The yeast Gal4 transcription factor binds to the promoters of several S. cerevisiae GAL genes through its N-terminal DNA-binding domain (reviewed in 25 ). This domain recognizes the CGGN₁₁CCG consensus sequence and, when attached to the yeast Spo11 protein, it is able to tether Spo11 to Gal4 recognition sites leading to the formation of Gal4BD-Spo11 specific recombination hotspots 26 27 28 . To evaluate if such tethering is possible in mice we used gene targeting in embryonic stem cells to introduce the DNA fragment coding for the DNA binding domain of the yeast Gal4 protein upstream of the start codon of the mouse Spo11 gene (Additional file 1: Figure S1). The expression and correct splicing of the resulting gene was confirmed by sequencing PCR fragments generated from cDNA of the Gal4BD-Spo11 homozygous (Spo11 ^Gal/Gal ) mice. We confirmed that both major isoforms of SPO11 – alpha and beta – were transcribed (data not shown). Nevertheless, testes of the Spo11 ^Gal/Gal mice were reduced in size and germ cells beyond the spermatocyte stage were absent (Figure 1A). Furthermore, no sperm was detected in the epididymus (Figure 1B) indicating that Spo11 ^Gal/Gal males were infertile. Similarly, Spo11 ^Gal/Gal ovaries were smaller than in wild type and showed a greatly diminished number of follicles (Figure 1C). Young Spo11 ^Gal/Gal females did produce progeny, however no viable litters have been obtained from mice older than 6 months, suggesting a premature cessation of fertility. Females carrying only one copy of the Gal allele and one null copy of Spo11 (Spo11 ^Gal/– ) showed a further reduction in ovarian size, no visible follicles and an underdeveloped uterus, indicating that a single Gal4BD-Spo11 allele is insufficient to maintain fertility (Figure 1C and data not shown).

The yeast counterpart of the chimeric GAL4BD-SPO11 gene complements the SPO11Δ mutation, indicating that the fusion Gal4BD-Spo11 protein is proficient at DSB formation 26 . To determine whether DSBs are formed in the Spo11 ^Gal/Gal mice we immunostained chromosomal spreads prepared from spermatocytes (Figure 2) or oocytes (Figure 3) with antibodies to the DMC1 protein, known to bind to the single stranded DNA tails of DSBs 29 30 . Although DMC1 foci were observed, their number was significantly reduced compared to wild type. Precise counting of DSBs is complicated by the dynamic nature of DSB formation and repair. While one can estimate the steady state level of DMC1 foci on each chromosomal spread, the DSBs that have already been repaired and the DSBs that have not yet formed will be under-counted. This problem is particularly pronounced when mutants with delayed or defective DSB repair are compared to wild type. To eliminate the effect of possible timing differences in the formation and/or repair of DSBs between Spo11 ^Gal/Gal and Spo11 ^+/+ mice, we compared the number of DMC1 foci on a Hop2 ^−/− genetic background, i.e., in Spo11 ^Gal/Gal Hop2 ^−/− versus Spo11 ^+/+ Hop2 ^−/− cells. The HOP2 protein is directly involved in the repair of meiotic DSBs 31 32 33 , accordingly, Hop2 ^−/− mice undergo meiotic arrest at the stage when all DSBs are already produced but none are repaired 34 . Therefore, the analysis of the mutants on the Hop2 ^−/− background will reduce counting bias resulting from DSB formation/repair dynamics. We found that the number of DMC1 foci in Spo11 ^Gal/Gal Hop2 ^−/− spermatocytes was approximately 2 fold lower than that in the Spo11 ^+/+ Hop2 ^−/− spermatocytes (Figure 4). Moreover, in Spo11 ^Gal/– spermatocytes, where one Gal and one null allele of Spo11 were present, the number of DMC1 foci was reduced even further – to approximately 1/6^th of the wild type level (data not shown). The reduction of the DSB number in Spo11 ^Gal/Gal Hop2 ^−/− females was more profound than in males, with only 1/4^th of the normal number of DMC1 foci being formed (Figure 4).

