Histone H3.3 regulates dynamic chromatin states during spermatogenesis

During spermatogenesis, a complex interplay of histone post-translational modifications (PTMs) take place in the nuclei of immature germ cells as they develop into mature spermatozoon (Chen et al., 1998; Godmann et al., 2007; Lewis et al., 2003; Payne and Braun, 2006; Rathke et al., 2013; Tachibana et al., 2007; van der Heijden et al., 2006; Zheng et al., 2008). Together with such histone PTMs, histone variants play a major role in the protamine transition and chromatin reorganization process during spermatogenesis, and many histone variants of histone H1 (H1t, H1T2 HILS1), H2A and H2B (TH2A, TH2B, ssH2B, H2A.B.bd, H2BL1, H2BL2, H2AL1-H2AL3) and H3 (H3t, H3.3) become expressed in specific germ cell types or are testes-specific (Lewis et al., 2003; Rathke et al., 2013). The N-terminal tail of H3 acquires specific PTMs during spermatogenesis, including methylation of residue Lys9, a mark involved in chromatin condensation and gene repression, although other residues, such as Lys27 of H3 and Lys20 of H4 also undergo PTM and play roles in establishing repressive chromatin (Jenuwein and Allis, 2001; Kouzarides, 2007; Payne and Braun, 2006). Aberrant regulation of H3 PTMs leads to germ cell apoptosis, defective spermiogenesis and infertility (Iwamori et al., 2011; Liu et al., 2010; Okada et al., 2007; Peters et al., 2001; Tachibana et al., 2007).

Histone variant H3.3 is encoded by two genes in Mus, H3f3a and H3f3b, although both H3f3a and H3f3b transcripts contain divergent 5′- and 3′-untranslated regions and regulatory sequences that mediate at least partially distinct expression patterns (Akhmanova et al., 1995; Szenker et al., 2011; Witt et al., 1997). H3.3 is incorporated into chromatin in transcriptionally active regions undergoing nucleosome displacement; this is in contrast to the canonical histone H3.1 and H3.2 proteins, which are deposited into chromatin in a replication-dependent manner (Ahmad and Henikoff, 2002; Ray-Gallet et al., 2011; Tagami et al., 2004). This difference is due in part to the unique structure of H3.3, which differs from H3.1 and H3.2 by five and four amino acids, respectively. Of particular importance, H3.3 contains unique Ala87, Ile89 and Gly90 residues, which grant H3.3 the ability to destabilize nucleosomes in transcriptionally active regions (Ahmad and Henikoff, 2002; Jin et al., 2009).

Loss of H3.3 through disruption of either H3f3a or H3f3b causes defects in fertility (Bush et al., 2013; Couldrey et al., 1999; Hodl and Basler, 2009; Sakai et al., 2009; Tang et al., 2013). H3.3 is expressed throughout the mouse seminiferous tubule (Bramlage et al., 1997) and accumulates in the sex body of spermatocytes (van der Heijden et al., 2007). H3.3 is also not completely removed from chromatin during spermiogenesis (Erkek et al., 2013; Hammoud et al., 2009), and the relatively small remaining H3.3 fraction is enriched in genomic regions important for zygotic development, potentially transmitting a parental epigenetic memory between generations (Erkek et al., 2013; Hammoud et al., 2009). Thus, H3.3 appears to play a crucial role in male spermatogenesis, although the cellular and molecular mechanisms by which H3.3 influences fertility largely remain an open question.

We recently generated an H3f3b null mouse model (Bush et al., 2013). Surviving H3f3b null males (hereafter referred to as null) displayed complete infertility. Here, we explore the mechanisms involved in the male infertility phenotype. Null male mice have abnormal testes and sperm morphology. Null males also exhibit increased rates of apoptosis in specific germ cell populations and in sperm, and aberrant levels of histone PTM and spermatogenesis-related gene expression within the testes. Finally, null testes display defective germ cell chromatin reorganization and a failure of normal protamine incorporation. These studies overall indicate that H3.3 plays an important role in regulating chromatin states, which are involved in proper male mouse germ cell development.

To examine the extent of H3.3 histone protein reduction upon homozygous targeted disruption of H3f3b, we first examined whole-testes cell lysate (WTCL). Loss of H3f3b did not lead to significant changes in total H3 levels in WTCL (supplementary material Fig. S1A,B) but resulted in a >2-fold reduction in H3.3 protein levels (supplementary material Fig. S1C-E, P=0.005), similar to previous reports following H3f3b loss in mouse embryonic fibroblasts (MEFs) (Bush et al., 2013). Notably, several faint, but consistently higher molecular-mass H3.3 bands were evident in wild-type (WT) samples but not in those from null mice (supplementary material Fig. S1C,D), indicating the involvement of larger, uncharacterized H3.3-specific PTMs in testes, because these PTMs were not observed in other cell types.

