutational spectrum analysis using the CAN1 reporter. The hom3-10 allele enables a reversion assay that reports single T⋅A base pair deletions in a run of 7 T⋅A base pairs, and has been considered diagnostic for defects in MMR (41). In contrast, forward mutation at CAN1 shows a wide variety of inactivating mutations, the spectra of which can be determined by DNA sequencing (41). Finally, because MLH3 is partially redundant with PMS1 for mutation avoidance (21–23), we examined the mutational spectra of relevant pms1 ATPase mutant strains in an mlh3Δ background.

Exo1p Interacts with Mlh1p by Yeast Two-Hybrid Assay.

Using full-length Mlh1p as a bait in a two-hybrid screen, we recovered a COOH-terminal fragment of Exo1p (residues 400–702). This Exo1p fragment was retested directly, and was shown to interact with LexAp-Mlh1p as depicted in Fig. 1a. Using deletion constructs, we mapped the minimal region of Mlh1p required to interact with this COOH-terminal fragment of Exo1p to residues 501–761 of Mlh1p (Fig. 1a). To address the question whether ATP-binding or ATP-binding-dependent conformational changes by Mlh1p were necessary for interaction with Exo1p, we examined LexAp-mlh1-G98A for interaction with Exo1p. As shown in Fig. 1a, this Mlh1p mutant, which is predicted to have compromised ATP-binding activity, retained ability to interact with Gal4p-Exo1p-(400–702).

Figure 1.

Exo1p and MMR proteins interact in a yeast two-hybrid assay. (a) Boxes correspond to Mlh1p “bait” constructs tested for interaction. The residues of Mlh1p included in the fusions are indicated above each respective construct. The amino acid substitution G98A made in Mlh1p is designated by a black bar within the construct box. Interaction is scored as growth on −histidine media and >0.5 β-galactosidase units with the substrate o-nitrophenylβ-d-galactosidase (ONPG) as described in Materials and Methods. (b) Full-length LexAp-Pms1p fusion alone or in a “three-hybrid” assay with native Mlh1p coexpressed was tested for interaction with the Gal4p-Exo1p-(400–702) fusion using the same analysis as in a.

We also tested the Exo1p clone against a panel of other MMR proteins. This COOH-terminal fragment of Exo1p interacted with full-length LexAp-Msh2p (data not shown) as in a previous report (36), but not with full-length LexAp-Msh6p (data not shown) or LexAp-Pms1p fusions (Fig. 1b). However, we found that the COOH-terminal fragment of Exo1p did interact with LexAp-Pms1p in a “three-hybrid” assay in which native Mlh1p was coexpressed (Fig. 1b). Taken together, these results suggest that the COOH-terminal fragment of Exo1p interacts with the Mlh1p subunit of MutLα.

Epistasis Analysis of hom3–10 Reversion Rates for mlh1 and pms1 “ATPase” Mutations and exo1Δ.

To detect genetic interactions between components of MutLα and EXO1, we determined mutation rates in a series of single and double mutants (Table 1). The most striking result was that either the pms1-E61A or the pms1-G128A mutation, when combined with exo1Δ, produced greater-than-additive mutation rates (with 95% CI) by using hom3–10 (Table 1; PTY204 > PTY105 + PTY201, and PTY304 > PTY105 + PTY301). Likewise, using hom3–10 reversion, the mlh1-E31A mutation appeared to synergize for spontaneous mutation in combination with exo1Δ (Table 1; PTY207 > PTY105 + PTY200). In contrast, the small effect of exo1Δ at hom3–10 relative to the large effect produced by the mlh1Δ, pms1Δ, or mlh1-G98A mutations prevented us from making conclusions regarding epistasis with the corresponding double mutants. Using 95% CIs at CAN1, we observed a synergistic interaction for mutation rates only for the pms1-G128A and exo1Δ mutations (Table 1). However, as presented below, determination of CAN1 mutational spectra suggested that similar to findings at hom3–10, exo1Δ interacted with mlh1-E31A, pms1-E61A, or pms1-G128A in a greater-than-additive fashion for FS mutations in short mononucleotide runs.

