A Highly Polymorphic Meiotic Recombination Mouse Hot Spot Exhibits Incomplete Repair

Abstract

The recent mapping of recombination hot spots in the human genome has demonstrated that crossover is a nonrandom process that occurs at well-defined positions along chromosomes. However, the mechanisms that direct hot-spot turnover in complex mammalian genomes are poorly understood. Analyses of the human genome are impaired by the inability to genetically dissect and molecularly manipulate recombinogenic regions to test their roles in regulating hot spots. Here, using the BXD recombinant inbred strains as a crossover library, three new recombination hot spots have been identified on mouse chromosome 19. Analyses of a highly polymorphic recombination hot spot (HS22) revealed that approximately 4% of recombinant molecules display complex and incomplete repair with discontinuous conversion tracts, as well as persistent heteroduplex DNA at crossover sites in mature spermatozoa. Also, sequence analysis of the wild house mouse revealed instability at the center of this hot spot. This suggests that complete repair is not required for completion of mammalian meiosis, a scenario that leaves duplex DNA containing mismatches at crossover sites.

Meiotic recombination is necessary to generate diversity, and recombination does not occur at random but rather concentrates in permissive regions called hot spots (21). Interestingly, while the rates of recombination can vary widely between hot-spot regions (21), the rest of the genome appears to be “in the cold.” The mouse provides the ideal model system for defining the rules and mechanisms regulating recombination hot spots. However, only two hot spots, E_β and Psmb9 (11, 43), have been analyzed in detail for the mouse, and both are localized in the MHC locus on chromosome 17. Identifying new recombination hot spots in strains where embryonic stem cells (e.g., C57BL/6J and 129/SvJ) and a vast catalog of knockouts are available simplifies genetic manipulation of hot spots and also allows rapid crosses without having to homogenize the strain background. However, while population studies of humans have identified potential recombination hot spots by linkage disequilibrium (13, 28), this strategy cannot be applied to laboratory inbred mouse strains.

To circumvent this hurdle, a facile method was developed using recombinant inbred (RI) mouse strains as a crossover (CO) library to quickly and precisely locate potential recombination hot spots, and this method was applied to hot spots on mouse chromosome 19. Three CO sites in the BXD RI strains were confirmed to be bona fide recombination hot spots, validating this strategy. Finally, analyses of a highly polymorphic hot spot revealed that 4% of CO molecules display discontinuous conversion tracts, as well as persistent heteroduplex DNA in mature spermatozoa.

MATERIALS AND METHODS

Mouse strains, breeding, and wild house mouse DNAs.

Mouse strains C57BL/6J (hereafter called B6) and DBA/2J (hereafter called DBA) were purchased from the Jackson Laboratory (Bar Harbor, ME). Male mice analyzed were all offspring of four independent DBA females crossed with a B6 male. All crosses were generated at the Animal Research Center of St. Jude Children's Research Hospital under Institutional Animal Care and Use Committee guidelines. Wild house mouse DNAs were kindly provided by François Bonhomme and Annie Orth (University Montpellier II, Montpellier, France).

Localization of recombination hot spots.

BXD recombinant inbred DNAs were purchased from the Jackson Laboratory (Bar Harbor, ME). Localization of hot spots in chromosome 19 among the first 26 BXD strains was performed by sequentially narrowing in on the interval known to contain the crossover in a given strain. The BXD RI strain distribution patterns were used as a starting point (http://www.informatics.jax.org/searches/riset_form.shtml). The Entrez mouse single nucleotide polymorphism (SNP) database was used as a first choice. However, in most cases, the identification of novel SNPs was required. Novel SNPs were identified by designing oligonucleotides across 1-kb intervals, amplifying the region in the B6, DBA, and BXD RI strains of interest, and sequencing. DNA sequence was obtained from ENSEMBL (http://www.ensembl.org/Mus_musculus/index.html, Ensembl release 34); repeats were identified using RepeatMasker (http://www.repeatmasker.org/cgi-bin/WEBRepeatMasker), and oligonucleotides were designed using Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi; a 21-nucleotide primer, a melting temperature of 55°C, and 50% GC content were found to be optimal; all other values were left unchanged) in regions free of repeats. We designed 253 oligonucleotide pairs. Eight percent failed to amplify, and 33% did not yield any SNP between the B6 and DBA strains. The HS9, HS22, and HS37 recombination hot-spot B6 and DBA sequences, as well as the description of all the oligomers, are available upon request.

