The eutherian-specific histone H3.4 promotes germ cell development and reproductive fitness

Abstract

Many genes encoding chromatin proteins are subject to evolutionary selection driving reproductive fitness. In mice and men, the H3f4 / H3-4 gene encoding the histone H3.4 variant (formerly known as H3t) is essential to spermatogenesis. Here we define the evolutionary origin and molecular-physiological roles of sequence variation in H3f4 for male germ cell development in mice. Our phylogenetic analyses indicate that eutherian H3f4 orthologs originate from an ancestral H3.2 gene existing prior to the divergence of eutherian and marsupial mammals over 100 million years ago. Positioned in small histone gene clusters, eutherian H3f4 orthologs show increased non-synonymous and synonymous substitution rates compared to orthologous marsupial H3.2 loci located in prototypal large histone clusters. To determine the impact of sequence divergence on reproductive fitness, we revert non-synonymously substituted residues in H3.4 to those present in canonical H3.1 (H3f4^V24A, H3f4^H42R, H3f4^S98A). Expression of such a triply reverted H3f4^H3.1 allele on a H3f4-deficiency background causes an >40% reduction in testis weight associated with impaired meiotic progression, death of pachytene spermatocytes, impaired differentiation of spermatids and aberrant expression of thousands of genes during spermatid elongation. Hemizygous expression of individual residue substitution alleles reveals residues V24 and H42 of H3.4 to promote spermatogenesis, while residue S98 is neutral. Together, our study shows that H3f4 has been subject to positive evolutionary selection, promoting male reproductive fitness.

Here the authors explore the evolutionary origin of the histone H3.4 variant which is essential for spermatogenesis in mice and human. Mutating residues so they match canonical H3.1 sequences leads to impaired meiosis.

Introduction

Spermatogenesis is a complex developmental process that yields to the daily formation of millions of haploid spermatozoa competent for supporting embryonic development. While the genome is packaged by histones during most of this process, histones become replaced by protamines during spermatid elongation, thereby compacting their chromatin, which is required for fertility^1,2. Contrasting somatic cells, germ cells express many histone variants, some serving prominent chromatin remodeling roles complementary to canonical histones^1,3,4. In mammals, testicular histone variants exist within all families except for histone H4^5,6. Multiple H1, H2A, H2B, and H3 genes are subject to rapid evolution, including pseudogenization^7–10.

Among H2A and H2B proteins, the TH2A and TH2B histones are highly expressed during male meiosis and replace over 60% of canonical H2A and H2B^11,12. TH2A is required for meiotic progression^13,14 and TH2B was shown to prime chromatin for histone-to-protamine replacement during spermatid elongation¹¹. In accordance, TH2A/TH2B-containing nucleosomes make fewer contacts with DNA and exhibit reduced stability, compared to canonical nucleosomes^15,16. Both proteins are also maternally expressed in oocytes and pre-implantation embryos, where they contribute to transcriptional activation of the paternal genome¹⁷. Further, specialized short H2A variants lacking the C-terminal part (H2A.B and H2A.L) are expressed almost exclusively in the testis^7,18. In mouse, histone H2A.B.3 is the most highly expressed H2A.B variant that is incorporated at transcription start sites (TSS) of active genes in spermatocytes and round spermatids^19,20. However, all three H2A.B paralogs serve biparental contributions to embryonic development of mice²¹. The H2A.L.2 histone is incorporated during spermatid elongation and is required for the loading of transition proteins (TNPs) onto chromatin^22,23. The short H2A variants (H2A.B, H2A.L, H2A.Q, and H2A.P) have been identified only in eutherians⁷. They likely originate from H2A.R, a regular H2A variant present only in marsupials and monotremes⁷. Importantly, short H2A genes underwent pseudogenization or further duplication during eutherian evolution⁷.

Next to TH2B, H2B.L is expressed during spermatid elongation and plays a role in pericentric heterochromatin organization²². Another variant, H2B.W.1, is expressed in human spermatogonia²⁴, and has been implicated in male fertility^25,26. Like H2A variants, the testis expressed eutherian-specific TH2B, H2B.W, and H2B.L-encoding genes, present in all mammals, also underwent duplication and/or pseudogenization during evolution⁸. For example, H2B.L exists only as a pseudogene in the human genome⁸.

For H1 histone variants, H1T (H1.6) is expressed from pachytene spermatocytes until late spermatids²⁷ in which it controls chromatin de-condensation at genic regions²⁸ and heterochromatin maintenance at repeat elements²⁹. The H1T2 (H1.7) and HILS1 (H1.9) variants are highly expressed during spermatid elongation^30,31. H1T2 is required for fertility in mouse^31,32 and human³³. In human, HILS1 is retained only as a pseudogene³⁴.

Within the H3 family, primates express novel H3 variants, mostly originating from replication-independent (RI) H3.3-encoding genes and accumulating non-synonymous mutations. For example, H3.X and H3.Y are specific to primates^35,36, while H3.5 exists in hominids³⁷. H3.5 is expressed in the spermatogonia and primary spermatocytes^38,39, and reduced H3-5 transcripts were associated with non-obstructive azoospermia³⁹. In mice, thirteen H3 genes were predicted to have evolved from  H3.3-encoding genes, with some being specifically expressed in the testis¹⁰.

Next to H3.3 variant-related genes, the H3.4 variant (formerly known as H3t)³⁴ is mainly expressed in spermatogenic cells^{10,22,40–44}. H3.4 is encoded by a single H3f4 gene possessing a 3′-UTR stem loop^10,45, and is incorporated into chromatin in a replication-coupled (RC) manner¹⁰. The nuclei of c-KIT positive differentiating spermatogonia, meiotic spermatocytes, and haploid round spermatids are widely labeled by H3.4⁴². The protein is removed from sex chromosomes during meiotic sex chromosome inactivation (MSCI) process^42,43,46. The precise developmental timing of H3.4 deposition and eviction, and its genomic distribution remain unexplored. In mouse, the H3f4 locus is essential to spermatogenesis, with spermatogonial differentiation failing in its absence⁴². In human, the c190C>T single nucleotide polymorphism (SNP) encoding the R64C substitution in H3-4 has been associated with a Sertoli cell only syndrome (SCOS)⁴⁷.

Given the pivotal role of H3f4 and H3-4 in male fertility of mice and human, it is intriguing why spermatogenesis became dependent on the gene, despite the presence of multiple canonical RC H3 genes in eutherian genomes. In human, H3-4 is a member of the small HIST3 gene cluster that also contains the H2AC25, H2BC26, and H2BC27 histone genes and which is flanked by the TRIM11, TRIM17, and RNF187 genes (Fig. 1). The composition of HIST3 clusters is largely conserved in other eutherian species, including apes, dog, cat, pig, horse, and mouse³⁴. The origin of H3-4 and its conservation among other mammals are, however, unknown.

Fig. 1. Sequence divergence of histone H3.4 in the mammalian lineage.

a Amino acid sequence alignment of histones H3.1, H3.2, and H3.3 in mice and of H3.4 in indicated species. Amino acid residues conserved relative to H3.1 are represented as dots while variations are highlighted in yellow. H3.2 and H3.3 specific residues are highlighted in blue and green, respectively. Eutherian species are ordered according to their position in a phylogenetic tree determined by TimeTree⁵⁴, representing over 100 million years of divergence. The panel includes eutherian H3-4 genes linked to TRIM11/TRIM17 genes. The marsupial putative H3-4 orthologs are ordered as in Supplementary Data 1. b Schematic representation of the postulated ancestral histone protocluster harboring a TRIM11-linked H3.2 gene (top) and its H3-4 derivatives in eutherian species (bottom). Ensembl genome browser views of histone clusters in human and marsupial species are shown in Supplementary Figs. 1–3.

In human, four amino acids differ between H3.4 and canonical H3.1 (V24A, M71V, S98A, and V111A). In vitro, H3.4-specific residues M71 and V111 reduce the stability of nucleosomes^48,49. Moreover, H3.4 interacts preferentially via its residue V111 with the histone chaperone hNAP2 over hNAP1 thereby promoting nucleosome formation in vitro⁵⁰. The Tudor domains of PHF1 and PHF19, two accessory components of the Polycomb Repressive Complex (PRC) 2 complex, preferentially bind H3.4K27me3 over canonical H3.1K27me3 via residue V24^44,51, suggesting a role for H3.4 in PcG-mediated gene repression. In mouse, while V24 and S98 are present, M71 and V111 residues are not (Fig. 1a). Instead, mouse H3.4 harbors a histidine rather than an arginine at position 42 which promotes open chromatin in vitro⁴². The in vivo impact of the evolutionary acquired amino acid substitutions in H3.4 for spermatogenesis and reproductive fitness are unknown.

In this study, we investigated the evolutionary origin of H3-4 orthologous genes and assessed whether acquired amino acid substitutions may regulate fertility traits. We identify that the eutherian-specific H3f4 evolved from a H3.2 gene existing over 100 million years ago, prior to separation of eutherian and marsupial lineages. Following the high expression of H3f4, H3.4 replaces progressively canonical H3 histones during spermatogonial expansion and differentiation. In proliferating spermatogonia, RC-deposition of H3.4 occurs genome-wide, whereas in post-replicative spermatocytes and spermatids, H3.4 remains located at transcriptionally silent and intergenic regions characterized by low nucleosome turnover. In turn, at transcriptional start sites (TSSs) of active genes and along the X chromosome in pachytene spermatocytes, H3.4 is gradually evicted and replaced by H3.3. Through gene editing, we generated a mouse model expressing H3.1 from the H3f4 locus. Under conditions of reduced gene dosage, our results show that H3.4 supports spermatogenesis better than H3.1, with V24 and H42 amino acids playing key roles in the function of H3.4.

Results

The H3-4 gene is conserved in the eutherian lineage

To determine the phylogeny of the H3-4 gene and the level of its amino acid conservation, we searched for H3-4 orthologs in a wide set of eutherian species (Fig. 1a and Supplementary Data 1). We performed TBLASTN search using the human H3.4 protein sequence as a query against the genomes available at the Ensembl website (www.ensembl.org, release 112) and defined H3-4 orthologs as those H3 genes localized within the “HIST3” gene cluster containing TRIM11 and/or TRIM17, H2A and several H2B genes³⁴ (Supplementary Fig. 1a). The presence of a valine residue at position 24 (V24) is the distinctive feature among 60 H3.4-coding orthologs identified in over 100 eutherian genomes analyzed, contrasting with alanine at position 24 (A24) universally present in canonical H3.1, H3.2 and variant H3.3 proteins (Fig. 1a, Supplementary Fig. 2a–c, and Supplementary Data 1). Only in the Upper Galilee mountains blind mole rat (Nannospalax galili), we identified an H3 gene next to TRIM17 that encodes A24. Given its sequence similarity to other H3-4 orthologs (Supplementary Fig. 3a), we concluded it is also an H3-4 ortholog. Beyond H3 genes within orthologous HIST3 gene clusters, we could not identify H3-4 paralogs encoding H3 proteins with V24. In addition to protein-coding H3-4 orthologs in HIST3 loci, we also identified multiple pseudogenes that locate next to TRIM17 and TRIM11 genes (Supplementary Figs. 1b and 3b). Maximum likelihood phylogenetic analysis^52,53 revealed their clustering with protein coding H3-4 genes, confirming their common origin and separation from genes encoding H3.1, H3.2, and H3.3 proteins (Supplementary Fig. 3a). We conclude that H3.4 is encoded by a single gene localized within TRIM11/TRIM17, H2A, and H2B containing HIST3-like gene clusters in eutherian mammals, which emerged over 100 million years (Fig. 1b).

The eutherian HIST3 cluster likely originated from a large histone proto-cluster still existing in marsupial mammals

To further identify the origin of the HIST3 cluster, we expanded our search for putative H3-4 orthologs in marsupial species, requiring linkage with a TRIM gene as a selective genome anchor. In opossum, we identified a large canonical replication-dependent histone cluster consisting of many H3, H2A, H2B, H4, and H1 histone genes, either protein-coding or as pseudogenes. It also contains TRIM11 with an adjacent H3.2 gene in the same orientation (Fig. 1b and Supplementary Fig. 4). This cluster is flanked by ZFP293 and ZFP496 genes on one side and BTN2A2 and BTN1A1 genes on the other. Using these genes as anchors, we identified large histone clusters harboring TRIM11 and an adjacent H3.2 gene in koala, common wombat and Tasmanian devil (Supplementary Fig. 5). Cladogram analysis showed that such TRIM11-linked H3.2 genes in different marsupial species form a separate gene group suggesting a common origin (Supplementary Fig. 3a). The clade also includes two paralogous H3.2 genes in Tasmanian devil and one in wombat that are not linked to TRIM11 (Supplementary Figs. 3a and 5). In eutherians, BTN2A2 and BTN1A1 orthologs are located next to HIST1, which is a large replication-dependent histone gene cluster.

Based on these data, we conclude that in eutherians, the small HIST3 and large HIST1 clusters evolved from a large ancestral histone proto-cluster that still exists in some marsupials (Supplementary Fig. 6). Placental H3-4 genes and their putative marsupial orthologs likely originate from a common TRIM11-linked ancestor and have evolved independently upon the separation of marsupial and placental lineages.

Increased sequence variation among H3-4 orthologs

H3-4 orthologs display amino acid polymorphisms conserved between evolutionary related species (Fig. 1a). For example, residues M71 and V111 are found in closely related human, chimp, and bonobo, Q134 is restricted to drill, sooty mangabey, gelada, and olive baboon and F54 is common to bovine, sheep, goat, and deer species. Residue H42 is specific to muroids (mice, rats, and hamsters). These differences from canonical H3.1/2 proteins may reflect purifying selection of H3.4 polymorphisms reflecting functional adaptation to germ cell development.

