MOF influences meiotic expansion of H2AX phosphorylation and spermatogenesis in mice

Three waves of H2AX phosphorylation (γH2AX) have been observed in male meiotic prophase I: the first is ATM-dependent and occurs at leptonema, while the second and third are ATR-dependent, occuring at zygonema and pachynema, respectively. The third wave of H2AX phosphorylation marks and silences unsynapsed chromosomes. Little is known about H2AX phosphorylation expands to chromatin-wide regions in spermatocytes. Here, we report that histone acetyltransferase (HAT) MOF is involved in all three waves of H2AX phosphorylation expansion. Germ cell-specific deletion of Mof in spermatocytes by Stra8-Cre (Mof cKO) caused global loss of H4K16ac. In leptotene and zygotene spermatocytes of cKO mice, the γH2AX signals were observed only along the chromosomal axes, and chromatin-wide H2AX phosphorylation was lost. In almost 40% of early-mid pachytene spermatocytes from Mof cKO mice, γH2AX and MDC1 were detected along the unsynapsed axes of the sex chromosomes, but failed to expand, which consequently caused meiotic sex chromosome inactivation (MSCI) failure. Furthermore, though RAD51 was proficiently recruited to double-strand break (DSB) sites, defects in DSB repair and crossover formation were observed in Mof cKO spermatocytes, indicating that MOF facilitates meiotic DSB repair after RAD51 recruitment. We propose that MOF regulates male meiosis and is involved in the expansion of all three waves of H2AX phosphorylation from the leptotene to pachytene stages, initiated by ATM and ATR, respectively.

Introduction Meiosis, the process that generates haploid daughter cells, is critical for sexual reproduction in eukaryotes. The key to normal meiotic progression is the fidelity of chromosome interactions, during which homologous chromosomes undergo pairing, synapsis, recombination and proper segregation. The formation and repair of programmed DNA double strand breaks (DSBs) are essential for chromosome dynamics in meiotic prophase I. These DSBs are repaired through two pathways, a synthesis-dependent strand annealing (SDSA) pathway to form noncrossover products and the classical pathway through double Holiday Junction (dHJ) formation to create both crossover and non-crossover products [1,2]. The DSBs that persist until pachynema are more likely to be repaired through the classical pathway to generate crossovers [3]. The mismatch repair complex MLH1-MLH3 is required for the resolution of dHJs to form the class I crossovers that account for approximately 90% of crossovers in mice [4].
The dynamics of γH2AX distribution is one of the major characteristics of meiosis and is commonly used as a marker for meiotic progression. In somatic cells, the phosphorylation of H2AX after DSB formation is first initiated by ATM and/or DNA-PK [5], then expanded by the recruitment of MDC1, which is facilitated by MOF mediated H4K16ac [6][7][8]. Despite its well-known role in the DNA damage response (DDR), many reports highlight that γH2AX should not be considered simply as a DSB marker [9,10]. In mammalian meiotic prophase I, three waves of H2AX phosphorylation can be observed, all of which are initiated and expanded through pathways that are different from those in somatic cells. During leptonema, γH2AX signals are initiated solely by ATM in a DSB-dependent manner and are distributed throughout the nucleus. This wave of γH2AX signals can be found in wild type (WT) and Atr knockout mice [11], but not in mice deficient for Atm or Spo11 [12]. The second wave is initiated by ATR at DSBs. This wave of H2AX phosphorylation can be observed in Atm-deficient spermatocytes at early-mid zygonema and is lost when Spo11 or Atr is deleted [11][12][13]. The third wave is also mediated by ATR and marks chromatin associated with unsynapsed axes at the zygotene/pachytene transition [14,15]. Unlike the first two waves of H2AX phosphorylation expansion at leptonema and zygonema, the third wave can be observed in Spo11-deficient spermatocytes [12]. This wave is responsible for meiotic silencing of unsynapsed chromosomes (MSUC), and occurs on the unpaired regions of the sex chromosomes, resulting in meiotic sex chromosome inactivation (MSCI) and formation of the sex body [16]. Mutations causing defective expansion of H2AX phosphorylation in pachytene spermatocytes lead to MSCI failure. This triggers the MSCI checkpoint before late pachynema, which subsequently leads to apoptosis of pachytene spermatocytes [16]. This later wave of H2AX phosphorylation is reported to be initiated by the localization of BRCA1 to the unsynapsed axes [17], which recruits ATR for H2AX phosphorylation and MDC1 for γH2AX signal expansion [11,18]. MDC1 also contributes to the DDR in somatic cells [7,19]. However, spreading of H2AX phosphorylation in leptotene and zygotene spermatocytes is unaltered in Mdc1 -/spermatocytes [18], and is therefore regulated by unknown mechanisms.
MOF is a major HAT for histone H4 lysine 16 (H4K16) in somatic cells, and belongs to the MYST family of acetyltransferases [20]. Unlike most histone modifications, acetylation of H4K16 is unique in regulating higher order chromatin structures beyond the level of nucleosomes, and removes the natural barrier to damaged DNA sites to facilitate repair protein access to DNA breaks [21,22]. Consistent with this, MOF is essential for the recruitment of MDC1 [23], a key step in the expansion of H2AX phosphorylation after DSB formation in somatic cells. It has also been reported that MOF is indispensable for the recruitment of BRCA1 and 53BP1 to damage sites after DSB formation [23]. In addition to its known function in recruitment of DDR factors, MOF is essential during DSB repair. In somatic cells, Mof knockout results in persistent DSBs and reduces homologous recombination (HR) efficiency, which can be rescued by HDAC inhibition [23,24], demonstrating a crucial role for H4K16ac in HR. H4K16ac is also detected in meiotic cells, and deacetylation of H4K16ac during oocyte maturation is critical for proper chromosome segregation [25]. However, the meiotic function of MOF and H4K16ac in male germ cells has not been reported.
In this study, we generated male mice with germ cell-specific Mof deletion mediated by Stra8-Cre (Mof cKO mice). We show that MOF is involved in the expansion of chromatinwide H2AX phosphorylation in leptonema and zygonema. MDC1 recruitment and MSCI are also disturbed in Mof deleted pachytene spermatocytes, with absence of XY-H2AX phosphorylation expansion, which may partly contribute to the germ cell loss observed in Mof cKO mice. RAD51 foci persist in Mof cKO spermatocytes until diplonema and are associated with a significant reduction in crossover markers and chiasma formation. Together, these data indicate that the expansion of different waves of H2AX phosphorylation, efficient DSB repair, and the formation of crossovers during male meiosis are affected by MOF, thus highlighting novel roles for this protein.