Sperm DNA extraction, quantification, and crossover analyses.

DNA was prepared as previously described (43). DNA was dissolved in 5 mM Tris-HCl (pH 7.5) and digested with the SacII restriction enzyme to cleave outside the tested intervals. All PCRs were performed with a reaction buffer previously described (17) with 0.1 U of recombinant Taq polymerase (Invitrogen), 0.01 U of turbo Pfu polymerase (Stratagene), and 0.4 μM of each oligonucleotide in a volume of 10 μl. For Poisson amplification or the second round of CO detection, Cresol Red (Sigma) and sucrose (Sigma) were added to the PCR mix to allow direct gel loading of the samples.

The number of amplifiable molecules was determined using Poisson analysis. Typically, a serial dilution ranging from 10⁻² to 10⁻⁵ was performed to estimate the overall number of amplifiable molecules. Once determined, 96 Poisson reactions were performed at the optimal dilution to obtain 30%- to 50%-positive reactions. The number of amplifiable molecules was then calculated as described previously (44). Primers used for Poisson amplifications were 19-7000640F (5′-GAC ACC CTC CAG GAC AGA ATC −3′) and 19-7001899R (5′-GGG CTC TGA TGA AGT GGG AAG −3′). PCRs were performed at 96°C for 1 min, followed by 32 cycles at 96°C for 30 s, 60°C for 40 s, and 70°C for 1 min. Poisson reactions were run on a 1.5% agarose gel in 0.5× Tris-borate-EDTA containing ethidium bromide and visualized under UV light.

Optimal annealing temperature for each primer set was determined using a gradient PCR block by amplifying either a 1:1-ratio mix of B6 and DBA DNA as a negative control or the BXD RI strain DNA (BXD1 for HS9, BXD16 for HS22, and BXD21 for HS37), where the CO was located as a positive control. Sperm and brain DNA from a DBA-B6 heterozygous mouse confirmed the correct conditions. Despite multiple attempts, we were unable to detect any amplification in the DBA-to-B6 CO orientation at the HS37 hot spot, most likely due to a genomic duplication. To detect CO, allele-specific primers were designed to amplify across the HS9, HS22, and HS37 hot spots. Nucleotide differences between B6 and DBA are labeled in bold (see below). Nucleotides indicated in lowercase letters were added to raise the overall GC content of the oligomer (HS9-2F and HS22-6R).

HS9 5′ allele-specific primers were HS9-B-1F (5′-AGC CCC CTT TAA AAG ACT TTT-3′) and HS9-B-2F (5′-ggg gTT CAA GCT ACC CTC-3′) for B6, HS9-D-1F (5′-GCC CCC TTT AAA AGA CTT TAA-3′) and HS9-D-2F (5′-ggg gTT CAA GCT ACC CTT-3′) for DBA. The 3′ primers were HS9-B-4R (5′-TTG CCT GGG CAG GAG TAG-3′) and HS9-B-3R (5′-TCT CCC TGC TGT GGA GTA-3′) for B6 and HS9-D-4R (5′-TTG CCT GGG CAG GAG TAT-3′) and HS9-D-3R (5′-TCT CCC TGC TGT GGA GTG-3′). The first round was performed using the HS9-1F and HS9-4R primer pair (3.5 kb) and the second round with the HS22-2F and HS22-3R primer pair (2.3 kb). First-round PCRs were performed at 96°C for 1 min, followed by 24 cycles at 95°C for 30 s, 60°C for 40 s, and 63°C for 3 min. Second-round PCRs were performed at 96°C for 1 min, followed by 28 cycles at 95°C for 30 s and 63°C for 2 min.

