Tsga8 is required for spermatid morphogenesis and male fertility in mice

Production of the mammalian spermatozoa begins with proliferation and differentiation of spermatogonial stem cells (SSCs) in postnatal testis. Spermatogonia undergo rapid mitotic divisions, which ultimately enter the meiotic phase of spermatocytes to produce haploid spermatids. Finally, spermatids undergo the phase of spermiogenesis, which, in mice, comprises 16 steps of morphological maturation: steps 1-8 are round spermatids and steps 9-16 are elongating spermatids (Clermont, 1972). All of these spermatogenic processes occur in seminiferous tubules that contain four or five layers of stratified germ cells developing towards the internal lumen. Owing to the periodic nature of development, each cross-section of a tubule shows different composition of germ cells, which defines stages of seminiferous epithelium cycle (I to XII) (Clermont, 1972).

During differentiation, male germ cells undergo massive nuclear and cytoplasmic remodeling, which are thought to require unique temporal gene expression patterns regulated by epigenetic mechanisms. In particular, during the meiotic phase, X and Y chromosomes, which contain a large number of reproductive genes, undergo meiotic sex chromosome inactivation (MSCI) followed by RNF8-dependent partial reactivation of escapee genes (Adams et al., 2018; Mueller et al., 2008; Sin et al., 2012). RNF8 is an E3 ubiquitin-protein ligase that is translocated to the nucleus after release from cytoplasmically localized PIWI and ubiquitylates histones (Gou et al., 2017). In the meantime, spermatid nuclei undergo histone H4 hyperacetylation (H4ac) by factors such as TIP60 and EPC1 (Dong et al., 2017) as well as CBP/P300, which are dependent on NUT (NUTM1) (Shiota et al., 2018). This acetylation provides a binding platform for BRDT, a testis-specific double bromodomain-containing protein, which initiates histone eviction by interacting with SWI/SNF proteins (Gaucher et al., 2012; Govin et al., 2006; Miller et al., 2016; Morinière et al., 2009; Pivot-Pajot et al., 2003; Sasaki et al., 2009). Histone eviction and exchange to protamines (PRM1 and PRM2) are also promoted by transition proteins encoded by Tnp1 and Tnp2 (Oliva, 2006; Wang et al., 2019). In spermatids, over tenfold compaction of DNA is achieved by the global exchange of histones with the basic protamines, and a species-specific elongated head shape is formed by the manchette and acroplaxome (O'Donnell, 2014). Imprecision in these processes presumably accounts for a significant proportion of cases of idiopathic male infertility (Krausz and Riera-Escamilla, 2018; Neto et al., 2016; O'Flynn O'Brien et al., 2010).

Because the testis is the organ that expresses the largest number of genes of all organs in mice (Kimmins and Sassone-Corsi, 2005; Xia et al., 2020), the roles of the vast majority of the genes expressed during spermiogenesis for morphological remodeling remain unknown. We recently showed that a small subset of genes that are upregulated during spermiogenesis are potentially programmed for expression in undifferentiated spermatogonia by histone H3 lysine 4 trimethylation (H3K4me3) (Tomizawa et al., 2018). Chromatin priming and later activation is a phenomenon suggested in other systems such as pluripotent stem cells and macrophages (Glass and Natoli, 2016; Harikumar and Meshorer, 2015). In spermatogonia, it has been suggested that a similar phenomenon is partly mediated by KMT2B, which conveys H3K4me3 at promoters. KMT2B mainly targets two groups of genes in undifferentiated spermatogonia: one activated during spermiogenesis and the other activated during embryonic development. The former group, comprising over 100 genes, is enriched for spermiogenesis or fertility functions (e.g. Adam3, Wbp2nl, Acrbp, Trp63, Adad1 and Piwil1), suggesting a possible role for KMT2B in modulating a defined subset of gene expression programs crucial for spermiogenesis.

Inspired by these findings, we decided to search for a novel gene crucial for spermiogenesis by exploiting the target preference of KMT2B. We identify testis-specific gene A8 (Tsga8) (Lu et al., 2019; Uchida et al., 2000) as being targeted by KMT2B for its specific upregulation in round-to-elongating spermatids. Tsga8 is encoded on the X chromosome, is one of the most rapidly evolving genes and has no known protein domain except for two types of alanine-rich repetitive motifs (Good et al., 2011; Taketo et al., 1997; Uchida et al., 2000). The specific expression pattern of Tsga8 in spermatids suggests its potential contribution to male reproductive development, but functions of Tsga8 remained unexplored. Here, by generating knockout mice using CRISPR-Cas9, we show that Tsga8 is essential for spermatogenesis and male fertility. Loss of Tsga8 leads to dysregulation of thousands of genes, including the X chromosome escapee genes in spermatids. Knockout causes reduced levels of TNP1 and PRM1 proteins, and severe deformity of elongating spermatids, resulting in a substantial decrease of epididymal spermatozoa. Collectively, these results present the first example of a KMT2B-targeted gene with crucial functions in spermiogenesis.