To understand what could be the reason for the reduced number of DSBs we examined the mRNA levels of the chimeric GAL4BD-SPO11 protein in testes and ovaries by quantitative PCR. Gal4BD-Spo11 male mice undergo meiotic arrest (Figure 1A and see below) and do not produce spermatids, which normally account for a large fraction of the testis germ cell population. Therefore, the cell type composition in testes of the Gal4BD-Spo11 and wild type mice is vastly different, creating a challenge for a meaningful comparison of gene expression levels. To minimize this difference we again employed Hop2 ^−/− mice, which undergo meiotic arrest at a similar stage 34 as the Gal4BD-Spo11 mice. We found that the mRNA level of the GAL4BD-SPO11 protein in the Gal4BD-Spo11 mice is significantly lower than the mRNA level of the SPO11 protein in wild type or in Hop2 ^−/− mice (Figure 5). More specifically, in Gal4BD-Spo11 males, the level of the beta isoform was reduced 3.3 fold compared to that in Hop2 ^−/− mice, and the level of the alpha isoform was reduced 5.3-fold (Figure 5A). In Gal4BD-Spo11 females the expression of the beta isoform was reduced 4.5 fold compared to wild type mice and the expression of the alpha isoform was reduced 3.8 fold (Figure 5B). These data indicate that the mRNA level of the chimeric GAL4BD-SPO11 protein is profoundly reduced and that the observed reduction in the number of introduced DSBs could be, at least in part, a consequence of this reduction. It is also conceivable that the addition of the Gal4 binding domain to the mouse SPO11 protein might compromise the activity of SPO11, contributing to the reduced DSB count in Gal4BD-Spo11 mice.

The formation of meiotic DSBs is followed by rapid phosphorylation of histone H2AX 35 . During break repair homologous chromosomes are brought together and align throughout their entire length within a proteinaceous structure called the synaptonemal complex (SC) (reviewed by 36 ). This coincides with the disappearance of the phosphorylated H2AX (γH2AX) and subsequent localization of the MLH1 protein to the sites of crossovers 37 . We analyzed homologous synapsis in Gal4BD-Spo11 spermatocytes by immunostaining the SCP1 protein, a component of the central element of the SC 38 (Figure 2). Although a number of chromosomes in Spo11 ^Gal/Gal spermatocytes appeared properly synapsed, no cells had undergone complete synapsis. A large fraction of chromosomes synapsed only partially or formed branched structures indicative of non-homologous synapsis. γH2AX staining was lost from those chromosomes that had undergone complete synapsis, indicating successful DSB repair (Figure 6). Partially synapsed and branched chromosomes retained substantial γH2AX staining. Synaptic defects were even more prominent in the Spo11 ^Gal/– spermatocytes, indicating that a further reduction in the number of DSBs further compromises homology search and homologous synapsis (Figures 2 and 6). Although some oocytes completed meiosis and resulted in a progeny, the majority showed defects in homologous synapsis. Nevertheless, unlike in spermatocytes (data not shown), MLH1 foci were evident in oocytes even when only a fraction of homologous chromosomes were successfully synapsed (Figure 7).

The X and Y chromosomes synapse and recombine at the short region of homology within the PAR 24 . Unsynapsed parts of X and Y trigger transcriptional silencing of sex chromosomes associated with their assembly into a specialized chromatin domain called the sex or XY body (reviewed by 39 40 41 ). At this stage the X/Y chromatin is decorated by a number of proteins including γH2AX. We found that a sex body was not assembled in the Spo11 ^Gal/Gal spermatocytes, as no area of dense and discrete γH2AX staining was observed (Figure 6). The failure of sex body formation and impairment of transcriptional silencing is common in recombination mutants with compromised DSB repair and homologous synapsis (reviewed by 42 ). Transcriptional silencing of sex chromosomes is required for meiotic progression beyond the pachytene stage in males 42 43 , potentially explaining the dimorphic phenotype (sterility of males and fertility or subfertility of females) in a number of mouse mutants with mild to moderate synaptic defects; e.g., Sycp3 ^−/− 44 45 , H2AX ^−/− 46 , Sycp2 ^−/− 47 , Dmc1 ^Mei11/+ 48 , Brca1 ^Δ11/Δ11 p53 ^+/− 49 , Spo11β-only 12 . Similar mechanisms are likely to be the reason for the dimorphic phenotype observed in Gal4BD-Spo11 mice.

