Meiotic Recombination in Drosophila Msh6 Mutants Yields Discontinuous Gene Conversion Tracts

Abstract

Crossovers (COs) generated through meiotic recombination are important for the correct segregation of homologous chromosomes during meiosis. Several models describing the molecular mechanism of meiotic recombination have been proposed. These models differ in the arrangement of heteroduplex DNA (hDNA) in recombination intermediates. Heterologies in hDNA are usually repaired prior to the recovery of recombination products, thereby obscuring information about the arrangement of hDNA. To examine hDNA in meiotic recombination in Drosophila melanogaster, we sought to block hDNA repair by conducting recombination assays in a mutant defective in mismatch repair (MMR). We generated mutations in the MMR gene Msh6 and analyzed recombination between highly polymorphic homologous chromosomes. We found that hDNA often goes unrepaired during meiotic recombination in an Msh6 mutant, leading to high levels of postmeiotic segregation; however, hDNA and gene conversion tracts are frequently discontinuous, with multiple transitions between gene conversion, restoration, and unrepaired hDNA. We suggest that these discontinuities reflect the activity of a short-patch repair system that operates when canonical MMR is defective.

CROSSOVERS (COs) generated through meiotic recombination are essential to the correct segregation of chromosomes during meiotic divisions in most eukaryotes. An understanding of the events required to generate COs is central to the understanding of this crucial phenomenon. Several models of meiotic recombination have been proposed to describe the molecular steps required to generate COs. These models must account not only for CO formation, but also for the formation of noncrossovers (NCOs) and the association of gene conversion (GC) with both COs and NCOs.

More than 40 years ago, Robin Holliday proposed a model for meiotic recombination to account for these observed phenomena (Holliday 1964). One prominent feature of this model is the suggestion that GC associated with both COs and NCOs is generated by the formation and repair of heteroduplex DNA (hDNA), DNA in which each strand of the duplex is derived from a different parental chromosome. Although several other models for meiotic recombination have been proposed over the years, invoking different initiating lesions and different recombination intermediates, formation and repair of hDNA as a mechanism for GC has remained a constant (Meselson and Radding 1975; Szostak et al. 1983).

Experimental evidence supports models in which the formation and repair of hDNA are important features of meiotic recombination. Physical characterization of recombination at a meiotic hotspot in Saccharomyces cerevisiae has revealed hDNA in recombination intermediates (Allers and Lichten 2001). Additionally, the products of meiotic recombination sometimes exhibit postmeiotic segregation (PMS), when the two parental alleles segregate from one another at the first postmeiotic mitosis. This is thought to arise from unrepaired heterologies in hDNA, since PMS occurs most frequently in mutants that abolish mismatch repair (MMR) (reviewed in Borts et al. 2000) or for heterologies that are not repaired efficiently, such as short palindromic insertions (Nag et al. 1989).

Although the formation and repair of hDNA is a common feature of models of meiotic recombination, the structure and arrangement of hDNA present in the proposed recombination intermediates and products differ in different models. Investigation of hDNA present in the PMS products of meiotic recombination in S. cerevisiae has provided insights into the mechanism of CO formation (Foss et al. 1999; Merker et al. 2003; Hoffmann and Borts 2005; Hoffmann et al. 2005), but the degree to which details of the structure and arrangement of hDNA can be determined through these experiments is limited. Detection of hDNA requires the presence of heterologies between homologous chromosomes, and it has been shown that high levels of heterology in S. cerevisiae decrease the frequency of recombination (Borts and Haber 1987). Levels of heterology ranging from ∼0.05 to ∼0.3% have been used in experiments in S. cerevisiae, but the highest levels have included, at most, only three polymorphisms (Borts and Haber 1987; Judd and Petes 1988; Symington and Petes 1988; Gilbertson and Stahl 1996; Merker et al. 2003; Jessop et al. 2005).

In Drosophila melanogaster, high levels of heterology (∼0.5%) do not affect the frequency of meiotic recombination (Hilliker et al. 1991). We have previously studied recombination between two highly polymorphic chromosomes (33 polymorphisms across 7420 bp, ∼0.45%) in Drosophila and have used the multiple heterologies to determine GC tract lengths with high resolution (Blanton et al. 2005). The study of unrepaired hDNA in the PMS products of meiotic recombination in Drosophila provides an opportunity to gain insight into the structure and arrangement of hDNA at a high resolution. In S. cerevisiae, even well-repaired markers, such as base–base mismatches or small insertion/deletion loops, sometimes go unrepaired, resulting in PMS (Fogel et al. 1981). In contrast, PMS is exceedingly rare in Drosophila (Chovnick et al. 1971), even though the same types of heterologies should be present in hDNA. To analyze the structure of hDNA, thereby gaining insight into the mechanism of meiotic recombination in Drosophila, we sought to decrease repair by removing the canonical MMR pathway.

