Maintenance DNA methylation in pre-meiotic germ cells regulates meiotic prophase by facilitating homologous chromosome pairing

Meiosis, the process by which haploid gametes are generated via two successive cell divisions, is essential for sexual reproduction in eukaryotes. In mammals, spermatogenesis is initiated by spermatogonia that undergo mitotic division to maintain homeostasis through self-renewal and progressive differentiation (Phillips et al., 2010). These spermatogonia consist of morphologically heterogenous populations and are classified into three fractions: type A, intermediate (In) and type B (according to their stage of differentiation) (de Rooij and Russell, 2000). Based on other criteria, spermatogonia are functionally classified into two fractions: one that retains stem cell capacity (A_s to A_al spermatogonia; or undifferentiated spermatogonia); and another that possesses only progenitor activity (A₁-A₄, In and type B spermatogonia; or differentiating spermatogonia) (de Rooij and Russell, 2000; Shirakawa et al., 2013). Type B spermatogonia, in contrast, enter meiotic prophase only after DNA replication.

During meiotic prophase, homologous chromosomes pair and connect along their entire length by their synaptonemal complexes and are locked to each other by recombination. Consistent with this, faithful synapsis and subsequent crossovers are prerequisites for segregation of homologs at meiosis I and for sister chromatid separation in meiosis II (Handel and Schimenti, 2010). Previous studies indicate a role for heterochromatin-related elements, such as centromeres, in facilitating pairing and recombination (Prieto et al., 2004; Dernburg et al., 1996; Ding et al., 2004; Kemp et al., 2004; Tsubouchi and Roeder, 2005). In particular, epigenetic modifications at pericentric heterochromatin (PCH) mediate pairing and synapsis of homologous chromosomes. Previous reports have revealed that histone H3K9 methylation mediated by SET-domain proteins, such as SUV39H1/2 and G9a/GLP, and recognition of these epigenetic marks by the reader protein HP1γ (CBX3), initiate homologous chromosome pairing by facilitating PCH clustering during early prophase (Peters et al., 2001; Tachibana et al., 2007; Takada et al., 2011). Dynamic regulation of PCH clustering may play a role in the early phase of homologous chromosome searching by constraining spatial movement of meiotic chromosomes (Takada et al., 2011).

In addition to H3K9 methylation, DNA methylation is another heterochromatin-associated epigenetic modification that regulates repetitive DNA elements, including PCH (Grewal and Jia, 2007). The role of DNA methylation during spermatogenesis in mammals has previously been examined by using a Dnmt3l (DNA methyltransferase 3 like) mutant (Bourc'his and Bestor, 2004; Hata et al., 2006; Webster et al., 2005; Zamudio et al., 2015). DNMT3L mediates de novo DNA methylation and is transiently expressed in pro-spermatogonia. Consistent with this role, Dnmt3l deletion causes meiotic failure, in association with demethylation and activation of repeated elements in meiotic spermatocytes. Intriguingly, DNA methylation and transcriptional silencing at PCH remain unchanged in Dnmt3l mutants (Bourc'his and Bestor, 2004).

While de novo DNA methylation is regulated by DNMT3A, 3B and 3L, DNMT1 [DNA (cytosine-5)-methyltransferase 1] mediates maintenance DNA methylation in cooperation with the SRA protein NP95 (nuclear protein 95 kDa, also known as UHRF1) (Sharif et al., 2007, 2016). Indeed, deletion of Np95 or Dnmt1 causes widespread demethylation in the genome, including PCH (Sharif et al., 2007; Dong et al., 2019). Consistent with this notion, we have previously shown that Np95 deletion in undifferentiated spermatogonia causes differentiation arrest during transition to the differentiating spermatogonia, indicating that DNA methylation controls differentiation from stem to progenitor cells (Shirakawa et al., 2013). Furthermore, conditional Np95 deletion by a Stra8-Cre driver, in differentiating spermatogonia, revealed that NP95 plays a role in silencing endogenous retrotransposons in spermatocytes (Dong et al., 2019). Despite these reports, the role of DNA methylation in pairing and synapsis of homologous chromosomes has not been fully elucidated.

