PHF7 is a novel histone H2A E3 ligase prior to histone-to-protamine exchange during spermiogenesis

Spermatogenesis is a complex process that can be roughly divided into the mitotic phase, meiotic phase and post-meiotic phase (i.e. spermiogenesis) (Rathke et al., 2014). During spermiogenesis, haploid round spermatids undergo extensive morphological changes to become flagella-containing motile sperm through a 16-step process, including round spermatids (steps 1 to 8), elongating spermatids (steps 9 to 11), elongated or condensed spermatids (steps 12-14), and spermatozoa (steps 15-16) (Carrell and Aston, 2013; Meistrich and Hess, 2013). This dramatic differentiation process is precisely regulated by a number of epigenetic factors. In spermiogenesis, DNA-binding histones are first replaced by transition proteins (Tnp1 and Tnp2) and transition proteins are then replaced by protamines (Prm1 and Prm2) constituting the nuclear proteins of spermatozoa and mature sperm (Bao and Bedford, 2016; Rathke et al., 2014). Histone-to-protamine exchange is crucial for chromatin compaction and packaging of genome DNA into sperm head, an essential step for the formation of functional sperm. Previous studies have established a link between defective histone-to-protamine exchange and male infertility (Casas and Vavouri, 2014; Meistrich et al., 2003; Oliva, 2006); however, the molecular mechanisms underlying chromatin compaction are still poorly understood.

Ubiquitylation of histones, especially H2A and H2B, in developing male germ cells (Baarends et al., 1999; Jason et al., 2002) has been considered to be one of the important epigenetic marks involved in chromatin remodeling during spermatogenesis (Bao and Bedford, 2016; Govin et al., 2004; Rathke et al., 2014; Sheng et al., 2014). A previous study indicated that the E3 ligase RNF8 mediates H2B and/or H2A ubiquitylation in elongating spermatids, which induces H4K16 acetylation (downstream of H2B ubiquitination) and nucleosome removal (Lu et al., 2010). Intriguingly, we recently found that ubiquitylation-deficient mutations in human PIWI genes (also known as Hiwi), identified in individuals with azoospermia, cause cytoplasmic sequestration of RNF8 and defective H2B ubiquitylation in late spermatids and subsequent impaired histone-to-protamine exchange during spermiogenesis (Gou et al., 2017). In addition to RNF8, a number of putative ubiquitin E3 ligases have been found to be expressed during spermatogenesis (Hou et al., 2012), implying a complex histone ubiquitylation process in male germ cell development.

PHD finger protein 7 (PHF7), one of the putative testis-specific E3 ligase in humans and mice (Hou et al., 2012), has been shown to act as a conserved epigenetic reader to specifically bind histone H3 N-terminal tails with a preference for dimethyl lysine 4 (H3K4me2), and to promote male sexual determination in the Drosophila germline (Yang et al., 2012b). Moreover, Phf7 (NYD-SP6 in humans) was found to be abnormally regulated in infertile males, and thus likely associated with human spermatogenesis (Xiao et al., 2002). However, the specific role of PHF7 in mammalian spermatogenesis has remained elusive. In this study, we characterized the function of PHF7 by generating Phf7-knockout mice using CRISPR-Cas9 technology and found that Phf7 mutant mice exhibit male infertility due to spermatogenic failure at late spermiogenesis. Mechanistically, we found that PHF7 is a novel E3 ligase for H2A ubiquitylation, and that depletion of PHF7 results in abnormal H2A ubiquitylation in spermatids and subsequently impairs histone-to-protamine exchange during late spermiogenesis. When the RING domain of PHF7 was inactivated, mutant mice exhibited similar male infertility phenotypes to Phf7-knockout mice. We further demonstrated that PHF7 plays dual roles in spermiogenesis, acting as both an epigenetic reader that can specifically bind histone H3 N-terminal tails with a preference for trimethyl lysine 4 (H3K4me3) and as an epigenetic writer that can mediate H2A ubiquitylation. Collectively, our findings indicate that PHF7 is a novel histone E3 ligase that is associated with histone-to-protamine exchange during spermiogenesis in mice.