The MutS and MutL proteins are central components of MMR in Escherichia coli (reviewed in Modrich and Lahue 1996); canonical MMR in eukaryotes involves several homologs of MutS (Msh proteins) and MutL (Mlh/Pms proteins) (reviewed in Kunkel and Erie 2005). The Drosophila genome has two genes encoding Msh proteins, spel1 (encodes ortholog of Msh2) and Msh6, and two genes encoding MutL homologs, Mlh1 and Pms2 (reviewed in Sekelsky et al. 2000). We generated deletions that remove most of the coding sequence of Msh6. We report here that mutation of Drosophila Msh6 greatly increases the incidence of PMS among both CO and NCO products of meiotic recombination. This suggests that hDNA is a common feature of meiotic recombination intermediates in Drosophila and that MMR homologs are important for the repair of this hDNA. Surprisingly, tracts of GC and unrepaired hDNA from Msh6 mutants are frequently discontinuous, a phenomenon that we did not observe among wild-type CO and NCO products (S. J. Radford and J. Sekelsky, unpublished results; Blanton et al. 2005; Radford et al. 2007, this issue). We propose that, as has been suggested in S. cerevisiae (Coic et al. 2000), in the absence of canonical MMR, mismatches can be repaired through a process that allows closely spaced mismatches to be repaired or left unrepaired, independent of one another. While this short-patch repair system obscures much of the information provided by the structure and arrangement of hDNA, analysis of the remaining unrepaired hDNA provides insights into the mechanism of meiotic recombination in Drosophila.

MATERIALS AND METHODS

Recovery of recombination events within the ry gene:

Thirty to 40 females of the genotype y; Msh6¹⁰ ry⁵³¹ cv-c/Msh6⁶⁸ kar ry⁶⁰⁶ were crossed to 10 to 15 males of the genotype y/Y, Dp(1;Y)y⁺; kar ry⁵⁰⁶ cv-c. Crosses were set up in bottles containing 25 ml of standard food medium and placed at 25°. After 3 days, adults were transferred to fresh media to establish a second brood, and purine was added to the first brood bottles. We used 0.75 ml of 0.18% (w/v) purine in water, which is the lowest level that effectively kills the majority of ry− progeny (Radford et al. 2007), maximizing the survival of ry−//ry+ mosaics. One of every 25 bottles was left untreated and adult progeny were counted to estimate the number of larvae screened. Recombinant progeny were scored for flanking visible markers (karmoisin and crossveinless) to determine whether a CO or an NCO event had occurred.

Detection of gene conversion and PMS tracts:

Recombinant progeny were mated to kar ry⁵⁰⁶ cv-c flies of the opposite sex to test for mosaicism via germline transmission. The recombinant fly was then homogenized in buffer containing proteinase K, as described (Gloor et al. 1993), to recover DNA for PCR. GC and hDNA tracts were analyzed by a combination of allele-specific and non-allele-specific PCR and sequencing of PCR products in bulk.

Allele-specific PCR primers were designed to specifically amplify each allele at each of several polymorphic sites (Figure 1A). Primers were designed such that the 3′ end corresponded to the allele-specific nucleotide, and an additional mismatch was engineered two or three nucleotides before the 3′ end to increase specificity. Annealing temperature and magnesium concentration were optimized for each primer pair. A list of polymorphic sites is given in supplemental data (supplemental Table S1 at http://www.genetics.org/supplemental/), along with allele-specific primer sequences and conditions (supplemental Table S2 at http://www.genetics.org/supplemental/). To prevent false-positive amplification primers were tested against positive and negative control fly preps with each use. Amplification with both allele-specific primers for a given site was taken as evidence for PMS at that site. Allele-specific PCR products were sequenced to determine the length and arrangement of hDNA tracts (Figure 1B). PMS was also detected by amplifying fragments with non-allele-specific primers, sequencing the bulk product, and examining the chromatogram for double peaks at polymorphic sites (Figure 1B).

Figure 1.—

Detection of PMS by allele-specific PCR and sequencing. (A) Allele-specific PCR. The image is from an agarose gel stained with ethidium bromide. DNA size markers were loaded in the leftmost lane. Other lanes contain products from PCR amplification of single-fly DNA preps. The sources of DNA were (left to right) a ry⁵³¹/ry⁵⁰⁶ fly, a ry⁶⁰⁶/ry⁵⁰⁶ fly, a 1:4 mixture of ry⁵³¹/ry⁵⁰⁶ and ry⁶⁰⁶/ry⁵⁰⁶, a noncrossover recombinant in which the ry⁵³¹ mutation was converted to wild type (GC), and a noncrossover recombinant with PMS at the site of the ry⁵³¹ mutation (PMS). Two PCRs were performed on each sample, one using an allele-specific primer for the ry⁵³¹ mutation (531) and another using an allele-specific primer for the wild-type sequence at that site (+). (B) Chromatograms from sequencing of PCR products, showing the region from 3731 to 3695 (reverse strand). Arrows below the chromatograms indicate the positions of two polymorphic sites; the site on the left (3723) is G in ry⁵³¹ and T in ry⁶⁰⁶, and the site on the right (3703) is A in ry⁵³¹ and T in ry⁶⁰⁶. For the chromatogram at the top, the template was a PCR product from a ry⁵³¹ fly. For the other three, the templates were PCR products from the same CO from an Msh6 mutant (boxed region of bottom CO in Figure 2, class IV). This CO had PMS at the ry⁵³¹ mutation site. PCR was done using a nonspecific primer (ns), a primer that amplifies the strand with the ry⁵³¹ mutation (531), or a primer that amplifies the strand with the wild-type sequence at that site (+). The chromatograms show PMS at the 3703 site and demonstrate that the base derived from the ry⁵³¹ chromosome (A) is on the strand carrying the ry⁵³¹ mutation. The polymorphism at 3723 has the base from the ry⁵³¹ chromosome (T) on both strands.