In the present study, we focus on the impact of maintenance DNA methylation on PCH-mediated regulation of homologous chromosome pairing and synapsis, by comparing of Np95 or Dnmt1 mutant meiotic spermatocytes. We show that NP95 and DNMT1 are co-expressed in the spermatogonia, and mediate maintenance methylation at repetitive elements, including PCH. We reveal that NP95 and DNMT1 are required for synapsis of homologous chromosomes and promote PCH clustering during meiotic prophase. Our work therefore reveals a previously unappreciated role for maintenance DNA methylation, mediated by DNMT1 and NP95, in spermatogenesis.

We investigated the expression of NP95 and DNMT1 during spermatogenesis by immunohistochemistry. We used several differentiation markers to identify cell types expressing NP95 or DNMT, or both, in adult seminiferous tubules. Undifferentiated spermatogonia were identified as PLZF positive (henceforth PLZF^pos)/KIT negative (henceforth KIT^neg), while differentiating spermatogonia were identified as PLZF^neg/KIT^pos (Shirakawa et al., 2013). STRA8 or synaptonemal complex protein SYCP3 (a component of the axial element) was used to detect morphological characteristics of preleptotene or meiotic spermatocytes, respectively (Zhou et al., 2008; Yuan et al., 2000). Round and elongating spermatids were distinguished by morphology after DAPI staining. We examined NP95 expression pattern in 8-week-old adult testes, and found that, while NP95 was expressed in PLZF^pos/KIT^neg undifferentiated spermatogonia and in PLZF^neg/KIT^pos differentiating spermatogonia, this expression was downregulated in pre-leptotene, leptotene and zygotene spermatocytes (Fig. 1A-C). Curiously, NP95 was re-expressed in pachytene spermatocytes (Fig. 1C, Fig. S1A-C), and modest NP95 expression was detected until the early round spermatid stage (Fig. 1A,B). We confirmed NP95 expression in meiotic prophase by immunofluorescence (IF) analyses on dried-down spreads of spermatocytes, which express SYCP3 as a stage-specific marker (Lammers et al., 1994), and found that NP95 was indeed expressed in pachytene and diplotene spermatocytes. Of note, strong accumulation of NP95 was detected at the telocentric ends of chromosomes, demarcating PCH (Fig. S1D).

Given that DNMT1 is a key partner of NP95 for regulation of maintenance DNA methylation, we examined whether DNMT1 was also co-expressed with NP95 during spermatogenesis. Consistent with previous reports, PLZF^neg/KIT^pos differentiating spermatogonia (representing type A₁-A₄ and up to type B cells) were mostly positive for DNMT1 (Fig. 1B-D) (de Rooij and Russell, 2000; Shirakawa et al., 2013; Handel and Schimenti, 2010). Preleptotene cells, however, exhibited modest expression of DNMT1, but not NP95 (Fig. S1C). These results show that NP95 and DNMT1 are co-expressed in undifferentiated and differentiating spermatogonia, but not in meiotic spermatocytes (Fig. 1E).

We next determined the role of NP95 for male meiosis. To bypass a previously identified checkpoint for NP95 activity at the transition from undifferentiated to differentiating spermatogonia, we used the Stra8-Cre deleter, which induces gene deletion in differentiating spermatogonia (Sadate-Ngatchou et al., 2008). We confirmed the decrease of NP95 expression in germ cells in the mutant (Np95^fl/fl:Stra8-Cre) mice by immunoblot (IB) and immunohistochemical analyses (Fig. 2A, Fig. S2A). Immunohistochemical analyses also revealed depletion of the NP95 protein in the stages later than the spermatogonia (Fig. S2A). Hematoxylin and Eosin (HE) staining revealed that spermatocytes and post-meiotic cells were severely depleted in the Np95^fl/fl:Stra8-Cre testis (Fig. 2B). In contrast, Sertoli cells that do not express Stra8-Cre appeared normal (indicated by arrows, Fig. 2B). We further determined the precise stage of spermatogenesis affected by Np95 deletion by analysis of dried-down spreads, and observed increase of leptotene and zygotene spermatocytes, but decrease of pachytene spermatocytes (Fig. 2C). Consistent with these observations, we noted that the testis size in Np95 mutants at 32 days postpartum (dpp) was reduced compared to the control (Fig. 2A).