HS22 5′ allele-specific primers were HS22-B-2F (5′-CCT CAA GGT CCT ACG-3′) and HS22-B-3F (5′-GCC AGA CAC TGT AGC-3′) for B6, HS22-D-2F (5′-CCT CAA GGT CCT ACC-3′) and HS22-D-3F (5′-GCC AGA CAC TGT AGT-3′) for DBA. The 3′ primers were HS22-B-7R (5′-TCG CCG ACT GAT GAT-3′) and HS22-B-6R (5′-ggc cgG CAT TTT AAT CTT CAT AG-3′) for B6 and HS22-D-7R (5′-TCG CCG ACT GAT GAC-3′) and HS22-D-6R (5′-ggc cgG CAT TTT AAT CTT CAT AC-3′). The first round was performed using the HS22-2F and HS22-7R primer pair (5.0 kb) and the second round with the HS22-3F and HS22-6R primer pair (4.6 kb). The first round of PCRs was performed at 96°C for 1 min, followed by 24 cycles at 96°C for 30 s, 60°C for 40 s, and 63°C for 4 min 30 s. The second round of PCRs was performed at 96°C for 1 min, followed by 28 cycles at 96°C for 30 s, 62°C for 40 s, and 63°C for 3 min 30 s.

HS37 5′ allele-specific primers were HS37-B-2F (5′-CCA ATC TGT TTA AGG CCC-3′) and HS37-B-3F (5′-TGA TGT TCC ATC CCA GCC-3′) for B6 and HS37-D-2F (5′-CCA ATC TGT TTA AGG CCT-3′) and HS37-D-3F (5′-TGA TGT TCC AAT CCA GCA-3′) for DBA. The 3′ primers were HS37-B-3R (5′-TAC TAA ACA CGG GGT CCC-3′) and HS37-B-2R (5′-TTA AAT ACA TCC AAC ACC TGC-3′) for B6 and HS37-D-3R (5′-TAC TAA ACA CGG GGT CCT-3′) and HS37-D-2R (5′-TTA AAT ACA TCC AAC ACC TGT-3′). The first round was performed using the HS37-2F and HS37-3R primer pair (8.6 kb) and the second round with the HS37-3F and HS37-2R primer pair (3.0 kb). The first round of PCRs was performed at 96°C for 1 min, followed by 24 cycles at 96°C for 30 s, 59°C for 40 s, and 63°C for 7 min. The second round of PCRs was performed at 96°C for 1 min, followed by 28 cycles at 96°C for 30 s, 62°C for 40 s, and 63°C for 2 min 30 s.

All primary PCR products for HS9, HS22, and HS37 were digested with S1 nuclease to remove single-stranded DNA as previously described (43). One-hundred-fold-diluted S1-treated DNA was then used to seed the secondary PCRs.

Secondary PCR products (70% of total reaction volume) were run on 0.8% agarose gel in 0.5× Tris-borate-EDTA and visualized with ethidium bromide and UV light. Positive reactions were diluted 50-fold and reamplified using internal primers. For the HS22 hot spot, the primers used were HS22-reamp/F) 5′-TGA GAG ACA GAG GCA GGT AGG-3′) and HS22-reamp/R (5′-CCG TCT GCC ACT GGT CTA TCT-3′). Reamplification PCRs were performed at 96°C for 1 min, followed by 24 cycles at 95°C for 30 s, 60°C for 40 s, and 63°C for 4 min, for both hot spots in a 7-μl reaction volume as described above. PCR products were purified using QIAGEN′s (Valencia, CA) QIAquick 96-well PCR purification system, following the supplier's instructions. Each product was sequenced. HS22 sequencing primers were HS22-D (5′-GGA GAG TTG GCC CAG CAG TTA-3′), HS22-E (5′-CCA TCT AGA CCA CCC TGC AA-3′), HS22-F (5′-AGA CGG TCT CGT CTG TGA CCT-3′), HS22-G (5′-CAG GAG CCA GAC CAA TCC TGT-3′), HS22-H (5′-CAC CAT CGG CAG TGT TAC CT-3′), and HS22-K (5′-GGT GGG GAA TCC ATC TTT CTT G-3′). All subsequent sequence analyses were performed using Sequencher software (version 4.7; GeneCodes, Ann Arbor, MI).

The sequences of each recombinant molecule were aligned with the two reference sequences of B6 and DBA, on which known polymorphism locations were marked. Due to the overall limited sequence background, calling SNP identity was straightforward. In the case of complex mixed strands (BD #7, DB #2, and DB #4 [with BD indicating B6-DBA recombination and DB indicating the reverse]), we used an arbitrary peak ratio (low/high) of at least 0.6 to positively call a base as present in both SNPs. However, as shown in Fig. 4, electropherograms for each base peak were of identical height. Moreover, after well over 1,000 recombinant molecules were sequenced (this study; also unpublished data), no early PCR misincorporation was detected anywhere in the surveyed HS22 sequence, suggesting that this type of PCR artifact is extremely rare. It is therefore highly unlikely to be involved in the generation of the complex recombinant molecules observed.