We first analyzed the expression profiles of Phf7 in different tissues from adult mice. Quantitative PCR and western blot analyses indicated that Phf7 is specifically expressed in testes. Further analyses showed that Phf7 was highly expressed in developing sperm, but not in spermatogonial stem cells (SSCs) or Sertoli cells (Fig. S1A-D). Immunofluorescence analysis showed that PHF7 was specifically located in the nuclei of haploid spermatids (Fig. 1A). To investigate the physiological role of Phf7 in mice, we generated Phf7 knockout mice using CRISPR-Cas9 technology. Two sgRNAs (sgPhf7-A and sgPhf7-B) were designed to target exons 2 and 4 of Phf7, respectively (Fig. S2A). We obtained mice with frameshift mutations via injection of Cas9 mRNA and sgPhf7-A or sgPhf7-B into zygotes with a C57BL/6 background. From these mice, females were backcrossed with wild-type (WT) males, leading to four stable mouse lines carrying mutant Phf7 (Phf7^−/−), termed P7A(−1) KO, P7A(−4) KO, P7A(+1) KO and P7B(−11) KO (Fig. 1B). Knockout efficiency was confirmed using western blots (Fig. 1C). Interestingly, Phf7^−/− mice were normal in growth and morphology, but males from all four lines displayed infertility (Table 1). The weights of the mutant testes and epididymes were similar to wild-type controls (Fig. S2B); however, histological analysis indicated a dramatic reduction of sperm count in mutant epididymes (Fig. 1D). Computation-assisted semen analysis (CASA) further confirmed a severe reduction of sperm count in the epididymes of Phf7^−/− mice (Fig. 1E). Moreover, CASA showed little motility and forward progression of Phf7^−/− sperm (Fig. 1F). These results show that Phf7 deletion results in spermatogenic failure in male animals.

We then carefully investigated the stage at which spermatogenesis failed in Phf7^−/− mice. Histological analysis of adult testes showed no obvious abnormal structure of seminiferous tubules in mutant mice (Fig. S2C). Immunostaining of testis sections for the germ cell markers MVH and γH₂A.X (Fig. S2D,E), the meiotic cell marker SYCP3 (Fig. S2F-G) and the acrosome marker PNA (Fig. S2H) excluded the possibilities that Phf7 plays roles in germ cell maintenance, meiosis or early spermiogenesis. Flow analysis of DNA content of testicular cells indicated comparable ratios of 1C, 2C and 4C cells in mutant and control testes (Fig. S2I). However, histological analysis indicated that the number of elongated spermatids at step 15 and 16 (stages IV-VI, and VII-VIII) was dramatically reduced in the seminiferous tubules of Phf7^−/− mice (Fig. 1G,H), concomitant with increased apoptotic signals detected in late spermatids (Fig. S2J). Taken together, these data indicate that depletion of PHF7 causes defective spermatogenesis at a later stage of spermiogenesis in mice.

To determine whether Phf7-deletion-induced spermiogenic defects are germ cell autonomous, we generated a mouse model carrying a Phf7 conditional knockout allele abolishing exons 4 and 5 (referred to as Phf7-Loxp; Fig. S3A) using haploid embryonic stem cell (haESC)-mediated semi-cloned technology (Yang et al., 2012a; Zhong et al., 2015). We generated Sertoli cell-specific Phf7 knockout mice [Phf7^f/−: Amh-Cre] by crossing Phf7-Loxp mice to a Sertoli cell-specific Amh-Cre transgenic mouse (Holdcraft and Braun, 2004) (Fig. S3B-D). We found that the resulting male mice exhibited normal testicular development, spermatogenesis and fertility (Fig. S3E-K), indicating that deletion of PHF7 in Sertoli cells does not impair spermatid development. In contrast, male germ cell-specific knockout mice generated by crossing Phf7-Loxp mice to germ cell-specific Stra8-Cre transgenic mice (Phf7^f/−: Stra8-Cre) (Sadate-Ngatchou et al., 2008) displayed male infertility and dramatically reduced spermatids at steps 15 and 16 (Fig. S4A-I), which resembled the phenotype observed in Phf7^−/− mice. These data demonstrate that spermiogenic defects in Phf7^−/− mice are most likely germ cell autonomous. We thus used Phf7^−/− mice [P7B(−11) KO] for the further studies.

To test whether Phf7 deletion impairs meiosis, we examined the genome integrity of round spermatids isolated from Phf7^−/− mice by round spermatid injection (ROSI) analysis (Kishigami et al., 2004; Yang et al., 2011). We found that mutant round spermatids could support offspring generation at an efficiency comparable with wild-type controls (Table S1). These results indicate that Phf7^−/− mice produce haploid spermatids with genome integrity, providing additional evidence that there are no significant meiotic defects in Phf7^−/− male animals. Next, we inspected the developmental potential of mutant sperm isolated from the epididymis by performing intracytoplasmic sperm injection (ICSI). The results showed that ICSI with Phf7^−/− sperm induced normal activation of mature oocytes, but few injected oocytes developed into blastocysts (Table S2). Taken together, these results indicate that round spermatids from Phf7^−/− mice maintain the genome integrity required to fertilize oocytes.