H3f3a transcripts are expressed throughout the testes (spermatogonia, spermatocytes and spermatids) at low levels, whereas H3f3b is most highly expressed specifically in spermatocytes (Bramlage et al., 1997). We utilized an H3.3-specific antibody for immunohistochemistry-based immunofluorescence staining (IHC-IF) in mouse testes sections from H3f3b WT and null male animals, along with a lectin peanut agglutinin (lectinPNA) conjugate that recognizes the sperm acrosome used to stage-match testes cross-sections (Aviles et al., 1997; Szasz et al., 2000). H3.3 protein accumulated in the nuclei of spermatogonia and leptotene spermatocytes (Fig. 1A,B) in WT tubules. Nuclear H3.3 levels peaked during mid-stage in WT pachytene spermatocytes and persisted throughout meiosis and most of spermiogenesis, with H3.3⁺ focal regions appearing within the nuclear interior of WT pachytene and diplotene spermatocytes (Fig. 1A,C; supplementary material Fig. S1F,G), corresponding to regions resembling the sex body (van der Heijden et al., 2007). WT step 1-8 round spermatids expressed H3.3 throughout the nucleus and at bright focal points within the nuclear interior, most likely to be the sex body described previously in Mus (van der Heijden et al., 2007) (supplementary material Fig. S1F,G). Relative H3.3 levels began to decrease in step 11 WT elongating spermatids (Fig. 1C) and were undetectable in step 12-16 spermatids (supplementary material Fig. S1F,G).

As expected, null tubules displayed substantially lower levels of H3.3 by IHC-IF analyses. H3.3 was nearly undetectable in spermatogonia, pre-leptotene, leptotene, zygotene and early pachytene (stage I-II) spermatocytes in null tubules. Faint nuclear expression of H3.3 became apparent in mid-stage (stage IV-V) null spermatogonia (supplementary material Fig. S1F) as opposed to those of WT, in which H3.3 expression was first detected in early spermatogonia populations. Weak nuclear focal expression of H3.3 was seen in early- to mid-stage null pachytene spermatocytes (stage III-V, supplementary material Fig. S1F,G), but the diffuse nuclear staining was never comparable in intensity to that observed in WT. Nuclear and focal areas of H3.3 protein localization in step 1-11 spermatids were also substantially fainter in null testes, and H3.3 was undetectable by IHC-IF in step 12-16 spermatids. Thus, knockout of H3f3b results in substantial decreases in H3.3 protein levels throughout spermatogenesis in a number of cell types within the seminiferous tubule (supplementary material Fig. S2), with the remaining low levels of H3.3 produced from H3f3a.

We examined WT and surviving null mature male mice (supplementary material Fig. S3 and Table S1 for age ranges, P=0.420) for defects. Nulls possessed smaller testes, both when compared with their WT littermates and when normalized to animal mass (supplementary material Fig. S3A-E). Compared with WT testes histological sections (supplementary material Fig. S3F), null seminiferous tubules exhibited abnormal architecture, with small and unusually open tubules and reduced numbers of germ cell types and early spermatozoa.

To determine the effect of H3f3b loss on spermatid quantity within a tubule cross-section, we measured the number of stage I round (step 1) and elongated (step 13) spermatids in WT and null tubules. Step 1 round spermatids are the direct result of meiotic divisions in spermatocytes that take place in the preceding stage XII tubules. We observed a significant 1.46-fold decrease in the total number of spermatids per null tubule area (supplementary material Fig. S3G), comprising a 1.36-fold decrease in round spermatids and a 1.6-fold decrease in elongated spermatids at this stage (supplementary material Fig. S3H,I). In order to further investigate potential changes in germ cell quantity, we used fluorescence-activated cell sorting (FACS) based on previously established protocols (Bastos et al., 2005; Gaysinskaya et al., 2014; Getun et al., 2011) on testes from two WT and four null aged males (see supplementary material Table S1). We utilized FACS to successfully separate spermatogonia, pre-leptotene spermatocytes, spermatocytes and spermatids with a high degree of enrichment (supplementary material Fig. S4A-H). The FACS profiles of WT and null populations showed high similarity in cell type distribution, although null testes contained substantially fewer cells overall (supplementary material Fig. S5A,B). Null testes exhibited comparable populations of spermatogonia and spermatocytes (supplementary material Fig. S5C,D) and a slight but significant increase in spermatid number (supplementary material Fig. S5E, P=0.033). These data suggest that despite the observed decrease in spermatid numbers per tubule cross-section, the overall proportion of spermatids in relation to other germ cell types is similar to that of WT. Null testes also exhibited a 1.9-fold decrease in the number of pre-leptotene spermatocytes (supplementary material Fig. S5F, P=0.020), indicating increased cell death in this population, or fewer spermatogonia transitioning into this cell type.

Loss of H3f3b has been reported to cause defects in chromosome segregation (Bush et al., 2013). We quantified the number of stage XII meiotic tubules in WT and null testes to determine whether H3f3b loss leads to a higher preponderance of stalled meiotic tubules by utilizing immunostaining of phosphorylated residue Ser10 on H3 (H3 S10P), a marker of mitotic and meiotic cells (Korhonen et al., 2011). Null testes exhibited a slight, but non-significant, increase in the proportion of meiotic tubules per absolute number of seminiferous tubules (12.3%) compared with WT (10.7%) (supplementary material Fig. S6A). We did not detect a difference in the number of meiotically dividing cells within a tubule when normalized to the tubule area (supplementary material Fig. S6B, P=0.923). Removal of H3f3b was previously shown to affect Cenpa incorporation and possibly contribute to centromere dysfunction in MEFs (Bush et al., 2013). We did not observe an increase in the number of Cenpa⁺ foci per spermatogonia or spermatocyte in null tubules (supplementary material Fig. S6C,D). These findings suggest that the observed changes in spermatid populations are not likely to result from defective meiotic processes.