To identify a novel factor involved in spermatogenesis, we focused on the genes targeted by KMT2B and upregulated during spermiogenesis (Tomizawa et al., 2018). To this end, we reanalyzed data previously obtained from cultured germline stem cells (GSCs) with chromatin immunoprecipitation followed by sequencing (ChIP-seq) and RNA-sequencing (RNA-seq) (Tomizawa et al., 2018). We compared levels of promoter H3K4me3 and mRNA of all protein-coding genes between the Kmt2b control carrying FLP-recombined floxed alleles (Kmt2bF/F) and the Cre-recombined knockout after tamoxifen-addition (Kmt2bFC/FC; Rosa26-CreERT2) (Glaser et al., 2006; Tomizawa et al., 2018). We found that the level of H3K4me3 at the promoter of the Tsga8 gene was the most affected (decreased by 80.5% in Kmt2bFC/FC) among the genes upregulated during spermiogenesis (Fig. 1A, Fig. S1A; see Materials and Methods for details). Although a low level of mRNA was detected in GSCs and spermatogonia, average Tsga8 expression in round and elongated spermatids was increased by 391-fold when compared with type A spermatogonia (Fig. S1A). Next, we analyzed single cell RNA-sequencing (scRNA-seq) data from mouse whole testis (Ernst et al., 2019) and ENCODE bulk RNA-seq data from 19 mouse tissues (Shen et al., 2012) to investigate whether mRNA expression of Tsga8 is specific to spermiogenic cells. Consistent with a previous report (Lu et al., 2019), specific expression of Tsga8 in spermatids as well as in the testicular tissue was confirmed (Fig. 1B, Fig. S1B).

To check whether the protein expression of TSGA8 parallels its mRNA upregulation in the germline, we performed immunohistochemistry with a specific antibody for TSGA8 (Uchida et al., 2000) using 12-week-old wild-type mouse testes. Consistent with a previous report (Uchida et al., 2000), we observed specific expression of TSGA8 in the nuclei of both round and elongating spermatids in adult testis with a lack of signals in spermatogonia, spermatocytes and somatic cells (Fig. 1C). Thus, these results suggest that, based on its highly specific upregulation in spermatids, Tsga8 is a target of KMT2B in undifferentiated spermatogonia.

To investigate the physiological role of Tsga8, we generated knockout mice using the CRISPR-Cas9 system targeted to Tsga8 exon 2. After injecting a vector that expresses single-guide RNA (sgRNA) into fertilized eggs, we obtained three male mice carrying a frameshift mutation at the target site (Fig. S2A), two of which were used for further experiments. Sequencing of the target genomic DNA of the two mice confirmed that mouse line 1 carried a 2 bp insertion and line 2 carried a 14 bp deletion (Fig. 2A). However, when we crossed the knockout males (F0) with wild-type females, no pups were obtained from either pair, which precluded further investigations. Because we were not able to obtain any mutant females, we used testes from these males for intracytoplasmic sperm injection (ICSI) to expand knockout colonies (Fig. 2B). After culture and transfer of the ICSI embryos, we obtained female pups from the two lines (1 and 2) carrying the expected mutations at the Tsga8 locus (Tsga8-/X). To promote efficiency of genotyping, we established a PCR-based screening approach for Tsga8 adapted from a single nucleotide substitution detection method (Mamotte, 2006). PCR with primers designed to overlap the mutation site at the 3′ end efficiently discriminated DNA from mutant 1, mutant 2 and wild type (Fig. 2C).

Using adult mice from the established mutant colonies, we assessed male fertility. Crosses of control (Tsga8+/Y) and knockout (Tsga8−/Y) males with wild-type females showed that knockout males from both lines were completely infertile (Fig. 2D). We observed no significant change in average body weights, but average weight of knockout testis with epididymis was 14.5% smaller than controls (Tsga8+/Y, 133.0 mg; Tsga8−/Y, 113.7 mg) (Fig. 2E). Apart from the weight difference, the appearance of knockout testes was indistinguishable from controls (Fig. S2B). To confirm the depletion of the TSGA8 protein in the knockout testes, we performed immunohistochemistry using a previously generated antibody specific for TSGA8 (Uchida et al., 2000). The staining of seminiferous tubules showed that the TSGA8 protein was absent in the knockout testes of both lines (Fig. 2F). Moreover, western blot analysis confirmed the lack of TSGA8 in the adult whole testis of both 1 and 2 knockout lines (Fig. S2C).

Tsga8 is located within intron 6 of the dystrophin gene (Dmd; 105.3 kb and 1.9 kb away from exons 6 and 7, respectively). A mutation in Dmd (mdx) is associated with moderate degeneration of seminiferous epithelium and sperm abnormality but does not affect fertility (Chen et al., 2017; Cox et al., 1993). We performed RNA-seq using 8-week-old whole testes and investigated whether the knockout affected Dmd expression. Our analysis confirmed a significant decrease of the Tsga8 mRNA, but the Dmd mRNA remained unchanged in the knockout testis (Fig. S2D). In support of this, Ensembl Variant Effect Predictor (McLaren et al., 2016) or RegRNA 2.0 (Chang et al., 2013) showed no matches for the Tsga8 mutation sites with transcriptional regulatory sequences or splicing regulatory motifs that might affect Dmd transcripts. These data suggest a low likelihood of the influence of Dmd for the phenotype. Collectively, these results indicated that Tsga8 is essential for male fertility.

