CXXC finger protein 1-mediated histone H3 lysine-4 trimethylation is essential for proper meiotic crossover formation in mice

During meiosis, a diploid cell produces haploid gametes with two consecutive rounds of division. Unique chromosomal events occur during the prophase of meiosis I, including genome-wide homologous recombination, initiated by programmed formation of double-strand breaks (DSBs). Successful homologous recombination is essential for proper chromosome segregation and fertility (Hunter, 2015).

Among several regulated factors, histone modifications play an important role in initiating and regulating chromosomal events during meiosis (Hinch et al., 2019). Evolutionarily conserved from yeast to human, trimethylated histone H3 at lysine-4 (H3K4me3) is a prominent mark of active meiotic recombination initiation sites (Borde et al., 2009). In Saccharomyces cerevisiae, the H3K4 trimethylase SET1 is required for DSB formation, which initiates homologous recombination in meiotic prophase (Sollier et al., 2004). In mammals, meiotic recombination occurs at 1- to 2-kb genomic regions termed hotspots (de Massy, 2013a; Paigen and Petkov, 2010), the positions and activities of which are determined by another DNA-binding histone methyltransferase, PRDM9 (Baudat et al., 2010; Berg et al., 2010; Diagouraga et al., 2018; Parvanov et al., 2010). Hypothetically, PRDM9 forms complexes with additional proteins, including CXXC finger protein 1 (CXXC1; also known as CFP1), thereby bringing hotspot DNA to the chromosomal axis, and providing the appropriate environment to allow hotspots to proceed into the next phase of recombination (Parvanov et al., 2017; Tian et al., 2018). However, in vivo evidence is lacking to support the role of CXXC1 in homologous chromosomal recombination.

CXXC1 is a DNA-binding component of the SETD1 methyltransferase complex. CXXC1 targets SETD1A/B to most CpG islands (CpGI) and mediates the generation of H3K4me3 (Thomson et al., 2010). Deficiency of Cxxc1 in mice leads to peri-implantation lethality (Carlone and Skalnik, 2001). To study the in vivo functions of CXXC1, several groups have independently generated Cxxc1^flox/flox mouse strains for CRE recombinase-mediated conditional knockout (Cao et al., 2016; Chun et al., 2014; Tian et al., 2018). Deletion of Cxxc1 in oocytes blocks the accumulation of H3K4me3 on chromatin during oogenesis and significantly impairs meiotic maturation and the subsequent maternal-to-zygotic transition (MZT) (Yu et al., 2017). Moreover, CXXC1 and H3K4me3 play transcription-independent roles in mediating meiotic and mitotic cell cycle progression, including G2-M phase transition, spindle assembly, and chromosome condensation and alignment, in oocytes and early embryos (Sha et al., 2018). These data all suggest an indispensable role of Cxxc1 in germ cell development and meiosis. In these studies, however, promoters of mouse Gdf9 and Zp3 genes were used to drive CRE expression and Cxxc1 knockout in female germ cells (Sha et al., 2018; Yu et al., 2017). Cxxc1 deletion was only accomplished by postnatal day(P) 3 and 5, respectively, when oocytes have developed to the diplotene stage of meiotic prophase and are enclosed in primordial and primary follicles. Therefore, it remains unclear whether CXXC1-mediated H3K4 trimethylation is also required for the earliest stages of meiosis in both male and female germ cells.

To answer this question, we selectively deleted Cxxc1 in pre-meiotic germ cells using a recently generated Stra8-Cre knock-in mouse strain (Cxxc1^fl/−;Stra8-Cre) (Lin et al., 2017). In contrast to previously used Stra8-Cre transgenic mice, in which the transgene is only expressed in male germ cells (Sadate-Ngatchou et al., 2008), CRE recombinase is expressed in both sexes of the Stra8-Cre knock-in mice. Another group has reported that conditional knockout of Cxxc1 using the transgenic Stra8-Cre did not affect spermatogenesis and male fertility (Tian et al., 2018). In contrast, our results indicate that Cxxc1 deletion in pre-meiotic germ cells using the Stra8-Cre knock-in mice causes a decrease of H3K4me3 levels in spermatocytes after the zygotene stage, and impairs DSB repair and crossover formation in meiotic prophase. Cxxc1-deleted spermatocytes failed to complete meiosis and were arrested at the secondary spermatocyte stage. RNA sequencing (RNA-seq) and H3K4me3 chromatin immunoprecipitation with sequencing (ChIP-seq) results further indicate that CXXC1 is required for H3K4me3 accumulation at the gene promoters of male germ cells and plays a key role in regulating the programmed gene expression that is essential for spermatogenesis.