We then carefully analyzed the morphology of the sperm isolated from mutant epididymes. Differential interference contrast (DIC) and scanning electronic microscopy (SEM) revealed aberrant heads in the majority of mutant sperm (Fig. S5A-B, Fig. 2A), whereas PNA staining revealed normal acrosomes in mutant sperm (Fig. S5C). Transmission electron microscope (TEM) analysis further revealed less condensed nuclei in most sperm and developing spermatids from Phf7^−/− epididymes and testes, respectively (Fig. 2B, Fig. S5D), suggesting defective chromatin compaction in Phf7^−/− sperm.

We next investigated the potential causes of abnormal chromatin compaction in mutant sperm. Given that impaired histone-to-protamine exchange can result in less-condensed nuclei in sperm (Gou et al., 2017; Lu et al., 2010), we examined histone and protamine levels in mutant sperm. Immunostaining analysis indicated abnormal retention of H2A, H2B, H3 and H4 in Phf7^−/− sperm (Fig. 2C). Western blot further confirmed increased levels of all core histones and TH2B (a testis-specific histone variant), and reduced levels of protamine PRM1 and PRM2 in Phf7^−/− sperm (Fig. 2D). In elongating spermatids, where histone-to-protamine exchange has been initiated and histones are first replaced by transition proteins, we found that TNP1 (Meistrich et al., 2003; Oliva, 2006), PRM1, PRM2 and TNP1 were dramatically decreased on chromatins, while TNP1 was increased in soluble fractions in mutant mice (Fig. 2E). These results indicate impaired incorporation of transition proteins and protamines into chromatins in mutant elongating spermatids. Given the importance of histone-to-protamine exchange for chromatin compaction during spermatogenesis, these data suggest that Phf7 deletion causes an incomplete histone-to-protamine exchange and results in defective spermiogenesis in male mice.

To reveal the mechanisms underlying abnormal histone removal, we compared the gene expression profiles of Phf7^−/− round spermatids with those of wild-type round spermatids. Surprisingly, clustering analysis based on RNA-sequencing (RNA-seq) data showed a high correlation between Phf7^−/− and wild-type round spermatids (Fig. 2F, Fig. S5E). The expression levels of histone-to-protamine exchange-related genes, including histones Tnp1, Tnp2, Prm1 and Prm2, were similar in Phf7^−/− and wild-type round spermatids (Table 2, Fig. S5F). Taken together, these results indicate that PHF7 deficiency may not directly impede histone-to-protamine exchange through regulating gene expression, implying that PHF7 may play a role as a post-transcriptional regulator during spermiogenesis.

We next asked how PHF7 regulates histone-to-protamine exchange during spermiogenesis in mice. To this end, we examined whether PHF7 physically interacts with histones by co-immunoprecipitation (co-IP) analysis. We first generated two transgenic haploid embryonic stem cell lines that stably expressed N- or C-terminal Flag-tagged PHF7 using PB-RFP-T2A-flag-Phf7 or PB-RFP-T2A-Phf7-flag vector transfection (Fig. S6A). Flag antibody could pull down PHF7 efficiently in both cell lines (Fig. S6B-D). We detected multiple peptides of H2A and H3 using proteomic analysis of IP pellets, which are indicative of an interaction between PHF7 and histones (Fig. S6E).

Previous studies have demonstrated that PHF7 contains the RING domain (Hou et al., 2012; Wei et al., 2017), an E3 ubiquitin ligase signature (Deshaies and Joazeiro, 2009). Ubiquitylation of H2A/H2B in elongating spermatids is an early step in histone-to-protamine exchange (Gou et al., 2017; Lu et al., 2010; Rathke et al., 2014), so we next examined the levels of ubiquitylated H2A and H2B in spermatids isolated from Phf7^−/− testes. Western blotting revealed dramatically decreased ubiquitylation of H2A (ub-H2A), but not H2B in mutant spermatids relative to wild-type controls (Fig. 3A). Immunostaining of spermatids at different stages showed a significant reduction of ub-H2A in the nuclei of round spermatids (Fig. 3B). In contrast, the levels of ub-H2B and H4K16Ac, an event downstream of ub-H2B (Lu et al., 2010), were not changed in mutant spermatids (Fig. S6F,G). Taken together, these data suggest that H2A, not H2B, is the ubiquitylation target of PHF7.