Table 1.

Mutation rates of exo1Δ strains in yeast MutL mutant backgrounds

Strain	Relevant genotype	Fold mutator rate (95% CI)*
		hom3-10†	CAN‡
GCY35§	Wild type	1  (0–2.3)	1  (0.1–2)
PTY100	mlh1Δ	1,118  (858–1378)	32  (22–41)
PTY101§	pms1Δ	1,212  (1017–1408)	28  (15–41)
PTY105	exo1Δ	8  (0–24)	9  (5–14)
PTY106	mlh1Δexo1Δ	1,227  (134–2320)	37  (30–43)
PTY107	pms1Δexo1Δ	1,097  (253–1940)	35  (11–60)
PTY200	mlh1-E31A	316  (170–461)	9  (3–14)
PTY207	mlh1-E31A exo1Δ	1,219  (779–1660)	46  (21–72)
PTY201§¶	pms1-E61A	19  (6–32)	1  (0.5–2)
PTY204¶	pms1-E61A exo1Δ	517  (163–872)	15  (9–20)
PTY300§	mlh1-G98A	725  (524–926)	22  (10–34)
PTY307	mlh1-G98A exo1Δ	922  (361–1482)	36  (22–50)
PTY301§¶	pms1-G128A	78  (33–122)	4  (3–5)
PTY304¶	pms1-G128A exo1Δ	611  (361–862)	24  (20–27)

^*From two to six determinations with 5–11 cultures per experiment. 

^† Relative to wild-type GCY35 rate of 9.9 × 10⁻⁹. 

^‡ Relative to wild-type GCY35 rate of 3.01 × 10⁻⁷. 

^§ These rates are taken from Tran and Liskay (23). 

^¶ Codeletion of MLH3 did not change the rates significantly (two-tailed Mann–Whitney test, P > 0.05). 

Analysis of CAN1 Mutation.

To elucidate further genetic interactions between MLH1, PMS1, and EXO1, we determined the CAN1 mutational spectra for a subset of single and double mutant strains. As shown in Table 2, all single mlh1 and pms1 mutant strains examined showed a CAN1 spectrum characterized by a FS to BS mutation ratio (FS/BS) of 2 or greater. A 2- to 3-fold preponderance of FS over BS mutations has been shown previously for msh2Δ strains (10). Based on other studies (10, 42) and our findings with mlh1 and pms1 null strains, we will consider an excess of FS over BS mutations as reflective of a defect in MMR. In contrast, the exo1Δ strain exhibited a spectrum that was different from a MMR-defective strain, e.g., mlh1Δ (P < 0.001), but not different from the wild-type strain (P > 0.1) (Table 2). Although no single pairwise comparison was statistically significant (P > 0.1), deletion of EXO1 combined with any of the mlh1 or pms1 single mutations produced, as a general trend, an apparent shift in spectrum toward more BS mutations (Table 2). Therefore, initial analysis of the CAN1 mutations suggested that exo1 deficiency alone did not produce a spectrum characteristic of MMR deficiency, and that combining exo1Δ with several of the pms1 or mlh1 mutations appeared to cause a shift in the FS/BS ratio toward one more characteristic of wild-type (or exo1Δ) cells.

Table 2.