To explore the role of CXXC1 in regulating H3K4me3 during spermatogenesis, we first examined the expression pattern of CXXC1 in mouse testes. A published single-cell RNA-seq dataset (Chen et al., 2018) (Fig. S1A) indicates that Cxxc1 remains highly expressed from spermatogonia to meiotic cells. We verified the high expression of CXXC1 in undifferentiated spermatogonia by co-staining with LIN28A (Fig. S1B). Immunohistochemistry (IHC) results showed that CXXC1 protein is mainly expressed in the nucleus of spermatocytes of all stages and round spermatids, but was undetectable in dividing spermatocytes and elongated spermatids (Fig. 1A). Western blotting results further confirmed the expression of CXXC1 in spermatocytes isolated by flow cytometry (Fig. 1B). The purity of isolated meiotic prophase I cells was verified with SYCP3-γH2AX staining (Fig. S1C,D) and the purity of isolated secondary spermatocytes was verified with X-Y chromosome fluorescent in situ hybridization (FISH) staining (Fig. S1E); 94% of diploid cells (n=109) isolated by fluorescence-activated cell sorting (FACS) were secondary spermatocytes with either X or Y chromosome labeling. To determine the role of Cxxc1 during mammalian meiosis, we used a conditional knockout (cKO) approach in which a Cxxc1 floxed line (Cxxc1^fl/fl, in which exons 4 and 5 of the Cxxc1 allele are flanked by loxP sites) (Cao et al., 2016) was crossed with Stra8-Cre knock-in mice expressing CRE from the A1 spermatogonia stage onward (Lin et al., 2017). We verified the efficiency of Cre-mediated recombination by separately genotyping somatic cells and Cre-mediated recombined germ cells (Fig. S1F). Western blot results also verified that CXXC1 was undetectable in spermatocytes purified from Cxxc1^fl/−;Stra8-Cre mice (Fig. 1B). The knockout efficiency was also verified by western blotting with another CXXC1 N-terminal antibody (Fig. S1G).

As the SETD1-CXXC1 complex is considered the main histone H3K4me3 methyltransferase in mammalian cells, we next analyzed the effect of Cxxc1 knockout on H3K4 methylation in mouse spermatocytes. In wild-type (WT) spermatocytes, H3K4me1, H3K4me2 and H3K4me3 all exhibited strong signals at the leptotene and zygotene stages and weaker (but still noteworthy) signals in the pachytene and diplotene stages as well as the secondary spermatocyte stage (Fig. 1B). Notably, H3K4me3 signals were significantly reduced at and after the pachytene stage in Cxxc1^fl/−;Stra8-Cre mice, whereas H3K4me1 and H3K4me2 levels were unaffected by Cxxc1 deletion. Taken together, these observations indicate that CXXC1 is required for maintaining the H3K4me3 level in spermatocytes. Immunofluorescence (IF) staining results on meiotic spreads also showed a significant decrease of H3K4me3 levels in Cxxc1 null spermatocytes at the pachytene and diplotene stages (Fig. 1C,D). We also performed IF staining on testicular sections and got the same results (Fig. S1H,I). Because CRE recombinase is also expressed in female pre-meiotic germ cells in the Stra8-Cre knock-in mice, we also examined H3K4me3 levels in oocytes at the meiotic prophase I stage (P17.5). Similar to the results in spermatocytes, H3K4me3 levels transiently decreased at the pachytene stage and then re-accumulated at the diplotene stage in WT oocytes; in contrast, the re-accumulation of H3K4me3 failed in Cxxc1-deleted oocytes (Fig. 1E,F). Collectively, these results indicate that CXXC1 is involved in the regulation of H3K4 trimethylation during meiotic prophase I, and is potentially important for meiotic cell cycle progression.

Although the Cxxc1^fl/−;Stra8-Cre males copulated normally, they were completely sterile (Fig. 2A). Testes from pubertal and adult Cxxc1^fl/−;Stra8-Cre mice were significantly smaller than those of the littermate controls (Fig. 2B,C). Immunofluorescence staining results of the germ cell marker mouse VASA homolog (MVH; also known as DDX4) and the DSB marker phosphorylated histone H2AX (γH2AX) indicated that the testicular seminiferous tubules of adult Cxxc1^fl/−;Stra8-Cre mice contained multiple layers of germ cells, which had entered meiosis and initiated the process of programmed DSB formation (Fig. 2D). Histological examination of 6-week-old germ cell mutant testes showed that the seminiferous tubules were devoid of elongating and elongated spermatids and contained exclusively multinucleated giant cells formed by apoptotic spermatids (Fig. 2E). Consistent with this, no sperm were found in the epididymis of adult Cxxc1^fl/−;Stra8-Cre mice (Fig. 2F). These results suggest that CXXC1 plays an essential role in spermatogenesis, and its loss leads to abnormalities in early spermatid development.