Recombinants were recovered in trans to the ry⁵⁰⁶ chromosome, which has a deletion beginning at base pair 1398 and extending 3.8 kb to the right; this deletion does not include the region around ry⁶⁰⁶, which is at −468. The ry⁵⁰⁶ deletion was induced on the same parental chromosome as ry⁵³¹, so both chromosomes have the same sequence differences relative to ry⁶⁰⁶. For sequences to the right of ry⁶⁰⁶, we analyzed PCR products where at least one primer was within the region deleted in ry⁵⁰⁶, to ensure that products were derived from the recombinant chromosome. For regions to the left of ry⁶⁰⁶, allele-specific PCR was done using a primer specific to the ry⁶⁰⁶ chromosome that was outside the region of GC or hDNA, as determined by sequencing.

Statistical comparisons:

Mean GC tract lengths and statistical comparisons were calculated as described previously (Blanton et al. 2005). PMS frequency comparisons were made using Fisher's exact test with two-tailed P-values, computed by Instat 3.05 (GraphPad Software). Comparison of crossover distributions in Table 2 was done by two-tailed G-test.

TABLE 2.

Crossover distribution in wild-type and Msh6 mutants

Interval	I	II	III	IV	V	VI	VII	VIII	IX	X	XI
Size (bp)	424	511	269	31	10	376	443	1057	304	191	164
Wild-type COs	10	12	7	0	0	8	13	32	8	7	1
Msh6 COs	9	3	8	0	1	3	3	16	4	2	2

The 3.8-kb region between the ry⁶⁰⁶ and ry⁵³¹ mutant sites was divided into 11 intervals defined by the 10 polymorphisms in this region (Figure 1, supplemental Table S1 at http://www.genetics.org/supplemental/). Intervals are listed in chromosomal order from left (ry⁶⁰⁶ mutation) to right (ry⁵³¹ mutation). The exchange positions of 105 COs from wild-type females and the 51 class I COs from Msh6 mutants were determined by PCR and sequencing. The number of COs with an exchange in each interval is given. The distributions are not significantly different from one another (P = 0.37).

RESULTS

Deletion of Msh6 coding sequence by P excision:

In yeast and mammalian systems, MMR is carried out by two functional heterodimers of MutS homologs, MutSα (Msh2 and Msh6) and MutSβ (Msh2 and Msh3) (reviewed in Kunkel and Erie 2005). Consequently, mutations in MSH2 and MSH6 do not elicit equivalent phenotypes in these organisms. The Drosophila and Caenorhabditis elegans genomes encode orthologs of Msh2 (SPEL1 in Drosophila) and Msh6, but not Msh3, suggesting that only MutSα is present in these organisms (Sekelsky et al. 2000; Denver et al. 2005). In support of this proposal, C. elegans msh-2 and msh-6 mutants have identical phenotypes with regard to mutation accumulation (Denver et al. 2005).

Flores and Engels constructed a synthetic deletion in Drosophila of spel1 by combining overlapping deletions in trans, together with a transgene to replace a second gene contained in the region of overlap between the deletions (Flores and Engels 1999). These mutants were shown to exhibit increased microsatellite instability (Flores and Engels 1999), a common feature of MMR mutants in other species, suggesting that MMR function is conserved in Drosophila. On the basis of the results from C. elegans, we hypothesize that SPEL1 and MSH6 are equivalently required for MMR in Drosophila. Because Msh6 is on chromosome 3, the same chromosome as the locus used for our recombination assay (Blanton et al. 2005), it is technically easier to use mutations in this gene for characterization of unrepaired hDNA in the products of meiotic recombination in Drosophila.

We obtained a Drosophila stock in which a P element was inserted 43 bp upstream of the Msh6 coding sequence and used it to generate deletions through excision of the P element. Mutations in mus309 increase the recovery of flanking deletions following P-element excision (Adams et al. 2003; McVey et al. 2004b), so we used a PCR-based strategy to screen for deletions into Msh6 coding sequence following excision in both wild-type and mus309 mutant backgrounds. Of 153 P-element excisions from wild type, 1 (0.6%) had a deletion into Msh6 coding sequence. In contrast, of 31 excisions from mus309 mutants, 4 (13%) had a deletion into Msh6 sequence, confirming that conducting excision in a mus309 mutant background is an effective method for increasing the recovery of deletion mutants from P-element excision screens.