To further explore the mechanisms of H3f3b null male infertility, we analyzed the sperm of WT-control and null males (Table 1). Null males exhibited >2-fold decrease in total sperm concentration. The null sperm that were present were quite abnormal, exhibiting >2-fold decreases in motility, >3-fold decreases in rapid cells and >4-fold decreases in progressive motility (supplementary material Movies S1, S2). Examination of WT and null sperm under high magnification revealed major abnormalities in morphology (Fig. 2A), with 99% of null sperm having abnormal heads (versus 30% of WT), indicating possible defects in meiotic chromatin condensation (Fig. 2B).

We also examined the potential effects that H3f3b loss would have on cell survival in the testes. We observed a 2.06-fold increase in the percentage of apoptotic tubules per null testis (Fig. 2C, classified as exhibiting one or more TUNEL⁺ nuclei per tubule) and a 1.63-fold increase in the absolute number of apoptotic tubules per null testis relative to WT (Fig. 2D). Null tubules also exhibited a significant 3.53-fold increase in apoptotic events (4.25% of null cells versus 1.20% of WT) (Fig. 2E). In null testes, the majority of apoptotic events took place in or around stage IX tubules, within one to two cell-layers from the tunica propria, where pre-leptotene spermatocytes transition into leptotene spermatocytes (Fig. 2F; supplementary material Fig. S2, Fig. S7A,D), with few apoptotic nuclei detected in early (stage I-V) and nearly no apoptotic nuclei detected during mid-stage (stage VI-VIII). Null tubules also exhibited very high levels of TUNEL⁺ nuclear fragments, which were essentially absent in the WT tubules (Fig. 2F, red stars). Examination of subsequent stages corroborated these findings because stage X tubules also contained increased numbers of apoptotic nuclei one to two cell-layers away from the tunica propria, corresponding to leptotene and pachytene spermatocytes (supplementary material Fig. S7B). Often, TUNEL⁺ null tubule sections in the stages following stage IX exhibited small and degraded apoptotic nuclei near the tunica propria, suggesting that these nuclear fragments are the remnants of earlier apoptotic nuclei (supplementary material Fig. S7B,C).

Because histone PTM plays key roles in spermatogenesis, we sought to determine whether aberrant histone methylation could play a role in the observed defects in H3f3b null testes. We utilized IHC-IF to examine the levels of H3 Lys9 trimethylation (H3K9me3), a mark associated with transcriptional silencing and heterochromatin. H3K9me3 levels were strikingly higher overall in null testes when compared with those of WT (supplementary material Fig. S2, Fig. S8A). WT mice exhibited patterns of H3K9me3 staining that were consistent with previous reports of H3K9me3 in normal mouse testes, in which H3K9me3 was found in spermatogonia, pre-leptotene, leptotene and pachytene spermatocytes, and in the chromocenters of round spermatids (Iwamori et al., 2011; Liu et al., 2010; Payne and Braun, 2006). By contrast, in nulls, nuclear H3K9me3 staining was relatively more intense in spermatogonia, pre-leptotene, leptotene, zygotene and early (stage I-V) pachytene spermatocytes (supplementary material Fig. S9, Fig. S10). H3K9me3 levels did gradually decrease from pachytene to diplotene stages of meiotic prophase (Fig. 3A,B), although H3K9me3 levels were still higher in null tubules. H3K9me3 nuclear foci with intensities much higher than those seen in WT persisted in null pachytene spermatocytes and step 1-13 spermatid chromocenters, and null tubules typically exhibited more brightly stained focal points in secondary or meiotic spermatocytes when compared with WT (Fig. 3C).

When we enriched for histone-containing chromatin proteins using acid extraction on whole testes, we observed a 1.85-fold increase in the levels of H3K9me3 relative to those of total H3 by immunoblotting samples from null males (P=0.001, Fig. 3D,E). By contrast, the amount of H3K9me3 relative to β-actin in WTCL from nulls was not elevated (P=0.287, supplementary material Fig. S8B-D). To determine whether H3K9me3 accumulates specifically in the cell types in which we observed TUNEL⁺ events, we stained pre-leptotene spermatocytes and spermatocytes for H3K9me3. Sorted pre-leptotene spermatocytes exhibited a 1.97-fold higher level of H3K9me3, through IHC-IF analyses, relative to WT (supplementary material Fig. S8E,F, P<0.001). Sorted null spermatocytes and null γH2A.X⁺ spermatocytes (primarily pachytene cells) generally displayed higher levels of H3K9me3, although this difference was not significant (supplementary material Fig. S8G,H).