To directly test the role of PHF7 in histone ubiquitylation, we established an in vitro PHF7 ubiquitylation assay using H2A or H2B as substrates. H2A could be consistently mono-ubiquitylated by purified PHF7 in vitro. Interestingly, in contrast with in vivo observations, PHF7 could also mono-ubiquitylate H2B in vitro (Fig. 3C). The potential explanation might be that the core histones (H2A, H2B, H3 and H4) constitute nucleosomes in vivo, which interact with PHF7, leading to different ubiquitylation outcomes compared with the in vitro assay, wherein a single core histone was used as a substrate. To further investigate this possibility, we constituted nucleosome core particles (NCPs) and subjected them to in vitro ubiquitylation assays, as described previously (Leung et al., 2014). As expected, PHF7 could only ubiquitylate H2A of NCPs in vitro (Fig. 3D). C-terminal K119 is the common ubiquitylation site of H2A (Huen et al., 2007; Kouzarides, 2007), so we reconstituted and purified H2A with mutant K119 for in vitro analysis. We found that PHF7 could not ubiquitylate mutant H2A (Fig. 3E). Finally, purified PHF7 with a mutant RING domain could not ubiquitylate H2A in vitro either (Fig. 3F). Taken together, these data demonstrate that H2A is the substrate of the PHF7 E3 ligase during spermiogenesis.

To directly test the relationship between PHF7 and H2A ubiquitylation in mice, we generated mice carrying Phf7 with a mutant RING domain (Phf7^RINGmut) (Fig. 4A, Fig. S7A) and western blot analysis revealed comparable expression of PHF7 in Phf7^RINGmut spermatids and wild-type spermatids (Fig. S7B). As expected, the mutant mice exhibited similar infertility phenotypes to Phf7^−/− mice (Fig. 4B-E, Fig. S7C,D). Consistent with the above in vitro results with PHF7^RINGmut (Fig. 3F), ub-H2A was almost undetectable in round spermatids from Phf7^RINGmut males (Fig. 4F,G, Fig. S7E,F). Taken together, these data demonstrate that PHF7 is a novel E3 ligase that ubiquitylates H2A via its RING domain in round spermatids during spermiogenesis.

A previous study indicated that the PHD domain of PHF7 binds histone H3 N-terminal tails with a preference for dimethyl lysine 4 (H3K4me2) and controls Drosophila male sex determination (Yang et al., 2012b). This information, coupled with our proteomic analysis of IP pellets showing direct interaction between PHF7 and H3 (Fig. S6E), led us to hypothesize that PHF7 might be an epigenetic reader that recognizes H3 modifications during spermiogenesis in mice.

To test this hypothesis, we first examined whether PHF7 functionally interacts with H3. Indeed, at least five independent pieces of evidence indicated a functional link between PHF7 and H3. First, glutathione S-transferase (GST) pull-down assays (Shechter et al., 2007; Shi et al., 2006) showed a direct interaction between PHF7 and histones (Fig. 5A), and western blot analyses of GST-PHF7-bound histones revealed significant enrichment of H3 (Fig. 5B). Second, we checked which modifications could be recognized by PHF7 and identified that H3K4 modifications, including H3K4me2 and H3K4me3, were PHF7-binding sites (Fig. 5C). Third, in far-western blot analyses (Wu et al., 2007), PHF7 strongly bound H3K4me3, with lower affinity for H3K4me2 (Fig. 5D). Fourth, we tested the recognition of methylated H3K4 by PHF7 in the context of physiological chromatin in cells using protein-protein chromatin immunoprecipitation (ChIP) assays (Ricke and Bielinsky, 2005). As expected, the results further confirmed that PHF7 was strongly associated with H3K4me3 and H3K4me2 (Fig. 5E). Fifth, purified PHF7 with mutant PHD domains could not bind H3K4me3 and H3K4me2 in vitro (Fig. 5F). Taken together, these data indicate that PHF7 specifically binds with H3K4me3 and H3K4me2.