We have mapped the distribution of DSBs in the Gal4BD-Spo11 mouse using anti-DMC1 chromatin immunoprecipitation (ChIP) followed by high throughput sequencing 19 50 . To increase the sensitivity of the hotspot detection we have also mapped DSBs in the Gal4BD-Spo11 mice on the Hop2 knockout background (Spo ^Gal/Gal Hop2 ^−/− ). This allowed identification of 13,445 hotspots. 98% of the hotspots detected in Spo11 ^Gal/Gal were also present in Spo11 ^Gal/Gal Hop2 ^−/− mice, reaffirming that the Hop2 mutation does not affect the distribution of meiotic DSBs genome-wide 23 . We have compared the distribution of Gal4BD-Spo11 hotspots to hotspots in wild type mice 19 and to those in Hop2 knockout mice. We found that 97.7% of the top 10,000 Gal4BD-Spo11 hotspots correspond to wild type hotspots. When all 13,445 Gal4BD-Spo11 hotspots are considered the overlap between Gal4BD-Spo11 and wild type hotspots is 94.7%. Therefore, the overall distribution of recombination hotspots in Gal4BD-Spo11 mice is not affected. The relative strength of the hotspots shared between wild type and Gal4BD-Spo11 mice is also highly correlated (Additional file 2: Figure S2).

There are 81,710 Gal4 consensus CGGN₁₁CCG sequences present in the mouse genome. Approximately 2% of DSB hotspots overlapped these sites in either wild type or Hop2 ^−/− mice (Table 1). However, when the Spo11 ^Gal/Gal Hop2 ^−/− mice were examined, the number of consensus-bearing hotspots increased to ~4%. These extra hotspots are weak, and although some of them are also apparent in the Spo11 ^Gal/Gal mice upon visual examination, they were beyond the detection threshold in this sample (Table 1). Out of 546 Gal4BD-Spo11 hotspots with a Gal4 consensus site 292 were also present in wild type mice (Figure 8). A prominent peak in the distribution of wild type hotspot motif 19 was found at the center of these hotspots, whereas the distribution of the Gal4 consensus appeared random. The remaining 254 Gal4BD-Spo11 hotspots with a Gal4 consensus were Gal4BD-Spo11-specific. We found that the Gal4 recognition sequence rather than the hotspot consensus motif is enriched at the centers of these hotspots. Furthermore, only 3% of these Gal4BD-Spo11-specific hotspots overlapped PRDM9-dependent histone H3 lysine 4 trimethylation marks (H3K4me3). In aggregate, these data indicate that a small fraction of DSBs in the Gal4BD-Spo11 mice are likely targeted through a PRDM9-independent mechanism by tethering of the GAL4BD-SPO11 protein to Gal4 binding sites. In addition to the H3K4me3 introduced by the PRDM9 protein, gene promoters and enhancers are also decorated with H3K4me3 51 52 53 . We have previously found that such sites represent preferred DSB formation loci in mice lacking Prdm9 19 . We now show that 84% of Gal4BD-Spo11-specific hotspots overlap transcription start sites. Importantly, the vast majority (96%) of Gal4BD-Spo11-specific hotspots overlap H3K4me3, indicating that H3K4me3 marks or/and events preceding H3K4 trimethylation are still essential when DSB targeting is GAL4-mediated.

The genotype of mouse strains is listed in the left column. Sequenced ssDNA tags indicate the number of ssDNA-derived fragments (e.g., those derived from ssDNA tails of DSBs) that have been sequenced for this analysis. The number of hotspots identified in each mouse strain is also indicated (column 3) and the overlap of these hotspots with the hotspots found in the wild type mice (“wild type” sample) is specified (column 5). The number of hotspots that contain Gal4 consensus motif (column 4, left) and the percentage of Gal4-containing hotspots in the total number of hotspots identified in the corresponding samples (column 4, right) are shown to demonstrate the increase in the Gal4-containing hotspots in the Spo11 ^Gal/Gal Hop2 ^−/− mice (see text for more explanation).