To determine whether H3f3b loss could affect other PTMs, we also examined levels of the euchromatic Lys4 trimethylation mark on H3 (H3K4me3) and the repressive Lys27 trimethylation mark on H3 (H3K27me3) in WT and null testes. We were able to detect a small, yet consistent decrease in the levels of H3K4me3 in null testes by IHC-IF (supplementary material Fig. S11A) but were unable to discern a difference in specific localization patterns between WT and null tubules. H3K4me3 levels were slightly decreased in null WTCL when compared with that from WT littermates, although overall this decrease was not significant (supplementary material Fig. S11B,C). H3.3 is enriched for euchromatic marks, such as H3K4me3 (Hake et al., 2006; McKittrick et al., 2004), and the observed decreases could reflect a loss of H3K4me3-bearing H3.3 histones. We were also unable to detect a significant difference in the amount of H3K27me3 between the two genotypes through analyses of WTCL (supplementary material Fig. S11D) or IHC-IF analyses (data not shown). We observed significant increases in the relative phosphorylation levels of the H3.3-specific Ser31 residue (H3.3 S31P) in null testes (supplementary material Fig. S11D, compared with the total H3.3 protein levels in the same samples shown in supplementary material Fig. S1C-E), suggesting that the remaining H3.3 protein in nulls might become hyper-phosphorylated on Ser31, as described previously (Bush et al., 2013). Notably, H3.3 S31P was found to localize primarily to meiotically dividing spermatocytes in stage XII tubules (supplementary material Fig. S11E) in samples from both WT and null mice. These results suggest that loss of H3.3 primarily affects H3K9me3 PTM at specific timepoints in spermatogenesis, particularly during the pre-leptotene to leptotene transition.

We isolated RNA from H3f3b WT and null testes and conducted gene expression microarray studies. Employing a 2-fold or greater expression change cutoff, only six genes were up- or downregulated by loss of H3f3b, with one of the downregulated genes being H3f3b itself (supplementary material Table S2). Using a 1.5-fold cutoff, 186 and 375 genes in total were up- and downregulated in testes by loss of H3f3b, respectively (supplementary material Table S3). Importantly, the expression changes in 45 of the up- and 111 of the downregulated genes were shared events in both samples from null mice (supplementary material Table S3). Database for annotation, visualization and integrated discovery (DAVID) gene ontology (GO) analysis (Huang et al., 2009a,b) revealed that the most prominent ontological cluster of genes downregulated using the 1.5-fold cutoff were genes associated with sperm and spermatogenesis. These genes included Catsper1 and Catsper3 (cation channel, sperm associated 1 and 3), Spaca3 and Spaca4 (sperm acrosome 3 and 4), the flagellum associated gene Cabyr, the spermatozoa associated Hook1 and the spermatogenesis-linked adipogenin (Adig) related cDNA (BC054059). Validation quantitative (q) PCR measurements of Catsper1, Spaca3, Adig/BC054059 and Hook1 levels confirmed that these fertility-linked genes were significantly decreased in null adult testis relative to that of WT littermates (Fig. 4A). GO analysis of expression data also indicated that genes found to be upregulated in the null testes were ontologically enriched for an ‘induction of apoptosis’ functional cluster. Downregulated genes tended to be localized away from CpG-rich regions, whereas 61% of upregulated genes trended the opposite way (supplementary material Table S4). Deregulated genes were not enriched on any particular chromosome (supplementary material Table S4).

For a more detailed analysis invoking a cell-type specific component, we utilized mouse transcriptome data from the Gene Expression Omnibus (GEO) database (GSE21749, GSE4193, GSE23119 and GSE21447) to determine the expression pattern of H3f3b null up- and downregulated genes in normal, staged germ cell populations, as performed previously (Gaucher et al., 2012; Tan et al., 2011). Importantly, the majority of genes found to be downregulated by 1.5-fold or more in null testes are normally, and predominantly, expressed in round spermatids and moderately in spermatogonia (Fig. 4B). This pattern is logical given that a large proportion of downregulated genes was enriched for functional processes involving sperm development. By contrast, the majority of genes upregulated by 1.5-fold or greater are normally expressed in spermatogonia and early spermatocyte (pre-leptotene, leptotene, zygotene) populations (Fig. 4C), the same populations of cells that exhibited an increase in apoptotic events upon TUNEL staining, although it is important to note that pre-leptotene spermatocytes were found to be significantly decreased through FACS analysis in the null.