To investigate the cause of infertility, we performed Hematoxylin and Eosin (H&E) staining of testes. Comparison of control and knockout testes showed no apparent abnormality in spermatocytes (Fig. 3A). In contrast, knockout round spermatids showed a biased staining of nuclei, although the size and shape of the cells remained unaffected: densely stained nucleoli and peripheral heterochromatin were present, but most nuclei were predominated by large, faintly stained regions (Fig. 3A). Furthermore, many knockout elongating spermatids appeared to show malformations in the head shape (Fig. 3A). The histological phenotypes were consistent between the two knockout lines. To identify in what step of spermiogenesis the spermatid abnormality starts to appear in the knockout, we performed fluorescence-based immunohistochemistry with peanut agglutinin (PNA), a widely used acrosome marker, and observed spermatids in each developmental stage of the seminiferous epithelium cycle. We found that spermatids at the spermiogenesis steps 10-11 onwards did not undergo proper elongation in the knockout testis: most of the steps 10-16 spermatids exhibited malformation with fatter nuclei (Fig. 3B). Quantification of the spermatids revealed that 1.83% of control and 97.7% of knockout elongating spermatids showed irregular shapes in the immunohistochemistry images (Fig. S2E). To examine whether testicular sperm are transported to epididymis in the knockout, we performed H&E staining of rete testis toward the end of epididymis. We observed only the rare presence of sperm in rete testis, efferent ducts and caput, and in corpus and cauda epididymal lumen in 12-week-old knockout mice (Fig. 3C, Fig. S2F). Instead, there was accumulation of abnormal cells in the knockout cauda epididymis (Fig. 3C). Spermatid head shaping and spermiation are tightly regulated by ectoplasmic specialization (ES), which forms actin-based adherens junctions between Sertoli cells and spermatids (Lee and Cheng, 2004). ES is also crucial for positioning and orientation of spermatids, which are normally pointed towards the basement membrane at steps 8-16 (O'Donnell, 2014). Our analysis using transmission electron microscopy (TEM) in adult testis showed that actin bundles at ES are present in Tsga8-depleted spermatids (Fig. S3A). Furthermore, although slightly decreased, likely due to deformity, knockout elongating spermatid heads showed a trend of right orientation relative to the basement membrane (73.7% of control and 61.0% of knockout spermatids were within 30° from the right angle to the basement membrane) (Fig. S3B). Instead, TEM revealed engulfment of deformed spermatid nuclei by Sertoli cells in the knockout, indicative of enhanced phagocytosis of malformed spermatids (Fig. S3C). Phagocytosis typically targets and removes apoptotic cells (Arandjelovic and Ravichandran, 2015). We investigated whether there is an increased rate of apoptosis in the Tsga8 knockout testis using the terminal deoxynucleotidyl transferase (TdT)-mediated d-UTP nick end-labelling (TUNEL) assay. We found that the average number of apoptotic signals increased by 7.2-fold in seminiferous tubules of Tsga8 knockout when compared with those of control (control, 1.4; knockout, 9.8; Fig. S3D,E). Moreover, a TUNEL assay of epididymis revealed that most of the accumulated abnormal cells found in cauda epididymis in the knockout (Fig. 3C) were apoptotic (Fig. S3F). Taken together, these data suggest a crucial role of Tsga8 for nuclear and head morphogenesis of spermatids, aberration of which does not cause ES disruption, but leads to apoptosis followed by either phagocytosis by Sertoli cells or a possible discharge into epididymis.

Given that Tsga8 depletion causes striking morphological abnormalities, we hypothesized that gene expression of relevant pathways may be affected. We analyzed our RNA-seq data from adult whole testis and identified 1028 upregulated and 1494 downregulated genes in knockout with false discovery rate (FDR)<0.05 (Fig. 4A). Gene ontology analysis revealed that the downregulated genes were most enriched for transcription (P=3.5e-19; n=199) (Fig. 4B, Table S1). Furthermore, 43 genes were associated with chromatin, which included Setd2 and Jmjd1c, genes crucial for spermiogenesis (Nakajima et al., 2016; Zuo et al., 2018). In addition, downregulated genes contained many spermatogenesis genes (P=0.04; n=33), including spermatid nuclear condensation factors such as Tnp1 and Prm1, and cytoskeleton genes (P=4.9e-8; n=105) (Fig. 4A,B, Table S1). In contrast, consistent with our TUNEL results (Fig. S3D-F), upregulated genes were enriched for cytokines and apoptotic pathways (Fig. 4B). These genes belonged to multiple apoptotic pathways induced by TNF- or Fas-dependent signaling cascades (e.g. Tradd, Fas, PUMA/Bbc3, Bad and Hrk) (Table S1).

