Role of Saccharomyces cerevisiae Msh2 and Msh3 repair proteins in double-strand break-induced recombination

Strains and Growth Conditions.

All strains were isogenic derivatives of either JKM111 (22) or tNR85 (23), both of which contain mutations of the HO cleavage site at MATa, so that only the introduced HO cutting site is cleaved. The two strains are not isogenic; however we have found that HO-induced recombination of the same substrate (plasmid pJF5; ref. 1) yielded a nearly identical efficiency and kinetics of DSB repair and the proportion of gene conversion events accompanied by crossing-over was the same (data not shown). Deletions of the RAD1, MSH2, and MSH3 were made by gene replacement (24) using the following constructs: rad1::LEU2 in plasmid pL962, from R. Keil; msh2::LEU2 in pRHB113 (25) and msh3::LEU2 (26). G. Marsischky and R. Kolodner constructed and provided strains RKY3111 (msh6::hisG) and RKY3112 (msh3::hisG) in tNR85. The GAL::HO gene was integrated at the ADE3 locus (27) in JKM146 and was carried on plasmid pFH800 (28) in the tNR85 derivatives. Induction of HO endonuclease was accomplished by adding galactose (2% final concentration) to cells grown in lactate-containing medium at 30°C, or by plating them on yeast extract/peptone-galactose.

Measurement of DSB Repair Efficiency.

To measure repair and retention of pFP122, pFP140, and pFP120, cells were plated on yeast extract/peptone/dextrose and yeast extract/peptone-galactose plates, and the resulting colonies replicated to plates lacking uracil, to determine plasmid retention. The percent plasmid retention, a measure of successful gene conversion, was calculated from the fraction of colonies retaining the plasmid on galactose medium divided by the same fraction on dextrose medium. Cell viability and deletion formation in the SSA assay were measured by plating cells on yeast extract/peptone/dextrose and replica plating cells to synthetic complete medium lacking tryptophan (29) to score for the retention of pFH800.

Recombination Substrates.

Plasmids pFP120, pFP122, and pFP140 were similar to the centromeric URA3-marked plasmids containing inverted copies of the Escherichia coli LacZ sequence used previously (30), except that they did not carry LEU2. Their structure is shown in Fig. 1A. In pFP122, the recipient LacZ sequence contains a 40-bp MATa cleavage site. The donor locus carries a mutant cut site differing by only a single bp substitution at position Z6 that prevents HO cleavage (31). Therefore, the DSB generates a cut plasmid whose ends are almost exactly identical to the uncleaved donor sequence that is used to repair the break by gene conversion. In pFP140, the recipient LacZ sequence carries a 60-bp MATa cleavage site, so that the DSB creates two ends with 30 bp of nonhomology that must be removed before gene conversion can begin. pFP120 contains a 40-bp cleavage site in the recipient LacZ sequence, and the donor lacZ copy is deleted for the 878-bp ClaI–SacI fragment; consequently, the HO cleavage site of the recipient is now surrounded by two large sequences (308 and 610 bp) without any homologous counterpart in the donor.

The SSA chromosomal substrates were derived from a triplication of URA3 sequences similar to tNS62 (8) except that the distal ura3–52 gene was completely deleted and replaced by a THR4 gene. The resulting structure on chromosome V has a 205-bp region of the 5′ end of the HindIII URA3 fragment duplicated on either side of an HO cleavage site. In Fig. 3 both duplicated regions were shown by sequencing to be identical to the URA3 sequence derived from strain +D4 (32), except that they both possess the 5-bp deletion found in strain FL100 (32) and a T to C substitution at nucleotide 57 of the +D4 sequence. In the strains shown in Fig. 4, the 205-bp segment to the left of the cleavage site is identical to that from FL100 and hence has seven single bp substitutions or frameshifts relative to the 205-bp region to the right of the cut site. In Fig. 5 similar substrates with 205, 415, or 1,170 bp of identical sequences were examined. Additional experiments were carried out using a LEU2-marked centromeric plasmid, pNSU208, similar to pJF6 (9), but carrying only 240 bp of directly repeated LacZ sequences.

DNA Analysis.