We performed ChIP-seq for H3K4me3 in WT and null testes samples (two biological replicates each). Globally, loss of reads mapping to the H3f3b locus was apparent in both null samples, validating the null genotype (supplementary material Fig. S12A). Both null samples had fewer H3K4me3 peaks than both WT samples, and the peaks were on average narrower in null samples than those of WT (supplementary material Table S5). We constructed consensus peak sets for WT and null samples by intersecting the peaks of each replicate. The WT consensus peak set contained 27% more H3K4me3 peaks than the null consensus peak set (Fig. 4D), and the WT peaks were an average of 39% broader than the peaks of the null samples (1680 versus 1210 bp, P<0.0001, supplementary material Fig. S12B). Nearly all of the peaks in the null samples were also found in the WT samples (97%), whereas 23% of the WT consensus peaks were absent from the null consensus peak set. The two WT samples were very similar to each other, with ∼90% overlapping peaks (supplementary material Fig. S12C). However, null samples were more divergent; only 81% of the peaks from null sample 73 were also found in null sample 67 (supplementary material Fig. S12D). We used the R package ‘DiffBind’ to cluster the ChIP-seq data and determine the correlation between samples. WT samples were highly similar to each other (correlation>0.94), whereas the null samples, although similar in an absolute sense, were relatively less alike to each other (correlation<0.90). The two null replicate samples specifically clustered together, as did the two WT samples with each other, demonstrating that loss of H3f3b might result in both overlapping, but also distinct gene expression changes (supplementary material Fig. S13A). DiffBind identified 281 peaks that were significantly reduced in null samples, and 90 peaks that were significantly increased in null samples compared with those of WT (supplementary material Table S6). No consistent pattern of changes in the frequency of satellite repeats were observed in the nulls (supplementary material Fig. S13B).

H3K4me3 peaks increased in null were enriched for embryonic development genes, whereas the decreased peaks were enriched for spermatogenesis genes (supplementary material Tables S6, S7). We found a significant overlap between differentially regulated genes in the nulls that were identified by microarray analyses and genes exhibiting unique or increased and/or decreased patterns of H3K4me3 through ChIP-seq analyses in the nulls (supplementary material Fig. S14A-D). qPCR analyses of overlapping genes between the microarray and H3K4me3 ChIP-seq datasets verified that changes in transcription were reflected in the upregulation or gain of H3K4me3 peaks (supplementary material Fig. S13C-E) or the downregulation or loss of peaks (supplementary material Fig. S13F-H). Notably, both null samples exhibited a loss of H3K4me3 peaks around the protamine domain cluster, comprising protamine genes 1, 2 and 3 (Prm1-Prm3) and transition protein gene 2 (Tnp2) (Martins and Krawetz, 2007) (Fig. 4E), and around the transition protein 1 gene (Tnp1) (supplementary material Fig. S15A). qPCR analyses on RNA from whole testes revealed significant decreases in Prm1-Prm3 expression in nulls (Fig. 4F) but surprisingly not in Tnp1 and Tnp2 expression (Fig. 4F; supplementary material Fig. S15B), although these results might be affected by differences in germ cell content.

To further explore a loss of expression from the protamine domain cluster in nulls, we examined WT and null spermatids, which had been sorted using FACS. Combined null samples exhibited lower levels of Prm1 and Prm2 when compared with combined WT samples (supplementary material Fig. S15C,D), although these differences were not statistically significant. This might be due, in part, to the presence of one outlier null male that exhibited aberrantly high levels of these transcripts (supplementary material Fig. S15E,F). All null males expressed lower levels of Prm3 (P=0.030) and Tnp2 (P<0.001) (supplementary material Fig. S15G,H). Null males did not differ from WT in Tnp1 expression (supplementary material Fig. S15I). Taken together, these data suggest that H3.3 is necessary for the appropriate expression of spermatogenesis-related genes.

We stained testes sections with lectinPNA to follow acrosome development and spermiogenesis in H3f3b WT and null tubules. Null and WT seminiferous tubules were capable of producing acrosomal granules and an acrosomal cap. However, during spermiogenesis, late-stage null tubules displayed asynchronous spermatid elongation. As the abnormal, mixed nature of late-stage null tubules made it difficult to distinguish stage IX tubules from stage X tubules, we classified stage IX as tubules in which the majority of the spermatids were step 9, whereas those at stage X were characterized as containing a majority of step 10 spermatids among a mixture of step 9-11 spermatids in the tubule lumen. Individual sections of late-stage null tubules frequently exhibited spermatids characteristic of a number of different stages.

Stage IX null tubules consistently exhibited aberrant mixed characteristics of (1) late stage VIII tubules, in which spermatozoa have exited the tubule lumen during spermiation and round spermatids persist, and (2) stage X tubules, in which early elongating spermatids begin to become bilaterally flattened (Fig. 5A). Similarly, stage X null tubules, normally containing bilateral flattening spermatids, frequently exhibited characteristics of stage IX and late-stage tubules (e.g. stage XI) (Fig. 5B), a phenotype that was only very rarely present in WT testes. These data suggest that null tubules do not exhibit synchronized chromatin reorganization, which is normally observed during spermiogenesis. It is also possible that this asynchronous development could lead to elevated spermatid numbers during certain stages of spermatogenesis due to delayed spermiation.

Tnp1 plays a key role in the transition to protamine-based chromatin that is essential for normal spermatogenesis and fertility (Balhorn, 2007; Braun, 2001; Carrell et al., 2007; Miller et al., 2010; Okada et al., 2007). Tnp1 first becomes strongly expressed in step 11 spermatids (Zhao et al., 2004) and persists until step 13-15 (Yu et al., 2000; Zhao et al., 2004). Null late-stage tubules contained step 8-9 spermatids (Tnp1^–) and step 10-11 spermatids (Tnp1⁺) within the same tubule, an exceedingly rare occurrence in stage-matched WT tubules (supplementary material Fig. S2; Fig. 5C). In addition, Tnp1⁺ spermatids were abnormally present in a small number of mid-stage null tubules.