Having demonstrated that RING domain-containing PHF7 can ubiquitylate H2A, and that its PHD domain can recognize H3K4 methylation, we next investigated the functional connection between PHD and the RING domain by performing chromatin immunoprecipitation sequencing (ChIP-seq) using antibodies against PHF7, H3K4me3 and ub-H2A. To this end, we generated a Phf7-Flag knock-in mouse line (Phf7-KI-Flag) by injection of haESCs carrying a Phf7-Flag knock-in into oocytes (Zhong et al., 2015) (Fig. S8A). Western blot and co-IP analyses showed that the anti-FLAG antibody could faithfully and efficiently recognize and pull down PHF7 in round and elongating spermatids (Fig. S8B,C). ChIP-seq analysis of spermatids isolated from the testes of Phf7-KI-Flag mice according to a previously reported protocol (Bastos et al., 2005) showed that the binding peaks of PHF7 and H3K4me3 were mainly enriched at the promoter (Fig. 6A). More than 96% (n=15,419) of all PHF7-binding peaks overlapped with those of H3K4me3 (Fig. 6B), providing supporting evidence for the specific binding of PHF7 to H3K4me3. Moreover, the common binding genes between PHF7 and H3K4me3 were associated with ub-H2A (Fig. 6C,D, Fig. S8D).

However, we observed even lower ub-H2A levels in the immediate vicinity of the TSSs of PHF7-binding genes compared with PHF7 non-binding genes (data not shown), suggesting that other E3 ligases might be involved in the process of binding and ubiquitylating PHF7 nonbinding genes. In future studies, it will be interesting to identify which histone E3 ligase is responsible for histone ubiquitylation in PHF7 nonbinding genes. Finally, H3K4me3 unique genes exhibited higher ub-H2A levels, implying that H3K4me3 could be one of the marks for H2A ubiquitylation during spermiogenesis (Fig. 6D). Taken together, these results demonstrate that PHF7 ubiquitylates H2A, likely by binding H3K4me3.

Ubiquitylation of histones is crucial prior to histone-to-protamine exchange during spermiogenesis. High-throughput analyses identified many factors that may be involved in this process (Hou et al., 2012). Of these factors, only RNF8 has been demonstrated to ubiquitylate core histones, especially H2B, in elongating spermatids (Gou et al., 2017; Lu et al., 2010). In this study, we demonstrate that PHF7 is a novel E3 ligase, which can specifically ubiquitylate H2A in round spermatids. PHF7 deletion results in reduced ub-H2A in round spermatids, retention of histones in sperm and male infertility. Moreover, we demonstrated that the RING domain of PHF7 ubiquitylates H2A, probably through binding H3K4me3 and H3K4me2 by its PHD domains.

Previous studies have shown that PHF7 is a histone code reader, which can specifically associate with H3K4me2/me3 and determine sex in Drosophila (Yang et al., 2012b). However, how PHF7 affects chromatin structure is still an unresolved issue. Our observations in mice further confirm that the PHD finger domain of PHF7, as the most conserved domain between Drosophila and mammals (Yang et al., 2012b), acts as the histone code reader to specifically bind H3K4me2/me3. Interestingly, Drosophila PHF7 associates preferentially with H3K4me2 over H3K4me3, whereas mouse PHF7 acts preferentially with H3K4me3 over H3K4me2, consistent with previous reports of the human ING2 PHD domain (Shi et al., 2006). Moreover, our results reveal that the RING domain of mouse PHF7 can also act as an epigenetic writer to ubiquitylate H2A in round spermatids, which is associated with histone-to-protamine exchange. Several pieces of evidences from co-IP, GST pull-down and protein ChIP strongly demonstrate that PHF7 can simultaneously recognize and bind both H2A and H3. Moreover, ChIP-seq experiments indicate that PHF7 specifically binds H3K4me3 in vivo, which associates with ub-H2A. These observations suggest that it is most likely that Phf7 acts as E3 ligase to ubiquitylate H2A by specifically binding H3K4me3 during spermatid maturation.

A large number of studies have shown that the correct dose of protamine is essential for genome compaction and formation of functional sperm (Cho et al., 2003, 2001). Interestingly, in this study, both Phf7^−/− and Phf7^RINGmut mice exhibited reduced levels of protamine in elongating spermatids and mature sperm (Fig. 3E,F, Fig. S7C,D). Together with evidence that PHF7 can specifically bind with H3K4me3/2 (an important marker for transcription in spermatogenic cells) in round spermatids (Siklenka et al., 2015), these observations imply that PHF7 may also act as a transcriptional regulator during spermiogenesis. However, RNA-seq and qPCR analyses showed that Phf7^−/− round spermatids exhibited a similar expression profile to wild-type controls, and that all tested histone-to-protamine exchange-related genes were similarly expressed in mutant and wild-type round spermatids (Fig. 2F, Fig. S5E,F, Table 2). In addition, given the nuclear location of PHF7 protein in mouse spermatids (Fig. 1A, Fig. S9A), it is less likely that PHF7 is directly involved in regulating the translation of protamine mRNAs. Additionally, we found that PHF7 binding peaks were not enriched in the promoters of histone-to-protamine exchange-related genes. Taken together, our results may exclude the possibility that PHF7 regulates the histone-to-protamine exchange through controlling the transcription of related genes. We thus speculate that the reduction of protamine proteins in Phf7^−/− spermatids is likely due to an indirect effect on translation and/or stability of protamine proteins during spermiogenesis. Future studies will be needed to understand the link between PHF7-mediated H2A ubiquitylation and histone-to-protamine exchange in the future.