Summary of mutational spectra at CAN1

Strain	Relevant genotype	Class of mutation				FS/BS†
		FS		BS	Complex
		Frequency (%)	Type*	Frequency (%)	Frequency (%)
GCY35‡§	Wild type	7/20 (35)	86:14	11/20 (55)	2/20 (10)¶	0.6
PTY100‡	mlh1Δ	8/10 (80)	100:0	2/10 (20)	0/10 (0)	4.0
PTY200‡	mlh1-E31A	12/20 (60)	75:25	8/20 (40)	0/20 (0)	1.9‖
PTY300‡	mlh1-G98A	17/20 (85)	99:12	3/20 (15)	0/20 (0)	5.7
PTY101‡	pms1Δ	8/10 (80)	62:38	2/10 (20)	0/10 (0)	4.0
PTY104‡	pms1Δmlh3Δ	8/10 (80)	62:38	2/10 (20)	0/10 (0)	4.0
PTY302‡	pms1-G128A mlh3Δ	18/25 (72)	94:6	7/25 (28)	0/25 (0)	3.3‖
PTY105	exo1Δ	9/20 (45)	67:33	11/20 (55)	0/20 (0)	0.8
PTY106	mlh1Δexo1Δ	13/19 (68)	100:0	6/19 (32)	0/19 (0)	2.2
PTY207	mlh1-E31A exo1Δ	15/19 (79)	87:13	4/19 (21)	0/19 (0)	3.8
PTY307	mlh1-G98A exo1Δ	9/20 (45)	89:11	11/20 (55)	0/20 (0)	0.8
PTY107	pms1Δexo1Δ	12/18 (67)	92:8	6/18 (33)	0/18 (0)	2.0
PTY205	pms1-E61A exo1Δmlh3Δ	15/20 (75)	100:0	5/20 (25)	0/20 (0)	3.0
PTY305	pms1-G128A exo1Δ mlh3Δ	13/20 (65)	100:0	7/20 (35)	0/20 (0)	1.9

^*Ratio of contractions:expansions. 

^† FS:BS ratio. 

^‡ Spectrum taken from Tran and Liskay (23). 

^§ The mlh3Δ strain is no different from wild type (R. D. Kolodner, personal communication) and pms1-E31A mlh3Δ is not a mutator at CAN1 (Table 1). 

^¶ Duplication events with direct repeats. 

^‖ This value is the FS/BS ratio with the wild-type spectrum contribution subtracted. 

Another way to analyze mutation spectrum data is to estimate the rates of specific types of mutations arising at the mutation-reporter locus (21, 43, 44). Using the CAN1 mutation rates (from Table 1) and the frequency of FS and BS mutations arising at CAN1 (from Table 2), we estimated the rates of FS and BS mutations at CAN1 for the various single and double mutant strains. We present these rates relative to wild type in Table 3. Because of the limited sample size of can1 mutants analyzed for each strain and because we were unable to perform statistical analysis on the values shown in Table 3 (see Materials And Methods), we could not make strong arguments regarding epistasis, additivity, or synergy. Despite this limitation, we observed two interesting trends. (i) Similar to the hom3–10 data described above, combination of the exo1Δ mutation with the pms1-E61A, pms1-G128A, or mlh1-E31A mutation appeared to result in greater than additive effects on FS mutation rates at CAN1. (ii) For mlh1Δ and exo1Δ, mlh1-E31A and exo1Δ, and pms1Δ and exo1Δ mutant combinations, there appeared to be a general trend of additivity for BS mutation rates at CAN1. Taken together, there appeared to be synergistic interactions between the mlh1-E31A and pms1 “ATPase” mutations and exo1Δ for FS mutation rates, which we interpret as indicative of MMR deficiency. One idea that is consistent with our observations is that the ATPase domains of MutLα are important for directing Exo1p and factors functionally redundant with Exo1p during MMR. In addition, the trend of apparent additivity for BS mutation rates suggested that BS mutations are being contributed by more than one pathway that act in parallel to one another.

The Majority of Spontaneous Mutation at CAN1 in an exo1Δ Strain Is REV3-Dependent.