We have previously demonstrated that the PAR carries a large cluster of overlapping DSB hotspots 23 that are most likely required to ensure an obligate DSB (and CO) in the PAR. Although individual hotspots cannot be resolved within the hotspot cluster it is clear that a large fraction, if not all, PAR hotspots are fundamentally different from those in the rest of the genome, because their formation does not depend on PRDM9 19 . The formation of DSBs in the PAR has been reported to occur in two rounds: one (early) set is introduced at the same time as the majority of the autosomal breaks and the second one is introduced at a relatively late stage, presumably, by the alpha isoform of the SPO11 protein 12 . We found that Gal4BD-Spo11 mice are severely deficient in DSB formation in the PAR (Figure 9). The contribution of the DSBs in the PAR hotspot cluster to the total number of DSBs introduced genome-wide in Hop2 ^−/− mice is estimated as 0.33%, but drops over 30-fold (to 0.01%) in the Spo11 ^Gal/Gal Hop2 ^−/− (data not shown). Although this defect may be explained to some extent by the extremely low expression of the SPO11 alpha isoform in the mutant, the complete lack of SPO11 alpha has been reported to result in only a 3-fold reduction of the DSBs in the PAR 12 . Reduced expression of the beta isoform in the Gal4BD-Spo11 mice could be another reason for the DSB reduction. However, the overall number of DSBs in the GAL4BD-Spo11 mice (presumably, introduced by the beta isoform) is reduced only 2.2-fold. It is therefore likely that, in addition to reduced protein level, a specific defect of the GAL4BD-SPO11 protein contributes to the drastic reduction of the DSBs in the PAR.

The PAR is located at the end of the X and Y chromosomes. We therefore examined whether DSB formation in Gal4BD-Spo11 mice is also affected at the ends of autosomes. Indeed, we found that the strength of DSB hotspots near chromosome ends is significantly reduced in Spo11 ^Gal/Gal and Spo11 ^Gal/Gal Hop2 ^−/− mice (Figure 10), and as a consequence, many subtelomeric hotspots that are present in wild-type mice are not detected. This phenomenon was not observed in Hop2 knockout mice indicating that the effect is not related to the meiotic arrest in Gal4BD-Spo11 spermatocytes.

We propose that specific suppression of homologous recombination at subtelomeres of the Gal4BD-Spo11 mice indicates the potential difference in the DSBs formed in these areas compared to DSBs in the rest of the genome. For example, it is conceivable that DSB formation close to chromosome ends is influenced by telomere attachment to the nuclear periphery 54 or by a specific chromosome organization and chromatin structure in the telomere-adjacent regions. Specific interactions may occur in these areas between the DSB machinery and proteins enriched at telomeres and/or subtelomeres. In fact, one such protein has recently been implicated in the regulation of transcription 55 . Incidentally, SPO11 itself was reported to bind to the telomeres of mouse embryonic stem cells 56 . It is therefore possible that the GAL4BD-SPO11 protein alone or as a component of the DSB formation complex is deficient in overcoming the inhibitory effects of the specific chromatin environment of subtelomeric regions. Such a deficiency can result from (i) the inability of GAL4BD-SPO11 to interact with a particular protein or protein complex, (ii) attenuated GAL4BD-SPO11 DNA binding in the context of such a chromatin environment or even (iii) a defect in GAL4BD-SPO11 removal from DNA in these regions that would prevent processing of meiotic DSB ends and loading RAD51 and DMC1. It is important to note that PRDM9-dependent H3K4me3 is introduced normally in hotspot-depleted regions of Gal4BD-Spo11 subtelomeres (Figure 10), therefore PRDM9 accessibility to these areas is not an issue.

All animal procedures have been approved by the USUHS Animal Care and Use Committee or were performed according to NIH Guide for the Care and Use of Laboratory Animals.