Through PCR and sequencing, we determined the breakpoints of four of five Msh6 deletions. In all four cases, the deletion was in one direction from the P-element insertion. The single deletion obtained from wild type (Msh6¹¹) removed ∼850 bp of the 3.8-kb coding sequence and retained ∼160 bp of P-element sequence. The three mapped deletions obtained from mus309 mutants were larger: Msh6¹⁰, Msh6⁵⁹, and Msh6⁶⁸ remove 2.2 kb (+2.0 kb of P element), 3.8 kb (+12 bp of P element), and 2.8 kb (+130 bp of P element) of coding sequence, respectively. This suggests that an excision screen performed in a mus309 mutant background not only increases the frequency of deletion, but also increases recovery of larger deletions.

To minimize the effects of other mutations that may have arisen during P-element excision, the experiments described below were performed in females heteroallelic for Msh6¹⁰ and Msh6⁶⁸ (the mus309 mutation was crossed off of each chromosome when ry alleles were crossed on). These alleles remove most of the coding sequence, so this genotype should represent a complete absence of Msh6 activity.

Recovery of COs and NCOs is increased in Msh6 mutants:

Several MutS and MutL homologs are required for the generation of COs in S. cerevisiae, including Msh4, Msh5, Mlh1, and Mlh3 (Hunter and Borts 1997; Wang et al. 1999; Borner et al. 2004; Guillon et al. 2005). To determine whether Drosophila MSH6 has a role in CO formation, we recovered meiotic recombination events within the rosy (ry) locus, using a procedure developed by Chovnick and colleagues (Chovnick et al. 1970, 1971). The ry gene encodes xanthine dehydrogenase (XDH), which is required both for the metabolism of purine and for normal eye pigmentation. Females trans-heterozygous for ry⁵³¹ and ry⁶⁰⁶, point mutations separated by 3.8 kb (Figure 2), were crossed to males homozygous for ry⁵⁰⁶, which deletes much of the gene. Rare rosy+ recombinants were selected by treating the larvae with purine; flanking markers were used to distinguish COs from NCOs (see materials and methods for details). Because an oocyte receives only one of the two chromatids involved in the recombination event (the other segregates into a polar body nucleus), we cannot determine whether the COs we recover have associated tracts of GC. NCOs, however, are recovered only when a GC (or PMS) tract spans one of the two ry mutations. Hence, we use the term “CO” to mean crossovers with or without gene conversion and the term “NCO” to mean gene conversion of a ry mutation (or PMS at that site) without an associated crossover.

Figure 2.—

Complex CO events from meiotic recombination in Msh6 mutants. The schematic at the top shows the rosy locus, except for the first exon. Solid regions are coding sequences. The positions of the selected sites corresponding to the ry⁶⁰⁶ and ry⁵³¹ mutations are indicated. Additional polymorphisms are shown as lollipops on the scale bar. These are all single-nucleotide polymorphisms, except for −1029 (not shown here, but see Figure 3) and −685, which are insertions of 1 and 4 bp, respectively, in ry⁵³¹ relative to ry⁶⁰⁶ (see supplemental Table S1 at http://www.genetics.org/supplemental/ for a list of polymorphic sites). The scale is in base pairs, using the coordinate system of Bender et al. (1983). The structures of the complex COs from Msh6 mutants are diagrammed below. Each bar represents an independent event, with the circles denoting the selected markers (ry⁶⁰⁶ or ry⁵³¹ mutant sites). Gray bars show the region surrounding the recombination event. Blue bars represent sequences derived from the ry⁵³¹ chromosome, and red bars represent sequences from the ry⁶⁰⁶ chromosome. Polymorphisms are marked with blue or pink lines to indicate alleles from the ry⁵³¹ or the ry⁶⁰⁶ parental chromosome, respectively. Polymorphisms that exhibited PMS are marked with half-pink, half-blue lines. Events are grouped according to the numbered classes described in the text. One structure was observed five times, as indicated by the (5) to the right of the structure. Gray asterisks mark events with multiple PMS patches; all were in the cis orientation. The chromatograms in Figure 1 are from PCR products corresponding to the boxed region on the bottommost CO. Class I COs are not pictured, but using this scheme they would be depicted as a simple transition from blue to red at a single point. The sites of these transitions are listed in Table 2.

We previously reported the results of an analysis of recombination events in wild-type females, in which we screened 2.3 million larvae and recovered 81 COs and 31 NCOs (Blanton et al. 2005). We have screened an additional 1.4 million larvae from wild type and recovered 31 COs and 22 NCOs (Table 1) (Radford et al. 2007). We screened 1.8 million larvae from Msh6 mutants and recovered 67 COs and 44 NCOs, a 25% rate increase over wild type in COs and a 65% rate increase in NCOs (Table 1). These results show that Msh6 is not required for the generation of COs.

TABLE 1.