Although previous data have shown that RNF8 is involved in the histone-to-protamine exchange process during spermiogenesis, the topic remains controversial (Lu et al., 2010; Sin et al., 2012). Our data demonstrate that PHF7 plays a crucial role in this process. However, exactly how PHF7 cooperates with RNF8 during the histone-to-protamine exchange process remains unclear. Our studies imply an orchestrated sequence of histone ubiquitylation by PHF7 and RNF8 during spermiogenesis. Our evidence suggests that PHF7 may begin ubiquitylation of H2A in round spermatids, followed by RNF8-induced ubiquitylation of H2A and H2B, with a preference for H2B in elongating spermatids. We therefore propose that histone ubiquitylation, a crucial step prior to histone-to-protamine exchange, is regulated by PHF7 and RNF8 sequentially during spermiogenesis. This work deepens our knowledge of the epigenetic regulation underlying histone-to-protamine exchange during spermiogenesis. However, future studies will be needed to understand the molecular mechanisms underlying PHF7 and RNF8 mediated histone ubiquitylation, and their link to histone-to-protamine exchange.

All animal procedures were approved under the ethical guidelines of the Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences.

Forceps were used to isolate different tissue samples from mice, which were placed in the corresponding Trizol tube. The samples were fully homogenized to extract high-quality RNA for reverse transcription followed by a Trizol-guided protocol. 1 μg RNA was first reverse transcribed into cDNA using the reverse transcriptase Rever Tra Ace qPCR RT Master Mix with gDNA Remover (Toyobo, FSQ-201) according to the manufacturer's instructions and the resulting cDNA was used as a template for subsequent PCR amplification using primers specific for one or more genes.

Mouse Sertoli, Leydig and spermatogenic cells were isolated from testes as previously described (Chang et al., 2011; Zhang et al., 2014). Briefly, mouse testes were harvested and the tunica albuginea was removed to expose the seminiferous tubules using surgical scissors. The testes were then cut into pieces and incubated in 3 ml collagenase type IV at 37°C for 15 min. After incubation, the tissue was resuspended with 3 ml of Trypsin for 10 min. After neutralization of the Trypsin with an equal volume of DMEM/10% FBS, the mixture was filtered with a 70 μm filter. The spermatogenic cell suspension was stained with Hoechst 33342 (5 μg/ml, Thermo Fisher Scientific, H1399) for 15-30 min at 37°C, and sorted according to DNA content and forward scatter parameter (FSC) (Bastos et al., 2005). The two types of spermatids exhibit distinguished FSC, which allows separation into two subpopulations. The FSC^high and FSC^low populations represent round spermatids and elongated spermatids, respectively. Isolated spermatogenic cells were confirmed by their distinct nuclear morphology (using DAPI staining of nuclei).

Sperm were harvested by dissecting the cauda epididymis in 37°C pre-warmed Eagle's medium (Sigma, D6046) supplemented with 25 mM HEPES and 4 mg/ml bovine serum albumin (BSA). Blood vessels were dissected away to prevent blood cell contamination of spermatozoa, and epididymis was washed in Ham's F10 medium (Sigma, N2147) containing 4 mg/ml BSA. Spermatozoa were released into 1 ml of Ham's F10 through puncture cut by scissors, gently filtered on ice through fine nylon mesh to remove tissue debris, and harvested by spinning at 500 g for 10 min at 4°C. Sperm-containing medium was dropped into a calibrated slide for CASA analysis. Accurate assessment of motion parameters for each species is dependent upon the computer settings and the concentration of spermatozoa analyzed, so each calibration parameter of the Cell Track/s System was optimized to track felid spermatozoa by evaluating prerecorded samples at various settings.

Plasmids used for expression of different proteins were constructed in the backbone of pGEX6P-1 or pET-51b. The procedures of GST-tagged or Strep-tagged protein purification were carried out according to the manufacturer's instructions (IBA, Manual Strep-Tactin Purification; Millipore, Protein purification and preparation). All primer sequences are listed in Table S3.

