Histone H3.5 forms an unstable nucleosome and accumulates around transcription start sites in human testis

Human histone H3.5 is a non-allelic H3 variant evolutionally derived from H3.3. The H3.5 mRNA is highly expressed in human testis. However, the function of H3.5 has remained poorly understood. We found that the H3.5 nucleosome is less stable than the H3.3 nucleosome. The crystal structure of the H3.5 nucleosome showed that the H3.5-specific Leu103 residue, which corresponds to the H3.3 Phe104 residue, reduces the hydrophobic interaction with histone H4. Mutational analyses revealed that the H3.5-specific Leu103 residue is responsible for the instability of the H3.5 nucleosome, both in vitro and in living cells. The H3.5 protein was present in human seminiferous tubules, but little to none was found in mature sperm. A chromatin immunoprecipitation coupled with sequencing analysis revealed that H3.5 accumulated around transcription start sites (TSSs) in testicular cells. We performed comprehensive studies of H3.5, and found the instability of the H3.5 nucleosome and the accumulation of H3.5 protein around TSSs in human testis. The unstable H3.5 nucleosome may function in the chromatin dynamics around the TSSs, during spermatogenesis.


Open Access
Epigenetics & Chromatin *Correspondence: kurumizaka@waseda.jp 1 Laboratory of Structural Biology, Graduate School of Advanced Science and Engineering, Institute for Medical-oriented Structural Biology, Waseda University, 2-2 Wakamatsu-cho, Shinjuku-ku, Tokyo 162-8480, Japan Full list of author information is available at the end of the article chromatin domains, such as heterochromatin, euchromatin, telomeres, and centromeres [21][22][23][24]. H3.X and H3.Y are expressed in normal and malignant tissues, including brain, and H3.Y is induced by stress stimuli, such as nutrient starvation [17]. CENP-A is an essential component of the chromosomal centromere [7], and forms the fundamental centromeric nucleosome [25]. H3T is highly expressed in the testis, suggesting that the H3T nucleosome may function during spermatogenesis [6,8,9]. H3T forms nucleosomes in vitro, and was incorporated into chromatin when ectopically expressed in human cells [26]. However, the H3T nucleosome was extremely unstable in vitro, and H3T tagged with green fluorescent protein (GFP) was rapidly exchanged in living cells [26].
H3.5 is conserved among great apes and Neanderthals, but not in non-hominid primates [18]. The H3.5 mRNA is highly expressed in the human testis [18]. In cells, ectopically expressed H3.5 is reportedly incorporated into chromatin, and predominantly localized in the euchromatic region [18]. Ectopic H3.5 expression complemented the growth defect of H3.3 knockdown cells, suggesting that it has an overlapping function with H3.3, as a replacement histone [18]. However, endogenous H3.5 has not been detected at the protein level, and the biochemical and cellular functions of the H3.5 nucleosome have not been clarified so far.
In the present study, we performed comprehensive studies of human histone H3.5, including structural and biochemical analyses with reconstituted nucleosomes, fluorescence recovery after photobleaching analyses with living cells, immunohistochemical analyses with human testis sections, and chromatin immunoprecipitation followed by sequencing (ChIP-Seq).
We next tested the stability of the H3.5 nucleosome, using a salt-titration assay. The reconstituted nucleosomes were incubated at 50 °C for 1 h, in the presence of 0.4, 0.6, 0.7, or 0.8 M NaCl, and the resulting nucleosomes were analyzed by native polyacrylamide gel electrophoresis. In this assay, the H3.1 and H3.3 nucleosomes were equally stable, and formed nucleosomes in 0.4-0.8 M NaCl (Fig. 1e). In contrast, the intact H3.5 nucleosome was only detected under the 0.4 M and 0.6 M NaCl conditions (Fig. 1f, lanes 9 and 10). At higher NaCl concentrations (i.e., 0.7 and 0.8 M), the bands corresponding to the H3.5 nucleosome disappeared, indicating that the H3.5 nucleosome was disrupted (Fig. 1f, lanes 11 and 12). Consistent with the previous study [26], the H3T nucleosome was disrupted in 0.6 M NaCl, and was the most labile ( Fig. 1f, lanes 5-8). We previously purified the complexes corresponding to the bands remaining after the H3T nucleosome disruption, and confirmed that these bands were non-specific H2A-H2B-DNA complexes (Fig. 1f, asterisks) [26]. These results showed that the H3.5 nucleosome is more stable than the H3T nucleosome, but is clearly unstable as compared to the H3.1 and H3.3 nucleosomes. The formation of unstable nucleosomes may be a common feature of the human testisspecific H3 variants.