Mof is highly expressed in testes particularly in spermatocytes
Measurement of mRNA levels by real-time PCR indicated that Mof is differentially expressed in mouse tissues, with much higher expression in testes than other organs ( Fig 1A). We further analyzed Mof expression in testes of mice at different developmental ages. Mof mRNA level started to increase at 12 dpp, then dramatically increased from 16 dpp onward, with the highest expression detected in adult testes (Fig 1B), indicating that Mof was highly expressed in meiotic and post-meiotic cells. To confirm this, we isolated meiotic prophase I spermatocytes, round spermatids and Sertoli cells from mouse testes (S1 Fig), and analyzed Mof expression by real-time PCR. The results showed that Mof was highly expressed in pachytene spermatocytes and round spermatids, but had much lower expression in Sertoli cells (Fig 1C).

Mof deletion in spermatocytes reduces testicular size and causes germ cell loss in mice
To determine the functional roles of Mof in spermatogenesis, we generated mice with a germ cell-specific Mof deletion, by crossing Mof-LoxP mice (Fig 2A) with Stra8-Cre mice, which express CRE recombinase in male germ cells from 3 dpp [26], before they initiate meiosis. Successful deletion of the Mof gene was confirmed by the absence of MOF protein in isolated cKO The Mof cKO males had very small testes (Fig 2C), with the ratio of testis to body weight decreased significantly when compared with controls (Fig 2D), suggesting a functional role for Mof in spermatogenesis. To further explore this function, we compared testis sections from 8 week old control and Mof cKO animals after H&E staining. As shown in Fig 3A, control testes have normal spermatogenesis and elongated spermatids in almost all tubules, while testes from  Fig 3B). Furthermore, TUNEL-positive cells were increased in testes from adult Mof cKO mice compared with the control (Fig 3C and S4 Fig). Meiotic progression is also disturbed in Mof cKO mice, with a significantly higher ratio of leptotene to zygotene spermatocytes and lower ratio of pachytene spermatocytes compared with the control (S5 Fig), which is consistent with the apoptosis of primary spermatocytes. The sperm count in epididymides of Mof cKO mice also dropped to about 10% of that in control animals (Fig 3D and 3E). Altogether, these data demonstrate that Mof is involved in the maintenance and development of germ cells at different stages, especially those in meiotic prophase I.