Intragenic recombination in wild-type and Msh6 mutants

		Crossovers		Noncrossovers
Genotype	Progeny screened	n (frequency)	PMS	n (frequency)	PMS
Wild typea	3,710,000	112 (3.0 × 10−5)	0	53 (1.4 × 10−5)	0
Msh6	1,775,000	67 (3.8 × 10−5)	14b	42 (2.4 × 10−5)	23b

^aIncludes data from Blanton et al. (2005) and Radford et al. (2007).

^bPCR failed for DNA preps from one CO and two NCOs, so the frequency of PMS was 21% among COs (14 of 66) and 58% among NCOs (23 of 40).

Increased recovery of rosy+ recombinants in an MMR mutant is not unexpected. If one of the ry point mutations is included in hDNA during recombination, MMR may either restore the ry− sequence or convert it to ry+ sequence; flies carrying conversions will survive purine selection, but those receiving a chromatid that underwent restoration repair will not be recovered. If repair does not occur, as expected in an MMR mutant, the oocyte will receive a chromatid in which one strand is ry+ and the other is ry−, resulting in a ry+//ry− mosaic larva. Because XDH is secreted and diffuses throughout the developing larva, many ry+//ry− mosaics survive purine treatment, developing into mosaic adults that are rosy+ in eye color (Romans 1980). Thus, loss of MMR allows us to recover some events that would have been lost otherwise.

In S. cerevisiae, recombination is initiated at hotspots, which are usually in promoter regions (Gerton et al. 2000). In contrast, recombination at ry is thought to initiate and terminate throughout the gene (Clark et al. 1988). Hence, hDNA may include either mutation (ry⁶⁰⁶ is 1.2 kb 3′ of the transcription start site; ry⁵³¹ is an additional 3.8 kb downstream, 0.5 kb from the 3′ end of the gene) or may be entirely between the mutations. In wild-type flies, we recover NCOs that include GC of either mutation, but in Msh6 we can also recover those that would have been repaired to restore the mutant allele, since this will result in ry+//ry− mosaics. The frequency of recovery of NCOs was increased by 65% in the Msh6 mutant, relative to wild type. This increase is consistent with MSH6 having a role in meiotic MMR and suggests that restoration repair often occurs in wild-type meiosis.

We recover COs that occur anywhere between the ry mutations so as to generate a ry+ chromosome. If a crossover includes hDNA spanning one of the mutant sites, repair of the hDNA to the mutant allele will prevent us from recovering the CO. As with NCOs, however, loss of MMR will allow us to recover these COs. Although we do not know the mean length of hDNA tracts associated with COs, the mean GC tract length among NCOs is 441 bp (Blanton et al. 2005), which is about one-ninth of the distance between the sites. Thus, it is probably a minority of COs that are lost due to repair of a mutant site, but are recovered in the Msh6 mutant. In agreement with this prediction, the elevation in recovery of rosy+ recombinants was only 25% for COs (Table 1).

Both COs and NCOs from Msh6 mutants exhibit PMS:

To directly assess the role of MSH6 in hDNA repair, we examined the incidence of PMS among recombination events. Previously, PMS was detected in Drosophila by observing mosaicism in the germline and staining fly sections for the mosaic expression of XDH activity (Carpenter 1982); however, these methods can detect only PMS of the ry point mutations. To investigate fully the structure and arrangement of hDNA, it is necessary to detect PMS of both the ry point mutations and the surrounding silent polymorphisms. To accomplish this, we developed molecular assays for the detection of PMS.

We used one genetic assay and two molecular assays to detect PMS. First, we mated the rosy+ recombinants to ry⁵⁰⁶ partners to look for evidence of mosaicism in the germline. The Drosophila germline is set aside from a few nuclei out of several hundred present early in embryonic development. In a ry+//ry− mosaic, these germline cells may include only ry+, only ry−, or both ry+ and ry− maternal chromosomes; the latter two cases are definitive evidence for mosaicism. Among the 88 ry+ recombinants from Msh6 mutants that produced progeny, two transmitted only ry− chromosomes, and none transmitted both ry+ and ry− maternal chromosomes. This is a large underestimate of the frequency of PMS, because this assay detects only a subset of ry+//ry− mosaics (Carpenter 1982) and can detect mosaicism only at the sites of the ry⁵³¹ and ry⁶⁰⁶ mutations, not at sites of silent polymorphisms.

We used allele-specific PCR to detect PMS at both the ry point mutation sites and the surrounding polymorphisms (Figure 1A). Using allele-specific PCR for the point mutation sites, we detected 19 instances of PMS. This included 6 flies that did not produce progeny, 11 that transmitted only ry+ through the germline, and both recombinants that transmitted only ry− through the germline. This demonstrates that allele-specific PCR is sensitive enough to detect somatic mosaicism even when the germline, which constitutes a significant fraction of the tissues in an adult fly, is entirely ry−. This finding, coupled with results from control amplifications in which we detect either allele when it is present in only 20% of the DNA molecules in a sample (the lowest concentration attempted, Figure 1A), suggests that allele-specific PCR detects most or all PMS.