Cells on glass and tissue sections were fixed with 4% PFA in PBS for 15 min at room temperature, followed by washing three times every 5 min with PBS. The samples were then blocked with permeabilizing buffer (10% FBS, 1% BSA, 1% Triton X-100, 88% PBS) for 1-2 h. All primary antibodies (Table S3) were diluted (if not specified, the antibody dilution ratio is 1:1000) by blocking buffer (10% FBS, 1% BSA, 89% PBS) and incubated with samples overnight at 4°C, followed by washing with PBS three times every 5 min. The samples were treated with secondary antibody (if not specified, the antibody dilution ratio is 1:1000) coupled with fluorescence for another 1 h at room temperature. Nuclei were stained with DAPI for 10 min. Laser confocal scanning images were captured with a Leica TCS SP5 inverted spectral confocal microscope. Immunostaining density was quantified using ImageJ software.

Mouse spermatogenic cells or other cells were lysed in the buffer [50 mM Tris-HCl (pH 7.4), 1% Triton X-100, 0.1% NP-40, 150 mM NaCl, 5 mM EDTA, 1× proteinase Inhibitor cocktail (Roche, 4693159001)]. Primary antibody-coupled Protein G beads (Life Technologies, REF10004D) were added to the precleared cell lysate and incubated for 4-6 h at 4°C. After washing the Protein G beads with the washing buffer [50 mM Tris-HCl (pH 7.4), 0.1% Triton X-100, 500 mM NaCl, 5 mM EDTA, proteinase inhibitor cocktail] three times, the IP pellets and extracts of cells were analyzed by standard SDS-PAGE and IB procedures.

MII-arrested oocytes were collected from 8-week-old superovulated B6D2F1 females and cumulus cells were removed using hyaluronidase. For ROSI, round spermatids were enriched from testicular cell suspension pretreated with Hoechst 33342 by FACS and injected into oocytes in HEPES-CZB medium containing cytochalasin B (CB, 5 μg/ml). The injected oocytes were activated by treatment with SrCl₂ for 6 h and then cultured in KSOM medium at 37°C under 5% CO₂ in air (Wu et al., 2015). For ICSI, mature sperm collected from the epididymis were injected into oocytes according to reported protocols (Kishigami et al., 2004; Yang et al., 2011). For ICAHCI, haploid ESCs were used in place of round spermatids for injection. The injected oocytes were cultured for 24 h or 3.5 days to reach the two-cell embryo or blastocyst stage for transplantation into the oviducts or uteri of pseudopregnant ICR females, respectively (Zhong et al., 2015).

In vitro ubiquitylation assays were performed as previously described (Gou et al., 2017). The ubiquitylation reaction mixture contained 50 mM (pH 7.5) Tris, 5 mM MgCl₂, 2 mM NaF, 2 mM ATP, 10 mM okadaic acid, 1 mM DTT, 0.1 mg E1, 0.2 mg UbcH5c or UbcH5a and 1 mg HA-ubiquitin (all from Boston Biochemisty). Histones (New England Biolabs, 1 μg) or nucleosome core particles (4 μg) were incubated in 30 μl ubiquitylation reaction buffer, and purified GST-PHF7 or RNF8 was added to the mixture for 2-4 h at 37°C. The proper amount of SDS loading buffer was then added to terminate the reaction before running an SDS-PAGE gel.

GST beads (10 μl beads/sample, GE Healthcare Life Sciences, 17513201) were washed three times every 5 min with PBS. Purified GST and GST-PHF proteins were incubated with GST beads overnight at 4°C and rotated gently. The bead-protein complexes were collected via centrifugation for 2 min at 1500 g at 4°C and washed three times with PBS. The bead-protein complexes were mixed with extracted histones and incubated for another 4-6 h at 4°C. After centrifugation (1500 g for 2 min at 4°C), the bead-protein-histone complexes were collected and washed three times before being subjected to SDS-PAGE analysis.

Far-western blotting used a non-antibody protein that is known to bind the protein of interest. Commercial histone peptides (H3, H3K4me1, H3K4me2 and H3K4me3) were separated using 20% native PAGE gel and were then transferred to a 0.1 µm nitrocellulose filter membrane (GE Healthcare Life Sciences, 10600000). The membrane was incubated with 2% BSA for 3 h at room temperature. For the purified protein GST-PHF, the membrane was incubated overnight at 4°C (GST protein was used as a negative control). After incubation, the membrane was washed gently in PBS before being incubated with GST antibody. Next, the secondary HRP-conjugated antibody was used to detect the protein bound to the histone peptides.