FIG. 4.

Sequence electropherograms of complex recombinant molecules. (A) A schematic representation of all seven BD complex recombinant molecules is shown, together with the sequence electrophorograms at the SNP locations. Only one set of electrophorograms is shown for BD #2 and #3, which were found twice each. B6 DNA is represented in black, while DBA DNA is displayed in gray. A double-line area indicates heteroduplex SNPs (BD #7). (B) Same representation as in panel A except for the four DB complex recombinant molecules.

DNA inputs per small-pool PCR consisted of 280 amplifiable molecules for the HS9 hot spot, 750 amplifiable molecules for the HS22 hot spot, and 500 amplifiable molecules for the HS37 hot spot. This was equivalent to 0.2 to 0.4 recombinant molecules per pool. We typically obtained 10% to 30% positive pools per experiment, to limit reactions with two recombinant molecules.

RESULTS AND DISCUSSION

Identification of recombination hot spots on mouse chromosome 19.

Using the BXD RI panel as a CO library (2), several CO sites of chromosome 19, which are indicative of potential recombination hot spots, were identified. The overall strategy and the validation step that allows the detection of recombinant molecules exclusively in sperm DNA are shown in Fig. 1. CO sites were defined by taking advantage of the high density of available polymorphisms for the BXD RI panel and by identifying additional SNPs (Fig. 1A). The analysis was restricted to the proximal 35 Mb of chromosome 19 and identified 11 CO (Fig. 1B). Such a strategy can be expanded to any chromosome region and recombinant inbred panel currently available. Due to the variable SNP density between DBA and B6 strains, only a few CO sites could be precisely mapped to a subkilobase region. Indeed, some regions were totally homozygous for well over 10 kb between the two strains, rendering CO detection poorly informative for profile analysis. Due to the common origin of laboratory strains, some regions between DBA and B6 display limited sequence divergence whereas others are highly polymorphic. Strain comparison information can be obtained from the Phenome Database website (http://phenome.jax.org/pub-cgi/phenome/mpdcgi?rtn=snps/distplot_pre). As a final step, the true hot-spot nature at three CO sites (HS9, HS22, and HS37) was confirmed by analyzing F₁ male offspring from DBA crossed with B6 mice (Fig. 1C to E and 2). HS9 (genomic location, Chr19:10734164.10737624; Ensembl release 43) contains the last three exons out of nine of the Pga5 (pepsinogen 5, group 1) gene (42). HS22 (Chr19:23026219.23030965) is located in an intron of the ubiquitously expressed Trpm3 (transient receptor potential cation channel, subfamily M, member 3) gene (24). Finally, HS37 (Chr19:38738042.38746570) is found in the third intron of the Plce1 (phospholipase C, epsilon 1) gene, which is also ubiquitously expressed (40).

FIG. 1.

Strategy used to identify recombination hot spots using the BXD RI panel. (A) The breeding scheme used to generate recombinant inbred strains (2). (B) Maps of chromosome 19 including band patterns (31) and the proximal region analyzed in this study. Identified CO are indicated by arrows. The three hot spots analyzed in detail are labeled in larger, bold font. (C) Batches of sperm DNA from a heterozygous mouse (white and gray circles) across the recombinogenic active region were amplified using two PCR rounds using allele-specific oligonucleotide pairs (primers are shown as black and gray arrows) to amplify specifically recombinant molecules in a given orientation. (D) Meiosis-specific recombination was confirmed using sperm and somatic (brain) DNA. Two thousand amplifiable molecules were used per reaction and were analyzed by agarose gel electrophoresis. (E) A representative small-pool PCR recombination detection assay (48 reactions). An average of 750 amplifiable molecules per reaction were used.

FIG. 2.

Examples of CO molecules at the HS9 and HS37 hot spots in the B-to-D orientation. (A) A representative small-pool PCR recombination detection assay using sperm DNA at the HS9 recombination hot spot (48 reactions). Two hundred eighty amplifiable molecules per reaction were used. (B) Same representation as in panel A at the HS37 recombination hot spot. Five hundred amplifiable molecules per reaction were used.