Chromatin-wide expansion of H2AX phosphorylation fails in Mof cKO spermatocytes
The three waves of H2AX phosphorylation expansion are key markers for progression of meiotic prophase I. To further understand how MOF functions in this stage, we analyzed H2AX phosphorylation in control and Mof cKO spermatocytes by immunofluorescence staining on surface spread nuclei. All three waves of H2AX phosphorylation in Mof cKO spermatocytes were normally initiated, but expansion of the first two expansion waves to the whole chromosomes was blocked in 85.0% of cells at leptonema and 84.5% of cells at zygonema (S6 Fig). The γH2AX signals were confined to the unsynapsed axes and proximal regions of sex chromosomes in these cells, instead of forming sex body as in control mice ( Fig 4C). However, this phenotype was not observed in late pachytene and diplotene spermatocytes ( Fig 4D and S6  Fig), suggesting that cells with insufficient XY γH2AX are eliminated prior to late pachynema. These results indicate that MOF is involved in all three waves of H2AX phosphorylation expansion during meiosis in male mice.

Mof deletion leads to defects in DSB repair and fewer crossovers in spermatocytes
One major function for the expansion of H2AX phosphorylation is to recruit proteins related to DSB repair, such as the strand exchange protein RAD51 [28]. Thus, the defect in the first two waves of H2AX phosphorylation expansion in Mof-deficient spermatocytes should disturb RAD51 recruitment. To test this, we investigated the accumulation of RAD51 on chromosome axes in Mof cKO spermatocytes. Interestingly, in Mof cKO leptotene and zygotene spermatocytes, almost the same number of RAD51 foci localized along the meiotic chromosome axes as in control spermatocytes (S7 Fig), indicating that the recruitment of RAD51 was not affected by Mof deletion. However, by late pachynema, when most RAD51 foci were lost from paired regions of both autosomes and XY chromosomes in control spermatocytes, Mof cKO spermatocytes still exhibited RAD51 foci on axes (Fig 5A and 5C). Some of these foci persisted until diplonema (Fig 5E). Statistical analysis revealed that the number of RAD51 foci in these stages was significantly higher in Mof cKO than in control spermatocytes ( Fig 5B, 5D and 5F). Similar results were also observed for DMC1 (S8 Fig), indicating that DSB repair in pachynema is compromised in Mof-deficient spermatocytes.
DSBs repaired in pachytene stage generate at least one crossover on each pair of homologous chromosomes [29]. To determine whether crossover formation was also disturbed by Mof deletion, we assessed the number of MLH1 foci in pachytene spermatocytes of Mof cKO mice. Consistent with previous reports [30], the average number of MLH1 foci per cell was 24.38 in control pachytene spermatocytes. However, it was reduced to 18.66 in Mof cKO spermatocytes (Fig 6A and 6B). As meiosis progresses, crossovers generated in pachytene spermatocytes develop into chiasmata in diplotene spermatocytes [31]. We thus analyzed chiasmata in late diplotene spermatocytes. In Mof cKO mice as many as 71.4% of late diplotene spermatocytes bore at least one pair of univalents, which was significantly higher than the 7.7% control level (Fig 6C and 6D). These results collectively show that Mof plays an important role in both DSB repair and crossover formation during male meiosis.