When different groups of spermatogenic cells were sorted by FACS, there were no haploid round spermatids in Cxxc1^fl/−;Stra8-Cre mice. Testes of Cxxc1^fl/−;Stra8-Cre mice contained more diploid spermatocytes than did control mice (Fig. 3A), indicating that Cxxc1-deleted spermatocytes were arrested at the secondary spermatocyte stage. We also found abnormal metaphase cells that contained lagging chromosomes on Hematoxylin and Eosin (H&E)-stained sections of Cxxc1^fl/−;Stra8-Cre mice (Fig. 3B). To examine the karyotypes more closely, we made spermatocyte chromosome spreads (Fig. 3C). The spermatocytes in Cxxc1^fl/−;Stra8-Cre mice exhibited an increase in premature separated univalent chromosomes during meiosis I (chromosome number >20) compared with those in control mice, indicating that meiotic metaphase I is also disordered upon Cxxc1 mutation (Fig. 3D).

Additionally, we performed IF staining on testis sections with FITC-conjugated peanut agglutinin (PNA), which marks the outer acrosomal membrane of spermatids. Surprisingly, PNA signals were still detected in Cxxc1^fl/−;Stra8-Cre tubules despite the lack of spermatids in them (Fig. 3E). We speculated that the PNA signals were on the arrested secondary spermatocytes. To confirm this, we performed FITC-PNA immunocytofluorescence staining on FACS-sorted haploid spermatids from control mice and the diploid MII spermatocytes from Cxxc1^fl/−;Stra8-Cre mice. The results showed that the MII-arrested Cxxc1-deleted spermatocytes were indeed positive for PNA signals (Fig. 3F). Moreover, at all three acrosomal differentiating phases (Golgi phase, cap phase and acrosome phase), the Cxxc1 null spermatocytes exhibited abnormal acrosome structures (Fig. 3E). Collectively, these results indicate that Cxxc1 null spermatocytes are partially arrested at MI stage. However, there were still a significant number of spermatocytes that could escape from the spindle checkpoint and these were able to develop into the secondary spermatocytes but failed to finish meiosis and initiated false acrosome biogenesis ahead of time.

To investigate further whether the DSB repair process and chromosomal synapsis are impaired after Cxxc1 knockout, we used co-immunofluorescence staining for phosphorylated H2AX (γH2AX), which marks unrepaired DNA lesions and the sex body in pachynema, and synaptonemal complex protein 3 (SYCP3), which marks the synaptonemal complex, to test for the processing of recombination repair. The pattern of γH2AX signals was largely unchanged in Cxxc1 knockout compared with the WT control spermatocytes, with γH2AX signal observed throughout the nucleus in leptonema and zygonema, when DSBs occur, and then restricted to the sex body in pachynema and diplonema, when the autosomal breaks are repaired (Fig. 4A). However, we detected unrepaired DSBs with γH2AX signals partially retained on autosomes in 62.7% of Cxxc1 null pachynema (Fig. 4B). The DSB number was further determined by immunostaining of DMC1 and RAD51, two markers reflecting different stages of DSB processing (Barlow et al., 1997; Pittman et al., 1998). Cxxc1 knockout did not change the numbers of DMC1 and RAD51 foci in early zygotene and late zygotene spermatocytes. However, whereas the numbers of DMC1 and RAD51 foci largely decreased at the early pachytene stage in WT spermatocytes, more foci persisted in Cxxc1 null spermatocytes (Fig. 4C,D).

We also measured primary spermatocyte proportions based on γH2AX and SYCP3 double staining and did not detect a difference in the proportions of spermatocytes at any stage of meiotic prophase I (Fig. 5A). Nevertheless, we observed precociously separated homologous chromosomes in 56.1% of Cxxc1 null diplonema, versus 12.8% in control (Fig. 5B). Co-immunostaining of SYCP1 and SYCP3 at the diplotene stage confirmed the dissociation of synapsis in these precociously separated chromosomes (Fig. 5C). This result suggested defects of crossover formation or resolution in Cxxc1 null spermatocytes. Therefore, we examined the number and location of potential crossover sites in pachynema using MLH1 as a marker. We observed a significant increase in potential crossover numbers in the Cxxc1 null pachynema compared with the controls (Fig. 5D). With further analysis of MLH1 distribution, we found that in Cxxc1 null pachynema, there were more chromosomes with two MLH1 sites, but at least one MLH1 site formed in almost all chromosomes (Fig. 5E,F). We also found more MLH1 sites on Cxxc1 null cell chromosomes located near to the centromere, which are more likely to cause chromosome segregation (Fig. 5E,G) (Mézard et al., 2015; Sherman et al., 1994; Vincenten et al., 2015). Above all, the whole spermatogenesis processes in WT and Cxxc1 null mice were briefly concluded (Fig. 5H).

