Meiotic recombination in Saccharomyces cerevisiae involves the formation of heteroduplexes, duplexes containing DNA strands derived from two different homologues. If the two strands of DNA differ by an insertion or deletion, the heteroduplex will contain an unpaired DNA loop. We found that unpaired loops as large as 5.6 kb can be accommodated within a heteroduplex. Repair of these loops involved the nucleotide excision repair (NER) enzymes Rad1p and Rad10p and the mismatch repair (MMR) proteins Msh2p and Msh3p, but not several other NER (Rad2p and Rad14p) and MMR (Msh4p, Msh6p, Mlh1p, Pms1p, Mlh2p, Mlh3p) proteins. Heteroduplexes were also formed with DNA strands derived from alleles containing two different large insertions, creating a large “bubble”; repair of this substrate was dependent on Rad1p. Although meiotic recombination events in yeast are initiated by double-strand DNA breaks (DSBs), we showed that DSBs occurring within heterozygous insertions do not stimulate interhomologue recombination.
IN Saccharomyces cerevisiae, the exchange of genetic information during meiotic recombination is a highly coordinated process that involves introduction of a double-strand break (DSB) and repair of that break with sequences derived from the homologous chromosome (Roeder 1997; Paques and Haber 1999). The enzyme responsible for DSB formation is the topoisomerase II-related protein Spo11p (Keeneyet al. 1997). Following DSB formation, the 5′ ends of the break are resected in a process requiring the Rad50p/Mre11p/Xrs2p complex, generating long, single-stranded DNA tails (Alaniet al. 1990; McKee and Kleckner 1997; Nairz and Klein 1997; Prinzet al. 1997). In the modified version of the double-strand break model (Szostaket al. 1983; Sunet al. 1991), the 3′-ended tails invade the homologous chromosome, creating a heteroduplex DNA molecule (Figure 1). The displaced strand is used as a template for repair of the break. Ligation of the products results in double Holliday junction intermediates (Schwacha and Kleckner 1994, 1995) that can be resolved as crossovers or noncrossovers.
If sequence heterologies are included in the heteroduplex region (either base pair changes or insertion/deletion heterologies), mismatches or unpaired loops will be generated (Figure 1). The repair of these mismatched templates can lead to gene conversion or restoration of Mendelian segregation (Peteset al. 1991). Failure to repair the mismatch will result in a post meiotic segregation (PMS) event. If the heterozygous marker involves an auxotrophic mutation, a PMS event is readily visualized as a sectored spore colony.
One aim of this study is to investigate whether large sequence heterologies can be incorporated into heteroduplexes during meiotic recombination in S. cerevisiae. Insertions (often involving transposable elements) and deletions are common in eukaryotic genomes. Gene conversions of large (>5 kb) heterozygous insertions and deletions have been observed (Fink and Styles 1974; Fogelet al. 1981; McKnightet al. 1981; Pukkilaet al. 1986; Vincent and Petes 1989), although these conversions have often been attributed to repair of a gapped DNA intermediate rather than repair of a loop within a heteroduplex (Szostaket al. 1983; Tranet al. 1996). Clikeman et al. (2001) recently showed that 2- to 3-kb heterologies can be incorporated into heteroduplexes during mitotic recombination events in yeast. As described below, we find that insertions of up to 5.6 kb can be incorporated into heteroduplex DNA, indicating that meiotic gene conversion of large heterologies can involve DNA loop repair.
A second aim of this study is to identify the proteins involved in the meiotic repair of large, unpaired DNA loops. In a previous study (Kirkpatrick and Petes 1997), we showed that meiotic repair of a small (26-base) DNA loop involved Msh2p, an enzyme involved in mismatch repair (MMR), and Rad1p, an enzyme involved in nucleotide excision repair (NER). In both mitotic and meiotic cells, the MMR enzymes function to repair DNA mismatches generated either by errors made during DNA replication or by heteroduplexes formed between DNA strands derived from nonidentical alleles (Crouse 1998; Harfe and Jinks-Robertson 2000). In S. cerevisiae, base-base mismatches are corrected by a complex containing the MutS homologues, Msh2p and Msh6p, and the MutL homologues, Mlh1p and Pms1p (Siaet al. 1997a; Kolodner and Marsischky 1999). In mitotic cells, DNA loops between 1 and 16 bases are corrected by a complex in which Msh3p is substituted for Msh6p (Siaet al. 1997b). Although these two MMR complexes have the major roles in the repair of DNA mismatches, other complexes that include the MutL homologues Mlh2p and Mlh3p are also involved in the repair of very small DNA loops (Flores-Rozas and Kolodner 1998; Harfe and Jinks-Robertson 2000). In addition, the frequency of large (~100 bp) deletions was increased in strains with msh3 or mlh2 mutations (Harfeet al. 2000); such deletions could reflect failure to repair large DNA loops formed by DNA polymerase slippage.
In addition to their roles in the repair of DNA mismatches, the MMR enzymes are involved in a number of other cellular processes. First, these enzymes reduce the frequency of recombination between diverged DNA sequences (Selvaet al. 1995; Dattaet al. 1996; Hunteret al. 1996; Nicholsonet al. 2000). Second, Msh2p and Msh3p are required for the removal of nonhomologous “tails” of DNA during single-strand annealing recombination events (Saparbaevet al. 1996; Sugawaraet al. 1997). Third, some of the MMR proteins (Msh4p, Msh5p, Mlh1p, and Mlh3p) promote meiotic crossing over in yeast (Ross-Macdonald and Roeder 1994; Hollingsworthet al. 1995; Wanget al. 1999).