MSCI does not occur in some Mof cKO spermatocytes
The third wave of H2AX phosphorylation expansion, which happens at pachynema, is a wellknown marker for MSCI [32]. To determine whether the failure in chromatin-wide expansion of H2AX phosphorylation in Mof cKO spermatocytes would also disturb MSCI, we first examined global transcriptional status in pachytene spermatocytes by immunostaining for trimethylated H3K4 (H3K4me3), an epigenetic modification enriched in transcriptionally activated regions [33]. As reported, the sex chromosomes formed the sex body, from which Essential roles of MOF in ƳH2AX signal expansion H3K4me3 was largely excluded, in controls. However exclusion of H3K4me3 was not observed in 36.7% of early-mid pachytene spermatocytes in Mof cKO mice (Fig 7A and 7B). To confirm the transcriptional activity of Mof cKO sex chromosomes, we measured the distribution of RNA polymerase II (RNA Pol II) in pachytene spermatocytes [34]. Like H3K4me3, RNA Pol II was excluded from the sex body in pachytene spermatocytes in control mice. However, in 38.3% Mof-deficient early-mid pachytene spermatocytes, strong RNA Pol II staining was observed around the sex chromosomes (S9 Fig). We also stained pachytene spermatocytes for alpha thalassemia/mental retardation X-linked (ATRX), a protein encoded by an X-linked gene that is silenced by MSCI and not detected in control spermatoyctes. In Mof cKO pachytene spermatocytes, ATRX was observed in 42.9% pachytene spermatocytes (Fig 7C and 7D), which further indicated that Mof deletion caused MSCI failure.
To further confirm these observations, we analyzed the expression of representative genes on autosomes and unpaired regions of sex chromosomes in isolated pachytene/diplotene cells by real-time PCR. Consistent with the cytological results, a remarkable increase in the expression of X-linked genes such as Tktl1, Usp26 and Atrx, was observed in Mof cKO spermatocytes when compared with control spermatocytes (Fig 7E). Similar results were found for the Essential roles of MOF in ƳH2AX signal expansion Y-linked genes Ube1y and Rbmy (Fig 7E), whereas expression of genes on autosomes such as Dazl and Setx remained unaltered by Mof deletion (Fig 7E). These results indicate that Mof has an important role in MSCI and further corroborate that expansion of H2AX phosphorylation during sex body formation and MSCI in pachytene spermatocytes is facilitated by MOF.

MOF facilitates the expansion of H2AX phosphorylation by recruiting MDC1 but not BRCA1 in spermatocytes
Facilitation of γH2AX expansion by MOF could occur through two mechanisms: up-regulation of genes related to the expansion or enhanced localization of specific required proteins. Levels of mRNA for genes involved in MSCI including BRCA1, ATR, ATM and MDC1 in isolated spermatocytes did not show significant differences between control and Mof cKO spermatocytes (S10 Fig). This indicated that MOF may facilitate sex body formation and MSCI through the recruitment rather than expression of sex body formation related components.
MSCI begins with recruitment of BRCA1 and ATR along the unpaired axes, followed by MDC1-mediated ATR and H2AX phosphorylation expansion over whole chromatin [18]. To determine whether defective H2AX phosphorylation expansion in Mof cKO spermatocytes were caused by the failure in BRCA1, ATR and/or MDC1 recruitment, we first performed immunostaining for these proteins on surface spread spermatocytes from control and Mof cKO mice. Surprisingly, BRCA1 was proficiently recruited to the axes of unpaired sex chromosome regions in all Mof deleted pachytene cells (Fig 8A), indicating undisturbed MSCI initiation. ATR signals were strong along the unpaired axes and weaker in the sex chromatin regions in almost all control early-mid pachytene spermatocytes (Fig 8B and 8C). In contrast, in more than 40% Mof cKO pachytene spermatocytes, ATR staining was confined to the unpaired axes (Fig 8B and 8C). These findings indicate that the defect in sex body formation in Mof-deficient spermatocytes is not due to axial recruitment of ATR but to failed expansion of ATR.
We then examined MDC1 localization, to determine whether its recruitment was disturbed by Mof deletion. Interestingly, no MDC1 staining was found in 37.1% of Mof deleted earlymid pachytene spermatocytes. In control spermatocytes all pachytene cells showed strong MDC1 staining of the sex body (Fig 8D and 8E), indicating that MSCI failure in some Mof cKO spermatocytes was due to failed MDC1 recruitment to the sex body.