To directly compare the degrees of sequence conservation among H3.4 and regular H3 histones, we selected several eutherian (human, lemur, mouse, rabbit, cow, camel, and elephant) and marsupial (wallaby, Tasmanian devil, opossum, koala, and common wombat) species across the evolutionary time axis. We retrieved sequences of all H3.1, H3.2, and H3.3 genes in these 12 species from the Ensembl database (release 112) (Supplementary Data 2). As an outgroup comparison, we included published evolutionary data on H2A and H2B genes^7,8.

We first measured the rates of protein divergence of H3 variants in different species relative to their human orthologs (Supplementary Fig. 3c and Supplementary Data 3). The divergence time reflects the evolutionary time that separates human from the species of interest, as determined by the TimeTree tool⁵⁴. The canonical RC H3.1 and H3.2 genes demonstrate very low divergence rates, even in comparison with the RC H2A and H2B proteins. The divergence rate of H3.4 is slightly higher (with 86 and 96% amino acid identity in camel and elephant, which are most distant species relative to human), yet it is still lower compared to those of H2A.B, H2A.L, H2A.P, and H2B.W variants^7,8.

To test if H3-4 genes are under purifying selection, we calculated the ratio of nonsynonymous to synonymous mutation rates (d_n/d_s) using the CODEML tool from the PAML package^55,56 (Supplementary Fig. 3d, Supplementary Data 4 and “Methods”). Genes with d_n/d_s values close to 0 are under purifying selection, while those with d_n/d_s close to 1 are under neutral evolution, which is typical for pseudogenes. For H3.1 and H3.2-encoding genes, the d_n/d_s ratios are close to zero (0.0002 for H3.1 and 0.005 and 0.001 for placental and marsupial H3.2, respectively), while for H3.3 genes, the d_n/d_s ratio is slightly higher (0.028). For protein-coding H3-4 orthologs, the d_n/d_s ratio of 0.024 is similar to that of H3.3, which is in striking contrast to those of H3-4 pseudogenes (0.46) and H2A.P, H2A.L, and H2B.W variant genes (0.58, 0.32, and 0.23, respectively)^7,8. Statistical analysis using the likelihood ratio test (LRT) strongly rejected the hypothesis of neutral evolution (d_n/d_s = 1), suggesting the need to maintain the conserved amino acid sequence in H3.4 (Supplementary Data 4). We were unable to identify amino acid residues subjected to positive selection (d_n/d_s > 1). These results argue for fixation of certain amino acid polymorphisms, possibly in adaptation to molecular and physiological functions of H3.4

We next investigated whether sequence variation at specific sites in H3f4 relative to canonical RC H3 variants reflects functional adaptation of H3.4 to spermatogenesis. To address this question, we studied H3.4 expression dynamics, cellular and genomic localization, and its functional role in male germ cell development.

Dynamic expression of H3f4 and H3.3 genes during spermatogenesis

We first profiled the expression dynamics of histone transcripts in seven FACS-purified cell types covering the entire process of spermatogenesis, from undifferentiated spermatogonia to elongated spermatids^57,58. Principle component analysis (PCA) on UCSC-annotated protein-coding genes demonstrated a clear separation of cell types, reflecting the developmental trajectory of developing male germ cells (Supplementary Fig. 7a). In undifferentiated and differentiating spermatogonia (SgU, SgD) and in early meiotic leptotene/zygotene stage spermatocytes (ScLZ), most canonical H2A, H2B, H3, and H4 genes were amply expressed (Fig. 2a and Supplementary Fig. 7b). Remarkably, H3f4 expression exceeded expression of H3.1- and H3.2-encoding genes, likely promoting H3.4 deposition during replication. Upon progression through meiotic prophase and in haploid spermatids transcript levels of all RC histones, including H3f4 decreased, while those of RI H3.3-encoding genes (H3f3a and H3f3b) were upregulated⁵⁹, thereby supporting nucleosome turnover beyond replication⁴⁶. Together, we measured a competitively high H3f4 expression during the replicative phase of spermatogenesis, while expression of H3.3 genes takes over during meiosis and spermatid development.

Fig. 2. Expression dynamics of histone H3.4 during spermatogenesis.

a Heatmap showing mRNA expression of H3 and H4 encoding genes during spermatogenesis^57,58. Gene names are according to current nomenclature³⁴. Abbreviations: SgU, SgD, ScLZ, ScPD, rSt, elSt, and eSt refer to populations of FACS-sorted undifferentiated and differentiating spermatogonia, leptotene/zygotene and pachytene/diplotene stage spermatocytes, and round, elongating, and elongated spermatids. b, c Western blots showing H3.1, H3.4, H3.2, and H3.3 proteins separated by TAU gel electrophoresis in FACS-purified spermatogonia (b) and spermatocytes and spermatids (c). We used mouse embryonic stem cells (mESCs) as controls and used a panH3 antibody recognizing the C-terminus of H3 to detect all H3 variants. ScL - leptotene spermatocytes, ScZ - zygotene spermatocytes. Data of one representative out of over three biological replicate experiments is shown. d Whole-mount immunofluorescence images of H3.4^MYC expression in different populations of spermatogonia from one H3f4^MYC/wt mouse (n = 1), marked by GFRα1, PLZF or c-KIT. H3K27me3 marks Sertoli cells. Scale bar: 10 μm. e Representative immunofluorescence images of stage IX to XII seminiferous tubules of one H3f4^MYC/wt mouse (n = 1), co-stained for H3.4^MYC, TNP2, and DAPI. Scale bar: 20 μm. f Quantification of H3.4^MYC and TNP2 signals in step 8–13 spermatids monitored in multiple seminiferous tubules at stages VIII to I. Data from one representative H3f4^MYC/wt mouse (n = 1) is shown. Numbers of spermatids quantified per stage are as follows: Stage VIII: 2584, Stage IX: 850, Stage X: 1018, Stage XI: 1518, Stage XII: 978. Center lines show the median; box limits indicate the interquartile range (25th–75th percentiles); notches represent the approximate 95% confidence interval of the median. Dashed lines represent the Q25 and Q75 intensity levels of respective signals in Sertoli cells in all tubules. Staging was performed manually as described before⁶⁶. a.u. - arbitrary units.

H3.4 is the main H3 protein in differentiating spermatogonia

To investigate the relative protein levels of different H3 variants during male germ cell development, we employed triton—acetic acid—urea (TAU) gel electrophoresis combined with immunoblot detection using a pan-H3 antibody, which enables separation and detection of different H3 variants based on their electric charge and hydrophobicity⁶⁰. Probing acid-extracted histone samples obtained from testis, various somatic tissues, and ESCs from mice, we observed comparable H3.1, H3.2, and H3.3 expression levels in most somatic tissues (Supplementary Fig. 7c). In ESCs, H3.1, and H3.2 were more abundant than H3.3, presumably reflecting their high proliferative nature. In testis, H3.1 and H3.2 levels were remarkably low compared to H3.3 and an abundant H3 protein migrating faster than H3.1. Mass spectrometry identified the faster migrating protein as H3.4, given the presence of the “KQLATKV₂₄AR” peptide and absence of the “KQLATKA₂₄AR” peptide diagnostic for other H3 proteins (Supplementary Fig. 7d). TAU gel electrophoresis of the H3.4 protein expressed in HEK293 cells confirmed the migration pattern of H3.4 (Supplementary Fig. 7e). These analyses revealed prominent H3.4 levels over H3.1 and H3.2 in testis and its absence in other somatic samples, underscoring its testis related expression, as previously noted^{10,22,40–42}.

We next profiled H3 protein levels in FACS-purified testicular male germ cells. While somatic H3 variants were prevalent, H3.4 was readily detectable in SgU cells (Fig. 2b). As spermatogonial proliferation and differentiation proceeded, H3.4 became the main H3 protein in c-KIT expressing SgD cells (Fig. 2b). In post-replicative meiotic prophase cells, H3.3 levels majorly increased from pachytene stage onwards, possibly reflecting increased transcription-coupled nucleosome turnover⁶¹, as well as nucleosome replacement at sex chromosomes during MSCI^46,62 (Fig. 2c). Together, the TAU-WB data argues for a RC replacement of most canonical H3.1 and H3.2-containing nucleosomes by H3.4-bearing nucleosomes during spermatogonial amplification.

Low expression of C-terminally tagged H3.4^MYC is compatible with male germ cell development

To further study the developmental dynamics and genomic localization of H3.4, we aimed for tagging it with small exogenous tags, commonly used in chromatin studies in somatic cells (Supplementary Fig. 8a). Using CRISPR/Cas9 gene editing transgenesis, we successfully added a MYC and an HA tag to the N-terminus (^MYCH3f4 and ^HAH3f4) and a MYC and a 3 × FLAG tag to the C-terminus of H3.4 (H3f4^MYC and H3f4^3xFLAG). To avoid disruption of protein function we added a flexible linker between H3.4 and the tags (Supplementary Fig. 8a). Nonetheless, except for the H3f4^MYC allele, spermatogenesis and male fertility were severely impaired in all transgenic mouse strains, even when the tagged H3.4 proteins were heterozygously expressed (H3f4^TAG/wt) (Supplementary Fig. 8b). In contrast, two independently generated H3f4^MYC alleles showed overall normal progression through spermatogenesis when heterozygously expressed, even though testis weights were ~20% reduced (Supplementary Fig. 8c). Homozygous H3f4^MYC/MYC expression impaired spermatogenesis during meiotic prophase with a phenotype less severe than observed for the other three tagged transgenic strains showing impaired spermatogonial development (Supplementary Fig. 8b).

The reduction in testis weight in H3f4^MYC/wt heterozygotes could be due to elevated cell death of spermatogonia and spermatocytes, compared to control animals (Supplementary Figs. 9a, b and 10c). In H3f4^MYC/wt FACS-sorted ScPD spermatocytes, we detected premature upregulation of genes associated with spermatid development while developmental genes were upregulated in elongating spermatids (Supplementary Fig. 9c–f). A substantially higher percentage of X-linked genes was upregulated compared to autosomal genes in ScPD and elSts possibly pointing to a partial derailment of MSCI (Supplementary Fig. 9g). This is surprising given that H3.4^MYC appears by immunofluorescence to be evicted and replaced by H3.3 along sex chromosomes during MSCI as reported before^42,46 (Supplementary Fig. 11). Western blot experiments of FACS-isolated spermatocytes from H3f4^MYC/wt mice showed that H3.4^MYC was expressed at a level of ~20% of non-tagged H3.1/H3.2/H3.4 proteins (Supplementary Fig. 8d, e). Despite these disturbances of spermatogenesis, we next used the transgenic H3f4^MYC/wt mice to investigate the global dynamics of H3.4 deposition, removal, and its genomic distribution.

H3.4 deposition into chromatin starts in spermatogonial stem cells and accumulates until entry into meiosis

To determine the onset of H3.4 deposition into chromatin, we recorded the presence of H3.4^MYC in populations of GRFα1 and PLZF-positive SgU cells and c-KIT-positive SgDs⁶³ by immunofluorescence microscopy on whole mount testicular tubules. H3.4^MYC is already detectable in nuclei of about 70% of GRFα1⁺ type A_single (A_s) and A_paired (A_pr) SgU cells and in all clones of type A_aligned (A_al) SgU cells containing 4 to 8 germ cells (Supplementary Fig. 12). The intensity of H3.4^MYC signals increased parallel to the clone size of GRFα1⁺ and PLZF⁺ SgU and reached a plateau in c-KIT⁺ SgD cells (Supplementary Fig. 12a, b and Fig. 2d). In contrast, H3.4^MYC was not detectable in somatic Sertoli cells marked by H3K27me3 (Fig. 2d and Supplementary Fig. 12a). In summary, these data suggest a progressive incorporation of H3.4 over multiple rounds of replication into spermatogonia starting already in undifferentiated GFRα1-positive A_s spermatogonia^64,65.

H3.4 is evicted from chromatin in step 11 and 12 elongating spermatids

From c-KIT⁺ SgD onwards, nuclear H3.4 levels remained high in spermatocytes and spermatids until their nuclear elongation. To identify the timing of H3.4 eviction from chromatin, we co-stained paraffin-embedded H3f4^MYC/wt testis sections for MYC and TNP2, a marker of elongating spermatids¹. We manually staged seminiferous tubules⁶⁶ and quantified intensities of anti-MYC and anti-TNP2 signals in step 8–13 spermatids. While TNP2 incorporation started in stage IX and gradually increased, H3.4^MYC signals remained high until stage XI, where H3.4 eviction began (Fig. 2e, f). At stage XII, H3.4^MYC signal was lost in step 12 elongating spermatids to a level measured in somatic Sertoli cells (Fig. 2e, f). We confirmed the timing of H3.4 eviction using an antibody recognizing all replication dependent H3.1, H3.2, and H3.4 proteins⁶⁷ (Supplementary Fig. 15d). Together, our data indicate that most RC-H3.4 proteins are removed at the end of nuclear elongation, in stage XII spermatids.

H3.4 is localized throughout the genome and gets replaced by H3.3 at specific genome regions in post-mitotic germ cells

Next, we investigated the genomic distribution of H3.4^MYC in spermatogonia, meiotic spermatocytes, and haploid spermatids that we had FACS-purified from MYC-expressing H3f4^MYC/wt animals, by using ultra-low input^68,69 and regular ChIP-seq approaches (Supplementary Figs. 10 and 13 and “Materials and Methods”). We first studied the occupancy of H3.4^MYC within regions ±25 kb around transcriptional start sites (TSS) of genes that we had classified into nine k-mean clusters according to having a promoter with or without a CpG island, occupancies of H3.4^MYC, H3.3, H3K4me3, and H3K27me3. Associated mRNA expression was included following the cluster definition (Fig. 3a–c and Supplementary Fig. 14a–c). Consistent with global protein and imaging data (Fig. 2), we readily observed H3.4 occupancy in SgU, which increased at TSS in SgD, both at CGI and non-CGI type promoter genes. Along the developmental trajectory from SgD to rSts, H3.4 became gradually evicted and replaced by H3.3 around TSS of genes in a sequence, histone modification, and transcription related manner (Fig. 3a–c and Supplementary Fig. 14a–c).