To examine the impact of Np95 knockout in meiotic progression at the pachytene transition stage, we performed gene expression analysis by RNA-seq (Fig. 2D). Among the top 500 genes showing expression changes between control and Np95 knockout, about 80% (402 genes) were downregulated, while the rest (98 genes) were upregulated (Fig. 2D). Gene ontology (GO) analyses revealed enrichment of terms associated with reproduction, spermatogenesis and gamete generation in downregulated genes, and with metabolic processes in upregulated genes (Fig. S2B). This led us to hypothesize that Np95 knockout could cause a retardation of the meiotic program at the late spermatocyte stage, thereby causing the failure to activate the expression of a group of genes linked with pachytene and diplotene stages. To investigate this, we measured the expression of Sox30 and Ddx25 genes, which are expressed at pachytene and diplotene stages (Zhang et al., 2018; Tsai-Morris et al., 2004). Indeed, both Sox30 and Ddx25 were downregulated in the absence of Np95, supporting the notion that the meiotic program to late spermatocyte stage might be affected in Np95-deficient germ cells (Fig. 2E). Consistent with this idea, we also observed a failure of silencing of the X-chromosome-associated genes in the Np95 knockout (Fig. 2F), indicating that meiotic sex chromosome inactivation (MSCI) is perturbed in Np95-deficient spermatocytes.

Given that Np95-deficient spermatocytes exhibit blockage in early meiotic prophase, we wondered whether loss of Np95 affected synapsis of homologous chromosomes. We performed IF of SYCP3 and SYCP1 (de Vries et al., 2005), two components that represent synapsis of homologous chromosomes, using dried-down spermatocytes. As expected, this analysis showed that SYCP3 and SYCP1 (de Vries et al., 2005) colocalized at autosomes but not at unsynapsed regions of sex chromosomes in pachytene spermatocytes (Fig. 3A). Strikingly, a similar analysis in Np95^fl/fl:Stra8-Cre mice revealed total or partial unsynapsed regions and non-homologous associations in about 87% of pachytene-like spermatocytes (Fig. 3B). Consistent with this observation, phosphorylated histone H2A.X (γH2A.X), a marker of unsynapsed chromosomes, was exclusively localized to the regions of sex chromosomes that do not stably pair during pachytene, in control spermatocytes (Fig. 3A) (Turner et al., 2004, 2005). In contrast, in about 90% of pachytene-like Np95^fl/fl:Stra8-Cre spermatocytes, intensive accumulation of γH2A.X was detected over the unsynapsed chromosomal regions. The basal signal intensities of γH2A.X were also elevated throughout the nuclei in the mutant cells (Fig. 3A). Furthermore, although we detected strong γH2A.X staining in the sex chromosomal regions forming the XY bodies in the control, such accumulation was not observed in the Np95 knockout (Fig. 3A,B). This finding, and our observation that X-linked genes are not properly repressed in the Np95 knockout (Fig. 2F), indicate that ablation of Np95 leads to defects in XY body formation (Fig. 2F).

During meiotic recombination, double-strand breaks (DSB) are generated by Spo11. This is followed by localization of the meiotic recombinases RAD51 and DMC1 at the ssDNA overhangs, to mediate DSBs repair. We, therefore, performed immunostaining of RAD51 in Np95^fl/fl or Np95^fl/fl:Stra8-Cre mice, to determine whether meiotic recombination is disrupted in the absence of Np95. The number of RAD51 foci was reduced in Np95^fl/fl:Stra8-Cre zygotene spermatocytes, indicating that the localization of RAD51 was partly destabilized or delayed (Fig. 3C). This result indicates that a massive elimination of mutant spermatocytes, defective in pairing, takes place in the Np95-deficient mice. Given the canonical role of NP95 for maintenance methylation, our findings indicated that defects in DNA methylation may cause defective pairing in Np95-deficient spermatocytes.