Our protein-protein ChIP protocol was modified from the paper published in 2005 (Ricke and Bielinsky, 2005). Specifically, cells constitutively expressing Flag-tagged PHF7 were crosslinked with 10 ml PBS containing 1% formaldehyde (freshly prepared by mixing 270 µl of 37% formaldehyde into 10 ml PBS) at room temperature for 10 min. A 1/20 volume of 2.5 M glycine was added to a final concentration of 0.125 M to quench the formaldehyde and mixed gently on a rocking station for 10 min. Cell pellets were collected and suspended in lysis buffer [0.5% NP-40, 10 mM EDTA, 50 mM (pH 8.0) Tris-HCl]. Chromatin was sonicated using a Bioruptor PLUS (Pico, 15 cycles, 15 s on/90 s off). After centrifugation at 13,400 g, Protein G beads were used to preclear the lysate. The sonicated chromatin was then incubated with Flag antibody, followed by Protein G beads. IgG was used as a negative control. Last, the binding complex was eluted with elution buffer (0.1 mM NaHCO₃, 1% SDS) and the formaldehyde crosslink reversed before detection of histone modifications.

We isolated round spermatids from the testes of wild-type and Phf7 knockout mice, according to a previously reported protocol (Bastos et al., 2005) for downstream analysis. Round spermatids mRNA profiles from 8-week-old wild-type and Phf7 knockout mice were generated by deep sequencing, in triplicate, using an Illumina HiSeq X10. Illumina Casava1.7 software was used for basecalling. RNA-seq reads were trimmed for adapter sequence using trim-galore v0.5.0, then mapped to mm9 with STAR v2.6.0. Raw read counts of mm9 RefSeq annotated genes were generated by feature Counts v1.5.3 with parameters -p -C, and only uniquely mapped reads were counted. Differentially expressed genes were analyzed by DESeq2 with default parameters. Genes with log2 fold change above 1 (upregulated in the KO) or below −1 (downregulated in the KO) and an adjusted P<0.05 were identified as differential expressed genes. Track files were generated using bedtools genomecov v2.27.1 with -split and then converted to bigwig format with bedGraphToBigWig. The values in bigwig files represent the raw read coverage without any normalization.

Haploid spermatids were isolated from the testes of 8-week-old males as described above. Spermatids were fixed in 1% formaldehyde at room temperature for 10 min and neutralized in 0.125 M glycine for 10 min at a rocking station, followed by cytoplasm cell lysis [10 mM Tris-HCl (pH 8), 10 mM NaCl, 0.5% NP-40, 1×protease inhibitor cocktail] on ice for 10 min, then nuclear lysis buffer [1% SDS, 10 mM EDTA, 50 mM Tris-Cl (pH 8.1), 1× protease inhibitor cocktail] on ice for 10 min. Samples were diluted in 20 mM Tris-HCl (pH 8), 2 mM EDTA, 150 mM NaCl, 0.01% SDS and 1% Triton X-100, and then sonicated using a Bioruptor PLUS (12 cycles, 30 s on/30 s off). The whole-cell extract was precleared with 50 µl protein G beads. After centrifugation at 1500 g, the supernatant was incubated overnight at 4°C with 30 µl of Protein G agarose beads and 5 µg of the appropriate antibodies. Beads were washed four times with ChIP buffer and once with TE buffer. Bound complexes were eluted from the beads and crosslinking was reversed by overnight incubation at 65°C. Immunoprecipitated DNA and whole-cell extract DNA (input sample) were then purified using Qiagen PCR purification kit (Qiagen, 28104). The ChIP-seq library was constructed using NEBNext Ultra DNA Library Prep Kit for Illumina (NEB, E7370). By obtaining over 20 M sequences from every sample according to the standard pipeline for chromatin immunoprecipitated DNA (Landt et al., 2012; Rozowsky et al., 2009), we generated genome-wide chromatin-state maps and mapped the PHF7 binding position of late spermatids. Every sample had one replicate and the ChIP-seq analysis was performed as followed: first basecalls were performed using CASAVA version 1.4, then ChIP-seq reads were aligned to the mm9 genome assembly using Bowtie version 1.2 (with configurations -m 1 -X 500). Only uniquely aligned reads were kept for downstream analysis. Peaks were called using MACS 1.3.7.1 with default parameters. Finally, signal tracks were generated using this pipeline, and the fold-change track was used for further analysis.

We thank Linyu Lu for Rnf8-deficient mice and Ronggui Hu from SIBCB for discussion. We also thank Dr Qinghua Shi from USTC for providing related genetic information on male infertility in humans.