DNA was extracted at intervals after HO induction as described (8) and appropriate restriction endonuclease-digested DNA was analyzed on Southern blots. Gene conversions of plasmid pFP122 both with and without crossing-over were analyzed by Southern blots probed with LacZ sequences (30). SSA was analyzed by probing with a HindIII–BamHI DNA fragment centromere proximal to the URA3 gene (8). The efficiency of SSA was measured by dividing the intensity of the product band after 5 hr of induction by the intensity of the parental band prior to induction. Values were normalized to LEU2 or HIS3 sequences to control for the amount of total DNA per sample.

MSH2 and MSH3 Are Required to Remove Nonhomologous Ends During Gene Conversion.

The site-specific HO endonuclease can be expressed under the control of a galactose-inducible promoter (33), to initiate DSB-mediated recombination in plasmid or chromosomal substrates containing an HO recognition site (31). We have examined gene conversion in a centromeric plasmid where a 60-bp HO endonuclease recognition site was introduced into one of a pair of inverted repeated LacZ sequences (pFP140; Fig. 1A). The terminal 30 bp of the cut ends are not homologous to the intact donor. In wild-type strains, 69% of the cells complete recombination and retain the repaired plasmid. DSB repair is reduced 12-fold in the absence of Rad1 (Fig. 1B). These results are similar to those reported in similar plasmids with a 117-bp HO recognition site (9).

Efficient DSB repair also depends on Msh2 and Msh3. Completion of HO induced gene conversion was reduced 6-fold in msh2 or 3-fold in a msh3 deletion (Fig. 1B). A rad1 msh2 and a rad1 msh3 double mutant was reduced 7- to 10-fold. Similar results were found with plasmid pFP120 (Fig. 1A), with nonhomologous ends of 308 bp on one side and 610 bp on the other. msh2, msh3, and rad1 reduced gene conversion by 33-, 20-, and 23-fold, respectively, relative to a wild-type control (Fig. 1B).

In contrast, msh2, msh3, and rad1 had no effect when the donor also contained a (mutant) HO cut site and the HO-cleaved ends of the recipient were therefore homologous to the donor, except for a 1 bp difference. A similar plasmid (pFP122; Fig. 1A) was constructed, where HO-cut ends are almost perfectly homologous to the donor DNA, by placing a 40-bp HO cut site in the recipient copy of LacZ and a mutant 40-bp HO cutting site, differing by a single bp substitution, into the donor. Neither Rad1, Msh2, nor Msh3 was required for efficient recombination (Fig. 1B). We conclude that Msh2 and Msh3 are required for the removal of nonhomologous tails in conjunction with the Rad1/Rad10 endonuclease and are not required when the DNA ends match their donor template. We have only used plasmid-borne substrates to analyze the requirement for MSH2 and MSH3 on gene conversion. However, using chromosomal substrates we have also demonstrated that RAD1 is required when the ends are nonhomologous (data not shown) and we presume that MSH2/MSH3 would also be required for chromosomal events.

Lack of Effect of MSH2 and RAD1 on Later Steps in Gene Conversion.

Since rad1 and msh2 had little or no effect on viability when the HO-cut ends were homologous to the donor, we more closely examined these strains by Southern blots of DNA extracted at intervals after HO induction, to discern whether these mutations had an effect on the kinetics of recombination or on the proportion of gene conversion events associated with crossing-over. Southern blots for pFP122 in wild-type, msh2, and rad1 strains (Fig. 2) show the kinetics of appearance and disappearance of HO-cleaved DNA fragments and the time of appearance of the two PstI restriction fragments representing gene conversions associated with crossing-over. PhosphorImager analysis of these Southern blots showed that the proportion of gene conversions associated with crossing-over was ≈23.5% in the wild-type strain. This proportion was not significantly affected by msh2 (19%) or rad1 (20.5%). Moreover, there was no change in the kinetics of appearance of the crossover products in the msh2 or rad1 strains. Thus, Msh2 and Rad1 are required at the same step: to remove nonhomologous single-stranded DNA from the end of recombining DNA. Neither msh2 nor rad1 affects any of the other steps that should be common to all gene conversion events (e.g., DNA pairing, strand invasion per se, branch migration, or Holliday junction resolution).

Role of Mismatch Repair Genes in SSA.