To understand further the role of CXXC1 in spermatogenesis at the molecular level, we performed RNA-seq analyses in WT and Cxxc1 null spermatocytes at pachytene, diplotene and MII stages. Gene expression levels were assessed as fragments per kilobase of transcript per million mapped reads (FPKM). All samples were analyzed in duplicate, and the replicates showed high correlations (Table S1, Fig. S2A). The numbers of upregulated genes in Cxxc1 null spermatocytes were similar at the three stages (649 for pachytene, 845 for diplotene and 829 for MII; Fig. 6A, Fig. S2B,C). As shown in Fig. 6B, 71% of upregulated genes at the pachytene stage (463 out of 649 genes) were also upregulated at the diplotene stages. Gene ontology analysis revealed that these genes were significantly enriched in terms relating to the spermatogenesis pathway, including positive regulation of acrosome reaction (Fig. S2D,E). We then examined the expression of upregulated genes within the spermatogenesis pathway across different stages in WT spermatocytes and found that they were upregulated at the MII stage (Fig. 6C, left), suggesting an earlier expression of these genes in Cxxc1 null spermatocytes. Indeed, the expression level of these genes in Cxxc1 null pachytene/diplotene spermatocytes was comparable to that in WT MII-stage cells (Fig. 6C, right). This observation is similar to the early formation of the acrosome in Cxxc1 null spermatocytes. Furthermore, we analyzed the expression level of acrosome-reaction genes at different stages in WT and Cxxc1 null spermatocytes, and the results were consistent with an advanced abnormal acrosome phenotype (Fig. S2F,G). By contrast, the number of downregulated genes in Cxxc1 null spermatocytes at the MII stage (n=744) was much higher compared with those at pachytene (n=143) or diplotene (n=135) stages (Fig. 6D, Fig. S2B,C). The downregulated genes in Cxxc1 null spermatocytes at the MII stage were significantly enriched in terms relating to chromatin organization, meiotic cell division, and DNA repair pathways (Fig. 6E), which may partially explain the abnormal homologous separation and cell division arrest observed at the MII stage. Note that we verified the knockout efficiency at Cxxc1 exons 4 and 5 (Fig. S2H). No reads of these two exons were detected, implying that the knockout strategy worked as expected.

In WT spermatocytes, the global H3K4me3 level dropped at the pachytene stage and recovered at the diplotene stage; however, the H3K4me3 level was even lower and could not be re-established at the diplotene stage in Cxxc1 null spermatocytes (Fig. 1C,D). To explore further the genome-wide H3K4me3 changes, we performed H3K4me3 ChIP-seq in WT and Cxxc1 null spermatocytes at the pachytene stage (Table S1). The results showed high correlations between biological replicates (Table S1); therefore, we pooled the biological replicates for further analyses. As shown in Fig. 7A, most genic regions exhibited similar H3K4me3 intensity between WT and Cxxc1 null spermatocytes, with the exception of promoter and 5′-UTR regions, which showed significantly decreased H3K4me3 in Cxxc1 null spermatocytes. Indeed, H3K4me3 levels were decreased at almost all transcription start sites (TSSs) in Cxxc1 null spermatocytes compared with WT (Fig. 7C). The general decreases of H3K4me3 at the TSS regions in Cxxc1 null pachytene spermatocytes prompted us to find out whether H3K4me3 levels were different between WT and Cxxc1 null spermatocytes at earlier spermatogenesis stages. Therefore, we next performed H3K4me3 ChIP-seq at the leptotene/zygotene stage (Table S2). Correlation of the normalized signal intensity between biological replicates was calculated on genome-wide H3K4me3 depositions (Table S2). Similar to the pachytene stage, H3K4me3 intensity was globally decreased at the promoter and 5′-UTR regions (Fig. 7B, Fig. S3A). In addition, other genic regions, such as exons, introns and 3′-UTRs, also showed decreased H3K4me3 levels although to a lesser extent (Fig. 7B). Because H3K4me3 has been implicated in spermatogenesis by forming DSBs at the leptotene stage, we analyzed changes of H3K4me3 in leptotene/zygotene spermatocytes at known DMC1-binding sites (Tian et al., 2018), a hallmark of meiotic DSBs (Fig. 7D). The result showed significant decreases of H3K4me3 levels at the DMC1-binding sites. We also observed a similar trend of a decrease of H3K4me3 at PRDM9-binding sites (Grey et al., 2017) at leptotene/zygotene stage (Fig. 7E). Consistent with the formation of the DSBs at the leptotene stage, H3K4me3 enrichment at the DMC1-binding sites almost completely disappeared in both WT and Cxxc1 null spermatocytes at the pachytene stage (Fig. S3B), indicating a correlation between H3K4me3 and DSB formation.