To further examine whether H3f3b loss affected chromatin reorganization and proper sperm function by interfering with protamine replacement into chromatin, we stained WT and null testes for Prm1. The highest amount of Prm1 protein staining in WT tubules was found in step 11-16 spermatids (stage XI-VIII tubules) (supplementary material Fig. S2; Fig. 6A,B), as previously reported (Zhao et al., 2004). By contrast, Prm1 staining in step 11-16 null spermatids was dramatically reduced (Fig. 6A,B and supplementary material Fig. S16A,B), perhaps a result of, at least in part, the decrease in total spermatids per tubule area and reduced Prm1 expression. Levels of Prm1 in null tubules was also very weak or undetectable in late, mixed-stage tubules with delayed step 8-9 spermatids present within the luminal area.

Prm1 levels were also decreased in null decondensed sperm (Fig. 6C). Decondensed null sperm also exhibited higher overall levels of H3K9me3 than those of WT (Fig. 6C; supplementary material Fig. S16C) and in a significantly higher proportion of sperm (supplementary material Fig. S16D, P<0.001). Null sperm were also found to have a high number of apoptotic nuclei by TUNEL staining (supplementary material Fig. S16C) with 23% of null sperm being TUNEL⁺, 12.1-fold higher than WT (P<0.001, supplementary material Fig. S16E). Increased levels of H3K9me3 in null sperm were moderately correlated to the increase in sperm apoptosis, as 18.8% of H3K9me3⁺ sperm were also TUNEL⁺ (P<0.001, supplementary material Fig. S16F). Taken together, these data indicate that H3.3 is necessary for the switch to protamine-based chromatin and for normal chromatin restructuring during spermatogenesis.

Here, we found that reduced H3.3 levels, as a result of H3f3b knockout, disrupted male reproductive function through developmental defects, resulting in dysfunctional spermatozoa and reduced total spermatozoa numbers. One possible explanation for the decreased sperm concentration could be defects arising in meiosis, although we were unable to discern significant differences in the number of stage XII meiotic tubules or in the number of meiotically dividing spermatocytes in WT and null testes. However, the null cells with the strongest meiotic defects could well be those most prone to apoptosis and hence be lost for analysis. Another explanation is that reduced sperm levels could be due to elevated levels of apoptosis that are not directly related to meiosis; indeed, a substantial proportion of null epididymal sperm exhibited apoptotic nuclei, suggesting that lethal defects arise as a result of earlier events in the generation of the sperm that are disrupted in the null.

Notably, the majority of apoptotic events in null testes occurred in or around stage IX (supplementary material Fig. S2), a period where spermatogonia stem cells differentiating into spermatocytes cross the tightly regulated blood-testis barrier (BTB) (Bordlein et al., 2011; Cheng and Mruk, 2012; Morales et al., 2007; Mruk and Cheng, 2004; Oishi et al., 2004). Downregulated genes in null testes are expressed normally in spermatogonia, a cell type that in null mice also displayed increases in H3K9me3, indicating that an aberrant repressive chromatin state established in the early stages of null spermatogonia development might impede the expression of genes necessary for differentiation into pre-leptotene spermatocytes and the BTB transition.

Strong decreases in H3.3 and increases in H3K9me3 were observed in late-stage null tubules (supplementary material Fig. S2), particularly in spermatogonia and spermatocytes. Western blotting for chromatin-bound H3K9me3 from acid extract revealed a significant increase in levels compared with those of WT. Although this change was not detectable in WTCLs, this discrepancy might be due to the variation in germ cell size within the testes, demonstrated by the fact that the increase of H3K9me3 varied amongst distinct cell types, or because WTCLs include concentrations of non-chromatin bound histones that probably have distinct PTM patterns. A previous study (Okada et al., 2007) has demonstrated that knock out of the histone demethylase Jhdm2a leads to elevated levels of H3K9me3 in round spermatids, although these changes were undetectable using western blot.

H3.3 is normally depleted for H3K9 methylation (Hake et al., 2006; McKittrick et al., 2004) and is thought to inhibit the spread of heterochromatin, in part through eviction of histones carrying H3K9 methylation (Nakayama et al., 2007). Thus, the gains in the amount of H3K9me3 in H3f3b null testes cells might be a direct consequence of decreased H3.3 protein. The H3K9me3 methyltransferase Suv39h2 is most highly expressed in B-type spermatogonia and leptotene spermatocytes and appears to regulate H3K9 methylation at these stages (O'Carroll et al., 2000; Peters et al., 2001), suggesting that the general loss of H3.3 might result in hypermethylation of canonical H3 by Suv39h2. Supporting this hypothesis, H3.3 levels were significantly lower in null germ cell populations where an increased H3K9me3 signal was also observed (supplementary material Fig. S2). Additionally, nucleosome replacement through H3.3 during early pachytene results in a loss of most histone PTMs, leading to the normal de novo reappearance of H3K9me3 during stages VI-XII (van der Heijden et al., 2007). We predict that the normal processes by which H3.3 evicts histones bearing repressive PTMs (i.e. H3K9me3) in spermatocytes (Fig. 7A) might be disrupted in null testes, resulting in the overabundance of H3K9me3 (Fig. 7B).