The testis is a complex tissue containing at least seven somatic cell types and at least 26 morphologically distinct germ cell classes (Hess and de Franca, 2009; Jung et al., 2019). Furthermore, possible changes in cellular composition in the knockout whole testis may lead to mRNA bias, which can affect differential gene expression analysis to investigate the effect of the knockout on spermatids. To dissect out the effect of Tsga8 depletion on spermiogenesis at a higher resolution, we collected 8-week-old testicular tissues enriched for spermatids using laser capture microdissection (LCM) and performed RNA-seq (Fig. 4C, Fig. S4A). The tissues were divided into two groups containing different stages of the seminiferous epithelium cycle according to cell morphology: one containing early-to-mid stages (Early-Mid) and another containing the rest of the stages (Late-Early). Roughly, Early-Mid should be enriched for round spermatids (steps 4-8), where Tsga8 expression starts (Fig. 1B); Late-Early should be enriched for initial round spermatids (steps 1-3) and elongating spermatids (steps 9-16). Consistent with the whole-testis data, comparison of the control and knockout showed a decrease of Tsga8 mRNA, and no change in the level of Dmd mRNA in the knockout (Fig. S4B). To assess the quality of the LCM-collected tissues, we analyzed the expression of marker genes for major cell types. Our analysis showed that both Early-Mid and Late-Early LCM tissues expressed high levels of spermatid marker genes and low levels of Sertoli, Leydig and spermatogonia markers (Fig. S4C). Furthermore, each LCM tissue was enriched for spermatid genes known to be expressed at the corresponding step of spermiogenesis (Hermann et al., 2018), indicating effective enrichment and separation of spermatids (Fig. 4D). This analysis also showed higher enrichment of some early markers (Speer4e, Ssxb1 and Ssxb2) in Late-Early spermatids in the knockout, suggesting that expression was affected from early steps of spermatids. Next, we focused on the genes known to be associated with nuclear condensation, an event crucial for morphological maturation of spermatids. We observed consistent downregulation of multiple genes leading to nuclear condensation in both fractions of spermatids: epigenetic factors [p300 (Ep300) Rnf8, L3mbtl2 and Cdyl], transition proteins (Tnp1 and Tnp2) and protamines (Prm1, Prm2, Prm3) showed reduction in the knockout (Fig. 4E). In contrast, acetyltransferases required for H4 hyperacetylation in spermatids remained less affected or showed changes in only one of the two spermatid fractions [Epc1, Tip60 (Kat5) and Gcn5 (Kat2a)]. Moreover, when we highlighted the escapee genes on the X chromosome, which we identified using published datasets (Gan et al., 2013), a large proportion of these, including Tsga8 itself, were significantly downregulated in Tsga8 knockout in Early-Mid spermatids (Fig. 4F). In contrast, the effect of Tsga8 knockout on escapee genes was less specific in Late-Early spermatids. In both cell types, the downregulated genes were enriched on the X chromosome when compared with autosomes (Fig. S4D). A previous report suggested that Rnf8, one of the downregulated genes, is required for X chromosome reactivation through epigenetic regulation (Adams et al., 2018; Sin et al., 2012). To investigate whether TSGA8 and RNF8 share the same pathway for X chromosome reactivation, we analyzed round spermatid RNA-seq data from the Rnf8 knockout mice (Adams et al., 2018). We found that 16% of Tsga8-affected and 68% of Rnf8-affected X chromosome genes overlapped with each other (Fig. S4E). Next, we performed the cell spreading method followed by immunohistochemistry to investigate the influence of Tsga8 knockout on RNF8-regulated chromatin features. In contrast to the situations found in Rnf8 knockout spermatids, Tsga8 knockout round spermatids did not show reductions of H3K4me2/3, H2AFZ or H3K27ac at post-meiotic sex chromosomes (PMSCs) (Fig. S4F). However, the global level of histone lysine crotonylation (Kcr), which normally starts to accumulate in round spermatids (Liu et al., 2017; Tan et al., 2011), was reduced in the knockout. The significant downregulation of p300 (Fig. 4E), which shows histone crotonyltransferase activity (Sabari et al., 2015) may have led to the decrease of Kcr. Thus, the extent of Rnf8 reduction in Tsga8 knockout was not sufficient to induce changes in RNF8-regulated chromatin, despite the partial overlap of their affected genes. Together, these results suggest a significant impact of Tsga8 on a wide range of spermatid maturation processes, which include expression of nuclear condensation factors and X-chromosome reactivation.

The observed effect of Tsga8 depletion on mRNA suggests that spermatids cannot undergo gene activation that is crucial for spermiogenesis. To further investigate the influence of nuclear condensation on deregulation of genes in Tsga8 knockout, we performed immunohistochemistry using TNP1, PRM1 and pan-H4ac antibodies. Testicular staining and signal quantification showed that the levels of TNP1 and PRM1 in the knockout spermatids at late stages (X-XII) decreased significantly (Fig. 5A-D). Western blot using 8-week-old whole testis confirmed the decrease in the protein levels of TNP1 and PRM1 (Fig. S5A-D). The reduction of PRM1 was more striking in the western blot when compared with the immunohistochemistry. This is likely due to the enrichment procedure of DNA-associating proteins in the PRM1 western blot experiment. In contrast, the H4ac level was not reduced in the knockout spermatids (Fig. S5E,F). The immunohistochemistry signal intensity of H4ac showed a small but statistically significant decrease only in stage XI, but an increase in stage X suggests that these differences reflect a possible alteration in developmental timing (Fig. S5E,F).

To analyze the nuclear abnormality at a higher resolution, we performed TEM using testicular tissues. We observed no structural difference in spermatocytes between control and knockout (Fig. S5G). In contrast, we noticed a biased distribution of nuclear DNA with the presence of low-density patches in round spermatid nuclei. Moreover, highly malformed and occasionally branched heads of elongating spermatids were observed in the knockout (Fig. 5E). When we measured nuclear areas of elongating spermatids, Tsga8 knockout showed a 1.8-fold increase in the average areas compared with controls (Fig. 5F). The thinning of nuclei in elongating spermatids as assessed by aspect ratio of the longest to shortest sides was incomplete in knockout spermatids, which showed a 50.2% decrease in the average value (Fig. 5G). Furthermore, density quantification of the TEM images revealed a 16.0% decrease in nuclear density in the knockout (Fig. 5H). Together, these data indicate that Tsga8-depleted spermatids cannot undergo key processes required for nuclear condensation.

Spermatid maturation requires extensive morphological changes of the nucleus and cytoplasm, which are represented by polarization of the nucleus to one side of the cell, nuclear condensation, head thinning and elongation. These changes occur in parallel with dynamic gene expression on a global scale. By investigating KMT2B target genes, we have successfully identified an X chromosome gene, Tsga8, as being required for spermatid development and fertility, thus providing supportive evidence for a suggested link between KMT2B and spermiogenesis (Tomizawa et al., 2018).