The third assay we used to detect PMS was to sequence non-allele-specific PCR products and examine chromatograms for the presence of double peaks at polymorphic sites (Figure 1B). With this assay, we detected PMS in each of the 19 cases detected by allele-specific PCR of the ry mutation sites, as well as in 18 additional recombinants that exhibited PMS at sites of silent polymorphism but not at the mutant sites. There were no cases in which results from one assay were inconsistent with results from another assay.

We did not detect PMS in any of the 165 recombinants (112 COs and 53 NCOs) from wild-type females (Table 1) (Blanton et al. 2005; Radford et al. 2007), confirming results of earlier experiments that suggested that PMS is exceedingly rare in Drosophila, at least among NCOs from wild-type females (Chovnick et al. 1971). In contrast, we detected PMS in 37 of 106 recombinants from Msh6 mutant females, a significantly higher frequency (P < 0.0001). The frequency of PMS was significantly lower (P = 0.0003) among COs (14 of 66, 21%) than among NCOs (23 of 40, 58%). This result was unexpected, because it suggests a difference in either MMR or hDNA formation between COs and NCOs. Nonetheless, the finding of PMS in both CO and NCO products from Msh6 mutants confirms that hDNA is indeed an important feature of meiotic recombination in Drosophila, and that MSH6 is involved in the repair of mismatches within this hDNA.

GC and hDNA tracts from Msh6 mutants are discontinuous:

The structure and arrangement of hDNA in recombination products can be used to infer the nature of recombination intermediates. Having successfully created a Drosophila MMR mutant in which we can recover unrepaired hDNA, we sought to examine the recombination events from this mutant to gain insight into the recombination process. We examined the extent of unrepaired hDNA and GC tracts in the COs and NCOs from Msh6 mutants through PCR and sequencing (Figure 1, materials and methods). Sequence analysis was considered complete when at least one unconverted polymorphism was found on both ends of the GC or hDNA tracts. In some cases, the end point identified by this method may actually be a patch of restoration repair. This would lead to an underestimation of the extent of GC and hDNA tracts and an underestimation of the level of discontinuity associated with COs and NCOs, but these limitations do not substantially affect the conclusions drawn from this analysis.

Among the 66 COs from Msh6 mutants, we observed four structural classes (Figure 2): (I) “normal,” with a single exchange point (51 COs, 77%); (II) an hDNA tract at a single point of exchange (7 COs, 10.5%); (III) an hDNA tract separated from a single point of exchange (3 COs, 5%); and (IV) multiple points of exchange, with or without hDNA (5 COs, 7.5%). Class I COs are indistinguishable from the COs we recovered from wild type in that they lacked discontinuities. In addition, the distribution of exchange points across the 3.8-kb region between the ry mutations was not significantly different between COs from wild-type females and class I COs from Msh6 mutants (Table 2). It is evident from this result that most COs do not require MSH6 for hDNA repair; however, it is not clear whether this reflects the existence of an MSH6-independent MMR pathway or whether hDNA is not formed, or is not extensive enough to be detected, in these COs. The remaining three types of COs from Msh6 are distinct from COs from wild type both in the appearance of unrepaired hDNA (classes II, III, and IV) and in multiple exchange points or discontinuities (class IV).

We also observed four structural classes among 40 NCOs from Msh6 mutants (Figure 3): (I) normal, with a single, continuous GC tract (9 NCOs, 20%); (II) a single, continuous hDNA tract (4 NCOs, 10%); (III) a single GC tract interrupted by an hDNA tract (10 NCOs, 25%); and (IV) a discontinuous GC tract, with or without hDNA (17 NCOs, 45%). Class I NCOs are similar to NCOs observed from wild type (Blanton et al. 2005; Radford et al. 2007) in that the GC tracts are continuous. Furthermore, all GC tract lengths observed from Class I NCOs from Msh6 mutants have also been observed in NCOs from wild type, although the small number of Class I NCOs from Msh6 mutants does not allow statistical comparison of mean GC tract lengths. The remaining three classes of NCOs from Msh6 mutants, which represent the majority of events, are distinct from NCOs from wild type due to the appearance of hDNA (classes II, III, and IV) and the discontinuity of tracts (classes III and IV).

Figure 3.—

NCO events from meiotic recombination in Msh6 mutants. A schematic of the ry locus is shown as in Figure 2, but covering a larger region. GC and hDNA tracts of NCOs recovered from Msh6 mutants are shown below. Shaded bars represent the minimum tract length for each event, and dotted lines the maximum possible length based on the location of the adjacent unconverted polymorphism. The selected sites (ry⁶⁰⁶ for tracts on the left, ry⁵³¹ for tracts on the right) are indicated with circles. Coconverted sites are marked by open vertical bars, and sites within a tract that were not converted (i.e., restored) are marked by solid vertical bars. Asterisks mark events with multiple patches of hDNA; those with solid asterisks have trans hDNA. Numbers in parentheses indicate multiple independent events with the same structure.