The NER enzymes are required for the efficient repair of a variety of types of DNA damage (reviewed by Prakashet al. 1993). Following recognition of a damaged substrate (usually a photoproduct) by Rad14p, the DNA surrounding the lesion is unwound by a helicase to form a small “bubble.” The damaged strand is removed by single-strand incisions made both 5′ (Rad1/10p) and 3′ (Rad2p) to the lesion. Repair synthesis of the excised region completes the reaction.
In S. cerevisiae, Rad1p and Rad10p are also required in single-strand annealing (Fishman-Lobell and Haber 1992) and the meiotic repair of small DNA loops (Kirkpatrick 1999). Mitotic recombination between inverted repeats containing multiple small heterozygous insertions is inhibited by the Rad1p and Rad10p (Nicholsonet al. 2000). In Schizosaccharomyces pombe, homologues of the S. cerevisiae Rad1p, Rad10p, and Rad14p function in short-patch meiotic repair of DNA mismatches (Flecket al. 1999). The mei9 gene product, homologous to Rad1p, is required for meiotic mismatch repair and crossover resolution in Drosophila (Carpenter 1979, 1982; Sekelskyet al. 1995). In the studies described below, we show that the MMR proteins, Msh2p and Msh3p, and the NER proteins, Rad1p and Rad10p, are involved in the meiotic repair of large DNA loops.
MATERIALS AND METHODS
Strains: All haploid yeast strains are derivatives of AS13 or AS4 (Stapleton and Petes 1991). The constructions and genotypes of haploid and diploid strains are described in Tables 1 and 2, respectively. All alterations were introduced by one-step (Rothstein 1983) or two-step (Shermanet al. 1982) transplacements. One-step transplacements using PCR-generated DNA fragments were done as described in Wach et al. (1994). Primers used to generate PCR cassettes are shown in Table 3. Deletions made with the kanMX4 or hygB cassettes (Wachet al. 1994; Goldstein and McCusker 1999) removed most or all of the coding sequence. All constructions were verified by PCR, Southern analysis, or DNA sequencing. To generate diploids homozygous for mutations in MSH2, MSH3, MSH6, PMS1, and MLH1 that have a mutator phenotype, we mated haploids overnight and sporulated the resulting diploids the next day without purification of the diploids. This procedure prevents accumulation of mutations that would reduce spore viability.
Genetic techniques: Standard media and genetic methods were used (Shermanet al. 1982). Plates for kanMX4 and hygB selection were YPD + 150 mg/liter geneticin and YPD + 300 mg/liter hygromycin B, respectively. We selected ura3 strains on medium containing 1 gm/liter 5-fluoroorate (Boekeet al. 1984). As in our previous studies (Fanet al. 1995), diploid strains were sporulated at 18° on plates containing 1% potassium acetate, 0.1% yeast extract, 0.05% dextrose, 2% agar, and supplemented with 6 μg/ml of adenine. Sporulated cells were dissected onto YPD and, following full colony growth at 30°, spore colonies were replica plated to appropriate omission and drug-containing media to score the segregating markers. Spore colonies were then examined for sectored growth patterns by microscopy.
Physical analysis of double-strand breaks: We performed Southern analysis by standard methods (Maniatiset al. 1982), using a PCR-generated probe to HIS4. Primers used to generate the probe were f: 5′ CCACTTGGAGACCATGTCTTG and r: 5′ CAATGGAACATAGAGCTTGAGTG, resulting in a fragment containing HIS4 sequences from +690 to +1768. For most of the DSB measurements, strains with the rad50S mutation were grown in liquid sporulation media at room temperature as described previously (Nag and Petes 1993). In some experiments, we also isolated and analyzed DNA derived from cells sporulated at 18° on plates. These different sporulation conditions had no effect on the relative levels of DSBs (wild-type vs. mutant chromosome) as assayed physically or genetically (data not shown). Levels of DSBs were quantitated using the PhosphorImager (Molecular Dynamics, Sunnyvale, CA) and Image-Quant software.
Statistical analysis: Comparisons were made using the Fisher exact test with two-tailed P values or by chi-square analysis (for comparisons involving more than two experimental parameters). Results were considered statistically significant when P < 0.05. Instat 1.12 (GraphPad Software) was used for statistical analysis.
Experimental rationale: In our genetic background, heterozygous markers near the 5′ end of HIS4 have an extraordinarily high rate of non-Mendelian segregation, ~50% of unselected tetrads (Naget al. 1989). This high frequency of aberrant segregation reflects a high level of meiosis-specific DSBs located about 200 bp upstream of the HIS4 initiating codon (Nag and Petes 1993; Fanet al. 1995). Heteroduplexes initiated in the HIS4 upstream region are efficiently extended through the HIS4 coding region, a distance of ~2.4 kb (Detloffet al. 1992; Porteret al. 1993). Heterozygous mutations within the coding sequences lead to mismatches in the heteroduplex; repair of these mismatches results in gene conversion or restoration events, whereas failure to repair the mismatches results in PMS events (Figure 1).