To test the role of DNA methylation for meiotic recombination and pairing, we first examined global DNA methylation levels in Np95^fl/fl:Stra8-Cre testes at 19 dpp, by using an anti-5-methyl-cytosine (5mC) antibody (Fig. S4A). In the Np95^fl/fl tubules, spermatogonia (indicated by arrowheads, Fig. S4A) and leptotene spermatocytes, demarcated by a spotted distribution of SYCP3 (indicated by arrows, Fig. S4A), exhibited strong 5mC staining. DNA methylation was maintained in pachytene spermatocytes, although the signal intensity of 5mC was weaker than leptotene spermatocytes (Fig. S4A). As expected, 5mC staining in meiotic spermatocytes expressing SYCP3 (arrows) was much reduced in Np95^fl/fl:Stra8-Cre mice, indicating widespread loss of DNA methylation. In contrast, 5mC levels were unchanged in spermatogonia (arrowheads), which expressed NP95 (Fig. S4A).

To investigate the loss of DNA methylation in Np95-deficient spermatocytes in detail, we carried out whole-genome bisulfite sequencing (WGBS). We confirmed efficient bisulfite conversion (99%) of control or Np95 knockout (two each) samples, by using lambda DNA spike-in controls. The average level of global CpG methylation in Np95^fl/fl:Stra8-Cre germ cells was substantially lower (replicate 1, 39.8%; replicate 2, 40.1%) than control (replicate 1, 72.1%; replicate 2, 76.0%). The decrease in CpG methylation in Np95^fl/fl:Stra8-Cre germ cells was most evident in the CpG sites with the highest level of DNA methylation (Fig. S4B). Examination of CpG methylation status at respective genomic domains revealed substantial demethylation at gene bodies, in intergenic sequences, in repetitive elements [including IAP (intracisternal A particle) endogenous retroviruses], in SINEs, in LINEs and in Np95-deficient germ cells. CpG island (CGI) methylation was low, as is typical of most tissues (Fig. 4A). Importantly, CpG methylation at major and minor satellites was also significantly reduced in Np95^fl/fl:Stra-Cre germ cells (Fig. 4B, Fig. S4C).

Endogenous retroviruses (ERVs) are a major target of DNA methylation machinery in germ and somatic cells. We have previously shown that conditional ablation of Dnmt1 or Np95 leads to widespread demethylation of ERVs in ES cells (Sharif et al., 2007, 2016). We found that, in Np95-deficient germ cells, ERVs were severely demethylated, compared with control (Fig. 4C). We explored whether widespread demethylation at ERVs lead to transcriptional derepression of these parasitic elements in the absence of Np95. To this end, we measured the average expression level of each ERV family (>100 copies in the mouse genome), using the RNA-seq data (Fig. 4C,D). Although a handful of ERVs, such as IAPLTR1_Mm, IAPLTR4_I, MmERVK10C-int and RLTR10C showed modest (about two- to threefold) upregulation, the majority of these elements paradoxically showed weak (about onefold) downregulation of transcription (Fig. 4C,D, Fig. S4D). These results demonstrated that, although ERVs were globally demethylated in the absence of Np95, this did not lead to a general upregulation of these elements, in sharp contrast with observations in Dmnt3l or Dnmt3c mutants (Bourc'his and Bestor, 2004; Webster et al., 2005; Hata et al., 2006; Zamudio et al., 2015; Barau et al., 2016; Jain et al., 2017). This result is also in line with our previous report that, in the absence of Np95, ERVs are only weakly derepressed (Sharif et al., 2016). In addition, we did not observe transcriptional upregulation in major satellites in Np95 knockout germ cells (Fig. S4E), although these sequences were demethylated (Fig. 4B).