Using an oocyte-specific Cxxc1 knockout driven by Gdf9-Cre (expressed at the primordial follicle stage as early as P3) and Zp3-Cre (expressed in awakening primary follicles as early as P5), we have demonstrated that in postnatal oocytes that have passed the meiotic prophase CXXC1 is required for normal meiotic maturation and developmental competence after fertilization. However, the in vivo function of CXXC1 at the early phases of oogenesis, particularly in meiotic recombination, were unknown. In the Stra8-Cre knock-in mice, Cre recombinase is expressed in both male and female pre-meiotic germ cells. Results of immunofluorescence staining on postnatal ovarian sections confirmed that CXXC1 was deleted and the H3K4me3 level decreased in oocytes of Cxxc1^fl/−;Stra8-Cre females (Fig. 8A). The ovaries of Cxxc1^fl/−;Stra8-Cre females were consistently smaller than those of control mice at the same age (Fig. 8B,C). At P1, the Cxxc1^fl/−;Stra8^Cre/+ females had fewer oocytes than did the control mice, as determined by quantifying the numbers of MVH-positive oocytes on ovarian serial sections (Fig. 8D,E). Immunostaining on female germ cell spreads at embryonic day (E) 17.5 indicated that Cxxc1 deletion caused an increase of MLH1 foci (Fig. 8F,G). Double staining of SYCP1 and SYCP3 indicated that synaptonemal complex had formed and dissolved normally in Cxxc1 null primary oocytes, but precocious separation of homologous chromosomes was detected in nearly half of Cxxc1 null oocytes (Fig. 8H,I). These results indicated that CXXC1 is required for proper homologous recombination and crossover formation in the meiotic prophase of female germ cells.

In this study, we investigated the potential role and functional importance of CXXC1 in regulation of H3K4me3 during spermatogenesis. In contrast to recently published reports, our current results show that CXXC1 is required for maintaining normal H3K4me3 levels in mouse spermatocytes, particularly at the pachytene and diplotene stages. In addition, CXXC1 is required for proper DSB repair, crossover formation, and meiotic cell cycle progression during spermatogenesis.

Recently, Tian et al. reported that the knockout of CXXC1 in male germ cells with transgenic Stra8-Cre did not impair the meiotic process, and that the knockout mice were fertile (Tian et al., 2018). By contrast, with another knock-in Stra8-Cre and using a different knockout strategy, we obtained Cxxc1 knockout mice that were completely sterile. The H3K4me3 level decreased in Cxxc1 null spermatocytes, along with improper crossover formation, which caused precocious homologous chromosome separation at the diplotene stage and finally spermatogenesis arrested at MII.

In meiosis, crossover formation is genetically essential for creating new combinations of parental genetic information, and at least one crossover per homologous pair is essential to create a physical connection at the diplotene stage before chromosome alignment and accurate segregation in metaphase. Crossovers come from millions of DSBs generated at the leptotene stage. Only a small proportion of DSBs will go through the crossover resolution process; others are repaired during the zygotene to pachytene stages except on the sex chromosomes (Cole et al., 2012). Most recombinations happen at clustered DSB intervals, named hotspots. In mammals, the hotspots are reported to be potentially determined by PRDM9 through its tri-methyltransferase activity on histone H3 lysine-4. But how the histone-modified marked sites are associated with DSB occurrence is still unknown.

In Saccharomyces cerevisiae, Spp1, a member of the histone H3K4 methyltransferase Set1 complex, reads H3K4 methylation through its PHD finger domain (Adam et al., 2018), and interacts with the DSB-promoting protein Mer2 on the chromosome axis, thus tethering the H4K4me3-marked chromatin loops to the chromosome axis where DSB formation occurs (de Massy, 2013b; Sommermeyer et al., 2013). In mammals, CXXC1 is the DNA-binding component of the SETD1 methyltransferase complex and also has a PHD finger domain, which is able to recognize H3K4me3 (Mahadevan and Skalnik, 2016; Tian et al., 2018); it has been shown that CXXC1 can interact with the Mer2 ortholog protein IHO1 (CCDC36) by yeast two-hybrid assay (Imai et al., 2017; Stanzione et al., 2016). This suggests that CXXC1 might play a similar function in mammalian meiosis.