The balance of opened versus closed chromatin during spermiogenesis (Rathke et al., 2013) might also be influenced by H3K9me3, which is involved in chromatin condensation and gene repression (Iwamori et al., 2011; Jenuwein and Allis, 2001; Kouzarides, 2007; Payne and Braun, 2006; Peters et al., 2001). Downregulated genes that are associated with male reproductive function and spermatogenesis in the null testes are normally expressed in round spermatids, suggesting that increased H3K9me3 in round spermatids inhibits their normal development (Fig. 7B). Null sperm also displayed high rates of apoptosis, coinciding with H3K9me3 expression, indicating that this repressive PTM might aberrantly persist throughout spermiogenesis in null testes, leading to sperm death.

Major replacement of nucleoproteins was also disrupted in null testes (supplementary material Fig. S2), as tubules expressing Tnp1 abnormally exhibited characteristics of both the early- and late-stages of spermiogenesis. Further, Prm1 expression was substantially reduced. Decondensed null epididymal sperm exhibited abnormally high levels of H3K9me3 and low levels of Prm1, suggesting that H3K9me3 inhibits the normal reorganization process by which protamines are incorporated into chromatin. These phenotypic effects in nulls could be a consequence of the lost H3K4me3 peaks around protamine and transition protein genes, the expression of which is essential for proper chromatin condensation and protamine replacement during spermiogenesis (Fig. 7A) (Cho et al., 2003; Lee et al., 1995; Okada et al., 2007). Another possibility is that H3.3 loss and H3K9me3 gain alters nucleoprotein exchange, as disruption of nucleoprotein displacement, particularly in regards to replication-independent histone variants (e.g. TH2B) (Montellier et al., 2013) or histone PTMs (Gaucher et al., 2012), can considerably impact the ability of spermatids to properly condense chromatin and incorporate transition proteins or protamines. In this regard, H3.3 plays an important role in chromatin condensation (Lin et al., 2013). In addition to H3.3, the presence of repressive (e.g. H3K27me3) PTMs and chromatin-associated factors contribute to histone retention in protamine-free regions in sperm (Erkek et al., 2013). Taken together, these observations suggest that H3.3 is necessary for normal protamine expression, and further, that the absence of H3.3 and in turn the aberrant presence of repressive PTMs (e.g. H3K9me3) might lead to abnormal chromatin domains in condensing spermatids that block Prm1 deposition in null mice (Fig. 7B).

It is interesting to note that the transcription-coupled incorporation of H3.3 is required to establish marks of active and inactive chromatin that are necessary during fertilization (Erkek et al., 2013); therefore, it might be possible that surviving, null Prm1⁺ spermatozoa face additional problems in activating zygotic transcriptional programs should fertilization occur. The majority of genes with decreased expression were not found in CpG-rich regions, which could suggest that H3.3 plays some regulatory role in maintaining the expression of genes found in CpG-poor areas. Meanwhile, upon H3.3 reduction, the remaining H3.3 histone might be redirected to CpG-rich regions to maintain the function of these genes. Future studies, particularly on H3f3a, will further clarify the normal role of histone H3.3 in regulating testes chromatin dynamics and sperm development.

Animal studies were approved University of California, Davis Institutional Animal Care and Use Committee (IACUC). Additional methods can be found in the supplementary material Methods.

For a comprehensive list of animals used in this study, consult supplementary material Table S1. For a detailed listing on the number of images, cells, testes, or seminiferous tubules used in this study, please consult supplementary material Table S8.

Testes (see supplementary material Tables S1, S8) were isolated, embedded in paraffin and specifically treated depending on the antibody (see supplementary material Methods). Primary antibodies against the following proteins were used in this study: H3K9me3 (Abcam 8898), H3K27me3 (Cell Signaling C36B11), H3K4me3 (Millipore 04-745), Tnp1 (Abcam 73135), Prm1 (Briar Patch Biosciences Hup1N), H3 S10P (Millipore 06-570), H3.3 S31P (Abcam ab92628), H3.3 (Abnova H00003021-M01), Sycp3 (from Neil Hunter, UC Davis) and γH2A.X (also from Neil Hunter). Secondary antibodies were used as follows: rabbit or mouse IgG (Alexa488 or Alexa546, Invitrogen) and lectinPNA (Invitrogen L21409). Slides were then mounted using Vectashield mounting medium with DAPI (Vector Labs) and imaged using a Nikon Eclipse 80i microscope (Nikon) with a Retiga 2000R camera (QImaging) or using a Biorevo BZ-9000 microscope system (Keyence).