Fig. 3. Genome-wide analysis of H3.4^MYC and H3.3 occupancy during spermatogenesis.

a Heatmap displaying chromatin states at TSS regions (± 25 kb; log₂CPM) and gene expression levels (log₂RPKM) of CGI- and nonCGI-containing genes measured in FACS-purified SgU, SgD, ScLZ, and ScPD cells. Regions were grouped into 9 gene clusters by k-means clustering on chromatin variables performed separately for CGI- and non-CGI-type promoters. The number of gene promoters is displayed for each cluster. CG density (± 1 kb around TSS) is indicated in percentage. The ChIP-seq data of H3.4^MYC in all four cell types originate from H3f4^MYC/wt mice. The ChIP-seq data related to H3K4me3, H3K27me3 in SgD cells, H3K27me3 in ScLZ and ScPD, as well as RNA expression in SgU, SgD, and ScLZ cells originate from H3f4^wt/wt mice and is from Bocker et al.⁵⁷. The RNA-seq data for ScPD from H3f4^wt/wt mice is from Gill et al.⁵⁸. b Metaplots showing quantification of ChIP-seq data at TSS regions (± 25 kb; log₂CPM) in clusters 1–9, as displayed in (a). c Boxplot showing quantification of expression of genes (in log₂RPKM) belonging to clusters 1–9 in (a). Center lines in box plots represent median values while lower and upper lines indicate interquartile range (IQR; from the 25th to 75th percentile). The numbers of genes (n) per each cluster are indicated in (a). d Density scatter plots displaying log₂CPM counts of H3.4^MYC (x-axis) and H3.3 (y-axis) on genomic tiles that contain 0, 1, 2, or 3+ transcription units with TSS in SgU, SgD, ScLZ, and ScPD cells. Red contour lines represent the density of tiles located on the X chromosome. The presence of a transcription unit with TSS was defined by an overlap of a given tile with a region ±1 kb around TSS.

We measured the highest level of H3.4^MYC to H3.3 replacement at CGI- and non-CGI-type promoters of actively expressed genes marked by H3K4me3 (Fig. 3a–c and Supplementary Fig. 14a–c, clusters 1, 2, and 5). Moderate levels of replacement were observed at promoters of H3K27me3-marked genes, and at one class of non-CGI promoter genes that are barely expressed (Fig. 3a–c and Supplementary Fig. 14a–c, clusters 3, 6, and 8). Lastly, cluster 7 of non-CGI promoter genes that are barely expressed displayed hardly any H3.4^MYC to H3.3 conversion across any cell types (Fig. 3a–c and Supplementary Fig. 14a–c).

Next, we investigated genome-wide localization of H3.4^MYC and H3.3. Hence, we calculated ChIP-seq signals of H3.4^MYC and H3.3 in 10 kb non-overlapping genomic tiles that we had classified according to the number of TSS-containing transcription units present and their chromosomal location (Fig. 3d and Supplementary Fig. 14d). On autosomes, genic and intergenic tiles showed comparable ChIP-seq signals for H3.4^MYC and H3.3 proteins in SgU and SgD spermatogonia. Upon entry into meiotic prophase and towards haploid round spermatids, however, H3.3 signals progressively increased at the expense of H3.4^MYC at many tiles, and such replacement was more pronounced for tiles containing one or more TSS-containing transcription units. The absence of replication-coupled H3.4^MYC deposition in post-mitotic cells, even within intergenic regions, is reflected by the progressive replacement by H3.3.

Contrary to autosomal genes, genes localized on the X chromosome displayed remarkably different temporal changes in chromatin composition at their promoters, irrespective of their sequence, histone post-translational modifications (PTMs), and expression states (Fig. 3a–c and Supplementary Fig. 14a–c, clusters 4 and 9). In SgU cells, X-linked gene promoter regions were packaged by H3.4^MYC and H3.3, alike autosomal genes. Then, while H3.4^MYC remained present in SgD and early meiotic ScLZ cells, it majorly decreased in ScPD cells, presumably as part of MSCI. In contrast, H3.3 levels already increased in highly pure populations of ScLZ (Supplementary Fig. 10d, e), and further to highest levels in ScPD and rSts (Supplementary Fig. 14a). Hence, X-linked genes initiate the exchange of H3.4^MYC to H3.3-holding nucleosomes at gene promoters already early in meiosis, several days prior to the chromosome-wide exchange observed in mid-pachytene spermatocytes (Supplementary Fig. 11)^46,62.

As seen at TSS regions, we also observed deposition of H3.3 at the expense of H3.4^MYC in all four classes of X-linked genomic tiles already in ScLZ cells (Fig. 3d), prior to the pachytene stage at which X-chromosome wide MSCI-associated eviction of replication-dependent nucleosomes had been reported to occur^46,62. These data underscore the sensitivity of ChIP-seq over immunofluorescence in detecting X-linked nucleosome turnover, as confirmed by the exclusion of H3.4^MYC signal along the X and Y chromosomes only in mid-pachytene but not in early-pachytene spermatocytes as seen by imaging (Supplementary Fig. 11).

H3.4 supports spermatogenesis more proficiently than H3.1

In mouse, H3f4 carries four non-synonymous substitutions relative to H3.2 genes. To understand the impact of these amino acid alterations on reproductive fitness, we performed CRISPR/Cas9 gene editing and reverted three of them to those present in H3.1 (H3f4^V24A; H3f4^H42R, or H3f4^S98A), which is initially abundantly expressed in spermatogonia. We left H3f4^C96 intact since this residue is already present in H3.1 (Fig. 4a and Supplementary Fig. 15a, b). TAU gel electrophoresis confirmed for H3f4^H3.1/H3.1 animals the absence of H3.4 and continued presence of H3.1 in male germ cells throughout their development from SgU to eSt (Fig. 4b). Importantly, by expressing H3.1 from the H3f4 locus, we did not alter global levels of histone H3, as measured by Western blot analysis (Fig. 4c and Supplementary Fig. 15c).

Fig. 4. Reduced efficiency of spermatogenesis in H3f4^H3.1/- substitution mice.

a Experimental design of CRISPR/Cas9-mediated editing of H3f4 resulting in the replacement of H3.4-specific amino acids with those present in H3.1. Sequences of sgRNAs and repair template are provided in Supplementary Data 6 and 7, respectively. b Western blot of TAU gel showing the presence of H3.1, H3.2, and H3.3 proteins and absence of H3.4 in FACS-purified germ cells from H3f4^H3.1/H3.1 mice. Mouse embryonic stem cells (mESCs) were used as control. A panH3 antibody recognizing the C-terminus of H3 was used to detect all H3 variants. Data of one representative out of two biological replicate experiments is shown. c Western blot of SDS-PAGE gel showing titration of H3 proteins present in 100, 50, and 25 K FACS-isolated rSts from indicated genotypes. Immuno-based detection was done as described for (b). Alpha tubulin was used as a loading control. Quantification of H3 to alpha tubulin ratios is shown in Supplementary Fig. 15c. One biological replicate per genotype was used for this experiment. d Violin plot showing testis weights of H3f4^wt/wt, H3f4^wt/− and H3f4^−/− mice. Each dot displays the mean weight of both testes per mouse. Bars represent median values of genotypes. The number of individual measurements (n) is indicated for each genotype. Statistical analysis was performed using one-way ANOVA followed by a two-sided Games–Howell post hoc multiple-comparison test to assess pairwise differences between genotypes. Significance levels are based on adjusted p-values from the post hoc test. NS not significant; * p < 0.05; ** p ≤ 0.01; *** p ≤ 0.001. e Violin plots showing testis weights (mg) of H3f4^w/wt, H3f4^H3.1/wt, H3f4^H3.1/H3.1, H3f4^wt/−, and H3f4^H3.1/− mice. Each dot represents the mean weight of two testes per mouse. Bars represent median values among mice per genotype. Number of individual males tested (n) per genotype is indicated. Statistical testing was performed as described for (d). f Sperm counts (cells/ml) of H3f4^w/wt, H3f4^wt/− and H3f4^H3.1/− males. Each dot represents sperm counts per mouse. Bars represent median values among mice per genotype. The number of individual males tested (n) is indicated for each genotype. Statistical testing was performed as described for (d). g Comparison of breeding efficiency of H3f4^w/wt, H3f4^wt/− and H3f4^H3.1/− males. Each dot represents the number of pups born per litter. Bars represent median values among mice per genotype. The number of males (n) used in breeding tests are indicated for each genotype. Statistical testing was performed as described for (d). h PAS-haematoxylin-stained testicular cross-sections obtained from H3f4^w/wt, H3f4^H3.1/H3.1, H3f4^wt/− and H3f4^H3.1/− mice. Scale bars: 100 µm for cross-section overview and 50 µm for stage V–VI zoom-in. i Quantification of indicated germ cell types at stage V and stage X tubules in H3f4^w/wt, H3f4^H3.1/H3.1, H3f4^wt/−, and H3f4^H3.1/− mice. Each dot indicates the number of cells counted per tubule per genotype. For each genotype, counts of 3 biological replicates (n = 3) are shown. For each cell type, pairwise comparisons between genotypes were performed using two-sided t-test with Bonferroni correction for multiple comparisons. * P < 0.05, ** P ≤ 0.01, *** P ≤ 0.001, NS not significant.

Interestingly, expressing H3.1 instead of H3.4 did not affect spermatogenesis with testicular weights being unchanged between the H3f4^H3.1/H3.1 and H3f4^wt/wt males (Fig. 4e). To further challenge the function of H3f4^H3.1, we reduced its gene dosage during spermatogenesis. To this end, we generated a H3f4 deficient allele (H3f4⁻, Supplementary Fig. 16) and combined it with the H3f4^H3.1 allele (H3f4^H3.1/−). By large, the overall proteins levels of H3.4 and H3.1 in FACS-purified rSts were comparable when being expressed from either one or two alleles (Fig. 4c). Nonetheless, in line with a previous report⁴², we observed gene dosage dependent effects of H3f4 expression on male germ cell development. Compared to control wildtype males, heterozygous H3f4^wt/− males showed a small reduction in testis weight of 10% (P < 0.001) and a ~ 25% reduction in sperm counts (P < 0.01), but did not lose fertility (Fig. 4f, g). In contrast, homozygous deletion of H3f4 (H3f4^−/−) caused a severe reduction in testis weight by 80%, extensive degeneration of germ cells and absence of epididymal sperm (Fig. 4d and Supplementary Fig. 16c). When combining the H3f4^H3.1 and H3f4^- alleles, the resulting H3f4^H3.1/− males displayed a further decrease in testicular weight by over 30% (P < 0.001) and an additional ~25% reduction in sperm counts (P < 0.05), compared to control heterozygotes (H3f4^wt/−). Nonetheless, H3f4^H3.1/− males were still able to reproduce (Fig. 4g, h and Supplementary Fig. 15e). Together, these data indicate that H3.4 is more proficient than H3.1 in supporting male germ cell development.

Substitution of H3.4 to H3.1 leads to the reduction of pachytene spermatocytes

Histologically, we next reviewed testicular sections and observed notably reduced numbers of spermatocytes and spermatids, only in H3f4^H3.1/− males (Fig. 4h). To determine quantitatively at what stage germ cell development was affected in H3f4^H3.1/− relative to control males, we developed a partially automated image analysis workflow to classify and count germ cells (Supplementary Fig. 17 and “Methods”). We quantified spermatogonia, spermatocytes, and spermatids present at stages V and X of the seminiferous epithelium cycle, in triplicate per genotype (Fig. 4i). While the number of SgU, SgD, and early meiotic ScZ cells were comparable between all genotypes, the number of spermatocytes at mid-pachytene (ScP, at stage V) was ~20% reduced only in H3f4^H3.1/− males compared to H3f4^wt/wt controls (P < 0.05), and H3f4^H3.1/H3.1 and H3f4^wt/− genotypes. The number of diplotene spermatocytes (ScD, at stage X) and round, elongating, and elongated spermatids were in H3f4^H3.1/− males also reduced, indicating an important role for H3.4 in meiosis and possibly spermatid differentiation. In H3f4^wt/− males, the number of elongated spermatids was ~35% reduced (P < 0.01). While the reason is unclear, it likely explains the reductions in testicular weights and sperm counts (Fig. 4e, f). Consistently with a reduced number of meiotic and post-meiotic germ cells, the sizes of seminiferous tubules in H3f4^H3.1/− testes were significantly reduced while the frequencies of stages of the seminiferous epithelium cycle were unaltered between genotypes (Supplementary Fig. 15f, g).

Meiotic defects in pachytene spermatocytes expressing H3.1 instead of H3.4

To investigate possible causes underlying the reduction of pachytene and diplotene H3f4^H3.1/− spermatocytes, we first performed immunofluorescence experiments on paraffin-embedded testicular sections for cleaved PARP1 (clPARP1), a marker of cells undergoing apoptosis⁷⁰ (Fig. 5a, b and Supplementary Fig. 18). Indeed, we observed elevated levels of clPARP1 signal in H3f4^H3.1/− late zygotene/early pachytene spermatocytes in stage XII—I tubules suggesting defects in chromosome synapsis and/or processing of programmed meiotic double strand DNA breaks (DSB). Remarkedly, even mid/late-pachytene and diplotene spermatocytes at stages IV–IX were labeled for clPARP1.