In line with many sex chromosome genes undergoing MSCI followed by reactivation, Tsga8 was not expressed in somatic tissues or spermatocytes but was specifically upregulated in round to elongating spermatids. Tsga8 does not appear to have a homolog on the Y chromosome, like other unsynapsed genes undergoing MSCI: our BLAST (Altschul et al., 1990) or BLAT (Burgoyne et al., 2009) search did not find similar sequences on the Y chromosome. In concordance with the expression pattern, Tsga8 was required for multiple events associated with spermatid morphogenesis and fertility. First, loss of Tsga8 caused abnormal distribution of DNA in round spermatid nuclei where Tsga8 starts to be expressed. This coincided with the dysregulation of a number of spermatid genes, which include a subset of escapee genes normally reactivated in round spermatids. Second, knockout caused insufficient nuclear condensation and nuclear malformation of spermatids. Here, activation of a set of genes required for nuclear condensation, transition proteins and protamines, was incomplete. In particular, we observed reduction of the protein levels of TNP1 and PRM1, deletion of which are linked to nuclear abnormality and reduced fertility (Cho et al., 2001; Shirley et al., 2004). Tsga8 loss led to a decreased expression of p300 but did not cause reduction of either Epc1 or Tip60, or global H4 hyperacetylation in spermatids. This is consistent with the observation that EPC1/TIP60, at least, is involved in global H4 hyperacetylation (Dong et al., 2017). Furthermore, the loss of p300 expression could have caused the global reduction of Kcr in the knockout spermatids. Tsga8 loss caused a partial reduction of the histone ubiquitin ligase Rnf8 and its downstream genes, but did not induce the loss of RNF8-associating chromatin features, such as H3K4me2/3 and H2AFZ on sex chromosomes (Sin et al., 2012). Thus, the phenotype is likely independent of the H4 hyperacetylation or RNF8-dependent pathways, but the nuclear condensation defect in the knockout involves simultaneous dysregulation of multiple genes involved in DNA packaging. Finally, Tsga8 depletion caused phagocytosis of elongating spermatids by Sertoli cells leading to epididymal azoospermia. Consistently, Tsga8-depleted testis showed induction of apoptosis, which is generally recognized by Sertoli cells for phagocytosis (Arandjelovic and Ravichandran, 2015; Nakanishi and Shiratsuchi, 2004). Because precise control of apoptosis safeguards formation of correct male germ cells (Print and Loveland, 2000; Shaha et al., 2010), spermatid malformation could have induced the apoptosis-based quality control system.

Whether the dysregulation of transcription factor expression is directly or indirectly caused by TSGA8 remains unknown. However, given the lack of known DNA-binding motifs, TSGA8 may not bind directly to chromatin but may instead associate with regulatory transcription factors required for spermatid maturation. The DNA distribution abnormality found in knockout round spermatids suggests that TSGA8 is required for nuclear structures needed for gene expression integrity. In support of this, similar spermatid head abnormality and nuclear condensation defects were also reported for knockout of chromatin factors Chd5 and Hanp1/H1T2 (H1f7) (Tanaka et al., 2005; Zhuang et al., 2014). Despite its crucial role in spermatogenesis, TSGA8 protein sequence is extraordinarily divergent across species and has not been found in humans (Good et al., 2011). Moreover, the only feature found in the TSGA8 amino acid sequence is an intrinsically disordered region (IDR) (Uchida et al., 2000), which may lack a fixed or an ordered 3D structure. The TSGA8 IDR, rich in hydrophobic residues, contrasts with some other nucleosome-associating IDRs containing hydrophilic residues (Mayanagi et al., 2019; Ohtomo et al., 2016), suggesting a low likelihood of direct association with basic histones. It will be also interesting to investigate whether Tsga8 plays a broader role: one limitation of the current work was that homozygous knockout females could not be obtained to analyze the influence on their development due to infertility in males.

After meiosis, spermatids undergo massive transcriptional activation followed by silencing and selective translation (Sassone-Corsi, 2002; Steger, 1999). Our model suggests that KMT2B programs a subset of crucial spermatid or fertilization genes, including Adam3, Wbp2nl, Acrbp, Trp63, Adad1 and Piwil1 for transcription in SSCs (Tomizawa et al., 2018). These genes may be transcriptionally activated by canonical transcription factors such as TATA-box binding protein (TBP)-associated factors (TAFs) and TAF variants that are abundant in spermatids (Klaus et al., 2016), but future studies will investigate how specific sets of genes are targeted and activated. Our study not only provides important knowledge in our understanding of germline morphogenesis and male infertility, but also offers insights about epigenetic regulation of the spermiogenic transcriptome.

Mouse experiments were approved by the Committee for Animal Care and Use at Yokohama City University, and were carried out in accordance with the institutional guidelines. The age of the mice used for each experiment is indicated in figure legends.

Oligo DNAs targeted for Tsga8 exon 2 (sense, 5′-caccGGTGCGAAGACCAACAAGCG-3′; antisense, 5′-aaacCGCTTGTTGGTCTTCGCACC-3′) (100 μM) were incubated in buffer containing 1 M Tris, 5 M NaCl and 0.5 M EDTA at 95°C for 2 min followed by 25°C for at least 45 min to hybridize. The pX330-U6-Chimeric_BB-CBh-hSpCas9 vector (a gift from Feng Zhang; Addgene plasmid #42230) (Cong et al., 2013) was digested with BpiI (BbsI), and ligated with the hybridized double strand DNA insert using T4 DNA ligase (Promega). The sequence of the ligated vector was checked by Sanger sequencing.