Gal4BD-Spo11 mice were generated on a C57Bl/6 background. The targeting construct was based on the pLoxpNEO vector 58 (a gift from Dr. Chu-Xia Deng, NIH). The Spo11 gene was obtained from the BAC clone RP23-52C7 (CHORI). A 4.3 Kb HincII/SmaI fragment of the 5′ region of the mouse Spo11 gene including the first exon was subcloned, and the DNA corresponding to the Gal DNA binding domain (from pGBKT7 plasmid, Clontech) was inserted upstream of the start codon of the Spo11 gene. The resulting fragment was then cloned into the pLoxpNEO vector upstream of the NEO gene cassette (Additional file 1: Figure S1A). A 3.8 Kb XmaI/XmaI fragment containing exon II of the Spo11 gene was cloned between the NEO and TK genes. Gene targeting and blastocyst injections were performed by Xenogen Biosciences Corporation (Caliper Life Sciences). Confirmation of targeting was done by Southern blot analysis (Additional file 1: Figure S1B and S1C). To excise the NEO targeting cassette Gal4BD-Spo11 heterozygous mice were crossed with a CRE-expressing mice (Jackson Laboratories stock # 003724) and the excision was confirmed by PCR. Expression of the correctly spliced Gal4BD-Spo11α and Gal4BD-Spo11β isoforms have been confirmed by RT PCR followed by sequencing. Gal4BD-Spo11 mice were genotyped by PCR using the following primers: Gal/dir: CTCAGAGCGGCTCCGCATCC; Gal/rev: GGCGCCACGAGGAACCTTCC. Spo11 ^−/− mice (strain deltaSpo.BC/B6) have been produced by the targeted deletion of exons 2–6 of the Spo11 gene resulting in the absence of the SPO11 protein (Romanienko and Camerini-Otero, unpublished). The phenotype of this knockout strain is identical to the phenotype of the previously described Spo11 ^−/− strain Spo11 ^tm1Rdco 8 . Hop2 knockout mice have been previously described 34 . C57Bl/6 was used as a wild type strain and all mutant strains were on a C57Bl/6 background. Adult (2–6 month old) mice were used for the analyses.

Testes or ovaries of adult animals were dissected in PBS solution and placed in 10% formalin. Tissue sections and hematoxylin/eosin staining was performed by American Histo Labs Inc.

Total RNA was extracted using RNeasy plus mini kit (Qiagen) and the cDNA was prepared using the Trancriptor First Strand cDNA Synthesis Kit (Roche). Quantitative PCR was performed using a Maxima SYBR Green/Rox Kit (Fermentas) according to the manufacturer’s instructions for a 7500 Real-Time PCR system. Gene copy number was calculated with ABI SDS Software. PCR amplification of the coding region of β actin gene was used for normalization. The following primer pairs have been used: Spo11 beta: CTCTAGTTCTGAGGTTCTTACAGCT, GGACAATACTTTCAGAATCAGAGCG; Spo11 alpha: GCGTGGCCTCTAGGTTTGATGATT, TCATCGATGGCGCTGTCCAC; ActB: CCAACTGGGACGACATGGAG, CCAACTGGGACGACATGGAG.

Rabbit anti-DMC1 antibodies to a full-length protein were generated by New England Peptide and affinity purified. Mouse anti-SCP3 antibodies were previously described 34 . Rabbit anti-SCP1 and anti-SCP3 antibodies were a gift from C. Hoog. The following commercial antibodies have been used: Anti-H3K4me3, Milipore (07–473); anti-γH2AX, Trevigen (4411-PC-100); mouse anti-MLH1, BD Biosciences (551092); rabbit anti-RAD51 and goat anti-DMC1, Santa Cruz (H-92 sc-8349 and C-20 sc-8973, respectively). Secondary antibodies were from Jackson IR Laboratories.

Spermatocytes: Seminiferous tubules were chopped in RPMI 1640 high-glucose media (GIBCO-BRL). The cells were released from the tubules by pipetting and filtered through a 40 μm cell strainer (Falcon). The cells were pelleted and washed with RPMI. The resulting pellet was resuspended in 0.5% NaCl, added to the glass slides, and allowed to adhere for 10–15 min. The slides were fixed in 2% paraformaldehyde with 0.03% SDS for 3 min, 2% paraformaldehyde for 3 min, washed 3 times in 0.4% Photo-Flo 200 (Kodak) for 1 min, and air dried.

Oocytes: Ovaries were dissected in PBS at E15.5 for DMC1 staining and at E18.5 for MLH1 staining. Ovaries were placed in 20 μl of 100 mM sucrose, disrupted with tweezers and pipetted up and down until cell suspension was formed. The cells were added to 100 μl of 1% paraformaldehyde, 0.1% Triton X-100 solution that was spread over the slide. Slides were kept 2-4 h in humidified chamber at room temperature, then air dried. After four 1 min washes in 0.4% Kodak Photo Flo the slides were air dried again.