Fig. 5. Impairment of meiotic progression in spermatocytes of H3f4^H3.1/− substitution mice.

a Cross-section immunofluorescent images showing cleaved PARP1 signal (green) in seminiferous tubules of H3f4^wt/− and H3f4^H3.1/− mice. γH2A.X (red) and acrosome (cyan) were used to mark spermatocytes and spermatids, respectively. Scale bar: 50 µm. b Quantification of testicular germ cells positive for cleaved PARP1 per stage comparing H3f4^wt/− and H3f4^H3.1/− mice. Staging was performed manually as described before⁶⁶. Each dot represents the number of clPARP1-positive cells per tubule. Data from three biological replicates are shown together. For each stage, two-sided t-test was performed between the indicated genotypes. * P < 0.05, ** P ≤ 0.01, *** P ≤ 0.001, NS not significant. c Types of nuclear distribution of γH2A.X (red) in spermatocytes of H3f4^wt/− control animals, labeled by SYCP3 staining (green). Scale bar: 10 μm. d Bar plot showing the distribution of γH2A.X nuclear patterns (as displayed in c) in spermatocytes during their developmental course of meiotic prophase, comparing H3f4^wt/− and H3f4^H3.1/− genotypes. Number of cells per stage is indicated. Data from two biological replicates are shown together.

To evaluate meiotic progression in more detail, we studied the processes of meiotic DSB formation and repair, and of synapsis between homologous parental chromosomes by staining dried-down nuclear spreads of H3f4^H3.1/− and control spermatocytes for phosphorylated H2A.X (γH2A.X) and SYCP3, a structural component of synaptonemal complexes^71–75. We assessed the level of H1t incorporation in meiotic chromatin to substage progression through pachytene⁷⁶. We observed comparable patterns of γH2A.X and SCYP3 in leptotene, zygotene and in mid-pachytene to diakinesis stage spermatocytes of H3f4^wt/− and H3f4^H3.1/− mice. However, early pachytene-like spermatocytes, characterized by (almost) complete autosomal chromosome synapsis and absence of H1t labeling, were more frequently labeled by γH2AX in H3f4^H3.1/− animals than in H3f4^wt/− animals (Fig. 5c, d and Supplementary Fig. 19). The data suggest that not all H3f4^H3.1/− spermatocytes manage to process meiotic DSBs on autosomes appropriately prior their entry into pachytene stage, consistent with the increased cell death of such early H3f4^H3.1/− spermatocytes (Fig. 5d and Supplementary Fig. 18). In summary, our data argues that chromatin containing H3.4-nucleosomes facilitates progression through early meiotic prophase more effectively than H3.1-comprising chromatin.

Transcriptional upregulation of Polycomb targets upon H3.4 to H3.1 substitution in meiosis

Next, we investigated the possible impact of H3.1 substituted expression on transcriptional regulation in the surviving cells. We therefore performed RNA-seq on ScLZ, ScPD, rSt, and elSt populations FACS-isolated from H3f4^wt/wt, H3f4^H3.1/H3.1, H3f4^wt/−, H3f4^H3.1/− males. PCA analysis revealed clear separation of cell types, recapitulating the developmental trajectory (Supplementary Fig. 20a). In addition, H3f4^H3.1/− elongating spermatids were in the first dimension well separated from H3f4^wt/wt, H3f4^H3.1/H3.1, and H3f4^wt/− elSts suggesting prominent changes in their gene expression upon H3.4 replacement (Supplementary Fig. 20a). When analyzing possible transcriptional changes in H3f4^H3.1/− ScLZ, ScPD, and rSt cells preceding spermatid elongation, we observed up to 159 genes upregulated in these cells, while only a few genes were downregulated (Fig. 6a and Supplementary Data 5). Interestingly, the three cell types shared several genes commonly upregulated (Fig. 6b), including transcription factors such as Six2, Foxg1, Nkx6-1, and a subset of Hox genes, as well as genes from the Vomeronasal receptor gene cluster (Supplementary Fig. 21a). In contrast, in H3f4^wt/− germ cells, barely any gene was misregulated in any cell type analyzed (Supplementary Fig. 20b). These data argue that transcription during meiosis and spermatid differentiation is substantially challenged by the combination of H3f4^H3.4 to H3f4^H3.1 substitution and reduced H3f4 gene dosage, rather than reduced H3f4 gene dosage per se.

Fig. 6. Differential gene expression in spermatocytes and spermatids of H3f4^H3.1/− substitution mice.

a, Scatter plots showing differential gene expression (log₂FC) in H3f4^H3.1/− versus H3f4^wt/wt FACS-isolated ScLZ, ScPD, rSt, and elSt cells plotted against gene expression levels in control H3f4^wt/wt cells. Numbers of significantly upregulated (red), downregulated (blue), and non-significant (grey) genes are displayed (|log₂FC| > 1, FDR ≤ 0.05). b Euler diagrams showing overlaps in DEGs between ScLZ, ScPD, and rSt cells. c Density plots showing the distribution of H3K4me3 and H3K27me3 enrichments at TSS (± 1 kb) for promoters of upregulated genes (colored fill) and for promoters of all genes (grey fill), as measured in indicated wildtype cell types. d Dot plot showing percentages of up and downregulated genes for each chromosome in ScLZ, ScPD, rSt, and elSt cells. Each dot has been derived from the edgeR estimation of DEGs (a) based on 3 biological replicates (n = 3) for H3f4^wt/wt and four biological replicates (n = 4) for H3f4^H3.1/− animals. The bars represent the mean values for the selected DEGs. e Scatter plot showing differential expression (log₂FC) of CGI-type promoter genes in H3f4^wt/wt elSts versus H3f4^wt/wt rSts (FDR ≤ 0.05) (x axis) compared to differential expression (log₂FC) in H3f4^H3.1/− versus H3f4^wt/wt elSts (FDR ≤ 0.05) (y axis). Genes up and downregulated in H3f4^H3.1/− elSts are highlighted in red and blue, respectively. DEGs are separated into nine groups, based on the |log₂FC| = 1 cutoff in both contrasts (dashed lines). The following numbers of genes (n) per clusters 1–9 were defined: 1397, 1133, 144, 1469, 3715, 2457, 16, 201, 876. f, g Boxplots showing absolute quantification of exonic and intronic gene expression in groups 1–9 defined in (e), comparing H3f4^H3.1/− to H3f4^wt/wt cells. Two-sided t-tests were performed between indicated genotypes: * P < 0.05, ** P ≤ 0.01, *** P ≤ 0.001. Center lines in box plots represent median values while lower and upper lines indicate the interquartile range (IQR; from the 25th to 75th percentile). h GO term search (Biological Process) for genes in groups 1, 2, 3, and 9 defined in (e). A one-sided overrepresentation hypergeometric test was used, followed by Benjamini–Hochberg correction for multiple testing. i Boxplots displaying ChIP-seq enrichments of different chromatin signatures (H3.4^MYC, H3.3, H3K4me3, and H3K27me3 at TSS ± 1 kb) in wildtype rSts for groups 1–9 defined in (e). Center lines in box plots represent median values while lower and upper lines indicate the interquartile range (IQR; from the 25th to 75th percentile).

To understand such shared mode of mis-regulation, we investigated chromatin signatures at promoters of the upregulated genes (± 1 kb around TSS) in wildtype cells using publicly available ChIP-seq data for H3K4me3 and H3K27me3^57,61. In all three cell types, we observed an overrepresentation of upregulated genes among genes marked by H3K27me3, with a majority being bivalently marked by H3K4me3 as well (Fig. 6c and Supplementary Fig. 21b). Hence, these data suggest that H3.4-containing nucleosomes promote PRC2-mediated repression in male germ cells.

Intriguingly, in H3f4^H3.1/H3.1 males, among the four sorted germ cell populations, 338 genes were upregulated in ScPD spermatocytes, of which only 18 were upregulated in H3f4^H3.1/− ScPD spermatocytes as well (supplementary Fig. 20b, c). Sperm specific functions in fertilization, acrosome reaction, and sperm-egg recognition are overrepresented among the 338 upregulated genes (Supplementary Fig. 22a). Most of them have non-CpG island promoters, are localized on autosomes and the X chromosome and become normally expressed in round spermatids, some even gaining H3K4me3 at their promoters (Supplementary Fig. 22b–e). Our data indicate that during meiotic prophase, as revealed under challenging H3f4 gene dosage and presumably overall H3 histone levels, H3.4-containing chromatin prevents precautious activation of post-meiotically expressed cells driving fertilization-related processes (as seen in H3f4^H3.1/H3.1 ScPD) and ensures repression of PRC2 target genes (as seen in H3f4^H3.1/− germ cells).

The H3f4^H3.1/− elongating spermatids transcriptionally resemble the preceding cellular stages with downregulated genes required for sperm maturation

Next, we focused our attention on H3f4^H3.1/− elSts, in which we observed dramatic up- and downregulation of 3867 and 2431 genes, respectively, compared to H3f4^wt/wt control cells (Fig. 6a and Supplementary Data 5). Interestingly, barely any X-linked genes were aberrantly upregulated (Fig. 6d), indicating that MSCI and PMSC-mediated gene repression mechanisms were not impaired. Given the prominent role of H3.3-nucleosomes in reprogramming chromatin of sex chromosomes during meiosis (Fig. 3 and Supplementary Figs. 11 and 14)^46,61,62, the observed mis-regulation at autosomes only may point to a direct effect of H3.4 substitution by H3.1 on transcription during spermatid elongation.

To better understand the phenomenon, we related the aberrant transcriptional changes observed in H3f4^H3.1/− versus control elSt cells to changes in gene expression that normally occur during the development of rSts into elSt in control mice. Given the differences in H3.4 and H3.3 occupancies at promoters of CGI-type and nonCGI-type genes (Fig. 3 and Supplementary Fig. 14), we analyzed these two gene classes separately (Fig. 6e and Supplementary Fig. 21c). For both classes, we observed a negative correlation between developmentally programmed expression changes versus changes aberrantly induced in H3f4^H3.1/− elSts, suggesting a defect in spermatid differentiation. Specifically, genes in group 1 that are normally downregulated upon differentiation of rSt into elSt were upregulated in H3f4^H3.1/− elSts (Fig. 6e, f and Supplementary Fig. 21c, d). Analysis of intronic reads showed that such genes failed to downregulate their transcription (Fig. 6g and Supplementary Fig. 21e). A similar response was measured for groups 2 & 3 genes that are normally highly expressed in spermatocytes and become downregulated in spermatids. These genes were even aberrantly upregulated in H3f4^H3.1/− elSts (Fig. 6e–g and Supplementary Fig. 21c–e). Gene ontology analysis indicated their involvement in housekeeping functions and meiosis (Fig. 6h), and in spermatid functions (Supplementary Fig. 21f). For genes belonging to groups 8 and 9 that normally sustain or even further increase their expression during nuclear elongation, we observed a downregulation which in turn likely abrogates at least in part spermatid differentiation, given their roles in spermatid differentiation and sperm function (Figs. 6e–h and Supplementary Fig. 21c–f). Together, these data demonstrate a dual impact of H3.1 expression at the expense of H3.4, preventing wide-spread changes in transcriptional regulation and programs in elSts.

Finally, we analyzed chromatin signatures of the promoters of differentially expressed genes (DEGs) (± 1 kb around TSS) as measured in wildtype rSt cells, using ChIP-seq data of H3.4^MYC and H3.3 (this study) and public H3K4me3 and H3K27me3⁶¹. In CGI-type promoter genes, transcriptional upregulation in H3f4^H3.1/− mutants mostly affected genes normally expressed being enriched in H3K4me3, the H3.3 variant and being devoid of repressive H3K27me3 (Fig. 6i). Concurrently, downregulated genes are normally marked and likely repressed by H3K27me3 in rSt (Fig. 6i). For non-CGI type promoter genes, the upregulated genes in groups 1–3 have normally higher H3.3 occupancy than unchanged or downregulated genes, suggestive for the occurrence of higher turn-over of nucleosomes at these genes (Supplementary Fig. 21g). Furthermore, H3K27me3 levels are indifferent among the groups of non-CGI genes (Supplementary Fig 21h) and low compared to CGI-type promoter groups with evident H3K27me3 enrichments (Fig. 6i). Finally, differential H3K4me3 occupancies largely reflect expression level differences among the different groups of non-CGI genes in rSts, with genes up- and downregulated in H3f4^H3.1/− elongating spermatids being more highly expressed and H3K4me3-marked in rSts compared to genes with unchanged expression (Supplementary Fig. 21g).

Next to a possible direct impact of altered nucleosome composition, changes in expression of transcription factors (TFs) may also have contributed to the massive transcriptional derailment observed in H3f4^H3.1/− elSts. To explore this notion, we first searched for the presence of transcription factor motifs in promoters of aberrantly regulated genes and secondly reviewed transcript levels of corresponding TFs (Supplementary Fig. 21h). Among downregulated genes, we measured significant enrichment of motifs for the immediate early gene response activating TFs FOS and JUN and observed reduced Jund and Fosb transcript levels specifically in H3f4^H3.1/− elSts. These two factors are critical for controlling spermatogenesis^77,78. Likewise, motifs for HMG box TFs, including SOX and HMG family members, were significantly enriched. While transcript levels of Hmgb2 and Sox17 were aberrantly increased, those of Sox18 and Sox30 were downregulated, particularly in H3f4^H3.1/− elSts. HMGB2 is normally highly expressed in spermatocytes and is essential for spermatogenesis⁷⁹. Sox30 is required for spermatid differentiation⁸⁰. An altered balance of such activating and repressive TFs may ultimately suppress spermiogenic genes during spermatid elongation⁸¹. Among upregulated genes, the motif of the transcriptional repressor NR6A1 (GCNF) was enriched and expression of its transcripts reduced, possibly explaining sustained expression of target genes normally downregulated during spermatid elongation^81,82. In contrast, expression levels of the Creb3l4 and Creb5, for which motifs were enriched in upregulated target genes, were downregulated, possibly affecting spermiogenesis as well⁸³.