In previous studies, we found that small (26 bp) heterozygous nonpalindromic insertions at position +469 in the HIS4 coding sequence had high levels of aberrant segregation (26% gene conversion and 4% PMS tetrads; Naget al. 1989; Kirkpatrick and Petes 1997). Mutations in the RAD1 or MSH2 gene increased the frequency of PMS events and decreased the frequency of gene conversion events, indicating that Rad1p and Msh2p were involved in the repair of 26-base DNA loops (Kirkpatrick and Petes 1997). In this study, we extend our analysis of the gene products required for the efficient repair of the 26-base loop. We also examined the meiotic segregation patterns of larger heterozygous insertions (1.1, 1.5, and 5.6 kb) at the same position in HIS4.
Meiotic repair of a 26-base loop: Aberrant segregation patterns of strains heterozygous for a 26-bp insertion in HIS4 are shown in Table 4. To determine whether various mutations have a significant effect on repair of the 26-base loop, we compared the relative numbers of tetrads with PMS and conversion in the wild-type and mutant strains by Fisher's exact test. By this criterion, in addition to the rad1 and msh2 mutations reported previously (Kirkpatrick and Petes 1997), strains homozygous for mutations in rad10 (DTK286) and msh3 (DTK488) had a significant repair deficiency (P values of 0.005 and <0.0001, respectively). The rad1 msh2 and rad1 msh3 double mutant strains (DTK230 and DTK500, respectively) had PMS frequencies similar to those observed in the single mutant rad1, msh2, and msh3 strains, indicating that these gene products are likely to function in a single repair pathway. The requirement for both Rad1p and Rad10p, which function as a heterodimeric structure-specific endonuclease, suggests that this endonuclease activity is required for efficient loop repair; several other NER enzymes, such as Rad2p and Rad14p, however, are not required for loop repair (Kirkpatrick and Petes 1997).
Neither mlh1 (DTK320) nor pms1 (DTK309) mutations had a significant effect on repair of the 26-base loop. Similarly, mutations in the MutS homologues Msh6p (DTK494) and Msh5p (DTK498) had no significant effect on loop repair. The msh4 mutation, however, had a small, but significant effect (P value of 0.005). Since epistasis analysis indicates that Msh4p and Msh5p function in a single pathway in crossover resolution (Ross-Macdonald and Roeder 1994; Hollingsworthet al. 1995), the differential effects of these two proteins on loop repair is unexpected. Mutations in EXO1 (DTK308) had a similar subtle defect in repair of the 26-base loop (Kirkpatricket al. 2000). It is possible that Msh4p and Exo1p are involved in a minor pathway of loop repair. Since none of the mutations examined in our study eliminate gene conversion, it is likely that there are pathways of loop repair that are independent of the Rad1p/Rad10p/Msh2p/Msh3p pathway.
The RAD27 gene encodes the S. cerevisiae homologue of the mammalian FEN-1 exo/endonuclease (Johnsonet al. 1995; Reaganet al. 1995). Rad27p is involved in the removal of single-stranded DNA “flaps” and the processing of Okazaki fragments during replication (Harrington and Lieber 1994a,b; Muranteet al. 1994). Deletion of RAD27 (DTK246) had no effect on repair of the 26-base loop (Table 4). Finally, we examined the role of the non-homologous end-joining repair pathway in loop repair. A strain homozygous for a deletion of HDF1/YKU70 (DTK341), a homologue of a component of the mammalian Ku DNA end-binding complex (Feldmann and Winnacker 1993; Troelstra and Jaspers 1994), had no effect on loop repair (Table 4).
Analysis of strains heterozygous for a 1.1-kb URA3 insertion: The his4::U1.1a allele was constructed by introduction of a 1.1-kb fragment containing the URA3 gene into the HIS4 coding sequence at position +469, the same position as the 26-bp insertion described above. Results from tetrad analysis of strains heterozygous for this insertion are shown in Table 5.
In the wild-type background (HMY100), we observed a high level of gene conversion events (22%) and no PMS events. One interpretation of this result is that heteroduplexes cannot accomodate the 1.1-kb insertion and, consequently, gene conversion occurs through a different pathway (for example, gap repair). Alternatively, heteroduplexes containing the heterozygous insertion are formed, and the resulting loop is repaired with complete efficiency. The second possibility is supported by the observation that the rad1 strain (HMY49) heterozygous for the 1.1-kb URA3 insertion had substantial rates of PMS tetrads (Table 5). Because the URA3 insertion confers the ability to grow on medium lacking uracil, PMS events were visualized as sectored colonies with His−Ura+ and His+Ura− halves (Figure 2a). The rad10 mutation had an effect on loop repair similar to that of the rad1 mutation (comparison of strains HMY49 and HMY151). The other mutations that resulted in PMS events for the 1.1-kb URA3 insertion were msh2 (HMY107) and msh3 (HMY113). The msh2 and msh3 mutations had significantly (P < 0.02) smaller effects than the rad1 and rad10 mutations. Since the efficiency of loop repair for the URA3 insertion is about the same in the rad1 msh3 double mutant (HMY184) and the rad1 single mutant (HMY49), Rad1p and Msh3p appear to act in a single pathway of loop repair. Mutations in msh4, msh5, msh6, mlh1, pms1, mlh2, mlh3, rad2, rad14, exo1, and rad27 had no effect on the repair of the 1.1-kb loop (Table 5).