Female C57BL/6J mice were administered with pregnant mare serum gonadotropin (PMSG) and human chorionic gonadotropin (hCG) at 48 h intervals by intraperitoneal injection to stimulate ovulation. The females were mated with C57BL/6J males to collect fertilized eggs from oviducts. Next, 5 ng/μl of the vector was injected into the pronuclei of one-cell embryos according to standard methods. Injected one-cell embryos were transferred into oviducts of pseudopregnant ICR mice prepared by sterile mating with vasectomized males. Tails of the pups obtained were used for PCR with following primers to amplify the Tsga8 genomic region: forward, ATTCTCAGGAGGTGAAGCCTATAAT; reverse, ACTTCTTAGCCACAGGTGGAGTAGT. Amplified products were sequenced with the same primers to screen for knockout mice.

Wild-type female mice received 7.5 units of equine chorionic gonadotropin (Teikoku-Zoki Pharmaceuticals) followed by 7.5 units of human chorionic gonadotropin (hCG; ASKA Pharmaceutical) 48-50 h later by intraperitoneal injection to stimulate ovulation. Mature oocytes were collected from oviducts at 15-17 h after hCG injection and were treated with 0.1% hyaluronidase (Sigma-Aldrich) for 1 min in Chatot, Ziomek and Bavister (CZB) medium to remove cumulus cells. The oocytes were transferred to fresh CZB medium and kept at 37°C in an 5% CO₂ incubator until they were used for microinjection.

Testicular spermatogenic cells were isolated as described previously (Ogonuki et al., 2010; Ogura and Yanagimachi, 1993). Briefly, testes from adult mice were gently dissociated in erythrocyte lysing buffer [155 mM NH₄Cl, 10 mM KHCO₃ and 2 mM EDTA (pH 7.2)], and seminiferous tubules were washed in ice-cold (5-10°C) PBS supplemented with 5.6 mM glucose, 5.4 mM sodium lactate and 0.1 mg/ml of polyvinyl alcohol (GL-PBS). The tubules were cut into small pieces and germ cells were dispersed into the GL-PBS by pipetting. The cell suspensions were filtered through a 38 µm nylon filter and rinsed in GL-PBS three times by centrifugation (200 g for 4 min). The cells were immediately used for ICSI and the remaining cells were cryopreserved.

ICSI was performed using a micropipette attached to a Piezo-electric actuator (PrimeTech), as described previously (Kimura and Yanagimachi, 1995a,b; Ogura et al., 1998). On a plastic dish lid (50×3 mm; Falcon 1006; Becton Dickinson), small drops of HEPES-buffered CZB with or without 10% PVP were placed and covered with mineral oil. Spermatogenic cells were placed in one or two PVP-containing droplets. Each morphologically identified sperm was picked up using a micropipette and injected into an oocyte. Some oocytes were injected with elongating spermatids found in the droplets. The injected oocytes were incubated in HEPES-CZB at room temperature (25°C) for 10 min followed by culturing in CZB at 37°C under 5% CO₂ in air.

Two-cell-stage embryos were cultured for 24 h in CZB and were then transferred into the oviducts (7-10 embryos each side) of pseudo-pregnant ICR females prepared by sterile mating with vasectomized males.

PCR primers were designed to overlap with the mutation sites at the 3′ ends to discriminate wild-type and mutated alleles. The primer pairs were as follows: wild type (110 bp) forward, TTGGCCACTTTGGACTTCTT; wild type (110 bp) reverse, GGTGCGAAGACCAACAAGC; knockout 1 specific (110 bp) forward, TTGGCCACTTTGGACTTCTT; knockout 1 specific (110 bp) reverse, GGTGCGAAGACCAACAAAA; knockout 2 specific (108 bp) forward, TTGGCCACTTTGGACTTCTT; knockout 2 specific (108 bp) reverse, TTATCATTACAGGGTGCGTGG. The PCR condition was 30 cycles of 94°C for 30 s, 60°C for 30 s and 72°C for 30 s, followed by a 7 min extension at 72°C with GoTaq DNA polymerase (Promega), which lacks 5′ to 3′ exonuclease activity.

Each of the adult male mice was caged with two wild-type ICR females for one week, and the number of newborn pups obtained from the females was counted. This examination was repeated three times per male mouse with new ICR females.

Staining of testis sections was performed as described previously (Ohmura et al., 2004; Shirakawa et al., 2013). Mice were fixed by perfusion with 4% paraformaldehyde (PFA) and sliced frozen sections were used for staining as follows. For PRM1, sections were treated with HistoVT One (Nacalai Tesque) at 70°C for 20 min. For TNP1, sections were permeabilized with 0.5% saponin followed by incubation with 0.25% Triton/PBS for 30 min. Sections were then blocked using 1% bovine serum albumin (BSA/PBS) (TSGA8, H4ac and TNP1), the Mouse on Mouse Polymer IHC Kit (Abcam) (PRM1) or 5% non-fat milk (PNA) for 30-60 min at room temperature, and incubated with the relevant primary antibody at 4°C overnight. Next, samples were washed with PBS, the required secondary antibody and DAPI (Sigma-Aldrich) were applied, and samples were incubated for 1 h at room temperature. Sections were mounted with ProLong Gold (Invitrogen) and observed using a confocal laser scanning microscope (Olympus FV-1000). For the comparison of control and knockout samples, we placed Tsga8+/Y and Tsga8−/Y slices on the same glass slide and performed reactions in the same drop to avoid biased staining.