Our results showed that knockout of Cxxc1 in germ cells caused more than half of diplotene spermatocytes to undergo precocious homologous separation, which may result from abnormal crossovers (Fig. 5). The Cxxc1^fl/−;Stra8-Cre mice formed more potential crossover sites at the pachytene stage than did WT, but these crossovers were unable to connect as homologous pairs, so we inferred that these crossovers were partially wrong or invalid. For example, Cxxc1^fl/−;Stra8-Cre mice formed more potential crossovers close to the centromere. Since crossovers come from DSBs and the whole process along with highly regulated gene expression, there are several possible reasons for the phenotype of Cxxc1 deletion.

First, CXXC1 may influence the crossover formation and resolution process. After the DSBs are generated, following strand invasion and double Holliday junction (dHJ) formation and resolution, only a small proportion of DSBs finally become crossovers (Jones, 1984). Other DSBs on autosomes are gradually repaired before mid-pachytene. Errors in crossovers of Cxxc1^fl/−;Stra8-Cre mice might come from the dysregulation of the crossover formation and resolution process, along with DSB repair delay (Fig. 4). As Holloway et al. reported, defects in Class II crossover formation also caused MLH1 foci accumulation and delayed DSB repair, which is similar to the Cxxc1 null spermatocyte phenotype (Holloway et al., 2008).

These improper crossovers might come from improper DSBs generated at the leptotene stage. PRDM9 deposits H3K4me3 at specific sites in the genome, and these sites are recruited to the chromosome axis to form DSBs, especially hotspots. We suggest that the lack of CXXC1 might cause loss of this selective mechanism. Thus, other chromatin loops without H3K4me3 modification were randomly tethered to form DSBs resulting in disorder of the canonical hotspots. Crossovers come from DSBs, especially hotspots. So hotspots disorder might cause errors in crossover number and distribution. Consistent with this hypothesis, our ChIP-seq data showed that CXXC1 depletion led to the failure of H3K4me3 enrichment at the known binding sites of DMC1, a marker of canonical DSBs and hotspots.

Moreover, loss of CXXC1 at leptotene/zygotene stage also caused a remarkable decrease of H3K4me3 level at TSSs, suggesting that the gene expression pattern in Cxxc1 null spermatocytes could be altered as early as the leptotene/zygotene stage, or even earlier. This may also affect the expression of genes related to DSB formation and repair, thereby leading to a delay in DSB repair and a disproportionate crossover distribution. The altered gene expression in pachytene and diplotene stages caused by Cxxc1 knockout could also affect the distribution of MLH1 and the formation of crossover products after the loading of MLH1, which may explain the failure of chiasma formation despite an increased number of MLH1 foci. Disturbed gene expression at the onset of meiosis may also have contributed to the phenotypes observed at the later stages. Especially at MII, downregulation of genes related to chromatin organization and cell division is likely to cause abnormal MII arrest.

CXXC1 is generally considered to influence H3K4me3 establishment by recruiting the SETD1 complex to DNA-binding sites. However, in this study, at leptotene/zygotene stages we detected no significant change in the expression of components of the SETD1 complex and DNMT1 by RNA-seq of WT and Cxxc1 null spermatocytes, which is different from its known function that loss of CXXC1 in embryonic stem cells leads to Dnmt1 transcript levels being significantly elevated (Butler et al., 2009). Moreover, we observed that H2AK119ub1 decreased in Cxxc1 null spermatocytes (Fig. 1B) and H3K4me3 has been reported to interact with ubH2A in histone-to-protamine exchange during spermiogenesis (Wang et al., 2019). Thus, it is possible that H3K4me3 may have as-yet-unknown interactions with other epigenetic modifications in the spermatogenic process.

The conditional mutant alleles for Cxxc1 (Cxxc1^flox/flox) and the Stra8-Cre knock-in mouse strain were generated previously (Cao et al., 2016; Lin et al., 2017). All mice had a C57BL/6J genetic background, and were maintained under specific pathogen-free conditions in a controlled environment of 20-22°C, with a 12/12 h light/dark cycle, 50-70% humidity, and food and water provided ad libitum. The experimental protocols involving mice were approved by the Zhejiang University Institutional Animal Care and Research Committee (Approval #ZJU20170014), and mouse care and use were performed in accordance with the relevant guidelines and regulations.