Testes cell isolation was performed as previously described (Gaysinskaya et al., 2014; Getun et al., 2011) with minor modifications, stained with Hoechst 33342 (Invitrogen #H21492) and propidium iodide (Roche #11348639001), and sorted on a 16 color inFlux v7 high speed cell sorter (Becton-Dickinson-Cytopeia) in a HEPA enclosure. Flow data analysis was performed using FlowJo software (Treestar Incorporated). For qPCR, 75-100% of sorted populations were used for RNA extraction (Macherey-Nagel 740955), concentrated using a Nucleospin RNA XS kit (Macherey-Nagel 740902), and 4-8 ng of RNA was used to amplify cDNA using the CellAmp Whole Transcriptome Amplification Kit (Real Time) Ver. 2 (TaKaRa Biosciences 3734). For staining, sorted cells on slides were treated with antibodies against H3K9me3 (Abcam 8898), Sycp3 and γH2A.X (from Neil Hunter, UC Davis), rabbit or mouse IgG (Alexa488 or Alexa546, Invitrogen), and lectinPNA (Invitrogen L21409) and imaged using a Biorevo BZ-9000 microscope system (Keyence).

Whole cell extracts and acid extracts were prepared from one half- to one quarter-testis. Western blotting was performed as described previously (Bush et al., 2013) with a few modifications. Antibodies against the following proteins were used: H3K9me3 (Abcam ab8898), H3K4me3 (Millipore 04-745), β-actin (Sigma A1978), H3.3 (Abnova H00003021-M01), H3.3 S31P (Abcam ab92628), H3K27me3 (Cell Signaling C36B11), H3 (Upstate no. 05-499), anti-rabbit IgG horseradish peroxidase (Jackson 111-035-003), anti-mouse IgG horseradish peroxidase (Jackson 115-035-003), anti-rabbit IgG IRdye800CW (Licor no. 827-08365) and anti-mouse IRdye680RD (Licor no. 926-68170).

SYBR Green-based qPCR for genes was performed using Absolute Blue QPCR Master Mix (Thermo Fisher Scientific) and run on a LightCycler 480 (Roche) with included software. Each reaction utilized 10 ng cDNA, and separate reactions containing 10 ng RNA were used to check for genomic DNA contamination. Fold changes were normalized to Ppia or Ubc and analyzed using the 2^-ΔΔCt method.

Microarray analysis was performed on n=2 WT and n=2 null whole testis RNA extracts, as described previously (Bush et al., 2013). RNA from constitutive H3f3b null and WT testes was submitted in biological duplicates. A total of 500 ng RNA from each sample was used on an Illumina MouseWG-6 v2 Expression BeadChip containing over 25,000 probes. Amplification and hybridization were performed at the UC Davis Expression Analysis Core according to the Illumina protocol. The data have been deposited with accession number GSE35303 in the GEO database. For CpG analysis, deregulated gene lists were cross-referenced to CpG island tracks on the UCSC Genome Browser and intersected using Galaxy.

We conducted ChIP-seq in WT and null testes tissue using two replicates of H3K4me3 (Millipore 04-745) and matched input controls. Tissue preparation and ChIP were performed as described previously (Barrilleaux et al., 2013). Libraries were prepared as described previously (Bush et al., 2013) and sequenced using an Illumina HiSeq, yielding at least 6 million reads per sample (supplementary material Table S5). The data have been deposited with the accession number GSE60207 in the GEO database.

Staging of mouse seminiferous tubules was performed as described previously (Ahmed and de Rooij, 2009; Hess and Renato de Franca, 2008; Meistrich and Hess, 2013) using lectinPNA acrosomal staining as a guide. For quantification of meiotic tubules, meiotic spermatocytes and spermatids, testes were stained with an antibody against H3 S10P (Millipore 06-570), lectinPNA (Invitrogen L21409) or lectinPNA alone. Staged tubules were then quantified for the presence of H3 S10P⁺ tubules and H3 S10P⁺ cells or round (step 1) and elongated (step 13) spermatids.

Male animals were submitted to the UC Davis Mouse Biology Program for sperm functional testing. Sperm were isolated using the swim-out method from the epididymis and assessed using an IVOS computerized sperm analyzer (Hamilton Thorne). Approximately 5×10⁶ spermatozoa were placed on slides, decondensed, stained and imaged using a Biorevo BZ-9000 microscope system (Keyence).

Paraffin-embedded testes sections or spermatozoa slides were used for TUNEL staining (DeadEnd Fluorometric TUNEL System, Promega G3250). TUNEL⁺ tubules were counted when exhibiting one or more apoptotic events. Quantification of apoptotic nuclei was performed using ImageJ software.

Student's two-tailed, unpaired t-tests were used to determine significance between null and WT animal measurements. P-values for Venn diagram overlaps were calculated in R using the hypergeometric test. Error bars represent s.e.m. unless otherwise noted.

We thank the UC Davis Mouse Biology Program for aid in production of the targeted H3f3b mice and sperm analysis and Stephen Griffey and Amanda Koehne for analysis of the knockout mice organ histology. We thank the UC Davis Expression Analysis and DNA Technologies Cores for their assistance with expression array studies and sequencing support, and Jonathan Van Dyke from the UC Davis Flow Cytometry Facility for his assistance and expertise. We thank Rod Balhorn for advice on testes analysis, Henriette O'Geen for aid in ChIP-Seq studies, Neil Hunter for advice and antibodies, and Richard Yan for performing blinded seminiferous tubule quantification and analysis.