Although the rad27 strain (HMY150) exhibited no defect in repair of the 1.1-kb loop, we observed a significant decrease in the level of aberrant segregants in this background (P < 0.0001). This result is specific for the his4::U1.1a allele, since we did not observe a similar effect in a rad27 strain heterozygous for the 26-bp insertion (Table 4) or on other heterozygous point mutations in HMY150 (data not shown). It is possible that the flap endonuclease plays a role in heteroduplex formation between alleles with large insertion/deletion heterologies.
In most models for meiotic recombination (Figure 1), DNA replication is involved in formation of heteroduplexes and in repair of DNA mismatches. S. cerevisiae has a number of nonessential DNA polymerases including β (encoded by POL4), η (encoded by RAD30), and ζ (catalytic subunit encoded by REV3; Woodgate 1999). We examined meiotic segregation of the his4::U1.1b allele (identical to his4::U1.1a except for a single T to A base pair change in the insertion, 74 bp downstream of the URA3 stop codon) in strains homozygous for null mutations in these three DNA polymerase genes (Table 5). Although the rad30 and rev3 mutations had no significant effect, the pol4 mutation significantly (P = 0.001) reduced the frequency of aberrant segregation of his4::U1.1b. We also analyzed the effects of the same three DNA polymerase mutations on the frequency of aberrant segregation of a point mutation located in the HIS4 initiating codon. No significant effect was observed for any of the three mutations (data not shown) and the pol4 mutation did not significantly affect the frequency of aberrant segregation or the repair of the 26-base loop (Table 4). In addition, DSB formation at the HIS4 hotspot (monitored in a rad50S strain) was unaffected by the pol4 mutation (data not shown). Although these results suggest that DNA polymerase β may have a role in heteroduplex formation across large heterologies, we found that the pol4 mutation did not significantly reduce the aberrant segregation frequency of a different larger insertion (as described below). Our results are in contrast to those of Leem et al. (1994) who found elevated levels of both DSBs and meiotic recombination in pol4 strains.
Genetic analysis of strains heterozygous for a 1.5-kb kanMX insertion or a 5.6-kb URA3 insertion in HIS4: To generalize the results obtained with the 1.1-kb URA3 insertion, we also examined strains heterozygous for other insertions in HIS4. The his4::k1.5 allele was generated by inserting the 1.5-kb kanMX4 gene (Wachet al. 1994) at position +469 of the HIS4 coding sequence. Strains with this insertion are His−, but resistant to geneticin. A wild-type strain heterozygous for this insertion (HMY190) had 12% conversion tetrads, but no PMS tetrads (Table 6). In a derivative homozygous for the rad1 mutation (HMY219), we observed 12% conversion events and 17% PMS events. This result confirms our conclusion that heteroduplexes can be formed that include large heterologies and that the repair of the resulting DNA loop is, at least partly, Rad1p dependent.
We also constructed strains heterozygous for a 5.6-kb insertion of URA3 at position +469 of the HIS4 coding sequence (his4::U5.6). A wild-type strain heterozygous for this insertion (HMY192) had 13% gene conversion and no PMS tetrads (Table 6). In strains with mutations in either rad1 (HMY203) or msh3 (HMY199), however, we observed a small number of PMS tetrads (five tetrads in the two strains). Thus, we conclude that heterologies as large as 5.6 kb can be included within heteroduplexes, although the efficiency of heteroduplex formation with these very large loops is reduced. The pol4 mutation did not significantly reduce the aberrant segregation frequency of his4::U5.6 (comparison of HMY192 and HMY258 in Table 6).
DSB formation in strains with heterozygous large insertions: In wild-type strains heterozygous for his4::U1.1a, his4::U1.1b, his4::k1.5, and his4::U5.6, we found that 2:6 gene conversion events were more common than 6:2 conversions. One interpretation of such a bias is that the chromosome containing the insertion has fewer DSBs at the upstream HIS4 hotspot than does the chromosome lacking the insertion (Figure 1). Consequently, we measured DSB formation in rad50S derivatives of strains heterozygous for his4::U1.1b (JG30) and his4-k1.5 (HMY-210); rad50S strains accumulate unprocessed DSBs, simplifying their quantitation (Caoet al. 1990).
In addition to the DSB associated with the upstream HIS4 hotspot (Fanet al. 1995), the JG30 strain had a novel minor DSB band that mapped to the 5′ end of the URA3 insert (Figure 3a). The frequency of the insert-associated DSB was about one-third of that associated with the wild-type hotspot. The hotspot-associated DSB on the chromosome with HIS4 was ~1.3-fold more intense than the upstream DSB on the his4::U1.1b-containing chromosome (average of three experiments). The observed conversion bias in HMY98 was 2-fold (Table 5). Thus, the difference in DSB formation on the two homologues accounts for much of the conversion bias observed in HMY98, assuming that DSB formation within the insertion does not contribute to gene conversion (an assumption discussed below).
A very strong DSB was also found within the kanMX4 insertion (HMY210; Figure 3b). The activity of the upstream HIS4 hotspot on the mutant chromosome was reduced 3.6-fold when compared to the upstream hotspot on the wild-type chromosome (average of three experiments). This bias in DSB formation accounts for part, although not all, of the 11-fold excess of the 2:6 class seen in the RAD50 companion strain HMY190 (Table 6). We also found DSBs associated with the insertion of the his4::U5.6 allele in the rad50S strain HMY211 (data not shown).