We used male mice for preparations of different types of spermatogenic cells as reported (Chen et al., 2018). After removal of the tunica albuginea, testes were incubated in 5 ml PBS with collagenase type I (120 U/ml) at 32°C with gentle agitation for 10 min. The dispersed seminiferous tubules were further digested with 5 ml 0.25% trypsin, plus 0.1 ml DNase I (5 mg/ml) at 32°C for 8 min, and then terminated by adding 0.5 ml fetal bovine serum (FBS). The resulting suspension was passed through a 70 μm cellular filter. After centrifugation at 500 g for 5 min, cells were resuspended at a concentration of 10⁶ cells/ml in DMEM with Hoechst 33342 (5 μg/10⁶ cells) and 5 μl DNase I, followed by rotating for 30 min at 32°C at a low speed. Immediately prior to sorting, cells were stained with propidium iodide (PI; 1 μg/10⁶ cells) at room temperature. Cell populations were collected based on their fluorescent label with Hoechst 33342/PI staining by FACS using a Flow Cytometer ARIA II (BD Biosciences). Hoechst 33342 was excited using a 355 nm UV laser, and the dye's wide emission spectrum was detected in two distinct channels: the ‘Hoechst Blue’ (DAPI, 450/50 nm band-pass filter) and the ‘Hoechst Red’ [Indo-1 (Blue), 530/30 band-pass filter].

Testes were fixed in Bouin's buffer or 4% paraformaldehyde (PFA), embedded in paraffin and sectioned at 5 μm thickness using an ultra-thin semi-automatic microtome (Leica). Sections were deparaffinized, rehydrated, and stained with H&E. For immunofluorescence analysis, sections were boiled in 10 mM sodium citrate buffer (pH 6.0) for 15 min, brought to room temperature, washed and blocked with 10% donkey serum for 60 min, and later incubated with primary antibodies (see Table S1) overnight at 4°C. Then slides were washed and incubated with Alexa Fluor 488- or 594-conjugated secondary antibody (712-545-150 and 611-585-215, Jackson ImmunoResearch Laboratories) for 1 h. Slides were mounted with DAPI (Molecular Probes) and then imaged using a Zeiss LSM710 confocal microscope.

For immunohistochemical analyses, the above primary antibody procedure was followed by incubation with biotin-labeled secondary antibodies (65-6140, Invitrogen) then Vectastain ABC kit and 3,3′-diaminobenzidine peroxidase substrate kit (Vector Laboratories) were used for color development. Slides were then counterstained with Hematoxylin and mounted.

Spermatogenic cells (5×10⁴) were lysed in 2-mercaptoethanol-containing loading buffer and heated at 95°C for 5 min. SDS-PAGE and immunoblots were performed following standard procedures using a Mini-PROTEAN Tetra Cell System (Bio-Rad). The antibodies used are listed in Table S1.

Testes were collected from adult male mice, and female ovaries were obtained from embryonic mice at E17.5-E18.5. Spreads were made as previously reported (Dia et al., 2017). Seminiferous tubules or embryonic ovaries were treated with hypotonic buffer [30 mM Tris, 5 mM EDTA, 50 mM sucrose, 17 mM trisodium citrate dehydrate and 0.5 mM dithiothreitol (pH 8.2)] a maximum of 30 min and then smashed in 100 mM sucrose buffer (pH 8.2). The suspension was then gently spread onto slides containing fixative buffer [1% PFA and 0.15% Triton X-100 (pH 9.2)]. After 2 h incubation in a humidity box, the slides were air-dried and washed in PBS three times before immunofluorescence staining. To stain, the slides were blocked with 10% donkey serum and incubated with primary antibody (see Table S1) for 1 h at room temperature. Then slides were washed and incubated with Alexa Fluor 488- or 594-conjugated secondary antibody (712-545-150 and 611-585-215, Jackson ImmunoResearch Laboratories), mounted, and imaged using a Zeiss LSM710 confocal microscope. Semiquantitative analysis of the fluorescence signals was conducted with the NIH Image program ImageJ.

We made a small improvement to the previously reported method (Evans et al., 1964). A suspension made in isotonic (2.2%) sodium citrate solution of a whole testis was centrifuged (500 g for 8 min) and resuspended in 1% sodium citrate solution for 12 min, sedimented again and fixed in a 3:1 mixture of absolute ethyl alcohol:glacial acetic acid. This was followed by two brief rinses in fresh fixative. Air-dried preparations were made from the final fixed suspension and stained with Giemsa.