Confirmation of the bona fide nature of identified recombination hot spots.

PCR strategies were developed at each of these three potential hot spots to detect recombinant molecules in both orientations at the HS9 and HS22 hot spots. For the HS37 hot spot, analysis was possible only for the B6-to-DBA orientation, due to a genomic duplication. These three CO were chosen due to their high-enough SNP density surrounding the CO site. These analyses confirmed the bona fide nature of these recombination hot spots with CO rates of 4.77 × 10⁻⁴, 2.17 × 10⁻⁴, and 3.45 × 10⁻⁴ for HS9, HS22, and HS37, respectively (Fig. 1 and 2; Table 1). This also validated this strategy as a facile and cost-effective method for identifying novel recombination hot spots in mice. Indeed, the rapidly increasing densities of SNPs that are now available render this approach extremely quick and efficient. Also, it alleviates the requirement for any extensive breeding of mice. A recent study, also using the same BXD RI panel, was used to derive a genetic recombination map of the mouse genome (34), emphasizing the tremendous utility of using RI panels to identify sites of recombination hot spots. The two main limitations to this method are as follows: (i) the finite number of CO present in the RI panel analyzed and (ii) the presence of a suitable SNP density across any putative recombinogenic region to allow PCR strategies to be developed for their analysis.

TABLE 1.

Sperm CO data for mouse HS9, HS22, and HS37 hot spots^a

Hot spot and cross (age of mouse [mo])	Orientations	No. of amplifiable molecules	No. of CO detected	Rate (10−4) [95% CIb]
HS9
D/B	B→D	110,400	57	5.16 [3.82-6.50]
	D→B	73,000	32	4.38 [2.86-5.90]
	Both	183,400	89	4.85
HS22
D/B no. 1 (2)	B→D	143,000	38	2.66 [1.79-3.53]
	D→B	138,000	32	2.32 [1.50-3.14]
	Both	281,000	70	2.49
B/D no. 2 (2)	B→D	142,500	34	2.39 [1.54-3.24]
	D→B	142,500	28	1.96 [1.13-2.79]
	Both	285,000	58	2.04
D/B no. 3 (12)	B→D	250,000	50	2.00 [1.39-2.61]
	D→B	250,000	44	1.76 [1.20-2.32]
	Both	500,000	90	1.80
B/D no. 4 (12)	B→D	221,000	49	2.22 [1.59-2.85]
	D→B	221,000	52	2.35 [1.69-3.01]
	Both	442,000	101	2.29
Total	B→D	756,500	171	2.26
	D→B	751,500	156	2.08
	Both	1,508,000	327	2.17
HS37
D/B	B→D	267,000	92	3.45 [2.74-4.15]

^aMixed pools containing at least two simple recombinant molecules were easily identified by sequencing, since they displayed a central heteroduplex DNA region surrounded by clean edges where each CO occurred. With an average of 30% of positive pools, 13.7% (39 pools out of 288 pools sequenced) of HS22 pools (12.3% expected) had at least 2 CO molecules present.

^bCI, confidence interval.

The HS22 hot spot was analyzed in detail due to its unusually high level of polymorphisms between B6 and DBA haplotypes, with 87 bp differences (49 simple SNPs and 5 insertions/deletions) across the 1.28 kb of the hot spot, equivalent to 7% of DNA being polymorphic between the two strains, including the presence of a large 50-bp insertion/deletion. In comparison, HS9 has no polymorphisms between B6 and DBA across the hot-spot region and HS37 displays only five SNPs across the 2 kb of the recombinogenic region (0.25% of polymorphism or 28-fold lower than at the HS22 locus). This variability in sequence and length at HS22 is particularly interesting in view of the repair pathways identified in yeast, which can resolve heteroduplex structure following resolution of double Holliday junctions (dHJs), such as simple mismatches or loops of various sizes (20, 39).

A PCR strategy allowed the detection of recombinant molecules in both orientations, from B6 to DBA and from DBA to B6, at the HS22 hot spot. A total of 1,508,000 amplifiable molecules were scanned, which detected a total of 327 recombinant molecules in both orientations for four different heterozygote males of 2 and 12 months of age. All four mice showed similar recombination rates at HS22, averaging 2.17 × 10⁻⁴ (Table 1), irrespective of age (2 months, mice no. 1 and 2, or 12 months, mice no. 3 and 4) (Table 1) or the chromosome of origin (female DBA crossed to a male B6, D/B cross, mice no. 1 and 3, or the reverse B/D cross, mice no. 2 and 4) (Table 1). The B6-to-DBA orientation recombination frequency was consistently higher than that for DBA to B6, 2.26 × 10⁻⁴ versus 2.08 × 10⁻⁴, respectively, but this most likely reflects a slight reduction in PCR efficiency, as previously observed (43).