Crystal structure of the H3.5 nucleosome
To understand the structural basis for the instability of the H3.5 nucleosome, we determined the crystal structure at 2.8 Å resolution ( Fig. 2a; Table 1). The overall structure was similar to that of the H3.3 nucleosome [27], as expected. H3.5 contains two residues, Asn78 and Leu103, which are not conserved in H3.3. Both residues do not directly interact with either the H2A-H2B dimers or the DNA, which could possibly affect nucleosome stability. Leu103, however, is located at the interface of H3.5 and H4, and may possibly exhibit reduced hydrophobic interactions compared with that of H3.3 (Fig. 2b, c). In H3.3, the corresponding residue is Phe104, which fills the pocket created by the α1 and α2 helices of H4, and apparently forms hydrophobic interactions with the side chains of the H4 Ile34, Ile50, and Thr54 residues [27]. In contrast, such close hydrophobic interactions are not observed around the Leu103 residue in the H3.5 nucleosome, because Leu has a smaller side chain than Phe (Fig. 2b). These data suggested that this structural difference may account for the instability of the H3.5 nucleosome.

Higher mobility of GFP-tagged H3.5 in living cells
To examine whether H3.5 is incorporated into nucleosomes less stably than H3.3 in living cells, we performed fluorescence recovery after photobleaching (FRAP), using HeLa cells expressing GFP-H3.5 and GFP-H3.3 [26,29,30]. One-half of the nucleus was bleached, and the fluorescence intensity was measured in the presence of cycloheximide, to suppress the fluorescence recovery due to new protein synthesis. As shown in Fig. 5a, both GFP-H3.5 and GFP-H3.3 exhibited slow recovery, consistent with their incorporation into chromatin as H3.1-GFP [31]. Quantitative measurements then indicated that GFP-H3.5 recovered substantially faster than GFP-H3.3 ( Fig. 5b), suggesting that nucleosomal H3.5 exchanges more rapidly than H3.3. FRAP analyses of the mutants further revealed that the GFP-H3.5 L103F mutant recovered more slowly than the reciprocal GFP-H3.3 F104L mutant (Fig. 5c, d). These results indicated that the H3.5-specific Leu103 residue is critical for the rapid exchange of H3.5 in living cells, in good agreement with the in vitro salt-titration data.

Presence of human histone H3.5 in testicular cells within seminiferous tubules
Since the endogenous H3.5 protein has not been detected, due to the lack of a specific antibody, we generated a specific monoclonal antibody directed against H3.5, using a peptide containing H3.5 Thr22-Arg41 (MAB Institute, Inc.). A Western blotting analysis showed that the H3.5 antibody specifically reacted to H3.5, but not to other variants (i.e., H3.1, H3.2, H3.3, and H3T) (Fig. 6a). In addition, we performed a Western blotting analysis with human testicular cell extracts from three individuals, and confirmed that the H3.5 antibody specifically detected endogenous H3.5 with low background signals (Fig. 6b). By using this antibody in immunohistochemical analyses, we detected positive signals in human testis sections, indicating that H3.5 is expressed at the protein level in cells within seminiferous tubules (Fig. 7a). Major histone H3 variants, such as H3.1 and H3.3, were also detected in testis sections (Fig. 7b, c). Interestingly, H3.5 was clearly present in spermatogonia and/or primary spermatocytes, in which the first meiotic cell division is not completed (Fig. 7a). However, the endogenous H3.5 protein was not detected in mature sperm by Western blotting using the H3.5-specific antibody, although H3 was clearly detected when the C-terminus-specific antibody was used (Fig. 7d), suggesting that H3 variants other than H3.5 are present in mature sperm. These results prompted us to test the genomic localization of endogenous H3.5 in human testicular cells.

Histone H3.5 in human seminiferous tubules
During spermatogenesis, the haploid genome becomes tightly packed into the sperm nucleus. This process requires robust chromatin reorganization, and eventually the histones are largely replaced by protamines [33][34][35]. Some human histone variants, including TSH2B and H3T, are highly expressed in testis, at least at the mRNA level, and may perform specific functions in the chromatin reorganization during spermatogenesis. H3.5 is a relatively newly identified variant, and its mRNA is also highly expressed in testis, as compared to other tissues [18]. However, the endogenous H3.5 protein has not been detected, due to the lack of its specific antibody.
Our immunohistochemical analysis with human testis sections detected H3.5 at the protein level in seminiferous tubules, especially in the cells before or during the first meiotic cell division (Fig. 7). However, H3.5 may not be retained in mature sperm (Fig. 7). Therefore, H3.5 may play a role in preparing the proper chromatin landscape for events before or during the first meiotic cell division. Stage-specific production has been reported for the mouse testis-specific H2B variant, TH2B [36], which is an ortholog of human TSH2B [37]. However, the timing of TH2B expression is clearly different from that of H3.5. TH2B is barely detected in spermatogonia, but exists in spermatids. In contrast, H3.5 can be detected in spermatogonia, but not in mature sperm, although we cannot exclude the possibility that a trace amount of H3.5 is retained at limited genomic loci in human sperm. The