The slides were incubated with blocking solution (1% donkey serum, 0.3% BSA, 0.005% Triton X-100 in PBS) for 20 minutes at 37°C in a humidity chamber. Primary antibodies were diluted in blocking buffer and incubated under the same conditions for 1–2 hr. After two 5 min washes in 0.4% Photo-Flo/PBS solution, slides were blocked for an additional 5 min and incubated with secondary antibodies for 20 min at room temperature. The slides were washed twice with 0.4% Photo-Flo in PBS, rinsed twice with 0.4% Photo-Flo, and allowed to air dry.

Chromatin immunoprecipitation was done as previously described 19 23 with small modifications. Testes were fixed for 10 min in 1% formaldehyde. After quenching the tissue was homogenised, filtered through 40 μm cell strainer, and washed in the following buffers: 1) PBS (twice); 2) 0.25% Triton X-100, 10 mM EDTA, 0.5 mM EGTA, 10 mM Tris pH8; 3) 0.2 M NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM Tris pH8. Cells were lysed in 1.5 ml of the lysis buffer (1% SDS, 10 mM EDTA, 50 mM TrisCl pH8 with complete protein inhibitor cocktail (Roche) and the chromatin was sheared to ~1000 bp by sonication. The sample was dialyzed against ChIP buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM TrisHCl, 167 mM NaCl). Chromatin was incubated with appropriate antibodies overnight at 4°C followed by 2 h incubation with Protein G beads. Beads were washed in the following buffers: 1) 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM TrisHCl, 150 mM NaCl; 2) 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM TrisCl pH8, 500 mM NaCl; 3) 0.25 M LiCl, 1% Igepal, 1 mM EDTA, 10 mM TrisCl, pH8, 1% Deoxycholic acid; 4) TE (twice). The chromatin was eluted by 1% SDS, 0.1 M NaHCO3 pH9 at 65°C and crosslinking was reversed at 65°C overnight. DNA was deproteinized for 2 h at 45°C and DNA was purified with a MinElute Reaction Clean up kit (QIAGEN).

Sequencing library construction was done according to the SSDS protocol that was previously described 19 50 . Sequencing was performed on an Illumina HiSeq 2000 using the standard paired-end cluster generation kit and sequencing reagents.

Mouse strain

Replicate

Antibody

Method

Accession #

* Published data 19 .

Spo11 ^Gal/Gal

1

Dmc1

SSDS

GSM1179920

Spo11 ^Gal/Gal

2

Dmc1

SSDS

GSM1179921

Spo11 ^Gal/-

1

Dmc1

SSDS

GSM1179919

Spo11 ^Gal/Gal Hop2 ^−/−

1

Dmc1

SSDS

GSM1179922

Hop2 ^−/−

1

Dmc1

SSDS

GSM1179917

Hop2 ^−/−

2

Dmc1

SSDS

GSM1179918

Wild type (C57Bl/6)

*

Dmc1

SSDS

GSM869781*, GSM869782*

Spo11 ^Gal/Gal

1

H3K4me3

ChIP-Seq

GSM1179925

Spo11 ^Gal/Gal Hop2 ^−/−

1

H3K4me3

ChIP-Seq

GSM1179926

Hop2 ^−/−

1

H3K4me3

ChIP-Seq

GSM1179924

The list of samples used in this study

Hotspot identification and peak calling

We have previously shown that specific usage of ssDNA increases DSB detection sensitivity 50 . For all SSDS samples, the computational pipeline described in the aforementioned work was used to align paired-end reads to the mouse reference genome (mm9) and to identify ssDNAs. We subsequently discarded ssDNA fragments where either the first or second end read had a quality score < 30. We also retained only a maximum of 20 duplicate ssDNA fragments at any locus. Data for replicates were pooled and DSB hotspots were identified using the peak calling method described in 19 . H3K4me3 reads were aligned to the mouse mm9 genome using CASAVA 1.8. H3K4me3 peaks were called using MACS 1.3.7 and the parameters described in 19 .

The effect of using different overlap windows between hotspots has been described in 19 . Overlaps between hotspots were limited to the central 400 nt. Overlaps between hotspots and H3K4me3 were limited to the central 1 Kb. Overlaps between hotspots and Gal4 binding consensus were limited to the central 2 Kb.

GAL4 binding sites were identified using an exhaustive genomic search for the consensus CGGN₁₁CCG. No mismatches were permitted. Sites matching the putative C57Bl/6 PRDM9 binding site were identified with MAST, using standard parameters.