Complete sequencing of all 171 B6-DBA and 156 DBA-B6 recombinant molecules allowed the generation of recombination profiles across the HS22 hot spot in both orientations (Fig. 3A and C and 4). While CO appear to be reciprocal in their rate, their distribution displays a strong asymmetry and transmission distortion, with the B6-DBA orientation shifted 5′ relative to the DBA-B6 orientation. This shift averaged 530 bp, with the B6-DBA center located at 2,569 bp whereas the DBA-B6 counterpart is located at 3,102 bp (Fig. 3A and C and 5A). This results in a gross overtransmission of the DBA allele within the hot spot (Fig. 5B), indicating that the B6 haplotype is the most active in DNA double-strand break (DSB) initiation.

FIG. 3.

HS22 sperm CO profiles and heteroduplex structures. (A) Progenitor haplotypes of strains B6 (black) and DBA (gray) are indicated together with a B6-to-DBA-orientation recombinant molecule. Allele-specific primers that were used to recover recombinants from sperm DNA for this orientation are shown as black and gray arrows. Polymorphisms are indicated by vertical black lines (simple SNPs) or red boxes (insertion/deletion, with the width of the box representing its size; note the large 50-bp insertion/deletion, shown as a wide red box). The sequence of the polymorphism is provided above. Distribution of B-to-D CO across the analyzed interval, the number of recombinant molecules identified, the number of CO mapping to each interval, and the CO rates are indicated. The center of the HS22 hot spot is indicated with a red arrowhead. (B) Complex structures are shown and numbered BD #1 to BD #7 (note that BD #2 and #3 were found twice, as indicated). Discontinuous conversion tract molecules are represented with black lines for B6 DNA and gray lines for DBA DNA. Complex molecules containing mixed strands are represented with a double black and gray line. (C and D) Same as panels A and B but in the DBA-to-B6 orientation.

FIG. 5.

CO asymmetry and transmission distortion at the HS22 hot spot. (A) Cumulative frequencies of sperm CO in the B6-to-DBA (white) or DBA-to-B6 (gray) orientation are shown. (B) The transmission distortion of DBA (gray) SNPs is shown across the HS22 hot spot.

The Spo11 protein initiates meiotic recombination by creating DSBs with 3′ overhangs (30), which most likely undergo further resection. The 3′ overhangs then invade the undamaged chromosome, and repair is performed using information from the undamaged (donor) chromosome. However, if DSB initiation rates vary between strain haplotypes, then asymmetry and transmission distortion will occur. If a hybrid (in this case B6/DBA) carries a high-efficiency (B6) and low-efficiency (DBA) haplotype, then the majority of CO will arise from the active haplotype (B6). DSB repair will then result in the replacement of markers from the initiating chromosome, resulting in biased gene conversion in favor of the donor (DBA) haplotype. Indeed, the interval ranging from 2,019 bp to 2,704 bp in HS22 showed a transmission distortion consistently above 80%. These observations are similar to the asymmetry seen in human hot spots and in the mouse E_β CO hot spot (16, 21, 43). However, while the B6 chromosome is the most prone to DSB initiation in this hybrid combination, other haplotype combinations may show a different asymmetry in favor of the non-B6 chromosome, as has been shown at the Eb and Psmb9 hot spots (3, 43).

Identification of imperfect CO molecules at the HS22 hot spot.