The above data indicated that loss of DNA methylation at ERVs, caused by Np95 ablation, does not lead to rampant transcriptional derepression of these elements. Based on this observation, we surmised that defects in PCH clustering, rather than derepression of ERVs, likely play a major role in the spermatogenesis defects observed in the Np95 knockout mice. Given that DNMT1 partners with NP95 for maintenance of DNA methylation in mammalian cells, we wondered whether DNMT1 played a role to regulate PCH clustering, in a similar manner to NP95. To this end, we asked whether DNMT1 expression in the spermatogonia was affected in the absence of Np95. In the control testis, we observed that, although DNMT1 was variably expressed in undifferentiated PLZF^pos/KIT^neg spermatogonia, this expression was eventually downregulated in spermatocytes. In Np95 knockout, however, DNMT1 was expressed in a homogenous and robust manner in spermatogonia, but failed to be downregulated in spermatocytes. This indicated that NP95 might play a role in fine-tuning DNMT1 expression during spermatogenesis (Fig. S3A). As both NP95 and DNMT1 are expressed in the wild-type spermatogonia, we hypothesized that DNMT1 may regulate synapsis of sister chromatids in spermatocytes, similar to NP95. To test this model, we took advantage of a mutant mouse line bearing a Dnmt1 conditional allele and an ER^T2-Cre transgene (Dnmt1^fl/fl:ER^T2-Cre). We injected 4-hydroxytamoxifen (OHT) at 16 dpp and 19 dpp to generate Dnmt1^ex/ex, and analyzed the male mice at 25 dpp to confirm depletion of DNMT1 (Fig. S3B,C). Identical to the Np95-mutant mice, the Dnmt1-mutants also exhibited hypogonadism (Fig. S3C) and decrease of pachytene spermatocytes, and a reciprocal increase of leptotene and zygotene spermatocytes (Fig. S3D,E). We detected unsynapsed chromosomes, associated with strong γH2A.X staining, in about 80% of pachytene-like Dnmt1^ex/ex spermatocytes (Fig. 3A,B). As observed in Np95 knockout spermatocytes, we could not detect γH2A.X staining in the sex chromosomes in the Dnmt1^ex/ex spermatocytes, indicating a defect in XY body formation (Fig. 3A,B).

The above results prompted us to ask whether meiotic recombination was disrupted in the absence of DNMT1. To examine the localization of RAD51 in Dnmt1^ex/ex mice, we performed immunostaining and found that the number of RAD51 foci was reduced in zygotene spermatocytes, indicating partial destabilization or delay of RAD51 accumulation (Fig. 3C). In the absence of DNMT1, therefore, a massive elimination of mutant spermatocytes defective in pairing takes place, as seen in Np95 knockout. Based on these findings, we propose that NP95 and DNMT1 cooperate to maintain DNA methylation in mitotic spermatogonia, which in turn contributes to synapsis in meiotic spermatocytes.

We asked how the widespread loss of DNA methylation in Np95 or Dnmt1 mutant spermatocytes caused aberrant synapsis of homologous chromosomes. As mentioned above, PCH loci, and the repressive epigenetic marks associated with these sequences, play key roles for meiotic synapsis by dynamically interacting with each other to form PCH clusters to facilitate homology search of meiotic chromosomes in early meiotic prophase (Takada et al., 2011). We performed IF and found that H4K20me3, a repressive histone mark (Schotta et al., 2004), was decreased in PCH in Np95-deficient pachytene-like spermatocytes (Fig. 5A). In addition, H3K9me3, which is another characteristic repressive histone modification enriched in PCH (Schotta et al., 2004; Lehnertz et al., 2003), was also downregulated, but to a lesser extent, in leptotene spermatocytes upon NP95 depletion (Fig. S5A). Dnmt1-deficient germ cells, therefore, essentially show the same pattern as the Np95 knockout (Fig. 5A, Fig. S5A,B). We surmised that both NP95 and DNMT1 regulate PCH clustering, via regulation of repressive histone modifications.