Our results reveal that many recombination events from Msh6 mutants exhibit discontinuities or “patchiness.” We have identified a single discontinuous GC tract from wild type (Blanton et al. 2005) and two discontinuous tracts from mei-9 mutants (Radford et al. 2007). In each of these, a long GC tract contains a single site that appeared to undergo restoration repair. We observed many more “patchy” events from Msh6 mutants, and these were more complex, sometimes including unrepaired sites or multiple discontinuities (Figures 2 and 3), suggesting that this phenomenon is a consequence of Msh6 mutation.

DISCUSSION

Our current understanding of the molecular mechanism of meiotic recombination is derived mostly from experimental evidence that has been collected in fungi. Although fungi have many advantages that make them amenable to studies of meiotic recombination, and these studies have greatly increased our knowledge of this process, it is important to investigate this process in other model organisms to determine which details are specific to a certain organism and which are universal. For example, Spo11 generates double-strand breaks that initiate meiotic recombination in S. cerevisiae (Keeney et al. 1997), and orthologs of this protein are essential for meiotic recombination in Schizosaccharomyces pombe, Arabidopsis thaliana, C. elegans, D. melanogaster, and Mus musculus, indicating that the mechanism of recombination initiation is widely conserved (reviewed in Keeney and Neale 2006). In contrast, the Mus81 endonuclease is required to generate most meiotic crossovers in S. pombe and a subset of crossovers in S. cerevisiae, but does not appear to be required to generate crossovers in C. elegans, Drosophila, or mice (reviewed in Whitby 2005; K. Trowbridge, K. S. McKim, S. Brill and J. Sekelsky, unpublished results), revealing unexpected diversity in the late stages of crossover formation.

One important feature of molecular models of meiotic recombination is the presence of hDNA. Investigation of hDNA in fungi has proven fruitful in many instances; however, these experiments have been limited in the structural detail they provide due to limitations in the number of heterologies that can be used simultaneously (see Introduction). We have developed a system originally designed by Chovnick and colleagues (Chovnick et al. 1970, 1971) that allows us to analyze the molecular structures of recombination events with high resolution in the model organism D. melanogaster (Blanton et al. 2005; Radford et al. 2007). Unrepaired hDNA is almost never detected in wild-type Drosophila, however, so we set out to increase our ability to recover hDNA using our recombination system.

In S. cerevisiae, recovery of unrepaired hDNA as PMS is greatly increased in MMR-defective mutants (reviewed in Borts et al. 2000). We found that PMS is dramatically increased in Drosophila Msh6 mutants, which demonstrates that hDNA and the use of MMR to repair heterologies in hDNA are both conserved features of meiotic recombination in this organism. Molecular analysis of the structures of recombination events in Msh6 mutants revealed a surprising amount of discontinuity in GC and hDNA tracts. There are several possible sources of these discontinuities. If mutations in Msh6 abolish all meiotic MMR, then the discontinuities in recombination events from these mutants must reveal the underlying structure of hDNA and GC tracts in recombination intermediates and products. This suggests that recombination normally results in the formation of discontinuous hDNA and GC tracts, which are then masked by the action of MMR. Alternatively, loss of MMR may disrupt the normal recombination process in some manner that leads to generation of discontinuous tracts. In mitotically proliferating Drosophila cells, it is thought that repair of double-strand gaps involves multiple rounds of strand invasion, repair DNA synthesis, and dissociation of the nascent strand from the template (McVey et al. 2004a). Discontinuous meiotic hDNA and GC tracts could occur if there are multiple cycles, with the sister chromatid used in some cycles and the homologous chromosome in others. This could explain some of the simpler discontinuous events we observed, but additional processes would be necessary to explain the complex events we recovered that contain patches of conversion, restoration, and hDNA. For this reason, we do not believe that the discontinuities we observed are present in recombination intermediates, but rather are the result of some MSH6-independent hDNA repair.

The only other MutS homolog in the Drosophila genome is SPEL1 (MSH2). It is possible that SPEL1 can function in MMR as a homodimer or as a heterodimer with some MutS homolog that has not been identified in the genome sequencing projects; however, in the discontinuous tracts that we observed, closely spaced heterologies are sometimes treated independently (converted, restored, or left unrepaired; Figures 2 and 3), suggesting that canonical MMR is not responsible.

To account for the discontinuities in GC and hDNA tracts that we recovered, we propose that removal of the canonical MMR pathway reveals the action of an alternative repair pathway in which repair tracts are very short. The existence of a short-patch repair pathway in the absence of canonical MMR has been suggested in both S. cerevisiae and S. pombe, although the proteins involved may not all be conserved between these organisms (Fleck et al. 1999; Coic et al. 2000). Discontinuous hDNA and GC tracts were also seen in the products of meiotic recombination from mouse Mlh1^−/− mutants (Guillon et al. 2005), suggesting that short-patch repair may also occur in vertebrates. Some PMS was detected for each type of base–base mismatch, but we did not detect any PMS at sites that would produce an insertion/deletion loop in hDNA (supplemental Table S3 at http://www.genetics.org/supplemental/). It is possible that unpaired loops are efficiently recognized by a short-patch repair system; however, there were other sites, including some for each possible base–base mismatch, at which we never detected PMS (e.g., −679, −668, −667, −636, 1153, and 3723; Figures 2 and 3), suggesting that sequence context may play a role in recognition of mismatches by the short-patch repair system.