In summary, upon the H3f4^H3.1/− substitution, major transcriptional changes occur at the stage of spermatid elongation. The H3f4^H3.1/− elSt cells have a partial transcriptional profile of the preceding stage due to failed downregulation of genes that had been expressed during meiosis or in rSt, or even due to an aberrant reactivation of such genes. At the same time, we measured reduced activation of genes required for spermatid differentiation that normally become highly expressed in elSt cells.

Amino acids V24 and H42 are most important for the function of H3.4

Finally, we aimed to understand which of the H3.4-specific amino acids are most important to its function. Hence, we generated three additional H3f4 substitution mouse lines, in which only one of the H3.4-specific amino acids was converted to that of H3.1 (H3f4^V24A, H3f4^H42R, and H3f4^S98A, Fig. 7a and Supplementary Fig. 23a, b). As observed for the triple H3f4^H3.1 substitution allele (Fig. 4a), none of the single substitution alleles affected testis weights nor breeding efficiencies, if expressed homozygously (Fig. 7b–d). However, when combined with the H3f4 deficiency allele (H3f4⁻), expression of the H3f4^V24A and H3f4^H42R substitutions resulted in decreased testis weights and sperm counts, compared to heterozygous control animals, while the H3f4^S98A mutation did not impair spermatogenesis (Fig. 7b–d). Histological examination of testis sections confirmed the impairment of spermatogenesis, with H3f4^H42R/− mice having the most pronounced effect (Fig. 7e).

Fig. 7. H3.4^V24 and H3.4^H42 amino acids advance spermatogenesis.

a Experimental design of the creation of H3f4 alleles bearing H3.1-specific substitution mutations (H3f4^V24A, H3f4^H42R, and H3f4^S98A) using CRISPR/Cas9 gene editing. Sequences of sgRNAs and repair templates are indicated in Supplementary Data 6 and 7, respectively. b Violin plots showing testis weights (mg) of H3f4^wt/wt, H3f4^mut/wt, H3f4^mut/mut, H3f4^wt/− and H3f4^mut/− mice, where H3f4^mut represents the H3f4 allele bearing a H3.1-specific substitution mutation (H3f4^V24A, H3f4^H42R, or H3f4^S98A). Each dot displays the mean weight of both testes per mouse. Bars represent median values among mice per genotype. Number of individual measurements (n) per genotype is indicated. Statistical analysis was performed using one-way ANOVA followed by a two-sided Games–Howell post hoc multiple-comparison test to assess pairwise differences between genotypes. Significance levels are based on adjusted p-values from the post hoc test. NS not significant; * p < 0.05; ** p ≤ 0.01; *** p ≤ 0.001. c Estimation of sperm counts in H3f4^wt/wt, H3f4^wt/− and H3f4^mut/− mice. Each dot represents sperm counts per mouse. The bars represent the median value for each genotype. The number of individual measurements (n) is indicated for each genotype. Statistical testing was performed as described for (b). d Comparison of breeding efficiency in H3f4^wt/wt, H3f4^mut/wt and H3f4^mut/mut mice. Each dot represents the number of pups born per litter. The bars represent the median value for each genotype. The number of males (n) used in breeding tests are indicated for each genotype. Statistical testing was performed as described for (b). e PAS-haematoxylin staining of cross-section of testes in H3f4^wt/−, H3f4^V24A/−, H3f4^H42R/−, and H3f4^S98A/− mice. Scale bars: 100 µm (cross-section overview) and 50 µm (Stage V–VI tubules). For each genotype, data of one representative out of two biological replicate experiments is shown.

To investigate the impact of individual amino acid substitutions on gene expression, we performed RNA-seq analyses on FACS-isolated rSt and elSt populations from H3f4^V24A/− and H3f4^H42R/− males. While PCA revealed a clear separation of the rSt and elSt populations, it did not discriminate between genotypes (Supplementary Fig. 23c). Direct RNA-seq analyses showed a more pronounced effect of the H3f4^H42R substitution than the H3f4^V24A (Supplementary Fig. 23d), particularly in rSts. Comparison of DEGs between the H3f4^H3.1, H3f4^V24A, and H3f4^H42R datasets showed a small yet consistent overlap in rSts (Supplementary Fig. 23e), including homeobox genes Hoxa5, Hoxa6, Six2, and genes from the Vomeronasal receptor cluster. Hence, the individual H3f4^V24A and H3f4^H42R substitutions alter gene expression patterns slightly and affect testis weights and sperm counts mildly. Their combination, however, triggers significant responses during meiosis and spermatid elongation, ultimately impairing spermatogenesis. These experiments show that the A24V and R42H amino acids substitutions improved the function of H3.4 for male germ cell development.

Discussion

Through synteny analyses, we identified that among eutherians with analyzable genomes, all contain a small HIST3 gene cluster harboring the H3-4 locus and one H2A and two H2B genes flanked by OBSCN and RNF187 genes (Supplementary Fig. 1b). We observed only minor variations to this configuration, such as an inversion of HIST3 and RNF187 relative to OBSCN and a loss of one H2B gene in Cetacea. Overall, the HIST3 configuration is highly conserved, as seen from the genomes of elephant and armadillo, being the eutherian species evolutionarily most distantly related to mouse and human.

Our analyses further indicate that eutherian HIST3 and HIST1 cluster orthologs likely originate from a large ancestral proto-histone cluster, of which derivatives exist in marsupial mammals. The proto-histone cluster contains an H3.2 gene, multiple core histone genes, and a linker histone H1 gene (Supplementary Fig. 6). In the Tasmanian devil marsupial, this cluster also underwent fractionation, while in others like opossum and koala, it is intact. We were unable to identify an orthologous cluster in monotremes (platypus). Hence, the evolutionary origin of the HIST1/3 proto-histone cluster remains to be resolved.

In marsupials, we identified the TRIM11-linked H3.2 genes as putative orthologs of eutherian H3-4. TRIM11-linked H3.2 genes form a distinct clade, different from other marsupial H3.2 genes but also from eutherian H3-4 genes. Hence, it is the linkage to TRIM11 that points to a common ancestry of eutherian H3-4 and TRIM11-H3.2 genes in marsupials.

Since the emergence of eutherian HIST3 over 100 million years ago, H3-4 orthologous genes accumulated more synonymous and non-synonymous mutations than canonical RC H3.1 and H3.2 genes situated in large histone clusters. Nonetheless, given that their d_n/d_s ratios are similar to those of RI H3.3 genes individually located in the genome, H3-4 genes are still under strong evolutionary constraints, suggesting conserved functions among different species, presumably promoting spermatogenesis and/or embryonic development.

In mice, H3f4 is highly expressed in spermatogonia, driving H3.4 deposition already in undifferentiated spermatogonia, and causing an efficient and progressive accumulation of H3.4 throughout the entire genome during spermatogonial amplification until the entry into meiosis, at the expense of the other RC H3.1/H3.2 histones. As we demonstrated with the tagged H3.4^MYC allele and alike RC H3.1/H3.2 in somatic cells^84–86, H3.4 is successively replaced by H3.3 variant proteins at TSS and in genes following transcription-coupled nucleosome turnover in proliferating spermatogonia. Upon entry into meiosis, H3.4-bearing nucleosomes become further progressively exchanged by H3.3-nucleosomes within these regions as shown before⁶¹ as well as along the entire sex chromosomes, the later due to MSCI-related chromatin remodeling^46,62.

Through the H3.4 to H3.1 reconstitution experiment, we show that expression of a H3 protein from the H3f4 locus is critical to male fertility in mice, thereby refining the notion that the H3f4 locus is essential for spermatogenesis⁴². Moreover, by generating transgenic animals in which we substituted amino acids that are variant between H3.4 and H3.1, we obtained evidence for purifying selection of the H3f4 sequence, resulting in fixation of specific residues. Our data show that particularly the combination of H3.4^V24, H3.4^H42, and H3.4^S98 substitutions majorly promotes spermatogenesis, enhancing progression through meiotic prophase, gene regulation, and germ cell viability. Considering the positive yet limited impact of single amino acid substitutions on germ cell development, we conclude that the H3.4^V24 and H3.4^H42 substitutions synergize in some way in improving spermatogenesis. Surprisingly, the positive cellular impact of H3.4 over H3.1 sequence was noticeable only under reduced gene dosage conditions of RC H3 histones, even though we did not detect differences in total H3 levels by Western blot analyses between homozygous or hemizygous expression conditions of H3.4 or H3.1. Our data reveal that progression through meiotic prophase as well as nuclear elongation and condensation are particularly sensitive to H3.4-specific sequence acquisitions, while plain heterozygosity for H3.4 induced only limited cell death. The impact of H3.4^S98A is neutral. Together, our data show that the expression level and amino acid composition of H3.4 define its optimal function during spermatogenesis.

H3f4^H3.1/− males displayed a ~ 20% reduction of spermatocytes prior to stage V of the seminiferous epithelium, which coincides with more persistent labeling of meiotic chromatin by phosphorylated H2A.X on autosomes in pachytene-like spermatocytes. These data may point to a defect in repair of meiotically programmed DNA double-strand breaks (DSBs), followed by activation of the synapsis checkpoint and cell death^87–89. Nonetheless, we also observed an increased percentage of spermatocytes marked by cleaved PARP1 at following stages, suggesting persistent or newly formed DNA damage triggering cell death at later time points. Together, these data raise the intriguing possibility that the identity of RC H3 histones at meiotic recombination sites may regulate the formation, processing, or repair of DSBs. For example, H3.4 may modulate tri-methylation of H3K4 and H3K36 by PRDM9, thereby controlling efficacy of SPO11-mediated DSB formation^90–92. Alternatively, H3.4 and H3.1 may differentially regulate possible nucleosome turnover occurring at recombination sites during DSB repair. Future chromatin analysis in single cells is required to address such questions.

In H3f4^H3.1/− males, gene expression is only moderately affected during spermatogonial proliferation, meiosis, and round spermatid differentiation. We predominantly observed preferential upregulation of genes normally bivalently marked by PRC2-mediated H3K27me3 and H3K4me3. These data are consistent with residue H3.4^V24 being previously shown to promote recognition of tri-methylated H3.4K27 by PHF1 and PHF19, two accessory components of PRC2 complex, stabilizing its chromatin binding and promoting H3K27me3^44,51,93. Indeed, we observed a clear overlap among genes upregulated in round spermatids isolated from H3f4^V24A/− and H3f4^H3.1/− mice. Many of these genes are normally marked by H3K27me3, supporting an in vivo role for H3.4 in promoting PRC2-mediated gene repression.

Remarkedly, in H3f4^H3.1/H3.1 animals, we monitored a precocious upregulation of over 300 genes in ScPD cells that otherwise get upregulated in haploid spermatids. These germ cell-specific fertilization-associated genes are different from those upregulated in H3f4^H3.1/− germ cells. The reasons for such distinct H3f4 gene dosage related transcriptional responses are unknown. Contrasting H3f4^H3.1/− males, spermatogenesis in H3f4^H31./H3.1 males is, however, not impaired. Then again, expressing the tagged H3f4^MYC allele at 20% of total H3.4 protein level affects spermatogenesis. And peculiarly, 83% (280/339) of the genes upregulated in H3f4^H3.1/H3.1 ScPD cells are also upregulated in H3f4^MYC/+ ScPD cells (Supplementary Data 5). Our data demonstrate a genetic interaction between sequence and dosage of the H3f4 gene in regulating different sets of genes in a cell-type-specific manner. These remarkable findings warrant further mechanistic investigations into the dynamics and function of H3.4-nucleosomes considering the H3.4 protein sequence and level.

Finally, we observed thousands of mis-regulated genes during spermatid elongation. On the one hand, our data points to a failure in the timely downregulation or even an aberrant transcriptional  reactivation of genes normally expressed in round spermatids. On the other hand, H3f4^H3.1-expressing elongating spermatids show reduced activation of spermatid differentiation genes normally upregulated during spermatid elongation. The data indicate that particularly elongating spermatids undergoing nucleosome replacement by transition proteins and protamines are sensitive to possible transcriptional regulatory functions of H3.4. Next to direct effects of H3.4-bearing nucleosomes, changes in expression of regulatory TFs may also contribute to the transcriptional deregulation observed in H3f4^H3.1/− elSts.

All eutherian H3f4 orthologs encode for C96, which is absent in marsupial “H3.2 orthologs”. Hence, C96 represents an ancient substitution mutation acquired early by the ancestral H3-4 ortholog, like V24. Intriguingly, all RC H3.1 histones, specific to eutherians, encode also for C96, while H3.2 genes, expressed in all eukaryotes, including eutherians and marsupials, encode for S96⁹⁴. C96 has been proposed to regulate chromatin condensation by possibly forming intermolecular disulfide bonds between neighboring nucleosomes or between nucleosomes and other chromatin proteins⁹⁴. The significance of C96 in H3.1 and H3.4 proteins for chromatin biology and spermatogenesis remains to be explored.

In human and other eutherians, various substitutions other than V24, C96, and S98 became fixed in H3f4 orthologs. For human, chimpanzee, and bonobo, the M71 and V111 residues regulate stability of nucleosomes^48,49 and interactions with histone chaperones in vitro⁵⁰, analogous to residue H42 in mice⁴². The in vivo function of M71 and V111 in regulating spermatogenesis, however, remains unknown. This is particularly relevant given that spermatogenesis in humans is less efficient than in rodents. Likewise, it will be important to assess the in vivo relevance of the variant Q134 and F54 residues encoded by genomes of some primate and bovine species, respectively.