The following antibodies were used at dilutions indicated: rabbit anti-TSGA8 (Uchida et al., 2000) (1:800); FITC-conjugated anti-PNA (1:100; EY Laboratories F2301; lot# 030917); rabbit anti-TNP1 (1:500; Abcam ab73135, lot# GR3270070-1); mouse anti-PRM1 (1:100; Briar Patch Biosciences Hup1N); rabbit anti-pan-H4ac (1:500; Thermo Fisher Scientific PA1-84526, lot# VB2931362A); donkey anti-rabbit IgG Alexa555 (1:200; Thermo Fisher Scientific A31572, lot# 1818686); donkey anti-mouse IgG Alexa546 (1:200; Thermo Fisher Scientific A10036; lot# 1736962); and donkey anti-rabbit IgG Alexa488 (1:200; Thermo Fisher Scientific A21206; lot# 2072687) .

Spermatid angle relative to the basement membrane was measured using the ImageJ software (National Institute of Health). An area (<30 μm×30 μm) of a DAPI image containing elongating spermatids positioned inside horizontally oriented basement membrane at stages X-XII was selected, and spermatid nuclei were detected with the ‘analyze particle’ function. The angle of each spermatid was calculated using the ‘measure’ function. ImageJ software was also used to quantify fluorescence signal intensity of H4ac, TNP1 and PRM1 as follows. DAPI signal was used to automatically identify each nucleus using ‘analyze particle’ and ‘auto select’ functions. After selecting elongating spermatids according to nuclear morphology, non-specific signals were removed. Selected areas were applied to measure mean intensity of each staining. To correct for the batch effect between experiments, values were normalized using mean signal intensity of control samples.

For H&E staining, testes from adult mice were fixed in Bouin's solution at 4°C overnight and embedded in paraffin. The samples were sliced and used for H&E staining, followed by imaging with Keyence BZ-X800. Identification of stages was performed according to the morphology of germ cells (Meistrich and Hess, 2013; Russell, 1990).

The TUNEL assay for the detection of apoptosis was performed using ApopTag Plus Fluorescein In Situ Apoptosis Detection Kit (Chemicon International, S7111) according to the instructions. TUNEL-positive signals inside seminiferous tubules were automatically detected by the ImageJ software and signals of >10 μm² were selected for quantification.

Spermatid spreading was performed according to previous reports (El Zowalaty et al., 2015; Reynard and Turner, 2009) with modifications. Briefly, adult seminiferous tubules were dissociated using 1 mg/ml collagenase (Sigma-Aldrich) and 1 U/ml DNaseI (Promega) at 32°C for 15 min. Cells were washed once and incubated in 2 ml of sodium citrate solution [30 mM Tris, 50 mM sucrose, 17 mM trisodium citrate, 5 mM EDTA (pH 8.2)] for 25 min at room temperature. Samples were fixed using 4 ml of 4% PFA/sodium citrate solution for 10 min. After drying, cells were permeabilized with 0.5% TritonX-100 for 10 min and washed in PBS. Blocking was performed with 0.15% BSA for 1 h, and the primary antibody reaction was carried out overnight at 37°C followed by the secondary antibody reaction for 1 h at 37°C. The following antibodies were used: H3K4me2 (1:400; Abcam ab7766-100, lot# 103492), H3K4me3 (1:400; Abcam ab8580, lot# GR33087-1), H2AFZ (1:400; Abcam ab4174, lot# GR35527-1), H3K27ac (1:400; Abcam ab4729, lot# GR167929-1), pan-Kcr (1:800; PTM Biolabs PTM501, lot# ZD060HA30P2) and anti-rabbit IgG Alexa555 (1:200, Thermo Fisher Scientific A31572, lot# 1818686). Intensities of the signals were measured using the ImageJ software. The mean intensities of three randomly selected areas of nucleoplasm, not overlapping PMSC or nucleoli, were compared with those of PMSCs.

Detection of TSGA8, TNP1 and β-catenin was performed as described previously (Shirakawa et al., 2013). Briefly, whole-cell extract from 8-week-old testicular tissues were used for sodium-dodecyl-sulfate PAGE (SDS-PAGE) and the proteins were separated on a 5-20% gradient gel (ePAGEL, E-R1020L, Atto). The proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Immobilon-P; Merck Millipore). After blocking, membranes were incubated with either rabbit anti-TSGA8 (Uchida et al., 2000) (1:2000), rabbit anti-TNP1 (1:5000; Abcam ab73135, lot# GR3270070-1) or mouse anti-β-catenin (1:3000; BD Transduction lab 610153; lot# 67695) overnight at 4°C, and incubated with an HRP-conjugated secondary antibody for 1 h at room temperature. Signal Enhancer HIKARI (Nacalai Tesque) was used to detect signals, and quantification was performed using the ImageJ software.

For the PRM1 analysis, proteins were extracted from 8-week-old whole testis as described previously (Liu et al., 2013) with modifications. Briefly, tissue lysates were sonicated using Bioruptor (Cosmo Bio) and the proteins were precipitated in buffer containing 3 M guanidine and 288 mM DTT. The proteins were resuspended in 0.5 M HCl, precipitated using 25% trichloroacetic acid (TCA) and separated using the Laemmli SDS-PAGE system on a 15% polyacrylamide gel without SDS. Protein transfer and detection was performed as described above with mouse anti-PRM1 (1:5000; Briar Patch Biosciences Hup1N) and rabbit anti-histone H3 (1:5000; Abcam ab1791, lot# 0013-a).