The mutational spectrum at CAN1 observed for the exo1Δ single mutant and the general trend of additive interactions between MutLα mutations and exo1Δ for BS rates at CAN1 suggested that EXO1 also may be involved in MMR-independent pathways for mutation avoidance. Because rev3Δ strains are slightly hypomutable spontaneously (45), and because rev3Δ can suppress the mutator phenotype of strains in which certain mutation avoidance pathways are defective (46–50), we examined the effect of rev3Δ on the exo1Δ mutator phenotype. Interestingly, although the CAN1 mutation rate of the exo1Δ strain (PTY105) was ≈9-fold higher than wild type (GCY35), the rate of the double rev3Δ exo1Δ strain (PTY111) was not statistically different from the wild-type rate (Table 4). These results suggested that the exo1Δ mutator phenotype at CAN1 was largely REV3-dependent.

Table 4.

Effect of rev3Δ on mutation rates

Strain	Relevant genotype	Fold mutator rate (95% CI)*
		hom3–10†	CANR‡
GCY35§	Wild type	1 (0–2.3)	1 (0.1–2)
PTY100§	mlh1Δ	1,118 (858–1378)	32 (22–41)
PTY105§	exo1Δ	8 (0–24)	9 (5–14)
PTY110	rev3Δ	0.3 (0.1–0.5)	0.7 (0–2)
PTY111	rev3Δexo1Δ	2 (0–4)	1 (0.6–3)
PTY112	rev3Δmlh1Δ	911 (156–1668)	28 (3–53)
PTY204§	pms1-E61A exo1Δ	517 (163–872)	15 (9–20)
PTY210	pms1-E61A exo1Δ rev3Δ	197 (121–273)	13 (10–15)
PTY304§	pms1-G128A exo1Δ	611 (361–862)	24 (20–27)
PTY310	pms1-G128A exo1Δrev3Δ	281 (192–370)	19 (12–25)

^*Experiments repeated two to four times with 5–11 cultures per experiment. 

^† Relative to wild-type GCY35 rate of 9.9 × 10⁻⁹. 

^‡ Relative to wild-type GCY35 rate of 3.01 × 10⁻⁷. 

^§ Rates taken from Table 1. 

As described above, we observed either significant (for hom3–10) or apparent (for CAN1) synergism between the pms1-E61A and pms1-G128A mutations and exo1Δ for FS mutation rates (Tables 1 and 3). To determine whether this pattern of FS synergy observed between pms1-E61A and pms1-G128A mutations and exo1Δ depended on REV3 function, we analyzed the effect of rev3Δ mutation on the pms1-E61A exo1Δ and pms1-G128A exo1Δ double mutant strains (Table 4). Notably, we observed that the mutation rate at either hom3–10 or CAN1 of the double mutants was not reduced significantly by rev3Δ (Table 4). Importantly, we determined that mlh1Δ rev3Δ and msh2Δ rev3Δ strains had mutation rates essentially the same as mlh1Δ or msh2Δ strains, respectively, suggesting that the majority of MMR-dependent mutations are not REV3-dependent (Table 4 and data not shown). These results are consistent with the hypothesis that the synergy seen between the pms1 “ATPase” mutations and exo1Δ is reflective of a defect in MMR. Taken together, the results suggest that EXO1 can be involved in at least two mutation avoidance pathways, an Mlh1p/Pms1p-dependent MMR pathway and a MMR-independent but REV3-dependent pathway.

Discussion

In this study, we report evidence for physical and genetic interactions between S. cerevisiae Exo1p and the components of MutLα, Mlh1p and Pms1p. In brief, using yeast two-hybrid and three-hybrid analyses, we observed that a COOH-terminal fragment of Exo1p interacted with MutLα through a COOH-fragment of Mlh1p. Epistasis analyses revealed greater than additive effects for spontaneous FS mutation rates, characteristic of a MMR defect, for strains containing the mlh1-E31A, pms1-E61A, or pms1-G128A mutations, located in the putative ATPase domains of MutLα, and combined with exo1Δ mutation. In addition, and in contrast to our findings for either mlh1Δ or pms1Δ strains, the CAN1 mutational spectrum of an exo1Δ strain was not consistent with a defect in MMR. In agreement with these CAN1 spectra comparisons, we found that, in contrast to the mlh1Δ or msh2Δ mutator phenotypes, much of the mutator effect at CAN1 in an exo1Δ strain was, in fact, REV3 dependent.