It is likely that the bias in the activities of the upstream hotspots reflects a competitive interaction between the DSBs within the insertions and the upstream DSB. We and others previously found that if two DSB sites are found close together on the same chromosome, the activity of each individual hotspot is reduced when compared to their “solo” activities (Wu and Lichten 1995; Xu and Kleckner 1995; Fanet al. 1997). Consistent with this hypothesis, we found that the strong DSB in the kanMX4 insertion had a larger effect than did the weak DSB in the 1.1-kb URA3 insertion. In addition, if the insert-associated DSBs do not contribute to gene conversion events at the HIS4 locus, the effect of the DSB competition would be to generate gene conversion disparity.
Since homology is required to initiate heteroduplex formation, we expected that DSBs within the insertions were unlikely to contribute to gene conversion events at HIS4 that included the insertions. To test this hypothesis genetically, we constructed strains heterozygous for his4::U1.1a (HMY157) or his4::k1.5 (HMY234) in which the HIS4 upstream hotspot had been inactivated by mutations of a Rap1p binding site (the his4-51 allele; Devlinet al. 1991; Whiteet al. 1991). The results from this analysis are in Table 7.
Deletion of the wild-type hotspot reduced aberrant segregation of his4::U1.1a from 22% (HMY100) to 3% (HMY157). The same deletion reduced aberrant segregation of his4::k1.5 from 13% (HMY190) to 4% (HMY234). In addition, those conversion events that occurred in HMY157 and HMY234 were not biased in favor of the 6:2 class. Such a bias would be expected if DSB formation within the insertion stimulated interhomologue recombination (Figure 1). The deletion of the wild-type hotspot also reduced crossovers between HIS4 and LEU2 in HMY157 and HMY234 compared to HMY100 and HMY190, shortening the average map distance from ~32 to 20 cM. In previous studies in which the effect of the hotspot deletion was monitored in the absence of a large insertion at HIS4, a similar reduction in crossovers was observed (Whiteet al. 1991). We also observed a reduction in aberrant segregation and HIS4-LEU2 crossovers in strain HMY239, a strain heterozygous for his4::k1.5 and homozygous for his4-51 and rad1 (Table 7). This reduction was less than that observed for HMY234, presumably reflecting the recovery of PMS events that would have been repaired as restoration events (Figure 1) in the RAD1 strain.
To be sure that strains without the upstream HIS4 hotspot had insertion-associated DSBs, we examined DSB formation in rad50S derivatives of HMY157 and HMY234 (HMY183 and HMY237, respectively). Since DSBs were observed (Figure 3c), we conclude that DSBs that occur within heterozygous insertions do not effectively stimulate gene conversion between homologous chromosomes. It is likely that such DSBs are repaired by genetically silent sister-strand recombination.
Since the disparity of gene conversion in strains heterozygous for insertion-generated mutations is reduced in rad1 strains, we also analyzed DSB formation in rad50S derivatives of the rad1 strains, HMY49 (HMY229) and HMY219 (HMY230), heterozygous for the his4::U1.1a and his4::k1.5 insertions, respectively. The patterns of DSBs in these rad1 strains were identical to those seen in the RAD1 strains (data not shown). Therefore, we conclude that Rad1p does not influence patterns of DSB formation.
Ectopic recombination between the his4::U insertions and the ura3 locus on chromosome V: Although it is likely (as discussed above) that most DSBs within the insertions are repaired by sister-strand interactions, we also observed tetrads that had ectopic gene conversion events between the his4::U1.1a or his4::U5.6 alleles on chromosome III and the mutant ura3 gene on chromosome V. Two types of spore colonies are indicative of ectopic recombination events: class 1, in which part or all of the colony is His− Ura−, and class 2, in which part or all of the colony is His+ Ura+. There were a total of 12 class 1 spore colonies in strains that were heterozygous for his4::U1.1a and the frequency of tetrads with such colonies was 1.4% or less in all strains; there was only a single class 2 tetrad in these strains. Strains heterozygous for the his4::U5.6 allele had a higher frequency of ectopic events with an average frequency of 3.8% class 1 tetrads and 1.5% class 2 tetrads.
In all class 1 spore colonies examined (12 of 12), the his4 gene retained the URA3 insertion. The simplest explanation of this event is that mutant information derived from the chromosome V ura3 gene was donated to the wild-type gene located in HIS4 on chromosome III. In confirmation of this explanation, in 12 of 12 His− Ura− spore colonies examined, we showed that the mutant ura3 insertion in his4 had the same nonsense mutation (G to T change at position 721) as the mutant genes on chromosome V. In class 2 spore colonies, the HIS4 gene lacked an insertion. The simplest explanation for such colonies is that ectopic conversion between one of the his4::U alleles and one of the mutant chromosome V ura3 genes resulted in a wild-type URA3 gene on one copy of chromosome V. Cosegregation of this allele with the HIS4 allele lacking an insertion would produce a His+ Ura+ spore colony or sector. In summary, there was a fairly high frequency of meiotic ectopic recombination events in strains with his4::U alleles, as expected from previous studies (Jinks-Robertson and Petes 1985; Lichtenet al. 1987). Since these ectopic events do not directly contribute to aberrant segregation at the HIS4 locus, they were not included in the data shown in Tables 5 and 6.