The nucleosome containing histone H3.5 is unstable
We found that the H3.5 nucleosome is quite unstable, as compared to the H3.3 nucleosome in vitro (Fig. 1). Consistently, the mobility of H3.5 is remarkably faster than that of H3.3 in living cells (Fig. 5). In humans, the expression level of another histone H3 variant, H3T, is also high in the testis, but low in somatic cells [6,8,9]. The nucleosome containing H3T is quite unstable in vitro and in living cells [26]. Nucleosome instability was also reported with a mouse testis-specific H2A variant, H2AL2 [38]. Therefore, instability may be a common characteristic of the testis-specific nucleosomes. The unstable nature of the H3.5 nucleosome may be suitable for further replacement with transition proteins and protamines. H3.5 incorporation may also regulate the transcription of the genes required during the early stages of spermatogenesis. However, H3.3 appears to be more relevant for regulating transcription during spermatogenesis, as its incorporation is correlated with the gene expression level. In contrast, H3.5 incorporation may function to transiently mark TSSs to assist in the replacement with H3.3, depending on gene expression. Histone acetylation may also play an important role in global and/or local histone exchange in the human testis, as shown in the mouse [39].
Our present and previous [26] studies demonstrated that the H3.5 Leu103 and H3T Val111 residues are predominantly responsible for the instability of the H3.5 and H3T nucleosomes, respectively. These H3.5 Leu103 and H3T Val111 residues correspond to Phe and Ala in H3.3 (and canonical H3.1), respectively. In the crystal structure of the H3.5 nucleosome, the H3.5 Leu103 residue forms fewer hydrophobic interactions with H4, as compared to the corresponding H3.3 Phe104 residue, and does not induce substantial structural distortion around the residue (Fig. 2). In contrast, the H3T Val111 residue induces local structural distortion around position 111 [26]. Therefore, the H3.5 Leu103 and H3T Val111 residues induce nucleosome instability by different mechanisms.
The H3.5 Leu103 and H3T Val111 residues are both located in the vicinity of the nucleosomal dyad. Intriguingly, genetic and biochemical studies have identified a mutation at Arg116 (to His) that destabilizes the nucleosome [40,41]. This H3 mutation is known as a Sin mutation that alleviates the requirement for the Swi/ Snf nucleosome-remodeling factor [40,42]. The Sin phenotype has also been found in Saccharomyces cerevisiae with the H3 Ala111 to Gly mutation [43]. In addition, comprehensive alanine-scanning mutagenesis in S. cerevisiae suggested that the mutation of the 103rd or 104th residue of H3 may affect transcriptional regulation, probably through chromatin remodeling [44]. The crystal structure of the nucleosome containing the Sin mutations revealed that the H3 Arg116 mutation may allosterically destabilize the nucleosome, by reducing the number of histone-DNA and/or histone-histone interactions [45]. Therefore, the H3 C-terminal region, which is located near the nucleosomal dyad, is important for stable nucleosome formation, and amino acid substitutions within this region sensitively affect the nucleosome stability.
In addition to the testis, small amounts of H3.5 mRNA expression are also observed in ejaculate, leukocytes, and liver [18]. Proteins specifically produced in the testis have frequently been found as inappropriately overexpressed proteins in cancer cells. Furthermore, several missense mutations of the human H3F3C gene, encoding H3.5, have recently been found in tumors [46], including Val100 and Arg130. Like the L103F and Sin mutations, the mutations of these residues may also influence the nucleosome stability, by affecting the intra-nucleosomal interactions of amino acid residues near the nucleosomal dyad [40][41][42][43][44][45]. Together, these findings suggest that the inappropriate production of the H3.5 mutant may compromise proper chromosomal function in tumor cells. It is thus intriguing to study the correlation between H3.5 nucleosome stability and cancer predisposition.