According to the model discussed above, the unrepaired hDNA we recovered represents a subset of the hDNA that was formed during recombination, so examination of the PMS events can provide insight into the meiotic recombination process. Among the 14 PMS COs, we obtained 10 (Figure 2, classes II and III) with structures predicted by the two resolution types described in the double-strand break repair (DSBR) model (Szostak et al. 1983) (Figure 4). The remaining four PMS COs are not inconsistent with this model, but they cannot be unambiguously assigned to these classes because of the discontinuities that are present. This analysis supports the DSBR model as applicable to Drosophila meiotic recombination.

Figure 4.—

DSBR model for meiotic recombination. According to this model, recombination initiates with the introduction of a double-strand break (DSB) on one chromatid (shaded lines), followed by 5′–3′ resection of the ends to leave 3′ single-stranded overhangs. One 3′ end invades the duplex of a chromatid of the homologous chromosome (solid lines), base pairing with the complementary strand and displacing the other strand as a D-loop. Synthesis follows, primed by the 3′ end of the broken chromosome and using the invaded chromosome as a template. This strand may dissociate and anneal to the second broken end to generate an NCO by SDSA. Alternatively, the strand displaced by synthesis may anneal to the second broken end. Additional synthesis and ligation will then produce a double Holliday junction (DHJ) intermediate. The DHJ is resolved by cutting to generate CO or NCO products. Each Holliday junction may be cut in either of two orientations. Triangles indicate cut sites for one orientation that generates COs and for one that generates NCOs, corresponding to the pairs of products below each. The bottom-most pairs of products are generated by cutting each junction in the opposite orientation to that shown by triangles. Proposed NCO pathways are indicated by stippled arrows.

In the canonical DSBR model, COs and NCOs are alternate outcomes of resolution of the same recombination intermediate; however, several lines of evidence have led to the suggestion that many NCOs are derived from synthesis-dependent strand annealing (SDSA). Both resolution and SDSA predict that only cis hDNA may be formed, hDNA in which markers remain linked in the parental orientation (Figure 4). Evidence from S. cerevisiae, however, suggests an additional source of NCOs because some events with two tracts of hDNA in opposite orientations are observed, termed trans hDNA (Gilbertson and Stahl 1996). Many of the NCO PMS events we characterized had a single tract of hDNA, but there were eight cases with two or three separate regions of hDNA (Figure 3, class IV, asterisks). In six of these, hDNA tracts were in the trans orientation (Figure 3, class IV, solid asterisks), suggesting that, as in S. cerevisiae, there is a source of NCOs in addition to SDSA and resolution. We also observed three cases in which COs contained two or three separate regions of hDNA (Figure 2, class IV, asterisks); however, these were never in the trans orientation.

We also noted that COs exhibit PMS less frequently than NCOs in Msh6 mutants. We have hypothesized that, in Msh6 mutants, a short-patch repair system can act on mismatches in hDNA, reducing the level of PMS that we observe. While we cannot exclude the possibility that short-patch repair is more active on COs than on NCOs, a more plausible explanation is that less hDNA is present in recombination intermediates that lead to COs. In current models for CO formation, hDNA is formed during strand invasion and during second-end capture (Figure 4). In contrast, many NCOs are proposed to arise from SDSA, in which hDNA is formed during annealing (Figure 4). A possible explanation for our results, therefore, is that strand invasion and second-end capture generate less hDNA than the annealing step of SDSA. This model makes the prediction that, in Drosophila, GC tracts associated with COs are shorter than GC tracts associated with NCOs. Because we recover only one of the four products of meiotic recombination in our experiments, we are unable to analyze GC tracts associated with COs; however, the recovery of both chromatids involved in a recombination event has been accomplished in Drosophila (Chovnick et al. 1970), so an investigation of GC tracts associated with COs is possible.

The results reported here underscore the importance of exploiting the unique advantages of different model organisms to understand conserved processes. In particular, studying meiotic recombination in Drosophila provides an opportunity to investigate the length and arrangement of GC and hDNA tracts with high resolution and to use this information to gain insights into functions of various recombination proteins and the molecular mechanism of recombination. To this end, we created a Drosophila MMR mutant that allows recovery of unrepaired hDNA. We demonstrate that discontinuities in GC and hDNA tracts are common in an MMR mutant. The structure of hDNA tracts suggests that some NCOs are derived from a recombination intermediate that contains trans hDNA, a result that corroborates existing data in S. cerevisiae. Our results also suggest that associated hDNA tracts may differ in length between COs and NCOs. This supports the hypothesis that not all NCOs and COs are alternative outcomes of resolution of a common recombination intermediate.