Discussion
As a major HAT for H4K16, MOF is well known for its functional role in chromatin remodelling, the DDR and DSB repair. Previous work on HDACs found that H4K16ac deacetylation is critical for proper chromosome segregation and kinetochore function [25]. Mof has been found to be indispensable for embryonic stem cell development and its deficiency results in mouse embryonic lethality [35,36]. Our recent study also revealed a role for MOF and H4K16ac in oocyte development [37]. However, Mof function in male meiotic cells, especially in the three waves of meiosisspecific H2AX phosphorylation expansion and programmed DSB repair, has not been studied. In this study, we report that Mof deletion in spermatocytes using Stra8-Cre causes global loss of H4K16ac as well as defective spermatogenesis. In Mof cKO mice, the γH2AX signal failed to expand chromatin-wide in leptotene and zygotene spermatocytes, and to the sex body region in about 40% of early-mid pachytene spermatocytes. We also observed persistent RAD51 foci, as well as a significant reduction in crossover number, in Mof cKO spermatocytes, consistent with a defect in chiasma formation in Mof cKO mice. To be noted, we also observed some tubules with no spermatocytes in cKO testes, which may due to the loss of Mof in some spermatogonia, and consequently cause germ cell loss before the initiation of meiosis. Indeed, we did find that Mof specific deletion in spermatogonia leads to completely loss of differentiated spermatogonia (S11 Fig). Our study demonstrates for the first time that one major function of Mof in meiotic prophase I is to facilitate the three waves of H2AX phosphorylation expansion. Mof deletion blocked the chromatin-wide expansion of H2AX phosphorylation in most leptotene spermatocytes. However, γH2AX foci could still be observed, indicating that Mof is not required for the initiation of H2AX phosphorylation at this stage (Fig 4). A similar pattern of γH2AX staining was observed in Mof cKO zygotene spermatocytes (Fig 4). We suggest that MOF is involved in the expansion, but not initiation, of the first two waves of H2AX phosphorylation. These two waves of chromatin-wide H2AX phosphorylation expansion are preserved in Mdc1-deficient mice [18] and thus MDC1-independent.
The third wave of H2AX phosphorylation expansion is well known for its specific localization and essential role in pachytene spermatocytes. This wave is responsible for MSCI [13]. During the zygotene/pachytene transition, MSCI is initiated by the recruitment of BRCA1 to the axes of Synaptonemal Complexes (SCs) [18,38,39], which consequently recruits ATR and MDC1, leading to phosphorylation and expansion of H2AX [18]. We found that MOF influences the expansion of H2AX phosphorylation during pachynema by recruiting MDC1 to the sex chromosomes. Notably, some Mof cKO pachytene spermatocytes exhibit normal γH2AX / MDC1 recruitment, suggesting that Mof deletion does not occur in all germ cells. We suspect that some cells that complete pachynema are those with residual MOF. This possibility is supported by the finding that a comparably higher rate of positive H4K16ac staining in surviving Mof cKO diplotene cells compared with pachytene cells (S2C and S2D Fig). The rest of the cells that complete pachynema may come from those pachytene cells with MDC1 recruitment and the expansion of H2AX phosphorylation despite the loss of MOF. This possibility is supported by the fact that H3K4me3, RNA Pol II (S12 Fig) and MDC1 (S13 Fig)  Though H2AX phosphorylation has been reported to facilitate early DSB repair, especially the recruitment of RAD51, we found no defects in these processes in Mof cKO spermatocytes (S7 Fig). This result might be explained by the observation that formation of γH2AX foci remains undisturbed in leptotene and zygotene spermatocytes after Mof deletion. Notably, at pachynema more RAD51 and DMC1 foci were observed in Mof cKO spermatocytes than in controls, suggesting a defect in HR after strand invasion. This is consistent with the decreased number of crossovers in Mof cKO spermatocytes, indicating that MOF may participate in processing HR intermediates. One possible mechanism could be via migration of dHJs. It has been reported that H3K4me3 mediated by PRDM9 may facilitate the dHJ migration by creating an extended nucleosome-depleted region (NDR), which restricts the localization of crossovers [40,41]. MOF can recognize H3K4me3 by its chromodomain [42,43], raising the possibility that MOF may recognize the H3K4me3 sites modified by PRDM9 [41], and initiate H4K16ac around that region to stabilize the NDR. Despite their importance in expansion of H2AX phosphorylation, Mof and H4K16ac are present at comparably low levels at leptonema and zygonema. We suggest that low levels of Mof and H4K16ac during these stages is sufficient to ensure accessibility of DDR factors, as observed in somatic cells [23,42,44], and that higher levels are required at pachynema for DSB repair after stand invasion.

Ethics statement
Animal experiments were carried out under authority of the Institutional Animal Care Committee of the University of Science and Technology of China (Licence USTCACUC1301021). All animal studies were conducted in accordance with the "Guideline on Humane Treatment of Laboratory Animals" (MOST, 2006) and procedures approved by the Institutional Animal Care Committee of the University of Science and Technology of China.

Experimental mice
Mof fl/fl mice, homozygous for a floxed allele of Mof, and Stra8-Cre transgenic mice were described previously [26,35]. They were mated to generate male Stra8-Cre;Mof fl/+ mice, which are then crossed with Mof fl/fl females. The offsprings were genotyped by PCR and the primers were list in S1 Table. Only the males with Cre band, flox band and minus band are considered as Mof cKO (Stra8-Cre;Mof fl/-) mice. All mice were kept under controlled photoperiod conditions (lights-on 08:00-20:00) and supplied with food and double distilled H 2 O ad libitum.