Conclusions
We found that the H3.5 nucleosome is less stable than the H3.3 nucleosome, and the H3.5-specific Leu103 residue is responsible for the instability of the H3.5 nucleosome, both in vitro and in living cells. We discovered that the H3.5 protein is actually present in seminiferous tubules in humans. Although the sample was limited to a single donor, our ChIP-seq analysis suggests that the endogenous H3.5 specifically accumulates at transcription start sites in human testicular cells. These findings provide new important insights into the role of H3.5 during spermatogenesis. Future analyses using more specimens from various donors, including those suffered from diseases, will be required for fully understanding the function of H3.5 in the chromatin reorganization.

Preparation of nucleosomes
The nucleosomes were reconstituted using the 146 basepair human α-satellite DNA [2], prepared by the method described previously [48]. For the H3.5, H3.3, and H3.1 nucleosomes, the purified histone octamers were mixed with the 146 base-pair DNA in a solution containing 2 M KCl. For the H3T nucleosome, the H3T-H4 and H2A-H2B complexes were mixed with the 146 base-pair DNA in a solution containing 2 M KCl [26]. The nucleosomes were reconstituted by the salt-dialysis method, heated at 55 °C for 2 h, and further purified from the free DNA and histones by non-denaturing polyacrylamide gel electrophoresis (Prep Cell, Bio-Rad).

Salt resistance assay for nucleosome stability
The nucleosomes (240 ng/μl) were incubated in the presence of 0.4, 0.6, 0.7, and 0.8 M NaCl in 36 mM Tris-HCl (pH 7.5) buffer, containing 1.8 mM EDTA and 1.8 mM dithiothreitol, at 50 °C for 1 h. After this incubation, the NaCl concentrations of the samples were adjusted to 0.4 M, and the samples were analyzed by non-denaturing 6 % PAGE with ethidium bromide staining.
Diffraction data were integrated and scaled with the HKL2000 program [49]. The data were processed using the CCP4 program suite [50]. The structure of the H3.5 nucleosome was solved by the molecular replacement method, using PHASER [51] and the H3.3 nucleosome structure [PDB:3AV2] as the search model [27]. The structure of the H3.5 nucleosome was refined using PHE-NIX [52], and the model was built using COOT [53]. Following the rigid body refinement, iterative rounds of xyz-coordinate, real-space, individual B-factor, and occupancy refinements were performed, with optimizing the X-ray/stereochemistry and the X-ray/B-factor weights. Secondary structure restraints and non-crystallographic symmetry restraints were applied for the refinements. The Ramachandran plot of the final H3.5 nucleosome structure showed 99.2 % of the residues in the favored region, 0.8 % of the residues in the allowed region, and no residues in the outlier region, as validated with the Mol-Probity program [54]. A summary of the data collection and refinement statistics is provided in Table 1. All structure figures were created using the PyMOL program [55].

Thermal stability assay
The tetrasomes were reconstituted by the salt dialysis method, and were purified by non-denaturing polyacrylamide gel electrophoresis (Prep Cell, Bio-Rad), as previously described [28]. The reconstituted tetrasomes, containing the H3.5-H4 or H3.

FRAP analysis
DNA fragments encoding H3.5, H3.3, and their mutants were cloned into the pEGFP-C3 vector (Clontech). HeLa cells were transfected with the expression vectors using Lipofectamine 2000 (Life Technologies), and cultured in 1 mg/ml G418 (Nacalai Tesque) to select those stably expressing GFP-tagged H3 proteins. Cells were grown in Dulbecco's modified Eagle's medium, supplemented with 10 U/ml penicillin, 50 μg/ml streptomycin, and 10 % fetal calf serum, on a glass-bottom dish (Mat-tek). FRAP was performed using a confocal microscope (FV-1000; Olympus), featuring a heated stage supplemented with 5 % CO2 [29]. A confocal image of a field containing about 10 nuclei was collected with a 60× UPlanSApo NA = 1.35 lens (800 × 800 pixels, zoom 2, scan speed 2 μs/pixel, pinhole 800 μm, Kalman filtration for four scans, LP505 emission filter, and 0.2 % transmission of 488-nm Ar laser). Afterward, one-half of each nucleus was photobleached using 90 % transmission of the 488 nm laser (two iterations), and images were collected using the original setting at 1 min intervals for 120 min. Fluorescence intensities of the bleached, unbleached, and background areas were measured using Image J 1.46r. After background subtraction, the relative intensity of the bleached area to the unbleached area was determined and normalized to the initial value before bleaching.