We determined whether the degree of PCH clustering was affected in the absence of NP95 or DNMT1. Although fewer than five PCH clusters were observed in about 50% of the leptotene spermatocytes in control, the number of cells with five or fewer PCH clusters was reduced to 24%, and six or more PCH clusters were observed in about 76% of the leptotene spermatocytes in the Np95 knockout (Fig. 5C). A closer inspection revealed that these PCH clusters in Np95-deficient spermatocytes were smaller than control (Fig. 5B). As expected, PCH clustering in Dnmt1^ex/ex leptotene spermatocytes showed a clustering defect very similar to the Np95 knockout, in association with downregulation of H3K9me3 in PCH (Fig. 5B,C). Similar to Np95-deficient spermatocytes, the number of cells with five or fewer PCH clusters was observed in 17% of Dnmt1-deficient mice, and the number of cells with six or more PCH clusters was 83% (Fig. 5B,C). Taken together, these observations reveal crucial roles for NP95 and DNMT1 in maintaining proper epigenetic status of PCH and their clustering during early meiotic prophase. We, therefore, expect that defects in homologous chromosome pairing in Np95- and Dnmt1-deficient spermatocytes are indirect and a consequence of defects in PCH clustering. Co-expression of NP95 and DNMT1 in spermatogonia, and identical synaptic failures observed in Np95 and Dnmt1 mutants, thus indicate that NP95- and DNMT1-mediated maintenance methylation regulates key downstream events during meiosis in male germ cells.

We designed the present study to elucidate the mechanisms by which the DNA maintenance methylation pathway contributes to spermatogenesis in mice. We show that two essential components of this pathway, NP95 and DNMT1, are co-expressed in spermatogonia and are necessary for progression of meiosis in male germ cells, indicating the requirement of maintenance of DNA methylation for male meiosis. Consistent with this notion, Np95- or Dnmt1-deficient male germ cells exhibit spermatogenic defects during meiotic prophase that are characterized by synaptic failure. Assembly of PCH clusters in early meiotic prophase, a process that is required to facilitate subsequent homologous chromosomal pairing, is also disrupted in Np95- and Dnmt1-deficient spermatocytes. Based on these observations, we propose that DNA methylation established in pre-meiotic spermatogonia controls synapsis of homologous chromosomes, which in turn contributes to the quality control of male germ cells and production of mature sperm (Fig. 5D). Alternatively, abnormal and untimely gene expression revealed in Np95-deficient spermatocytes may also disrupt the meiotic program (Fig. 2D-F).

Our findings, and other reports, suggest that DNA methylation marks are monitored by various quality control mechanisms during spermatogenesis. Of note, a recent study by Dong et al. also revealed that NP95 depletion in differentiating spermatogonia induced meiotic arrest during prophase (Dong et al., 2019). In the previously reported Dnmt3l mutants, meiotic arrest has been shown to occur in association with demethylation and derepression of IAP elements (Bourc'his and Bestor, 2004). Although, in this case, meiotic defects did not accompany demethylation of centromeric repeats, unlike our observations in the Np95-deficient mice. NP95, therefore, could act as a quality control checkpoint during meiotic prophase to ensure the integrity and function of male germ cells, by monitoring pairing and synapsis of homologous chromosomes. Interestingly, we have previously shown that NP95 depletion disrupts the transition from undifferentiated to differentiating spermatogonia (Shirakawa et al., 2013). DNA methylation patterns, therefore, could be surveyed at multiple stages (e.g. undifferentiated spermatogonia, meiotic germ cells), and likely by multiple pathways. Consistent with this model, our present study also reveals that the maintenance DNA methylation pathway is linked with repressive chromatin modifications, such as H4K20me3 and H3K9me3, to facilitate PCH formation during spermatogenesis. In contrast, Dong et al. reported that NP95 is linked with PRMT5 that regulates repressive histone arginine methylation, and PIWI proteins that regulate generation of small RNAs, to mediate silencing of retrotransposons (Dong et al., 2019).