Unlike the case with previously described hot spots in the human and mouse genomes (8, 11, 15, 16, 18, 43), not all of the CO at HS22 were simple in this wild-type cross. Indeed, in both orientations an average of 4% of recombinants displayed evidence of partially repaired DNA and discontinuous conversion tracts (BD recombinant molecules 1 to 9 and DB recombinant molecules 1 to 4) (Fig. 3B and D and 4A and B). Mixed pools containing two (12.3%) or three (1.2%) simple recombinant molecules were easily identified by sequencing, since they show a central heteroduplex DNA region surrounded by clean edges. With an average 30% of positive pools, it is expected that 12.3% of them (among the positive pools) contain at least 2 CO. Indeed, 13.7% were observed. More importantly, increasing (by 2.5-fold) the number of amplifiable molecules to 2,000 using somatic (brain) DNA did not result in the generation of artifactual jump-PCR products (Fig. 1D). This experiment provided two important controls. First, it confirmed that the allele-specific amplification was very efficient. More importantly, it also demonstrated that potential jump-PCR artifacts were not generated at a detectable level. Finally, the average centers of these complex recombinant molecules were at locations almost identical to those of classic CO products, at 2,369 bp and 3,017 bp for the B6-DBA and DBA-B6 orientations, respectively; therefore, artifactual generation of these molecules is very unlikely. Indeed, in the case of jump-PCR products, one would expect a random distribution of CO across the 5 kb scanned, rather than the asymmetric and similarly centered products found for each orientation. These controls confirmed the bona fide nature of these complex recombinant molecules.

These partially repaired molecules were detected in both orientations (nine for B6-DBA and four for DBA-B6; Fig. 3B and D), and they could be separated into two classes. The first class of CO molecules still contained unrepaired heteroduplex regions (BD #7 and DB #1 and 3; Fig. 3B and D), as well as discontinuous conversion tracts. One molecule displayed a repaired region surrounded by heteroduplex DNA (DB #3; Fig. 3D), suggesting that repair could be initiated from the center of duplex DNA containing mismatches with an aborted repair tract. Such molecules have been observed in wild-type strains of yeast at recombination hot spots where artificial mismatches were introduced (9, 25); however, typically there was only one single heterology. The second class of complex CO molecules that were documented contained only discontinuous conversion tracts without any heteroduplex DNA present (BD #1 to #6 and DB #2 and #4; Fig. 3B and D). Two such complex recombinant molecules, BD #2 and BD #3, were independently identified twice. These discontinuous conversion tracts usually involved one or two closely spaced single polymorphisms (BD #1, #3, #4, #5, and #6 and DB #2 and #4; Fig. 3B and D). Only BD recombinant molecule #2 showed a longer discontinuous conversion tract, which spread over 130 bp. Finally, none of these imperfectly repaired CO molecules appeared to display any base scramble or microrearrangement in the CO intervals, suggesting that this process occurred normally during meiotic progression.

These findings imply that heteroduplex DNA is still present in spermatozoa well after the completion of meiosis. Thus, in addition to demonstrating the persistence of heteroduplexes in spermatid DNA at the HS22 locus, these molecules also provide insights into the repair mechanisms that are operational following resolution of dHJs in highly heterozygote recombinogenic regions. Specifically, the programmed induction of DSBs by Spo11 dimers in the very first stage of meiotic prophase provokes meiotic CO (22). Spo11 generates DSBs where the protein is covalently attached to the 5′ strand termini, before being released by endonucleolytic cleavage and subsequent strand invasion (30). In the case of CO pathway repair, resolution of DSBs proceeds through strand synthesis, formation of dHJs, and cleavage, leading to a CO (22, 29). In all cases, this leaves heteroduplex DNA regions, which are then thought to be repaired via the mismatch repair machinery. Indeed, the detection and characterization of these mixed molecules suggest that once the dHJs have been resolved, a patchy repair process occurs where some but not all of the heteroduplex molecules can be repaired in time to copy either a B6 template or a DBA template. This observation was true in both orientations where small patches within or flanking a heteroduplex region could be observed.

Sometimes the heteroduplex region in HS22 had been repaired yet generated recombinant molecules with discontinuous conversion tracts. Such molecules have been observed at the mouse Psmb9 hot spot but only in the context of the Mlh1 deficiency (10). It is likely that such incomplete repair is directly linked to the level and probably type (simple mismatch versus large insertion/deletion) of polymorphism observed at recombination hot spots, as has been observed for yeast (12, 20). These molecules can also be explained by mismatch repair during strand invasion, as previously observed for yeast (1, 35). In this case, the discontinuous tracts would be a consequence of two possibilities for mismatch repair with the two repair tracts in an opposite orientation. Along these lines, it may be important that the 50-bp insertion/deletion is unique to the core of the HS22 hot spot.

Conclusions.