Formation and repair of bubble structures: Diploid strains were constructed in which each homologue contained an insertion at the same position in HIS4, but the insertions had different sequences (his4::U1.1a and his4::k1.5). Heteroduplexes that include these insertions would be expected to form a large “open” bubble (Figure 2b). Spore colonies diagnostic for such a heteroduplex would be expected to be His−, but have GenR Ura− and GenS Ura+ sectors (Figure 2b).
In the wild-type background (HMY189), we observed only 6% aberrant tetrads, and all were gene conversion events (Table 6). Deletion of RAD1 (HMY215) resulted in a dramatic increase in aberrant segregants (26%) with the recovery of many PMS tetrads (73% of the aberrant tetrads). The msh3 deletion (HMY222) also elevated aberrant segregation frequencies, but in contrast to the results obtained with HMY215, very few PMS events were recovered (25% aberrant with 5% PMS/Ab; Table 6). Possible explanations for this difference will be discussed further below.
It has been reported previously that strains with mutations in MLH1 and PMS1 have elevated frequencies of PMS events involving the mating type locus (Wanget al. 1999). PMS events involving this locus would be expected to reflect formation of a bubble of ~700 bp. In our genetic background, the mlh1 mutation had no effect on bubble repair (HMY198, Table 6). It is possible that the difference in our results and those of the previous study reflect size or sequence differences in the insertions. Alternatively, the colonies sectored at the mating type locus might reflect an Mlh1p-Pms1p-dependent increase in mating type switch as a consequence of mitotic gene conversion (Chen and Jinks-Robertson 1999).
The main conclusions from this study are: (1) heteroduplexes formed during meiotic recombination can include large (5.6 kb) insertions; (2) heteroduplexes can be formed between alleles that include two different large insertions; (3) the efficient repair of DNA loops formed during meiotic recombination requires Rad1p, Rad10p, Msh2p, and Msh3p; (4) gene conversion events that involve large insertions usually duplicate, rather than delete, the insertions; and (5) DSBs within insertions do not stimulate recombination between homologues.
Accommodation of heterologies within heteroduplexes: The finding that meiotic heteroduplexes can include regions of heterology as large as 5.6 kb is somewhat surprising. There are three related mechanisms that could result in heteroduplex formation through a large heterology during meiotic recombination. First, there could be extensive degradation of the broken DNA ends, followed by invasion of the resulting single-stranded DNA into the other homologue (Figure 1). The inclusion of the heterology in the heteroduplex requires either a single strand to migrate through the inserted sequence in a homoduplex, or the ability of the single strand to invade the homoduplex simultaneously on both sides of the insertion. It should be noted, however, that the ability of Escherichia coli strand exchange proteins to bypass 1-kb heterologies in vitro is limited, even in the presence of the RuvAB complex (Iypeet al. 1994; Morelet al. 1994; Adams and West 1996). In the second model, the resection resulting in heteroduplex formation is limited to the DNA upstream of the insertion; the resulting Holliday junction is translated through the heterology by branch migration. The genetic evidence indicating that branch migration in vivo in yeast is limited or nonexistent (Fogelet al. 1981; Peteset al. 1991) argues against this model. In addition, formation of a heteroduplex that includes a bubble structure (Figure 2b) is unlikely to occur by branch migration. In our favored model, heteroduplex formation is driven by DNA replication rather than by branch migration (Figure 4). This model is similar to synthesis-dependent strand annealing models of recombination (Paques and Haber 1999), except that there is no requirement for migration of a D-loop.
In many previous yeast studies (reviewed in Peteset al. 1991), efficient meiotic conversions of large heterozygous insertions have been observed. Our results demonstrate that at least some of these conversion events are likely to reflect Rad1p/Rad10p/Msh2p/Msh3p-dependent repair of large DNA loops within heteroduplexes. We cannot rule out the alternative possibility that conversion in the presence of these proteins occurs through gap repair rather than the repair of a DNA loop in a heteroduplex. The observation of large (2 kb) DNA loops in heteroduplexes generated during mitotic recombination in RAD1 strains (Clikemanet al. 2001), however, argues against this interpretation.
Proteins required for DNA loop repair: We examined the effects of many different mutations on the efficiency of repair of the 26-base and 1.1-kb DNA loops (Tables 4 and 5). Repair of both types of loops involved the Rad1p, Rad10p, Msh2p, and Msh3p. These same four proteins are required for the removal of nonhomologous ends during certain types of mitotic recombination events (Paques and Haber 1999). We found that the requirement for the four proteins was approximately the same in the repair of the 26-base loop, but the requirement for Msh2p and Msh3p for the repair of the 1.1-kb loop was reduced. Similarly, all four proteins are required to approximately the same extent for single-strand annealing events when the interacting repeats are short (~200 bp), but the requirement for Msh2p and Msh3p is substantially reduced when the repeats are long (1 kb; Sugawaraet al. 1997). One interpretation of our results is that the Msh2p and Msh3p stabilize the interactions of Rad1p and Rad10p with the substrate, and this stabilization is more important with the smaller region of single-stranded DNA represented by the 26-base loop.