We made probes following a previously reported method (Kauppi et al., 2011). Plasmid DNA (Chromosome Y: pY353/B; Chromosome X: DXWas70) was labeled with SpectumRed dUTP (Vysis, 30-803400) or SpectrumGreen dUTP (Vysis, 30-803200), respectively, using a bioprime DNA labeling system (Invitrogen, 18094-011). FACS-isolated diploid cells were fixed on slides and then washed with 2×SSC (sodium saline citrate buffer) for 10 min at room temperature followed by a dehydration series of 80%, 90% and 100% ethanol for 2 min each. FISH probes described above were denatured on slides on a hot plate at 80°C for 8 min and hybridized at 37°C for 48 h. After washing, the slides were mounted with mounting media containing DAPI for DNA staining.

Spermatocytes at different stages were collected from the indicated genotypes (500 cells per sample) using FACS as described above. Each sample was directly lysed with 4 μl lysis buffer (0.2% Triton X-100, RNase inhibitor, dNTPs, oligo-dT primers) and immediately used for cDNA synthesis using the Smart-seq2 method as described previously (Picelli et al., 2014). Sequencing libraries were constructed from 500 pg of amplified cDNA using TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme, TD503) according to the manufacturer's instructions. Barcoded libraries were pooled and sequenced on the Illumina HiSeq X Ten platform to generate 150 bp paired-end reads.

Spermatocytes at different stages were isolated using FACS. H3K4me3 ChIP-seq libraries were generated from ∼10,000 spermatocytes using the ULI-NChIP-seq protocol as described previously (Brind'Amour et al., 2015). Briefly, spermatocytes were lysed with nuclear isolation buffer (Sigma-Aldrich), followed by MNase (NEB) digestion at 21°C for 7.5 min. The digested chromatin was then diluted in complete immunoprecipitation buffer, pre-cleared with 10 μl of protein A:protein G 1:1 Dynabeads (Thermo Fisher) and incubated with 58 ng of anti-H3K4me3 (Cell Signaling Technology, 9727, Lot6) and 10 μl of protein A:protein G 1:1 Dynabeads (Thermo Fisher) at 4°C overnight. ChIPed DNA was eluted in 100 mM NaHCO₃ and 1% SDS at 65°C for 1.5 h. Eluted DNA was then extracted using phenol:chloroform:isoamylol (25:24:1) and precipitated in 75% ethanol. The sequence libraries were generated using the NEBNext Ultra II DNA Library Prep Kit (NEB, E7645S) according to the manufacturer's instructions. Barcoded libraries were pooled and sequenced on the Illumina HiSeq X Ten platform to generate 150 bp paired-end reads.

RNA-seq was performed with two biological replicates for all samples. Raw reads were trimmed to 50 bp and mapped to the mouse genome (mm9) with TopHat (v2.1.1) with default parameters. Uniquely mapped reads were subsequently assembled into transcripts guided by reference annotation [University of California at Santa Cruz (UCSC) gene models] with Cufflinks (v2.2.1). The expression level of each gene was quantified as FPKM, and FPKM values of replicates were averaged. Only expressed genes (FPKM>1 in at least one sample) were considered in all analyses. Functional annotation was performed with Metascape (http://metascape.org). Statistical analyses were implemented with R (http://www.rproject.org). The Spearman correlation coefficient (rs) was calculated using the ‘cor’ function, and the complete linkage hierarchical algorithm was used to cluster the genes. Quality information for the RNA-seq data used in this study is summarized in Table S2.

ChIP-seq reads were trimmed to 50 bp and aligned to the mouse genome build mm9 using bowtie2 (v2.3.4.1) with default parameters. All unmapped reads, non-uniquely mapped reads and PCR duplicates were removed. H3K4me3 peaks were called using MACS2 (v2.1.1.20160309) with the parameters -q 0.05 --nomodel --nolambda --broad --extsize 300 -B --SPMR -g mm and signal tracks for each sample were generated with the wig To BigWig tool from UCSC. The correlation between biological replicates and heat maps were generated using deeptools (v2.5.4). Biological replicates were analyzed together for each stage and for downstream analysis. Promoters were defined as 2 kb up- and downstream of annotated TSSs. The H3K4me3 intensity of different genomic regions was calculated using bedtools (v2.26.0) coverage command and normalized to FPKM. Quality information for the ChIP-seq data used in this study are summarized in Table S1.

Results are presented as mean±s.e.m. Most experiments included at least three independent samples and were repeated at least three times. The results for two experimental groups were compared using two-tailed unpaired Student's t-tests (*P<0.05, **P<0.01, ***P<0.001; n.s., not significant).

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.183764.reviewer-comments.pdf