Immunohistochemistry
Human testicular samples were obtained, fixed in Bouin's solution for 2 h and embedded in paraffin. After deparaffinization, 5-μm sections were incubated with hydrogen peroxide to inhibit endogenous peroxidases. After non-specific binding was blocked by rabbit serum, the slides were incubated with the culture supernatant of the hybridoma producing the anti-H3.5 monoclonal antibody (1:100 dilution) for 24 h at room temperature, and endogenous H3.5 was visualized using the avidin-biotin complex method. The sections were then counterstained with haematoxylin. In the negative control slides, the primary antibody was omitted. Endogenous H3.1 and H3.3 were detected by the same method as that for the H3.5 detection, using the H3.1-specific monoclonal antibody [56] or the H3.3-specific monoclonal antibody [57].

Chromatin immunoprecipitation
Human testis homogenates were fixed with 1 % formaldehyde in PBS(+) buffer. The fixed human testis homogenates were precipitated then suspended in RIPA buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA, 1 % NP-40, 0.5 % sodium deoxycholate, 0.1 % SDS, and protease inhibitor cocktail; Nacalai Tesque Inc.), instead of ChIP buffer. The sample was sonicated twenty times for 5 s. The sheared samples were then centrifuged at 15,000×g for 10 min. The supernatant, containing the DNA, was incubated with magnetic beads conjugated with the anti-H3.3 rat monoclonal antibody [57] or the culture supernatant of the hybridoma producing the anti-H3.5 mouse monoclonal antibody, at 4 °C overnight with rotation. The immune complexes were pulled down, washed with RIPA buffer and TE buffer (both twice), and then eluted from the beads using 1 % SDS and 0.02 % Proteinase K (Nacalai Tesque Inc.) in TE. The cross-links were reversed by an incubation for 4 h at 65 °C, followed by an incubation for 1 h at 50 °C. The DNA samples were then purified with a Qiaquick PCR purification kit (Qiagen, Valencia, CA, USA). The ChIP library was prepared with the Illumina protocol, and the samples were sequenced on an Illumina HiSeq-1500 system.

ChIP-Seq data analysis
Sequenced reads of H3.3 and H3.5 ChIP-Seq were mapped onto the human genome (hg19) with Bowtie (version 0.12.8), with the parameters "-v 2 -m 1". The uniquely mapped and PCR duplicates-removed reads, obtained using SAMTools [58], were utilized for further analysis. The number of reads for input, H3.5, and H3.3 were 26,524,770, 14,818,772, and 24,356,280, respectively. The estimation of normalized ChIP-Seq signal intensities was calculated as follows. First, we counted the mapped reads throughout 1000 bp intervals (bins) on the genome, and then the counts were normalized as RPKM (Reads Per Kilobases per Million reads) [59]. Finally, the ChIP-Seq signal intensities were calculated as the RPKM differences between the ChIP and input DNA-control data (i.e., ChIP-control) on each bin. For the peak distribution analysis (percentages of peak localizations on each genomic category), the HOMER software [60] was utilized. Peaks were called using the MACS2 software (version 2.1.0) [61] with the following parameters: q value <0.1 for H3.5, p value <0.01 for H3.5.

mRNA-Seq and analysis
Total RNA was extracted from human testis homogenates. cDNA synthesis was performed with Primescript Reverse Transcriptase and a dT primer (Takara Bio Inc.). The preparation of the mRNA-Seq library and the sequencing were performed according to the Illumina protocol. Sequenced reads were mapped onto the human genome (hg19) with Tophat (version 2.0.8). The gene expression levels (FPKM; Fragments Per Kilobase of exon per Million mapped sequence reads) were estimated with Cufflinks (version 2.0.1), using the mapped reads. The default parameters of the software were employed. We defined ten expression groups, labeled q0-10 %, q10-20 %,…, q90-100 %, with respect to the FPKMs of genes, which define the 10 percentile intervals of all FPKMs; i.e., the genes were ordered by the FPKMs and then separated into ten groups with equal numbers of members. The genes with FPKM = 0 were excluded.

Data access
The atomic coordinates of the H3.5 nucleosome have been deposited in the RCSB Protein Data Bank, with the RCSB ID code [PDB:4Z5T]. Deep-sequencing data have been deposited in the DDBJ sequence read archive, with the accession number [DDBJ:DRA002604].

Authors' contributions
TU performed structural and biochemical analyses of the H3.5 nucleosome. TU, NH, AO, and HT prepared recombinant histones, and TU, NH, and WK collected the X-ray diffraction data, and determined the crystal structure of the H3.5 nucleosome. AH, KM, and YO performed the ChIP-Seq experiments and analyzed the data. KSato, YS, and HKimura performed the FRAP experiments. KShiraishi and NS performed immunohistochemical analyses with the human testis sections and analyzed the data. HKurumizaka, HKimura, and YO designed all experiments, and HKurumizaka wrote the manuscript. All authors read and approved the final manuscript.