The observed effect of mutating RAD1 or RAD10 in strains heterozygous for his4-lopd or his4::U1.1a was an increase in the frequency of both 5:3 and 3:5 PMS tetrads and a reduction in the frequency of 2:6, but not 6:2, gene conversion tetrads (Tables 4 and 5; Kirkpatrick and Petes 1997). The reductions in the frequencies of 2:6 tetrads relative to wild type were statistically significant for the rad1 strains TP1013 and HMY49 (P values of 0.02 and 0.01, respectively) and the rad10 strains DTK286 and HMY151 (P values of 0.02 and 0.003, respectively).
As illustrated in Figure 1, mispaired loops can be repaired either by loop removal or duplication of the loop. Repair of 5:3-type heteroduplexes by removal of the loop will result in a 6:2 tetrad. Duplication of the loop will result in restoration of 4:4 segregation. Alternatively, repair of a 3:5-type heteroduplex by removal of the loop will result in restoration of 4:4 segregation, while duplication of the loop will result in a 2:6 gene conversion. A strain deficient in the ability to repair mispaired loops by duplication would generate fewer 2:6 conversions (with a concomitant increase in the production of 3:5 PMS tetrads), without an effect on the 6:2 conversion class. One would still observe 5:3 PMS tetrads, however, due to the inability to complete restoration repair from this intermediate. Thus, the reduction of the 2:6 conversion class in rad1 and rad10 strains suggests that the Rad1p/Rad10p endonuclease is involved in loop repair events that require cleavage of the strand opposite the loop (Kirkpatrick and Petes 1997).
As shown in Figure 5a, Rad1p/Rad10p is a junction-specific endonuclease that recognizes single- to double-strand transitions (Bardwellet al. 1994). The human homologues of Rad1p/10p (XPF/ERCC1) incise duplex DNA 2–8 nucleotides 5′ of a junction with single-stranded DNA (Matsunagaet al. 1996; De Laat et al. 1998a,b). In one study, the polar binding of RPA to the single-stranded substrate affected the incision activity of XPF/ERCC1 (de Laatet al. 1998b). As shown in Figure 5b, RPA bound to 3′-protruding single-stranded arms inhibits incision by XPF/ERCC1, whereas binding of RPA to 5′-protruding single-stranded arms greatly stimulates incision. On the basis of these observations, we suggest that binding of RPA on the large DNA loops preferentially directs Rad1p/Rad10p cleavage to the strand opposite the DNA loop. The repair events subsequent to this cleavage would result in duplication of the insertion (Figure 5c).
Several additional points concerning this model are relevant. First, there must be a second, as yet unidentified, repair system that results in deletion, rather than duplication, of the insertion. Second, since in vitro, XPF/ERCC1 (in the presence of RPA) cleaves 30-base loop substrates on both strands (Matsunagaet al. 1996; T. Matsunaga and A. Sancar, personal communication), it is possible that Msh2p/Msh3p contribute to the specificity of strand cleavage. Third, Clikeman et al. (2001) found that DNA loop repair in mitotic yeast cells required both Msh2p and Pms1p. This result suggests that meiotic and mitotic repair of loops may have different genetic requirements. Alternatively, since the substrates examined by Clikeman et al. contained both point mutations and insertions, it is possible that their results reflect an interaction between two different repair systems. Fourth, the model is consistent with observation that the efficiency of targeted integration of transforming DNA is greatly reduced by mutations in ERCC1, the mammalian equivalent of RAD10 (L. Niedernhofer and R. Kanaar, personal communication). If integration of transforming DNA involves a heteroduplex intermediate (such as that shown in Figure 5c), the Rad1p/Rad10p endonuclease would be required for the cleavage event necessary to integrate the insertion.
We previously showed that loops composed of palindromic sequences frequently escape meiotic repair (Naget al. 1989; Nag and Petes 1991; Mooreet al. 1999). Since the Rad1p/Rad10p endonuclease requires a single- to double-strand transition for cleavage, it is likely that these hairpin-forming loops do not present an appropriate substrate for repair. This conclusion is consistent with the observations that the efficiency of incision in stem-loop substrates (Figure 5c) increases with the size of the loop and that no incisions are observed for loops eight bases or smaller (De Laat et al. 1998a). In preliminary studies, we find no effect of rad1 on the repair of a heterozygous palindromic insertion (H. M. Kearney and T. D. Petes, unpublished data).
The roles of Rad1p and Msh3p in the repair of bubble substrates are different from their roles in the repair of DNA loops. Although mutations in either RAD1 or MSH3 result in a fourfold elevation in the frequency of gene conversion (comparison of HMY215 and HMY222 with HMY189, Table 6), the rad1, but not the msh3, mutation leads to substantially elevated PMS frequencies. One interpretation of this result is that Rad1p and Msh3p reduce the formation of the bubble substrate. Since crossovers between HIS4 and LEU2 are not substantially affected by these mutations (Table 6), this reduction probably does not involve DSB formation, but a subsequent step. In addition to its role in preventing formation of the bubble substrate, Rad1p, but not Msh3p, is significantly involved in its repair. Although we favor this interpretation, we cannot exclude other possibilities. For example, the Rad1p and Msh3p may be involved in directing repair events to restorations rather than gene conversions.