Why must DNA methylation marks be monitored in such a strict manner? During mammalian development, germ cells undergo two waves of widespread DNA demethylation/methylation cycles: first during preimplantation; and second during primordial germ cell development. We speculate that re-constitution of DNA methylation marks followed by widespread demethylation may require both de novo and maintenance DNA methylation pathways to ensure the fidelity of epigenetic inheritance in undifferentiated spermatogonia and in pre-meiotic spermatocytes.

The Np95 and Dnmt1 conditional alleles have been described previously (Sharif et al., 2016). Tamoxifen-inducible Dnmt1^ex/ex lines were generated by mating Dnmt1^fl/fl mice with ER^T2-Cre transgene-bearing mice. ER^T2-Cre mediated deletion of Np95 or Dnmt1 was induced by repetitive intraperitoneal injection of 4-OH tamoxifen as previously described (Sharif et al., 2016). The Np95^fl/fl:Stra8-Cre line was generated by mating Np95^fl/fl mice with Stra8-Cre. ER^T2-Cre and Stra8-Cre transgenic mice were purchased from Taconic Biosciences and Jackson Laboratory, respectively. Although experiments were carried out in wild-type, Np95^fl/+, Np95^fl/fl, Np95^fl/fl:Stra8-Cre, Dnmt1^fl/fl, Dnmt1^fl/+ and Dnmt1^ex/ex mice, we only show images of oil-injected Dnmt1^fl/fl as controls to avoid redundancy in the figures, as the meiotic phenotypes of wild-type and heterozygous animals were identical. Mouse genotyping was determined by PCR using tail DNA and the primers listed in Table S1. This study was approved by the Institutional Animal Care and Use Committee (IACUC) of RIKEN Yokohama Branch [27-001(5)]. The mice were euthanized by cervical dislocation, and all efforts were made to minimize suffering.

Testes were fixed by Bouin fixation solution and embedded in paraffin wax. Sections were stained with Hematoxylin and Eosin (HE). Immunohistochemical analysis was performed as described previously (Shirakawa et al., 2013). Briefly, the mice were perfusion-fixed with 4% PFA and the excised testes were post-fixed in the same fixative at 4°C for 4 h. Testes were then embedded in Tissue-Tek OCT compound (Sakura Finetechnical), sectioned at 7 µm and mounted onto Silane-coated slides (Dako). After blocking with PBS containing 1% BSA (bovine serum albumin) with 0.1% Triton X-100 or MOM blocking reagent (Vector Laboratories) for 30 min at room temperature, the specimens were incubated with primary antibodies. The following primary antibodies were used for immunohistochemistry: DNMT1, NP95, PLZF, c-KIT, SYCP3 and STRA8 (Table S2). The sections were washed and incubated in appropriate secondary antibodies (Table S2) with DAPI staining (Sigma), before mounting with the ProLong Gold reagent (Thermo Fisher Scientific). Images were acquired on an Olympus FV1000 confocal microscope.

The testicular germ cells were prepared as described previously with the following modifications (Takada et al., 2011). In brief, testes were removed from 21-day-old mice and detunication was performed manually by tweezers. Materials dissected out were immersed in phosphate-buffered saline (PBS), and repeatedly passed through a 1 ml syringe without a needle, to dissociate seminiferous tubules. Dissociated tubules were manually picked by tweezers, immersed in accutase solution (Innovative Cell Technologies) and sheared with scissors, to dissociate germ cells. Cell suspension was filtrated with a 95 µm mesh to remove interstitial cells, including Leydig cells, blood vessels and lymphatic vessels. Immunostaining of germ cells obtained by the above procedure with SYCP3 antibody showed that more than 80% of the single cells were SYCP3 positive, suggesting that the meiotic prophase germ cells were enriched. Cell spreads for immunostaining were prepared, using the single cell suspension, following standard procedures. Antibodies used for immunofluorescence are listed in Table S2.