Cell isolation
Spermatogenic and Sertoli cells were isolated as we described previously [45,46]. Briefly, after the mice were euthanized by cervical dislocation, testes were de-capsulated under a dissection microscope. The seminiferous tubules were pooled and washed with PBS thrice and incubated with 1 mg/ml collagenase I (C7661; Sigma, MO, USA) and 0.5 mg/ml DNase I (D4527; Sigma) in Dulbecco modified Eagle medium (DMEM) (12800-017; GIBCO, CA, USA) for 15 min at 34˚C on a shaker, then washed twice with DMEM/F12 and further digested with 1 mg/ml collagenase I, 0.5 mg/ml DNase I, 1.5 mg/ml hyaluronidase (H3506; Sigma) and 1 mg/ml trypsin (T8003; Sigma) for 15 min at 37˚C. After dissociation, cells were filtered through a nylon mesh (70 μm) and washed with DMEM with 0.5% bovine serum albumin (BSA, A1933; Sigma). Cells were then suspended in 25 ml of 0.5% BSA/DMEM and loaded into the velocity sedimentation apparatus (STA-PUT; Proscience, ON, Canada), and separated on a 2-4% BSA gradient in DMEM for separation by sedimentation at unit gravity. Spermatogenic cells of different substages were identified by phase contrast microscopy and confirmed by immunostaining (S1 Fig), then stored at -80˚C until use. For the isolation of Sertoli cells, the final dispersed cells were placed into culture dishes in DMEM/F12 containing 10% fetal bovine serum (HyClone, SV30087-02) and after 1 day culture, the cells were treated with a hypotonic solution (20 mM Tris, pH 7.4) for 1 minute to remove germ cells and and confirmed by immunostaining (S1 Fig). Leptotene/zygotene spermatocytes and Sertoli cells were isolated from 12-14 dpp mice. Pachytene/diplotene spermatocytes and round spermatids were isolated from 8 week mice.

Hematoxylin and eosin (H&E) staining, immunohistochemistry and TUNEL staining
The control (Stra-Cre;Mof fl/+ ) and Mof cKO mice were euthanized by cervical dislocation and testes were immediately fixed in Bouin's solution for H&E staining or in 4% paraformaldehyde for immunohistochemistry. For immunohistochemistry, tissue sections were incubated overnight at 4˚C with the primary antibody of SOX9 (1:200; AB5535, Millipore), VASA (1:100; ab13840, Abcam, Cambridge, UK) and acetyl Histone H4K16 (1:100; 07-329, Millipore). For H4K16ac staining, sections were treated with metal-3,3'-diaminobenzidine (DAB) after introduction of horseradish peroxidase (HRP) with secondary antibody, and then counterstained with hematoxylin. For VASA staining, sections were incubated with secondary antibody AF488 Donkey Anti-Rabbit IgG (H+L) (1:100, A21206, Molecular Probes, OR, USA) and DAPI. TUNEL staining was performed on testicular sections according to the instructions provided with the cell death detection kit (11684795910, Roche). To reduce inter-experiment variations, testes from control and knockout mice were processed simultaneously. Images were captured using a Nikon Eclipse 80i microscope equipped with a digital camera (Nikon DS-Ri1) or a CCD camera (Hamamatsu) and analyzed using the 694 NIS-Element Microscope imaging software (Nikon).

Sperm counting
The epididymides and vasa deferentia were removed from 8 week old control and cKO mice, incised several times and incubated in 1 ml buffer containing 75 mM NaCl, 24 mM EDTA and 0.4% BSA (Sigma) at 37˚C with 5% CO2 for 30 minutes to allow sperm release from the epididymides. Sperm were collected after a nylon-mesh filtration and counted with a haemocytometer.

Real-time PCR
RNA isolation, RT-PCR and Real-time PCR were performed as previously described [46,47]. All PCR primers used were listed in S2 Table. For Real-time PCR analysis, Cm values of samples were normalized to the corresponding Cm values of Actb. Quantification of the fold change in gene expression was determined by the comparative Cm method.

Statistical analysis
All data are presented as mean ± standard deviation (SD). The Mann-Whitney test was applied to compare the number of RAD51, DMC1 and MLH1 foci between control and Mof cKO spermatocytes. The student's t-test was applied to compare the testis/body weight, sperm count, TUNEL and mRNA levels between control and Mof cKO mice. Cell population differences between control and Mof cKO mice were analyzed by the Chi-square test. Statistical significance was set at P < 0.05.