Adult testes were embedded in OCT compound (Tissue-Tek) and frozen in liquid nitrogen. The blocks were sectioned into 10 μm slices for staining. After washing with 75% ethanol and diethyl pyrocarbonate (DPEC)-treated water for a total of 1 min, the sections were stained with Histogene Staining Solution (Thermo Fisher Scientific) for 30 s. Next, the sections were washed in water, 75% ethanol, 95% ethanol and 100% ethanol for the total of 80 s. Early-Mid and Late-Early stage spermatids were identified by cell morphology and collected in separate tubes using a Leica LMD6000. The tissues were immediately lysed in Buffer RLT Plus (Qiagen) containing 1 U/μl SUPERase·In RNase inhibitor (Ambion), and snap-frozen in liquid nitrogen. The samples were used for RNA-seq library preparation.

Total RNA from 8-week-old whole testis from which tunica albuginea was removed and isolated using Isogen (Nippon Gene) according to the manufacturer's instructions. For the whole-testis samples, quality of purified RNA was evaluated using Bioanalyzer (Agilent) with RNA 6000 Nano Kit (Agilent), and genomic DNA was digested using RQ1 DNase (Promega) at 37°C for 30 min. About 500 ng of resulting RNA was used for library preparation using NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB) with 11 cycles of PCR according to manufacturer's instructions. Triplicate samples per genotype (one from mouse line 1, two from mouse line 2) were analyzed. For the LCM samples, tissue lysate from approximately 1500-2000 cells per sample was used for RNA extraction using Isogen (Nippon Gene). Purified total RNA was used to prepare RNA-seq libraries using the SMART-seq2 protocol (Picelli et al., 2013) with 18 cycles of PCR. Duplicate samples (one from each mouse line per genotype) from adult testes were analyzed. The libraries were sequenced using either Illumina HiSeq 2500 or NovaSeq 6000.

H3K4me3 ChIP-seq and RNA-seq from a previous report from GSCs (Tomizawa et al., 2018) were reanalyzed as follows. Sequence reads were checked for quality with FastQC (Babraham Bioinformatics) and trimmed with Trim Galore! (version 0.4.0) (Babraham Bioinformatics) with default parameters. ChIP-seq reads were mapped onto mouse genome mm10 using Bowtie (version 1.1.1) (Langmead et al., 2009) with the parameters -n 2 -l 30 --best --strata -m 1. RNA-seq reads were mapped onto mm10 using Hisat2 (version 2.1.0) (Kim et al., 2015) with default parameters. The number of sequencing reads and mapping rates are summarized in Table S2. Read counts were performed using SeqMonk (version 0.30.2) (Babraham Bioinformatics), correcting for total read count per million reads (log2 rpm) for ChIP-seq. For RNA-seq, reads were quantified using SeqMonk for read per kilobase per million mapped reads (rpkm). When we analyzed differential gene expression, we used edgeR (Robinson et al., 2010) to calculate count per million (cpm) values. Gene ontology (GO) analysis was performed using DAVID (Huang et al., 2009a,b). The full list of enriched terms (biological process) are provided in Table S1. For the identification of differentially expressed genes from RNA-seq data, edgeR-QLF test was used with a cutoff value of FDR<0.05 unless otherwise indicated.

To identify the genes targeted by KMT2B, protein-coding genes with a H3K4me3 peak at promoter (>1 rpm at ±0.5 kb of a transcription start site) in GSCs were ranked by decreased levels of H3K4me3 in Kmt2b knockout GSCs using published data (Tomizawa et al., 2018). Genes showing mRNA upregulation during spermatogenesis (>10-fold increase from type A spermatogonia to either round or elongated spermatids) were then selected using published data (Gan et al., 2013). The genes were further filtered by mRNA expression differences between Kmt2bF/F and Kmt2bFC/FC GSCs (>1 rpkm) to exclude genes showing little effect of KMT2B on transcription. Tsga8 mRNA expression analysis using scRNA-seq data was carried out using adult whole-testis datasets and analysis scripts published previously (https://github.com/MarioniLab/Spermatogenesis2018) (Ernst et al., 2019).

Transmission electron microscopy (TEM) sample preparation and image analysis were performed as described (Nakajima et al., 2019). Briefly, adult mice were fixed by perfusion-fixation using phosphate buffer (0.1 M, pH 7.4) containing 2% glutaraldehyde and 2% paraformaldehyde. Testis was dissected out, cut into small pieces and incubated in the same fixative for 2 h at room temperature. Samples were washed with 0.1 M phosphate buffer and fixed again with 1% OsO₄ for 1 h at 4°C. After washing with water, tissues were stained in 4% uranyl acetate for 1 h at room temperature and dehydrated in graded concentrations of ethanol. The tissues were placed in propylene oxide, embedded in Epon812 resin (TAAB Laboratories Equipment) and polymerized for 48 h at 60°C. Ultra-thin sections for TEM were prepared using Reichert Ultracut N Ultramicrotome (Leica Microsystems) and sections were stained with 2% uranyl acetate in 70% ethanol and 0.4% lead citrate. Images were obtained and analyzed by using a H-7500 TEM (Hitachi) operated at 80 kV. Areas, aspect ratio and density of spermatid nuclei were measured on the TEM images obtained using functions of the ImageJ software. All experiments were performed under the same conditions and images were obtained with the same microscope settings to enable fair comparison between samples.

The numbers of replicate samples for each experiment are indicated in the respective figure legends. At least two biological replicate samples for immunohistochemistry, TEM, RNA-seq and ChIP-seq were analyzed. No statistical method was used to predetermine sample size, and no randomization or blinding methods were used.

We thank the members of Histology and Cell Biology, Yokohama City University School of Medicine for discussion. We thank Ms Kuniko Nakajima, Ms Ikue Hoshi and Ms Ayako Iida for their support with experiments, Dr Rachel Fellows for critical reading of the manuscript, and Dr Yasuo Tsunaka for useful advice on the project.