DISCUSSION

Several conclusions may be drawn from this study. The first is that, in some situations, DNA sequence plays a major role in the regulation of recombination. Similar conclusions were drawn from studies using heterologous palindromic sequences (Naget al. 1989; Mooreet al. 1999) and homozygous repetitive sequences (Gendrelet al. 2000). In our experiments, DNA sequence seemed to affect two different steps of recombination: heteroduplex formation and loop repair. In both cases, the recognition of DNA by MMR proteins was the step sensitive to the sequence.

Our results indicate that the S. cerevisiae Msh2/Pms1 complex inhibits RS heteroduplex formation. This has been suggested by Alani et al. (1994) from the observation that Msh2/Pms1 proteins regulate heteroduplex DNA tract length. Consistent with this model, the MutS/MutL protein E. coli analog inhibits the strand exchange promoted by the RecA protein (Worthet al. 1998). Two MMR complexes, which bind to base-base mismatches (Msh2p-Msh6p) and a small loop insertion (Msh2p-Msh3p), have been found to interact with Holliday junctions (Marsischkyet al. 1999) and recombination intermediates (Evanset al. 2000). We found that RS large heterologous inserts inhibited conversion. Two models could explain these observations. One is that mismatch repair recognizes intermediates that are formed and destroys them or restores parental segregation as a byproduct of repair. The heterozygous region would bias the resolution of recombination intermediates toward restoring the sequence of the invaded duplex rather than copying the invading strand. In this hypothesis we would expect that the total number of recombinant tetrads (PMS and conversions) would increase in MMR-deficient strains since heteroduplex would be detected as PMS. In fact, we observed no significant difference in the frequency of recombinant tetrads in single MMR mutants, which can repair loops, and double mutants, which cannot. Alternatively, some combination of MMR enzymes could recognize the formation of the loop and block the formation of the mature recombination intermediate. In this hypothesis, the inhibitory effect would be due to binding of MMR proteins to the heteroduplex loops.

We found that RS sequences inhibited conversion whereas MS sequences did not. The difference between RS and MS sequences might be due to the poor recognition of repetitive loops by MMR proteins. Two hypotheses can account for the poor binding of MMR proteins to MS loops. Repetitive DNA can adopt Z-DNA structure and has been shown to modify the binding affinity of recombination proteins (Bietet al. 1999). However, the observation that defects in mismatch repair strongly increase microsatellite instability suggests that some MMR proteins bind to repetitive sequences. Since we demonstrated that the MMR complex involved in large loop repair is different from the complex involved in conversion inhibition, we cannot reject the possibility that the MMR complex involved in conversion inhibition does not bind efficiently to a loop with repetitive sequence and thus does not promote conversion inhibition of the microsatellite insert. Alternatively, because RS secondary structures are a mixture of single-stranded regions and a short duplex with “mismatched” regions, they could be recognized by MMR proteins (Figure 1b). Such structures are very likely to form on single-stranded DNA with random sequences as long as 107 bp but have been shown to not form in CA or TG dinucleotide repeat tracts (Bietet al. 1999). A mixture of single-stranded regions and a short duplex in the heteroduplex could be sufficient for the observed inhibition. In that case, the lack of recombination inhibition by large repetitive MS loops would be due to the poor binding of MMR proteins on single-stranded DNA with repetitive sequence. Actually, in the two models discussed above, primary binding to the single-stranded DNA in the loop would trigger the secondary binding to the junction. We rather favor this last hypothesis since recombination was not inhibited by the 26-bp insert forming a perfect palindrome, but was reduced twofold when mismatches were present in the palindromic structure (Naget al. 1989). The fact that the decrease of recombination by heterologous inserts in our study was slightly higher may be due to the larger size of the insert (107 bp instead of 26 bp) or to differences between recombination at ARG4 and HIS4 loci.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

A summary of the various interactions of Msh2 and Pms1 proteins during recombination and heteroduplex repair.

The observation that msh2 mutations greatly increased repetitive sequence instability demonstrates that Msh2p complexes bind to dinucleotide repetitive loops formed by DNA polymerase slippage (Siaet al. 1997). Two of the most common types of trinucleotide repeats associated with large expansions in several human diseases are poorly corrected when located in heteroduplexes formed during yeast meiotic recombination. This inefficient correction has been proposed to reflect the ability of the tested repetitive sequences to form hairpin secondary structures within the loop (Mooreet al. 1999). This hypothesis is in agreement with the observation that palindromic sequences in heteroduplex DNA are poorly repaired (Naget al. 1989). Our results confirm that loops with repetitive sequences can be efficiently repaired as long as they do not form hairpins.

We demonstrate that RS and MS large loops are similarly recognized and repaired by at least two different MMR complexes (Figure 2). The efficient repair of large loops was suggested by the lack of PMS in all MMR-proficient strains. Similar observations have been reported by others, using larger inserts (Fink and Styles 1974; Fogelet al. 1981). We demonstrated that this repair involved MMR proteins, using msh2 pms1 double mutants. We found that the repair of large loops was not sequence sensitive and was almost as efficient as repair of the 2-bp RV mismatch (Table 2). However, the MMR complexes involved in the repair of large loops and of RV mismatch were different. Two complexes, containing either Pms1 or Msh2, are capable of repairing large loops. In contrast, the RV mismatch correction requires the activity of both Msh2 and Pms1, probably present in a single complex (Kolodner and Marsischky 1999). It is not clear how MMR complexes interact with large loops to repair it since none of the proteins has been found to bind to these structures. The simplest explanation would be that the MMR complexes would recognize DNA structures formed at the junction of homologous and heterologous duplex. Actually, it has been suggested that Msh2/Msh3 complexes could bind to branched DNA structures (Sugawaraet al. 1997).

Numerous experiments have shown that MMR inhibits mitotic recombination between homeologous sequences (Bailis and Rothstein 1990; Selvaet al. 1995; Dattaet al. 1997; Coicet al. 2000). Inhibition of meiotic recombination between homeologous chromosomes was also reported (Chamberset al. 1996). However, the results were contradictory concerning the involvement of the PMS1 mismatch repair gene. For example, pms1 mutation had no effect on recombination between 23% diverged sequences, but increased 11-fold in recombination between 10% diverged sequences (Dattaet al. 1997). It consistently inhibited recombination to a lesser degree than did msh2. On the other hand, pms1 mutation was reported to increase the frequency of recombination between chromosomes from closely related species (S. cerevisiae and S. paradoxus), the genomes of which were estimated to have diverged by ~8–20% (Chamberset al. 1996). In our experiments, pms1 and msh2 mutations seem to have a similar effect on inhibition of recombination. In the experiments with homeologous DNAs, all kinds of heteroduplexes were probably formed during recombination, containing high or low density of mismatches, small and large loops with palindromic sequences, and repetitive sequences. Since the recognition by mismatch repair of these different structures seems to involve different protein complexes, the effect of the MMR mutations on the different strains probably reflects the distribution of the different kinds of mismatches that can be formed during recombination.

Our results suggest that different MMR complexes act in the recombination process and in heteroduplex repair (Figure 2). The complex formed by Msh2 and Pms1, together with other proteins, probably recognizes small mismatches and repairs them. In the case of large single-stranded DNA loops included in the heteroduplex, the binding of the complex is likely to inhibit conversion of the region. The correction of the large loop heteroduplex requires two other different MMR complexes, containing either Msh2 or Pms1.