Heretofore such a high level of polymorphism between strands in the vicinity of recombination hot spots has been found only at human hypermutable minisatellites, where sequence divergences between various repeats at a given locus range from 5% to 20% (Table 2). In this case, extremely complex repeat shuffling was observed to occur by an as yet undefined mechanism (4, 14). The discontinuous conversion tracts observed at the HS22 hot spot in 4% of CO products may explain this apparent complexity detected at hypermutable minisatellites, where the repair machinery's struggle to rapidly restore complex heteroduplex molecules either during strand invasion or following dHJ resolution leads to either conversion or CO events (7, 15). This would, in turn, provide a driving mechanism for the extreme tandem repeat shuffling observed at these loci (6, 19, 36). Interestingly, another example of discontinuous conversion tracts is recombination events at 17p11.2, which lead to either duplications or deletions associated with autosomal dominant Charcot-Marie-Tooth type 1A disease and hereditary neuropathy with liability to pressure palsies, respectively (26).

TABLE 2.

Levels of polymorphisms between tandem repeats of hypermutable human minisatellites^a

Locus	Reference	Male mutation rate (%)	Array length (bp)	No. of SNPs	Polymorphism level (%)
CEB1	37	12.9	38-42	8	19-21
B6.7	23	7.6	33-34	4	12
MS1	41	5.3	9	1	11
MS32	41	1.2	29	2	7
MS31A	41	0.75	20	2	10
plg3	33	0.65	36	2	5.5

^aThe average mutation rate is provided together with the array length (minimum to maximum), the number of SNPs between various repeats, and the percentage of polymorphisms. Note that the polymorphism level for most loci is above the 7% level observed at HS22.

Finally, the persistence of duplex DNA containing mismatches in mature spermatids at highly polymorphic recombination hot spots predicts that these genomic regions would be inherently unstable. To test this hypothesis, the HS22 locus was sequenced from 22 independent wild-mouse species and subspecies (5), including Mus musculus domesticus (7 isolates; from Algeria, Georgia, Israel, Morocco, Spain, Tunisia, and Denmark), Mus musculus musculus (7 isolates; 1 isolate each from Armenia, Georgia, Bulgaria, Denmark, and Austria and 2 isolates from Poland), Mus castaneus (2 isolates; from Kenya and Madagascar), Mus famulus (1 isolate; from India), Mus macedonius (1 isolate; from Bulgaria), and Mus spretus (4 isolates; from Spain, France, Morocco, and Tunisia). This analysis revealed that the center of the most active B6 hot spot shows a high level of genomic hypervariability (Fig. 6). Indeed, the HS22 core region was entirely deleted in Mus famulus (187-bp deletion). In addition, the G-rich tract present at the center of the HS22 hot spot is entirely replaced by a poly(A) stretch ranging from 23 to 29 bp in Mus spretus. Not surprisingly, though, expected length polymorphisms were observed at this tract. Interestingly, regions flanking the HS22 hot spot did not present such levels of instability, other than the usual base substitution already detected between B6 and DBA DNAs.

FIG. 6.

HS22 core sequence across various wild house mouse populations. The HS22 locus in 22 independent wild-mouse species and subspecies, including Mus musculus domesticus (6), Mus musculus musculus (6), Mus castaneus (1), Mus famulus (13), Mus macedonius (13), and Mus spretus (3). DBA and B6 (the 50-bp deletion is highlighted in magenta) are also shown at the top. The center of the most active B6 hot spot shows a particularly high level of rearrangement. In particular, the hot spot core region was entirely deleted in Mus famulus (187-bp deletion; magenta). In addition, the G-rich tract present at the center of the hot spot was replaced in Mus spretus by a poly(A) stretch ranging from 23 to 29 bp (shown in green). Interestingly, regions flanking the hot spot did not present such a level of instability, other than the usual base substitution already detected between B6 and DBA DNA molecules.

Collectively, these observations suggest that the maintenance of heteroduplex DNA following male meiosis, most likely originating from the difficulty in repairing recombination intermediates consisting of duplex DNA containing many mismatches, would render highly polymorphic and recombinogenic regions prone to instability either during strand invasion or following dHJ resolution. Such enhanced plasticity also supports recent observations where hot spots appear remarkably fluid between species (27, 32, 38). The posited hypothesis would place CO repair pathways at the very heart of this genomic instability.