A previous report demonstrated physical interaction between Exo1p and MSH2 (36). Here, using two-hybrid analysis, we show that a COOH-terminal domain of Mlh1p required for interaction with Pms1p (17, 52), Mlh2p, and Mlh3p (52) also interacted with a COOH-terminal fragment of Exo1p (Fig. 1a). Furthermore, using a three-hybrid scheme, we provide evidence that Exo1p interacts with MutLα through the COOH-terminal of Mlh1p (Fig. 1b). Similar to studies with yeast and human Exo1p showing interaction with yeast and human Msh2p (36, 51, 53), respectively, we identified interactions between a COOH-terminal fragment of Exo1p and Mlh1p. These findings raise the possibility that, in eukaryotes, a conserved COOH-terminal domain of Exo1p is responsible for interactions with several classes of MMR proteins. Whether these interactions can occur concomitantly or only independently is of interest and will require further study. In addition, we found that a mutant form of Mlh1p, predicted to have reduced ATP binding, retained ability to interact with Exo1p, suggesting that ATP-dependent conformational changes in Mlh1p may not be necessary for interaction with Exo1p. Findings with similar mutant forms of E. coli MutL suggested that the ATPase activity of MutL was not necessary for interaction with MutH and UvrD, but was required for activation of these components during incision and excision (7). The genetic analysis (see below) addresses the issue of whether the ATPase domains of MutLα might be required for activation of downstream candidates in MMR, such as Exo1p.

We performed epistasis analysis using MutLα “ATPase” mutations and exo1Δ because of the evidence for physical interaction discussed above and our previous studies, which suggested that MutLα undergoes ATP-dependent NH₂-terminal conformational changes (23, 24) that in turn may be important for coordinating downstream events. Using two mutation reporters, we observed genetic interactions for mutation avoidance between mlh1 or pms1 mutations and exo1Δ. Of primary interest, the two pms1 “ATPase” mutations and mlh1-E31A mutation each synergized with exo1Δ for FS mutation rates. Similar synergistic interactions for mutation avoidance between weak mutator alleles of mlh1 or pms1 and exo1Δ were identified by the Kolodner group in a genetic screen designed to identify second-site mutations that would synthetically enhance the weak mutator phenotype of an exo1Δ strain (73). One explanation for the synergy is that the Mlh1p ATP-hydrolysis motif, for example, is necessary to coordinate a factor(s) [e.g., another exonuclease(s)] that is functionally redundant with Exo1p. Similarly, the synergy observed between each of the two pms1 “ATPase” mutations and exo1Δ suggests that these Pms1p motifs are also important for coordinating a factor(s) redundant with Exo1p. In other words, as one hypothesis, we suggest that these specific MutLα ATPase motif mutations mimic defects in factors redundant to Exo1p, at least in terms of MMR function, such that the combination of these MutLα mutations and exo1Δ (e.g., pms1-E61A and exo1Δ) mimics inactivation of EXO1 and the putative “redundant” gene(s). This explanation is consistent with the idea that the ATPase activity of E. coli MutL helps to coordinate both incision (6–8) and excision (7).

The existence of redundant factors for Exo1p agrees with the lack of a strong mutator phenotype for exo1Δ strains (36, 54) and in vitro evidence for bidirectional repair capability for eukaryotic MMR (55). In light of the proposal that MMR and the replication machinery may interact directly (40, 56, 57), an alternative explanation is that the MutLα ATPase motif mutations may impinge on replication fidelity and, in conjunction with exo1Δ, result in synergistic increases in mutation similar to those proposed for DNA polymerase proofreading defects combined with exo1Δ (54). Another explanation for a synergistic interaction for FS mutation rates is that exo1Δ combined with specific MutLα ATPase mutations may result in the “structural collapse” of a complex required for MMR. We do not favor exclusively such a structural requirement for Exo1p during MMR because studies using an exonuclease-deficient exo1 allele (37) that appears to be structurally intact (P.T.T. and R.M.L., unpublished data) also resulted in a synergistic increase in FS mutation rates when combined with either of the pms1 ATPase mutations (P.T.T. and R.M.L., unpublished data). Although a structural role for Exo1p in MMR is possible based on the studies of others (73), we favor the idea put forth previously (37) that Exo1p can have a catalytic role in MMR.