Testes were fixed overnight in 4% buffered paraformaldehyde, embedded in paraffin and sectioned. After preparation, sections were processed for immunofluorescence using standard procedures. Briefly, paraffin-embedded sections were de-paraffinized, followed by antigen retrieval in Tris-EDTA buffer (10 mM Tris and 1 mM EDTA at pH 9.0) at 120°C for 5 min using an autoclave, and subsequent cooling for 20 min. The slides were then rinsed with PBS and blocked with 1% BSA in PBST (0.5% Tween 20 in PBS). Slides were incubated overnight at 4°C with anti-5-methylcytosine (5mC) and anti-SYCP3 antibodies diluted in 1% BSA in PBST, and then washed with PBST. Further incubation with anti-mouse IgG (H+L) Alexa 488 and anti-rabbit IgG (H+L) Alexa 568 was performed, followed by washes with PBST and counterstaining with DAPI (Sigma) before mounting.

Germ cell spreads were immunostained using anti-SYCP3 (1:250, a gift from Dr Melanie Hardman, Cancer Research Technology, London, UK), anti-SYCP1(1:300, Novus Biologicals, NB300-229) and anti-γH2A.X (1:300, Millipore, 16-202A) antibodies. Spermatocytes at each stage were counted based on their morphology and subnuclear distributions of SYCP3, SYCP1 and γH2A.X.

Immunoblot analysis was performed using adaptations of previously described methods (Shirakawa et al., 2013). Briefly, cell extracts were prepared from testis and separated by 10% sodium dodecyl sulfate PAGE (SDS-PAGE) under reducing conditions and then transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were probed using a mouse anti-NP95 antibody overnight at 4°C, followed by an anti-mouse secondary antibody conjugated to HRP. Membranes were stained using the Chemi-Lumi One reagent (Nakarai tesque).

Genomic DNA was isolated from testicular germ cells of Np95^fl/fl and Np95^fl/fl:Stra8-Cre mice by using the QIAGEN Allprep DNA/RNA Mini Kit according to the supplier's protocol. Two µg of genomic DNA was used for bisulfite conversion using the EpiTect Bisulfite Kit (QIAGEN). Whole-genome bisulfite DNA sequencing (WGBS) libraries were generated using the PBAT method (Miura et al., 2012). The libraries were sequenced using Illumina HiSeq X with the paired-end mode.

Illumina sequencing reads of WGBS samples were mapped onto the mouse mm10 genome using BMap. Quantification of DNA methylation levels of individual CpG sites was performed using MPTC and MethExport. The global CpG methylation trend was analyzed using methylation information of CpG sites with at least 10× coverage in both the control and knockout samples (25,972,255 CpG sites), and SmoothScatter function of R was used to generate the scatter plot. To analyze CpG methylation changes of major and minor satellite sequences, reads were mapped onto consensus sequences of each repeat sequence with BMap, and methylation levels were analyzed with MPTC and MethExport. CpG sites with at least 3× coverage were used to estimate methylation levels. Methylation levels at each genomic feature and retrotransposons (CGIs, gene bodies, intergenic regions, IAPs, SINEs and LINEs) were calculated for features having three or more CpG sites with 3× or higher coverage (n=22,998, 24,167, 17,917, 19,695, 471,885 and 399,793, respectively) using SeqMonk version 44 (https://www.bioinformatics.babraham.ac.uk/projects/seqmonk/). CGI coordinates were obtained from a previous report (Illingworth et al., 2010). Genomic coordinates of genes and retrotransposons were either obtained from the UCSC Genome Browser or annotation files of SeqMonk were used.

RNA isolation, RNA-seq library preparation and data analysis were performed as described in a previous report (Sharif et al., 2016).

We are grateful to Dr Rudolf Jaenisch (Whitehead Institute and Department of Biology, MIT) for the Dnmt1 conditional mice. We thank Melanie Hardman for the SYCP3 antibody. We also thank Dr Katsuhiko Hayashi (Kyushu University) for his valuable comments on data analysis and interpretation. WGBS analysis was performed with support from AMED-BINDS (T.I.).

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.194605