When a DSB is flanked by homologous regions, repair can occur by nonreciprocal recombination leading to a deletion. Most of these events arise by SSA (Fig. 3A) but could also occur by “one-ended” recombination events, in which one DNA end could invade the other flanking homologous region before it became single-stranded (1, 9). In contrast to efficient deletion formation between 205-bp URA3 sequences in wild-type cells (Fig. 3B), deletions were greatly reduced in a rad1 derivative (Fig. 3B). The effect of msh2 and msh3 deletion mutations was very similar to rad1 (Fig. 3B). Cell viability (those that repaired the DSB) went from 77% in the wild type to 3.7% in msh2 and 5.4% in msh3, very similar to a rad1Δ strain (1.0%).

Similar experiments were carried out with a deletion of MSH6, whose gene product has been shown to form heterodimers with Msh2, but which has been implicated principally in the repair of single bp substitutions rather than small loops in heteroduplex DNA (16–18). In these experiments (Fig. 3C), SSA was monitored on a centromeric plasmid containing 240 bp of LacZ homology, in direct orientation, flanking the DSB, similar to plasmids described (1, 9). As shown in Fig. 3C, msh6 does not impair deletion formation by SSA, while msh3 does. A similar result has been obtained using chromosomal substrates identical to those shown in Fig. 3A (data not shown). This result argues that Msh2/Msh3 recognizes nonhomologous tails during recombination in a fashion similar to their recognition of insertion/deletion heterologies, whereas Msh2/Msh6, which mainly recognizes bp substitutions, plays no significant role in SSA.

Mismatch repair of bp substitutions and frameshifts requires not only Msh2, Msh3, and Msh6, but also the Mlh1 and Pms1 proteins (19, 20). However, while both MSH2 and MSH3 are required to remove nonhomologous DNA tails, PMS1 and MLH1 are not (Fig. 4). We first showed this for pms1 when the 205-bp repeats were perfectly homologous (data not shown). It was then of interest to know if SSA would be differently affected if the two repeats were divergent, since previous studies had shown that homologous recombination was improved by mutations in msh2, msh3, and pms1 (25, 26, 34). We therefore examined SSA in substrates that carried seven heterologies in the 205-bp chromosomal regions flanking the DSB. SSA was reduced in the wild-type strain to ≈25% of the value for fully homologous sequences (Fig. 3B vs. Fig. 4). Neither pms1 nor mlh1 affected the efficiency of deletion formation suggesting that (i) they are not required to excise nonhomologous tails and (ii) they are not responsible for the discouragement of annealing by 3% heterology. In contrast, the absence of msh2 caused a dramatic decrease in deletion formation (Fig. 4). These results argue that MSH2 and MSH3 play a role in the removal of nonhomologous tails from the annealed intermediate that is distinct from their roles in mismatch correction or heteroduplex DNA formation. Whatever improvement msh2 might have caused in increasing recombination between diverged substrates was overwhelmed by its very strong negative effect on the completion of SSA.

Distinguishing the Roles of RAD1 and MSH2/MSH3 During SSA.

Msh2 and Msh3 might act in several different ways in removing nonhomologous DNA tails. Msh2/Msh3 may recognize and bind to branched intermediates and present them to Rad1/Rad10 for cleavage. They might also stabilize initially unstable strand annealing intermediates, allowing Rad1/Rad10 more time to act. Finally they might form a complex with Rad1/Rad10 to increase its endonuclease activity. Further experiments have shown that the requirement for Msh2 and Msh3 in SSA depends on the length of the regions that anneal. Three SSA substrates were constructed on chromosome V in which the lengths of the single-stranded DNA tails that must be removed remained constant, but the length of the annealed region increased from 205 bp to 415 bp or to 1,170 bp (Fig. 5). The absence of RAD1 had the same severe effect on deletion formation, regardless of the length of the flanking homologous regions. In contrast, while the requirement for MSH2 and MSH3 was as great as for RAD1 with 205-bp flanking regions, the absence of the mismatch repair proteins was less profound with longer flanking regions. With 1,170-bp flanking homology, msh2 and msh3 reduced SSA only 25% relative to wild type. The significance of this difference will be discussed below.