As discussed above, the CAN1 mutational spectrum for the exo1Δ strain was not characteristic of known MMR-defective strains, e.g., mlh1Δ. In addition, when we estimated rates of FS and BS mutations at CAN1 in various single and double mutants we observed two general trends. (i) Similar to the hom3–10 results, the mlh1-E31A mutation and both the pms1 ATPase mutations each appeared to synergize with exo1Δ for FS mutation rates at CAN1. (ii) Using a limited number of BS events at CAN1, estimated rates of BS mutations in the single and double mutant suggested an additive effect when exo1Δ was combined with any of several different MutLα mutations. Together, the results suggest involvement of EXO1 in a MutLα-dependent MMR pathway, based on FS mutation rates, and a MMR-independent mutation avoidance pathway, based on BS mutations rates.

What is the nature of the EXO1-dependent, MMR-independent pathway for mutation avoidance? Recent studies have shown that the mutator phenotypes observed in strains defective in several DNA repair pathways are largely dependent on the REV3 gene (46–50). One explanation for these findings is that when certain DNA repair pathways are blocked, spontaneously occurring DNA lesions are “funneled” into the REV3-dependent error-prone replication bypass pathway (48). Therefore, we characterized the effect of rev3Δ on the exo1Δ mutator phenotype. Interestingly, we found that CAN1 mutation rates in the exo1Δ rev3Δ strain were reduced to near wild-type, indicating that most of the CAN1 mutator phenotype of exo1Δ was dependent on REV3. Rev3p functions as a component of the error-prone polymerase ζ (Rev3p-Rev7p) to bypass DNA lesions that stall the replicative polymerases (45). Our data would therefore suggest that Exo1p assists in an error-free process acting on spontaneous DNA lesions. Because Exo1p has been implicated in several DNA metabolic pathways, such as repair of UV damage, recombination, and replication (58–64), further studies are required to clarify the proposed relationship between REV3 and EXO1. In contrast, rev3Δ did not significantly reduce the rate of mutation in either an mlh1Δ or msh2Δ strain, suggesting, not surprisingly, that the mutator phenotype of a MMR-defective strain is not REV3-dependent. Importantly, the synergistic interaction seen between exo1Δ and either of the two pms1 “ATPase” mutations was not REV3-dependent, consistent with our hypothesis that the synergy observed for FS mutation rates reflected a defect in MMR. As a whole, the results suggest that EXO1 can be involved in MMR-dependent and MMR-independent mutation avoidance pathways.

Until recently, mechanistic details of how the eukaryotic MutL homologues couple the mismatch binding activities of MutS homologues to downstream effectors in eukaryotes have been scarce. Studies identifying MutL homologues as members of an emerging ATPase superfamily have provided a framework with which to examine MutL homologue function during MMR-dependent mutation avoidance (6, 23, 29, 65). Based on the results presented here, we suggest that one function for the MutLα ATPase domains in S. cerevisiae is to coordinate Exo1p and redundant activities during mutation avoidance. Because MMR proteins also have been shown to function in other pathways, such as meiotic (52, 66–68) and homeologous recombination (69–72), the use of MutLα ATPase mutations may be important for a more complete understanding of the role of MMR proteins in other DNA transactions.

Abbreviations

MMR

mismatch repair

MutLα

Mlh1p and Pms1p

MSH

MutS homologues

MLH

MutL homologues

FS

frameshift

BS

base substitution

CI

confidence intervals

GHL

gyrase b/Hsp90/MutL