Disparity of gene conversion in wild-type strains: Gene conversion events involving heterozygous point mutations usually show no disparity (equal frequencies of 6:2 and 2:6) or subtle disparities (Fogelet al. 1981; Nagylaki and Petes 1982). In contrast, conversion events involving either large insertions or deletions often (although not always) show disparity in favor of the conversion events that duplicate the insertion or result in loss of the deletion (Fink and Styles 1974; Fogelet al. 1981; McKnightet al. 1981; Pukkilaet al. 1986; Vincent and Petes 1989). In our study, we also observed that conversion events tend to duplicate, rather than delete, the insertion. We suggest that this disparity reflects two factors. First, if the insertion contains a site susceptible to DSB formation, then competition for adjacent DSB formation (Wu and Lichten 1995; Xu and Kleckner 1995; Fanet al. 1997) reduces DSB formation at the normal HIS4 upstream site (Figure 3b). Since only DSB formation at the normal upstream region results in gene conversion involving the homologues (as discussed below), this effect results in disparity.
We suggest that a second factor in generating disparity is the relative efficiency of heteroduplex formation across the insertion. As shown in Figure 4, recombination events initiated on the chromosome with large (1.1–5.6 kb) insertions require extensive degradation (>1.1–5.6 kb) of one of the strands of the recipient chromosome in order to allow DNA loop formation. Events initiated on the chromosome with the wild-type allele require less degradation, although more extensive DNA synthesis. If strand degradation is limited, then one would expect a bias in favor of 2:6 conversions. This expectation assumes that heteroduplex formation events that are initiated, but not completed, can nonetheless give rise to viable spore products. In summary, the degree of conversion bias observed for heterozygous insertions is likely to reflect the strength of DSB formation within the insertion, the size of the insertion, and the position of the insertion relative to the initiating DSB.
HIS4-LEU2 crossovers in wild-type and mutant strains: A number of the mutations examined in our study significantly affected crossovers in the HIS4-LEU2 interval (Tables 4, 5, 6 and 7). Crossover distances were determined by measuring the number of PD:NPD:T tetrads (Perkins 1949). The relative numbers of these tetrads is a complicated function of the frequency of initiating recombination intermediates, the resolution of these intermediates as crossovers or noncrossovers, and the ratio of various classes of double crossovers. In previous studies, it has been shown that mutations in MLH1, MSH4, MSH5, and EXO1 reduce crossovers (Ross-Macdonald and Roeder 1994; Hollingsworthet al. 1995; Hunter and Borts 1997; Khazanehdari and Borts 2000; Kirkpatricket al. 2000). These same mutations significantly affected crossovers in strains heterozygous for his4-lopd (Table 4). In addition, the msh3 and rad2 mutations altered the distribution of the PD, NPD, and T tetrads. The effect of msh3 was not completely straight-forward. The msh3 strains had a reduction in the relative number of NPD tetrads (presumably representing four-strand double crossovers), but an increase in the relative number of T tetrads (presumably representing single crossovers and three-strand double crossovers; data not shown). The rad2 mutation reduced the relative numbers of both NPD and T tetrads. Since the level of aberrant segregation at HIS4 was also reduced in the rad2 strain (Table 4), the reduction in crossovers might reflect an effect of rad2 on the initiation of recombination.
In strains heterozygous for his4::U1.1a, we also observed significant reductions in crossovers in msh4 and exo1 mutant strains (Table 5). In addition, the rad27 mutation significantly reduced both crossovers and aberrant segregation. As suggested above, the endonuclease encoded by RAD27 may be involved in heteroduplex formation involving large heterozygous insertions or deletions.
DSBs formed within heterozygous insertions stimulate ectopic recombination, but not recombination between homologues: We found that DSBs were efficiently formed within the his4::U1.1 and his4::k1.5 insertions. These breaks, however, did not contribute to recombination between homologues, although the breaks in the URA3 insertions stimulated ectopic recombination with the ura3 genes on chromosome V. The DNA ends formed by DSBs within the insertion contain nonhomologous DNA sequences that presumably block interactions with the homologous chromosome. If the nonhomologous sequences were efficiently removed from both strands, as expected if conversion could result from DNA gap repair, it is likely that these DSBs would stimulate interaction with the homologues. Thus, our results argue that gap repair in yeast occurs rarely. It is likely that DSBs formed within the insertion, if not used for ectopic recombination events, are repaired by sister-strand interactions. One argument for this type of repair is that we find no loss in spore viability in strains heterozygous for the his4::k1.5 insertion, despite the DSBs that occur in the kanMX4 insertion (which shares no homology with the yeast genome).
Conclusions: Heteroduplexes can be formed during meiotic recombination in S. cerevisiae through very large heterologies. The resulting DNA loops are repaired by a process requiring Rad1p, Rad10p, Msh2p, and Msh3p. This mechanism duplicates, rather than removes, large loops. DSBs that occur within heterozygous insertions do not efficiently initiate interhomologue exchange.
We thank M. Dominska, N. Perabo, and E. Vitriol for assistance with the genetic analysis; L. Niedernhofer, R. Kanaar, T. Matsunaga, and A. Sancar for communicating unpublished information; and J. Merker, J. Stone, J. Hoeijmakers, and J. Sekelsky for helpful comments. The research was supported by National Institutes of Health (NIH) grant GM24110. H.M.K. was supported by NIH Training Grant (5 T32 GM07092-27), D.T.K. was a Special Fellow of the Leukemia and Lymphoma Society, and J.L.G. was supported by the American Cancer Society (Grant 5-39833).
Communicating editor: L. S. Symington
- Received March 30, 2001.
- Accepted May 24, 2001.
- Copyright © 2001 by the Genetics Society of America