The acquired role of H3f4 orthologs in germ cells is not only determined by their conserved amino acid substitutions but also by their levels of expression relative to those of canonical H3.1 and H3.2 histones. Contrasting to mouse and human in which H3f4 / H3-4 is critical to spermatogenesis^42,47, H3-4 orthologs underwent loss-of-function mutations in some other species, which may can be linked to different reproductive strategies. For example, H3-4 is truncated in gorilla but is retained in most other primates with annotated genomes. The social reproductive system of gorilla populations is polygynous, which has been associated with reduced sperm competition and reduced testis to body ratios, compared to other primates⁹⁵ and with relaxed selection for hundreds of genes implicated in germ cell development⁹⁶. We also observed pseudogenization or even a complete loss of H3-4 in New World monkey species, although here the link to reproductive strategy is unclear. More examples exist in Carnivora, where H3-4 encodes for a functional protein only in a subset of species. In Felidae, we identified a truncation of 13 nucleotides within the coding sequence of H3-4, resulting in a frameshift mutation (Supplementary Fig. 3b). Similarly, in Canidae, another truncation of 25 nucleotides was found in the open reading frame of H3-4 of several dog breeds and Dingo (Supplementary Fig. 3b). Given that these variants were measured in several independent genome assemblies of evolutionarily close species, their presence is unlikely to be due to sequencing artifacts. The reason why cats and dogs have truncations in H3-4 is enigmatic. In Felidae, polygyny is common, and the mean testis to body weight in Felidae is among the lowest among all Carnivora⁹⁷. Canidae, in turn, are characterized by social monogamy, yet with a high testis to body ratio, which has been linked to occasional intra-male competition. In any of these species, the absence of H3-4 ortholog expression strongly advocates for a role of regular H3.1 and H3.2 proteins in germ cell development, underscoring the opportunistic impact of histone evolution in reproduction.

Several published studies^11,43 and our own experiments (Supplementary Figs. 8 and 9) reveal the detrimental impact of expression of tagged core histones for germ cell development. Whereas such approach is commonly used in biochemical and cell biological experiments, our in vivo experiments raise awareness that too high expression levels are detrimental for development, a notion to be further investigated at the molecular level.

Methods

Phylogenetic analysis of H3-4 genes in the mammalian lineage

To search for H3-4 orthologs in the mammalian lineage, we performed a TBLASTN search using the human H3.4 protein sequence as a query against 109 mammalian genomes available at Ensembl (release 112). In each species, we looked for BLAST hits that encoded valine 24 (V24) and are located in the HIST3 gene cluster³⁴. In addition, an A24-encoding H3 gene was found next to TRIM17/11 in Nannospalax galili. If a hit was not annotated as a coding sequence but otherwise seems to be correct, we manually translated the DNA sequences for this region. Sequences orthologous to H3-4 and linked to TRIM17/11 that possessed truncations and insertions were annotated as pseudogenes and also included into analysis. If the genomic assembly was unable to provide information about the neighboring TRIM and/or H2A and two H2B genes, the hit was discarded. The results of this analysis are available in Supplementary Data 1. Pairwise divergence time trees were made using the TimeTree Database (http://www.timetree.org)⁵⁴.

Analysis of H3-4 gene paralogues in the selected species

The Comparative Genomics functionality of the Ensembl website (release 112) was used to identify H3-4 paralogues in human (GRCh38.p14), lemur (Mmur_3.0), mouse (GRCm38.p5), rabbit (OryCun2.0), cow (ARS-UCD1.3), elephant (Loxafr3.0), camel (CamDro2), wallaby (Meug_1.0), Tasmanian devil (mSarHar1.11), opossum (ASM229v1), koala (phaCin_unsw_v4.1), and common wombat (bare-nosed_wombat_genome_assembly). In species with an H3-4 gene annotated in the genome (human, lemur, rabbit, cow, elephant, camel), H3-4 was used as a query for the search for paralogues. In species where H3-4 was not annotated, but coding the sequence was identified in the genome (rabbit), canonical H3.1 was used as a query for the search for paralogues. For marsupial species, the H3.2 sequence was used as a query. The results of this analysis are available in Supplementary Data 2.

The information for Ensembl gene ID, number of transcripts, transcript ID, number of exons, cross-reference to RefSeq, and location was obtained using the Ensembl Perl api (release 112). For genes with more than one transcript, the transcript labelled with the flag “Ensembl Canonical” was initially selected for further analysis (Supplementary Data 2). Coding sequences and translations of coding sequences of all genes not annotated as pseudogenes were fetched as FASTA files via the Ensembl Perl API. In the case of rabbit, the genomic location, cDNA, and protein sequences were added manually.

Alignment of DNA and protein sequences

To generate Fig. 1a and Supplementary Figs. 2 and 3b, alignments of DNA and protein sequences were performed with the msa R package⁹⁸ with the default parameters (method = “ClustalW”, order = “input”). For protein sequences, the first amino acid (methionine) was excluded from the analysis. All other alignments were performed using ClustalW algorithm⁹⁹ using the tool available online (https://www.genome.jp/tools-bin/clustalw), or using MEGA12 software¹⁰⁰.

Maximum likelihood analysis of H3 genes

For maximum likelihood analysis (Supplementary Fig. 3a), we retrieved coding sequences of protein-coding orthologs of H3-4, DNA sequences of H3-4 pseudogenes (regions of homology to functional H3-4) from all eutherian species listed in Supplementary Data 1, as well as coding sequences of H3.1, H3.2, and H3.3 genes from the selected species (Supplementary Data 2). We aligned the sequences and downloaded the output in the Phylip format. Next, we uploaded the alignment file to the PhyML tool^52,53 available online (http://www.atgc-montpellier.fr/phyml/) to build a tree using the HKY85 substitution model with 100 bootstrap replicates⁵³. To edit the tree, we used the Interactive Tree of Life tool¹⁰¹ available online (https://itol.embl.de/). For rooting, we used coding sequences of H3.2 genes from Xenopus tropicalis, as an outgroup, being an evolutionarily distant species.

Analysis of variation rates of histone protein sequences

For each H3 variant (H3.1, H3.2, H3.3 and H3.4), we obtained their protein sequences from the genomes of the selected species (Supplementary Data 2). For each sequence, divergence rate was defined as percentage identity relative to human reference. When multiple protein sequences are available (due to several paralogs in the same genome), their mean value was taken. Species were ordered according to the time of their divergence from human determined by TimeTree⁵⁴. For external references, divergence rates of several H2A and H2B variants were added from previously published studies^7,8. The results of this analysis are available in Supplementary Data 3 and Supplementary Fig. 3c.

Determining the rates of synonymous and nonsynonymous mutations of H3 genes

Here, we used the same sequences as in the maximum likelihood analysis. First, using the MEGA12 software¹⁰⁰, we performed codon alignments for each gene group using ClustalW⁹⁹ algorithm and saved the results in Phylip format. In parallel, a neighbor-joining tree file was generated. The resulting alignment and tree files were analyzed using the CODEML tool from the PAML package^55,56 (4.10.6 release). We tested the M0 model (model = 0, NSsites = 0) that assumes that all sequences and sites evolve under the same evolutionary pressure. To test for purifying selection, we performed the likelihood ratio test (LRT) comparing the likelihoods between model M0 with fixed d_n/d_s at 1 (fix_omega = 1) and M0 with the estimated d_n/d_s ratio. We calculated twice the difference of log-likelihood between the null and alternative hypotheses (2*(L₀–L₁)) and compared it with the χ² distribution with the degree of freedoms determined as the difference between the number of parameters (np in the output file). For external references, we added the d_n/d_s rates of several H2A and H2B variants from previously published studies (same as in the analysis of variation rates)^7,8. Model M8 (NSites = 0) was used to search for possible sites with d_n/d_s rates >1. The results of this analysis are available in Supplementary Data 4 and Supplementary Fig. 3d.

Genome target selection and sgRNA design

The sequences of sgRNAs used in this study are available in Supplementary Data 6. The CRISPR design web tool was used to design sgRNAs. sgRNAs were chosen based on the proximity to the desired cut site (knock-out) or insertion site (knock-in) and the quality score was calculated as probability of on-target activity minus the sum of probability of off-target cutting¹⁰². A minimum of three sgRNAs per experiment were compared for cutting efficiency in vitro.

Generation of sgRNAs

sgRNAs were generated by one of the following three methods: (1) Oligonucleotides encoding for desired sgRNA (complementary, excluding the PAM and a BbsI restriction site added to the 5′) were annealed, ligated into the pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid¹⁰³, and transformed into 10-beta Competent E. coli (C3019I, NEB). Individual clones were expanded and miniprep was performed using the NucleoBond Xtra Midi Kit (740410, Macherey Nagel). Positive clones were selected based on Sanger Sequencing analyzed using the CLC Main Workbench. A T7 promoter was added, and in vitro transcription was performed using the MEGAshortscript T7 Transcription Kit (AM1354, Thermo Fisher Scientific). sgRNAs were purified using the MEGAclear Transcription Clean-Up Kit (AM1908, Thermo Fisher Scientific). (2) The EnGen sgRNA Synthesis Kit (E3322S, NEB) was used according to the manufacturer’s instructions. In short, the EnGen sgRNA Template Oligo Designer (https://nebiocalculator.neb.com/#!/sgrna, NEB) was used to design oligos for in vitro transcription of sgRNAs. Resulting oligos were diluted to a final concentration of 1 μM in nuclease-free H2O and mixed with EnGen 2× sgRNA Reaction Mix and EnGen sgRNA Enzyme Mix followed by incubation at 37 °C for 2 h. sgRNAs were purified using the RNA Clean & Concentrator-25 Kit (R1017, Zymo Research). (3) The Alt-R CRISPR-Cas9 System (IDT), following the manufacturer’s instructions, was used. In short, on day of injection, Alt-R CRISPR-Cas9 crRNA and Alt-R CRISPR-Cas9 tracrRNA were mixed, heated to 95 °C, and left to cool down to RT to anneal.

In vitro cleavage assay to test activity of sgRNAs

All sgRNAs were tested in vitro for cutting activity following guidelines by NEB with minor adaptations. In short, sgRNAs were mixed with Cas9 protein (M0386S, NEB) and preincubated at 25 °C for 10 min. A linearized pGEM-T plasmid (A3600, Promega) containing the H3f4 sequence +/− 200 bp was added and incubated at 37 °C for 60 min. Resulting fragments were analysed on 2% agarose gels and cutting efficiency was scored based on the amount of uncut product left.

Homology templates for generating knock-in mouse models

The sequences of homology repair templates are available in Supplementary Data 7. (1) For homology-directed repair of <100 bp, Ultramer DNA Oligonucleotides (IDT) complementary to the strand bound by the sgRNA (Supplementary Data 6) were used. To prevent cutting of the homology template and the newly generated allele, a part of the PAM sequence was altered and/or mismatches close to the PAM sequence were introduced. (2) To generate the H3f4^H3.1 allele, a plasmid was designed. The H3f4 sequence plus 1 kb up- and downstream were cloned into the pJET1.2/blunt plasmid (K1231, Thermo Fisher Scientific). The H3f4 sequence was mutated using the Q5 Site-Directed Mutagenesis Kit (E0554S, NEB) to encode for a H3.1 protein. The resulting plasmid was amplified in 10-beta Competent E. coli and purified using the Plasmid Plus Maxi Kit (12963, QIAGEN).

Preparation of injection mix

sgRNAs were mixed with Cas9 Nuclease (Cas9-TOO-50, Labomics) in IDTE buffer (11-01-02-02, IDT) and incubated at 37 °C for 15 min. Cas9 mRNA (CAS9MRNA, Sigma) and, pending on whether a knock in was made, the homology template was added. The mix was prepared freshly on the day of injection.

Mouse husbandry, embryo collection, CRISPR injection, and embryo transfer

Animal housing, handling, and procedures of mice conformed to the Swiss Animal Protection Ordinance (protocol numbers: 1235, 1236, 2670, 3183, and 3091) and are compliant with the FMI institutional guidelines. Mice were housed in Type II long cages containing aspen bedding on IVC racks in rooms with 15–20 air changes per hour and a 12 h light/dark cycle. Temperature was maintained within 20–24 °C with a relative humidity within 45–65%. Food and water were provided ad libitum.

C57BL/6 J female mice were super ovulated by injection of 5U PMSG (MSD) followed by 5U hCG (MSD) after 48 h. After natural mating with C57BL/6J males, zygotes were harvested, and injections were done into paternal pronuclei. On average, 25 injected embryos were transferred into each ampulla of pseudo pregnant B6CF1 female mice.

Sequence validation of mouse models

Genomic DNA of potential founder animals was isolated from ear biopsies and PCR was performed using the Q5 High-Fidelity DNA Polymerase (M0491, NEB). For the H3f4 knock out mouse model, the following primers were used: 5′-TCCAGAACTCAGGAAAACTATGCC-3′ and 5′-ACTGAGCAGATGCGTTTGGA-3′. To genotype knock-in mouse models, primers spanning the H3f4 gene were used: 5′-GGCGGACGATTCAGGAAG-3′ and 5′-TCCTGAGGAGACAGGACACC-3′. To exclude random integration of the plasmid in the H3f4^H3.1 mouse model, a primer combination with one primer outside and one inside the 5′ homology arm was used as well: 5′-GGCGGACGATTCAGGAAG-3′ and 5′-ATTTGAGGCCAGAGGGAGTT-3′. Resulting PCR fragments were ligated into the pJET1.2/blunt vector and amplified in 10-beta Competent E. coli. Plasmids were isolated using the NucleoSpin Plasmid Kit (740588, Macherey-Nagel), subjected to Sanger Sequencing, and analyzed using the CLC Main Workbench.

Genotyping of mutant H3f4 alleles

To identify mutant H3f4 alleles, two strategies were used. (1) Enzyme-based genotyping. Genotyping of mutant H3f4 alleles (H3f4^V24A, H3f4^H42R, H3f4^S98A, and H3f4^H3.1) is based on the presence or absence of a restriction enzyme cutting site in the H3f4 gene introduced by the homology repair template. PCR products were incubated with the restriction enzyme and the restriction fragments were analyzed on agarose gel. (2) Non-enzyme-based genotyping. To genotype the H3f4 knock-out allele and tagged H3f4 alleles, the genotyping assay is only based on the size of the PCR product. Sequences of primers and PCR protocols are provided in Supplementary Data 8 and 9, respectively.

Mass spectrometry analysis of putative H3 bands

To perform the MS analysis on the putative histone bands, we performed TAU gel electrophoresis on acid-extracted histones and stained the gel with Coomassie Brilliant Blue (Supplementary Fig. 7d). The migration patterns of bands corresponding to H3.1, H3.2, and H3.3 were described by Shechter and colleagues⁶⁰. An additional band (band 1) was observed in the testis sample. We excised the putative H3 bands (band 1 and 2) from testis along with bands from ESCs (band 3 and 4) and digested them with trypsin. An EASY-nLC 1000 Liquid Chromatograph (Thermo Fisher Scientific) and an LTQ Orbitrap Velos hybrid mass spectrometer (Thermo Fisher Scientific) were used to separate peptides. Mascot 2.3 (MATRIX SCIENCE) was used for database search and Progenesis LC-MS (Nonlinear Dynamics) was used for label free quantification. Peptides were normalized using eight H3 peptides conserved among H3 proteins (residues 54–63, 64–69, 73–83, 117–128, and 130–134). The additional testis band (band 1) is 8-fold enriched for the H3.4 peptide compared to the H3.1/2/3 peptide. Conversely, the H3.4 peptide is only detected at very low levels in band 2 (H3.3) and is nearly completely absent in bands 3 and 4 (H3.1 and H3.3), respectively.

Isolation of H3 variants and determination of their relative ratios using TAU gel electrophoresis

To perform tissue-wide profiling of H3 variants, small pieces of different tissues were incubated in tissue lysis buffer (15 mM Tris-HCl pH 7.5, 60 mM KCl, 11 mM CaCl2, 5 mM NaCl, 250 mM sucrose, 5 mM MgCl₂, 0.5 mM DTT, 5 mM sodium butyrate, 0.3% NP-40 (74385, Fluka) and 1 tablet of cOmplete, EDTA-free Protease Inhibitor Cocktail (000000011873580001, Sigma) and homogenized in a Dounce homogenizer. Nuclei were pelleted by centrifugation and resuspended in 0.2 M HCl, incubated for a minimum of 2 h at 4 °C, precipitated in trichloroacetic acid, and resuspended in H₂O. Histones from cells were isolated as described previously⁶⁰ with minor modifications. Briefly, pelleted cells were incubated in hypotonic lysis buffer (10 mM Tris-Cl pH 8.0, 1 mM KCl, 1.5 mM MgCl2, 1 mM DTT, 1 tablet of cOmplete, EDTA-free Protease Inhibitor Cocktail) for 50 min rotating at 4 °C. Nuclei were pelleted, resuspended in 0.4 N HCl, and incubated rotating o/n at 4 °C. Samples were centrifuged at full speed and histone containing supernatants were loaded onto Amicon Ultra-0.5 mL Centrifugal Filters (UFC500324, Millipore) to exchange HCl to H₂O. Finally, an aliquot was taken and dried in a SpeedVac chamber. The pellet was dissolved in ice-cold freshly prepared TAU gel sample buffer (8 M urea, 5% acetic acid, 5% β-mercaptoethanol and 1% methyl green), incubated for 30 min on ice and loaded on the TAU gel.

To perform profiling of H3 variants in different populations during spermatogenesis, FACS-isolated cells were collected in Eppendorf tubes in the following numbers: 50 K for spermatogonia, 25 K for spermatocytes, 100 K for rSt, and 1 M for elSt and eSt populations. The proportions for spermatogonia, spermatocytes, and rSt cells were chosen to compensate for the DNA content. elSt and eSt populations were collected 10 times more than rSt, in order to compensate the loss of histones during spermatid elongation¹⁰⁴. After sorting, cells were pelleted in a swing bucket centrifuge at 300 × g for 10 min and the volume of the sample was reduced to 10 µl. The sample was frozen in liquid nitrogen and stored at −80 °C until further processing. The sample was thawed on ice and 50 µl of lysis buffer was added (with a final concentration of 10 mM HEPES pH 7.5, 1.5 mM MgCl₂, 10 mM KCl, 0.05% NP-40, 0.05 mM DTT, 1× of cOmplete, EDTA-free Protease Inhibitor Cocktail (000000011873580001, Sigma), 1 mM PMSF (Roche 10837091001), and 10 mM sodium butyrate). Lysis was performed on ice for one hour and then HCl was added to a final concentration of 0.2 M and incubated overnight. For the gel, a 10 µl aliquot was taken and dried in a SpeedVac chamber. Pellets were dissolved in TAU gel sample buffer, as described above. Histone samples were loaded on freshly polymerized 15% gels (15% 60%:0.4% acryl:bisacrylamide, 6 M urea/5% acetic acid/0.37% Triton X-100). Proteins were run at 200 V at 4 °C until the bands of methyl green exit the gel (5% acetic acid was used as a running buffer). Following the electrophoresis, gels were incubated in TAU gel transfer buffer (0.5% acetic acid) for 15 min. Finally, proteins were transferred to the PVDF membrane for 20 min at 500 mA. After the transfer, a standard Western Blotting protocol was used. To visualize all H3 variants, we used the anti-H3 antibody that recognized the C-terminal epitope present in all H3 variants (Abcam ab1791, 1:10,000, Supplementary Data 10). The ratio between the intensities of bands was used to determine the relative ratio of H3 variants.

SDS-PAGE followed by western blotting

Histone extracts or whole cell pellets were resuspended in SDS sample buffer (50 mM Tris-HCl pH 6.8, 12.5 mM EDTA, 10% glycerol, 2% SDS, 1% β-mercaptoethanol, 0.02 % bromophenol blue), incubated for 10 min at 94 °C and separated on a 15% polyacrylamide gels. Antibodies used in Western blot experiments are indicated in Supplementary Data 10.

Fixation of tissues and preparation of microscopy sections

Preparation and analysis of histological samples of mouse testis and epididymal tissues was performed as described previously⁶⁹. In brief, freshly isolated testis or epididymis samples were fixed in 5–10 ml of Bouin’s solution (Sigma HT10132) or in 4% PFA in PBS, for the histological and immunofluorescence experiments, respectively. Fixation was performed overnight on a rotating mixer at 4 °C. After fixation, samples were washed twice with 70% ethanol. Fixed samples were dehydrated through a graded series of ethanol solutions (2 × 70%, 80%, 2 × 96% and 3 × 100%) and xylene and finally embedded in paraffin using an automated tissue processing center (TPC 15 Duo, Medite) with standard settings. Sectioning was done at 3 µm thickness using the automatic microtome (HM355S, Thermo Fisher Scientific). Sections were mounted onto Superfrost Plus Adhesion Microscope Slides (J1800AMNZ, Thermo Fisher Scientific) and dried at 37 °C overnight.

Periodic acid-schiff and haematoxylin staining of paraffin-embedded testicular sections

For histological examination, testis sections were deparaffinized by incubating in xylene solution (Sigma 534056) 2 times for 5 min and then rehydrated in a series of decreasing concentrations of ethanol (2 × 100%, 95%, 70%, 3 min each) to deionized water. Rehydrated tissues were immersed in Periodic Acid Solution (Sigma, 395132) for 5 min at RT, rinsed several times in deionized water, immersed in Schiff’s Reagent (Sigma 3952016) for 15 min at RT and then rinsed in tap water for 5 min. Samples were counterstained with Mayer’s Haematoxylin Solution (MHS32, Sigma) for 2 min and rinsed in tap water for 5 min. Finally, samples were dehydrated in a series of increasing concentrations of ethanol (70%, 95%, 2 × 100%, 3 min each), cleared in xylenes solution (2 × 3 min) and mounted with Permount^TM mounting media (ThermoFischer, SP15-100). mage acquisition was performed using motorized automated slide Scanner Zeiss Axioscan Z1 with 20× air objective and analyzed with ZEN blue software (version 2.3, Zeiss).

Immunofluorescence imaging of paraffin-embedded testicular sections

Testis sections were deparaffinized by incubating in xylene solution (Sigma 534056) 2 times for 5 min and then rehydrated in a series of decreasing concentrations of ethanol (2 × 100%, 95%, 70%, 3 min each) to deionized water. Antigen retrieval was performed for 20 min at 94 °C in 10 mM sodium citrate buffer (pH 6.0) containing 0.05% Tween-20. Finally, the slides were washed in PBS. Around each sample, a circle was drawn using DAKO hydrophobic pen (Agilent S200230-2), in order to perform staining reactions in smaller volume. Blocking was performed in 30-50 µl of PBS containing 5% horse serum, 1% BSA, and 0.05% Tween-20 for 1 h. Primary antibodies (Supplementary Data 10) were diluted in blocking buffer, added to the samples, and incubated overnight at 4 °C. After incubation with primary antibodies, samples were washed with PBS and incubated with secondary antibodies diluted in blocking buffer for 1 h. Finally, sections were incubated 1:1000 DAPI solution (Sigma D9542) for 10 min and then washed with PBS. Mounting was performed using VECTASHIELD Antifade Mounting Medium (H-1000). Images were obtained using a spinning disk confocal scanning unit Yokogawa CSU W1 Dual T2 with 40×/1.3 oil immersion objective. Image analysis was performed with Fiji software¹⁰⁵ and using the automated image quantification pipeline described below.

EdU labelling of replicating cells

Intraperitoneal injection of 10 mg/ml EdU solution (Thermo Fischer Scientific A10044) was performed in the proportion of 0.5 × g of pure EdU per 10 × g of mouse weight. 3 h after injection, animals were sacrificed, followed by collection, PFA fixation and paraffin-embedding of testes by a standard protocol. For imaging, testicular sections were prepared, rehydrated, and blocked by a standard protocol. Labelling reaction mix was prepared in the following proportions: 400 µl PBS, 5 µl 100 mM CuSO₄, 0.5 µl Alexa Fluor 488 Azide (Thermo Fischer Scientific C10632) and 100 µl 500 mM ascorbic acid (added last). After 1 h of blocking, 30–50 µl of labelling reaction solution was added to samples and incubated for 30 min at room temperature in dark chamber. After the completion of reaction, samples were washed several times with blocking buffer and processed into a standard immunofluorescence imaging protocol.

Preparation of meiotic spreads

Meiotic spreads were prepared as described previously^106,107. Prior to staining, slides were washed in PBS in order to remove the excess of PFA.

Whole-mount immunofluorescence of testicular tubules

After decapsulation of testes from adult mice, seminiferous tubules were gently detangled and washed 2–3 times with PBS to remove the interstitial cells. Then, tubules were fixed with 4% PFA in PBS for 1 h at 4 °C. After fixation, tubules were washed 3 times in PBS and then 3 times in PBS containing 0.04% Tween 20 (PBST) in a cell strainer at RT, each wash for 10 min. Then, the tubules were dehydrated through a graded series of 25, 50, 75% methanol containing PBST and 100% methanol at 4 °C for 7 min, respectively. The samples were stored at −80 °C for later usage. At the time of observation, the samples were rehydrated through a graded series of 75, 50, 25% methanol containing PBST for 7 min at 4 °C and then washed 3 times with PBST for 10 min each at RT. Rehydrated tubules were blocked in PBST containing 4% Normal Donkey Serum (Abcam ab7475) for 1 h at RT. After blocking, the tubules were incubated with primary antibodies for 3 h at RT. After this, the tubules were washed 3 times for 10 min in blocking buffer followed by incubation with secondary antibodies for 2 h at RT. Finally, tubules were incubated 1:1000 DAPI solution (Sigma D9542) in PBST for 10 min and then washed 3 times with PBST for 10 min. Acrosome labeling was performed with lectin PNA (Invitrogen L32460). For imaging, tubules were oriented between a microscope slide and a coverslip separated by a 0.12 mm thick SecureSeal™ Imaging Spacers (Grace Bio Labs), with PBST used as a mounting media. Images were obtained using spinning disk confocal scanning unit Yokogawa CSU W1 Dual T2 with 63×/1.3 oil immersion objective. Image analysis was performed with Fiji software¹⁰⁵.

Quantitative analysis of immunofluorescence data

The quantitative analysis of the immunofluorescence images is performed in five consecutive steps: (1) Projecting, stitching, and binning of acquired image tiles; (2) Segmentation of testes tubules; (3) Extraction of testes tubules; (4) Nuclei segmentation in extracted testes tubules, and (5) Cell classification. In detail: (1) Projecting, stitching, and binning: The raw input data consists of 3D stacks with 10% overlap in XY for all four acquired channels. These 3D tiles a first maximum intensity projected (MIP) along Z. Then the illumination correction is applied to the DAPI channel for which the illumination field is obtained by averaging all 3D stacks of the DAPI channel for a given acquisition, computing the maximum intensity projection along Z, and fitting a 2D Gaussian. The MIPs are stitched with the Fiji Image Stitching plugin¹⁰⁸. The stitched image is average binned by a factor of 2 and saved. Additionally, the 8 times averaged binned version is saved for segmentation. (2) Segmentation of testicular tubules: Testes tubules were segmented on the 8 times average binned DAPI channel with a custom StarDist (v0.7.3)¹⁰⁹ model. For the training four images where manually annotated with Napari (https://zenodo.org/records/8076424). The ground truth data consists of 14 training patches and 2 validation patches containing a total of 752 annotated tubules. The StarDist network uses default parameters except for n_rays = 128 and grid = (4, 4). It was trained for 400 epochs, with 100 steps per epoch, an initial learning rate of 0.0003, and a training batch size of 8. Training was performed on nVidia Tesla V100 GPU with 32GB of memory. (3) Extraction of testicular tubules: The StarDist model for tubule segmentation predicts a star-convex polygon centred on the tubules. The predicted polygons are upscaled from 8× binning to 2× binning and refined such that the label boundary is not cutting through nuclei. (4) Nuclei segmentation in extracted testes tubules: A second StarDist model was trained for the nuclei segmentation on the 2 times binned DAPI channel. The ground truth data was manually annotated with Napari and split into 12 training and 4 validation images. The StarDist network uses default parameters except for n_rays = 64 and gird = (1, 1). It was trained for 400 epochs, with 100 steps per epoch, an initial learning rate of 0.0003, and a training batch size of 4. Training was performed on an nVidia Tesla V100 GPU with 32GB of memory. (5) Cell classification: Segmented cell nuclei from tubules of stages of interest were classified using Ilastik software¹¹⁰.

Fluorescence activated cell sorting (FACS)

To isolate SgU, SgD, ScLZ, and ScPD cells for ultra-low native ChIP-seq experiments (ULI-NChIP-seq, Fig. 3), we used a protocol based on surface markers of spermatogonial populations (antibodies listed in Supplementary Data 10)⁶⁹. Gating parameters and purity analysis of ScLZ population are available in Supplementary Fig. 10. To isolate ScPD and rSt cells used for conventional ChIP-seq experiments (Supplementary Figs. 13 and 14) and ScLZ, ScPD, rSt, and elSt cells for RNA-seq (Fig. 6a), we used a protocol based on dual staining of DNA with Hoechst 33342 and SYTO 16¹⁰⁷.

Sperm counting

After euthanizing a mouse, two cauda epididymis were isolated, with surrounding fat carefully removed, and squeezed in total of 1000 µl PBS. The suspension was filtered through 30 um mesh. From this filtrate, an aliquot of 50 µl was diluted in 1000 µl of water to immobilize cells. From this solution, 10 µl were taken to manually count cells in a cell counter chamber.

Antibodies

All primary and secondary antibodies used in this study are indicated in Supplementary Table 10.

RNA sequencing

FACS-isolated populations of ScLZ, ScPD, rSt, and elSt cells were pelleted at 2000 × g for 10 min at 4 °C. Total RNA was extracted from cell pellets using the RNeasy Mini Kit (Qiagen 74104) according to manufacturer’s instructions. DNA removal was performed using the on-column RNase-Free DNase Set (Qiagen, 79254). RNA concentration was determined with Qubit RNA HS Assay Kit (Q32852). Quality of RNA was determined using Fragment Analyzer system. Libraries were prepared with the Illumina Stranded total RNA-seq protocol according to manufacturer’s instructions. Quality control of libraries was performed with Agilent 2100 Bioanalyzer System. After passing the QC, sequencing was performed on Illumina NovaSeq platform (50 bp, paired end).

Chromatin Immunoprecipitation followed by sequencing (ChIP-seq)

To profile the H3 variants in early cell populations (SgU, SgD, ScLZ, and ScPD), we followed an ultra-low ChIP protocol⁶⁸. Its adaptation to FACS-isolated spermatogonial populations with the detailed workflow was described previously⁶⁹. In brief, to profile the H3 variants, chromatin from 5 K SgU or 10 K SgD, ScLZ or ScPD cells were used with 0.5 µl of anti-H3.3 (Cosmo Bio CE-040B) or 0.5 µl anti-Myc (Abcam ab9132) antibodies per sample. ULI-ChIP-seq libraries were prepared from the eluted ChIP DNA using the NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB E7645L), as described before⁶⁹. Libraries were sequenced in 50-bp paired-end mode with Illumina NovaSeq 6000 using the manufacturer’s instructions.

ChIP-seq in post-replicative cell populations (ScPD and rSt) was performed as follows. Chromatin from FACS-isolated ScPD and rSt populations was prepared by incubating cells for 20 min on ice in lysis buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% Triton X-100, 0.2% sodium deoxycholate, 5 mM CaCl₂) containing 5 U MNase (Roche Nuclease S7 10107921001) per 1 M cells. After incubation on ice, chromatin was digested for 10 min at 37 °C. The MNAse reaction was stopped by adding EGTA to 50 mM final concentration. The IP reactions were performed overnight at 4 °C on a rotator in the following proportions. To profile H3 variants, chromatin from 50 K ScPD or 200 K rSt was incubated with 5 µl of antibodies (anti-MYC: Abcam ab9132, anti-H3.3: Cosmo Bio CE-040B). To profile H3K4me3 and H3K27me3 in ScPD cells, chromatin from 1200 K ScPD was incubated with 3 µl of antibodies (anti-H3K4me3: Millipore 17-614, anti-H3K27me3: Cell Signaling #9733). Per sample, 25 µl of 1:1 mix of protein A:protein G Dynabeads (Thermo Fisher Scientific, 10001D and 10003D) was washed once with 25 µl lysis dilution buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% Triton X-100, 0.2% sodium deoxycholate, 50 mM EGTA), then added to chromatin-antibody mix and incubated on a rotator for 4 h at 4 °C. The Chromatin-Antibody-Dynabeads complexes were washed twice with 200 µl of RIPA buffer (140 mM NaCl, 10 mM Tris-HCl (pH 8.0), 0.1% Sodium deoxycholate, 0.1% SDS, 1% Triton X-100, 1 mM EDTA), once with 200 µl high salt buffer (140 mM NaCl, 10 mM Tris-HCl (pH 8.0), 0.1% Sodium deoxycholate, 0.1% SDS, 1% Triton X-100, 1 mM EDTA), once with 200 µl LiCl wash buffer (250 mM LiCl, 0.5% NP40, 0.5% deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, (pH 8.0)), and once with 200 µl TE buffer (10 mM Tris-HCl (pH 8.0), 1 mM EDTA). Finally, chromatin was eluted and digested for 1 h at 63 °C in Elution buffer (10 mM Tris-HCl (pH 8.0), 1 mM EDTA, 0.1% SDS, 300 mM NaCl). DNA from eluted material was purified by 2X volumes of Ampure XP DNA purification beads (Beckman Coulter, A63881) according to manufacturer’s instructions and resuspended in 9 µl of nuclease-free water. The eluted ChIP DNA was used to prepare libraries using the NEB Ultra protocol according to the manufacturer’s instructions. The libraries were sequenced in a 50 pb single end mode on the Illumina HiSeq 2500 platform, according to the manufacturer’s instructions.

Analysis of publicly available genomic data

ChIP-seq data profiling H3K4me3 and H3K27me3 in wildtype SgD cells and H3K27me3 in wildtype ScLZ and ScPD cells prepared with ULI-ChIP method were downloaded from the GEO database file GSE214682⁵⁷. ChIP-seq data profiling H3K4me3 and H3K27me3 in wildtype rSt cells were obtained from GSE42629⁶¹. RNA-seq data from wildtype SgU, SgD, and ScLZ cells was downloaded from GSE214682⁵⁷. RNA-seq from wildtype ScLZ, ScPD, rSt, elSt, and eSt populations were downloaded from GSE214316⁵⁸.

Computational analysis of ChIP-seq data

ChIP-seq data were processed with TrimGalore (version 0.6.2) to perform adaptor trimming and remove low quality reads with settings (-clip_R1 2 –stringency 3). The trimmed reads were aligned to the mouse genome build mm10 using STAR (version 2.7.10a¹¹¹) with settings (–alignIntronMin 1 –alignIntronMax 1 –alignEndsType EndToEnd –alignMatesGapMax 1000 –outFilterMatchNminOverLread 0.85). The BAM files were de-duplicated using SAMtools (version 1.10) with standard settings and used for the downstream analysis.

To profile the occupancy of the H3 variants genome-wide, the mouse genome from the BSgenome.Mmusculus.UCSC.mm10 assembly was partitioned into 10 kb non overlapping tiles. Tiles overlapping with blacklisted regions¹¹² were excluded from the downstream analysis. The tiles were classified as containing a TSS if an overlap with ±1 kb around genic TSS (TxDB.Mmusculus.UCSC.mm10.knownGene) was reported. Further, the uniquely mapped reads were counted on genomic regions with the qCount function (mapqMin = 225 L) from the QuasR package¹¹³ and correction for reads originating from chrX was performed. log₂CPM values were calculated for each sample and normalized within each ChIP experiment (i.e., libraries from different cell types but the same antibody were normalized to each other) using normalizeBetweenArrays function from the Limma package¹¹⁴ (method = “cyclicloess”, cyclic.method = “fast”).

To profile the H3 variants and the H3 modifications at gene promoters, genic TSS coordinates were extracted as described above and reads mapping to ±25 kb around the TSS were calculated with the qProfile function (binSize = 201L) and log₂CPM values were calculated. To normalize these data, additionally, qProfile quantification (binSize = 201 L) was performed on 50 kb genomic tiles that do not contain TSS and do not overlap with chrX. For these tiles, log₂CPM values were calculated and tiles that commonly possess read counts below Q25 and above Q75 per ChIP were defined. These regions were added to the genic TSS, and normalization was performed on the merged data as described above. After normalization, non-TSS regions were removed, and downstream analysis was performed on normalized TSS data only. In Fig. 3a and Supplementary Fig. 14a, clustering was done on rowSums values of normalized log₂CPM qProfile data. The resulting data frame was scaled and clustered using the functions hclust (method = “ward.D”) and dist (method = “euclidean”) from the stats R package (version 4.2.1). The rows within each clusters were ordered according to RNA expression levels. TSS-regions in heatmaps in Supplementary Figs. 13g and 14a are displayed in the same order as shown in Fig. 3a. For heatmap plotting, normalized data was visualized using the ComplexHeatmap R package¹¹⁵.

In ScPD and rST cells of wildtype control mice, the anti-MYC antibody reveals a non-uniform pattern around TSS of CGI genes and partially along X-linked genic regions, opposite of the patterns observed in H3.4^MYC-expressing H3f4^MYC/wt transgenic cells (Supplementary Fig. 13g). Since H3.4^MYC enrichment in H3f4^wt/wt mice does not reflect the presence or absence of the DNA-sequence motif of the endogenous MYC protein at promoter regions, it may point to cross-reactivity with other epitopes. We consider the H3.4^MYC signal in H3f4^MYC/wt transgenic animals as representing genuine H3.4 occupancy.

Computational analysis of RNA-seq data

RNA-seq data were processed with TrimGalore (version 0.6.2) to perform adaptor trimming and remove low quality reads with settings (-clip_R1 2 –stringency 3). The trimmed reads were aligned to the mouse genome build mm10 using STAR (version 2.7.10a¹¹¹) with settings (–alignIntronMin 1 –alignIntronMax 1 --alignEndsType EndToEnd –alignMatesGapMax 1000 –outFilterMatchNminOverLread 0.85). The BAM files were de-duplicated using SAMtools (version 1.10) with standard settings and used for the downstream analysis. Quantification of gene expression (from TxDb.Mmusculus.UCSC.mm10.knownGene Bioconductor annotation) was performed with QuasR R package¹¹³.

Genes with at least 1 count per million (CPM) in three biological replicates were used in downstream analysis. The edgeR R package¹¹⁶ was used for statistical analysis of differential gene expression between the following separate comparison pairs: (1) H3f4^H3.1/− and H3f4^wt/−; (2) H3f4^H3.1/− and H3f4^wt/wt; (3) H3f4^H3.1/H3.1 and H3f4^wt/wt; (4) H3f4^wt/− and H3f4^wt/wt; (5) H3f4^MYC/wt, and H3f4^wt/wt; (6) H3f4^V24A/− and H3f4^wt/−; (7) H3f4^H42R/− and H3f4^wt/−. CPM values were normalized using trimmed mean of M values (TMM) normalization using function calcNormFactors(method = “TMM”). Generalized Linear Model (GLM) was fit using genotypes as covariates. The following statistical significance cutoffs were used: |log₂FC| > 1 and false discovery rate-adjusted (FDR) < 0.05. Gene Ontology (GO) search was performed using clusterProfiler R package¹¹⁷.

Author contributions

P.A.K., P.C.B., C.-Y.L. and A.H.F.M.P. designed experiments. P.A.K., P.C.B., and H.-R.H. performed evolutionary analysis of the H3 genes. P.C.B., C.-Y.L. and J.-F.S. generated H3f4 mouse models by CRISPR/Cas9 editing. H.K. performed FACS experiments. C.-Y.L. and P.A.K. performed TAU gel experiments. C.-Y.L. performed MS experiments. P.A.K., G.F. and S.C. performed genomic experiments. P.A.K. analyzed genomic data. P.A.K. performed imaging experiments. T.-O.B. developed the imaging quantification pipeline, with input from P.A.K.. P.A.K., P.C.B. and A.H.F.M.P. wrote the manuscript, with input from all authors. A.H.F.M.P. conceived and supervised the project.

Peer review

Peer review information

Nature Communications thanks Shunsuke Suzuki and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The NGS datasets used in this study are available at the Gene Expression Omnibus (GEO) database with the following accession number: GSE291274. Raw and processed imaging data generated in this study are available at the BioStudies repository under accession code S-BSST2851. Source data files for each are available at the Figshare repository (10.6084/m9.figshare.31829755). Uncropped raw Western blot images are available at the Figshare repository (10.6084/m9.figshare.31829722). Source FACS data are available at the Figshare repository (10.6084/m9.figshare.31830952). Raw mass-spectrometry data is available at the MassIVE repository (https://massive.ucsd.edu/) with data identifier MSV000101220.

Code availability

Code for evolutionary and imaging data analysis is available at the GitHub repository (https://github.com/fmi-basel/FMI_H3.4).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73954-9.