Background

In eukaryotic cells, the nucleosome is the basic unit of chromatin organization. Nucleosomes are composed of an octamer of four core histones, H2A, H2B, H3 and H4, wrapped by DNA [1]. An additional linker histone, H1, can be deposited near the DNA entry–exit points [2, 3]. The dynamic organization of chromatin impacts many cellular events, including the regulation of gene transcription, DNA replication and the maintenance of genome integrity through DNA repair mechanisms [4, 5]. These pathways signal to chromatin by different mechanisms including DNA methylation, non-coding regulatory RNAs, recruitment of remodelling factors, incorporation of histone variants and covalent modifications of histones [6,7,8,9,10,11,12]. Histones are decorated by many post-translational modifications, the most common of which are acetylation, methylation, phosphorylation and ubiquitination [13, 14]. Some of these modifications favour transcription activation, while others are associated with repression of transcription [15]. In addition to transcription, histone modifications are involved in numerous regulatory circuits, such as chromosome dynamics [16], DNA repair [17] or the establishment and maintenance of heterochromatin [18]. Furthermore, dedicated molecular machineries can load and mobilize nucleosomes along the DNA (for review, see [4]). These chromatin remodellers play an important role in the regulation of transcription by organizing the nucleosomal positions at critical regulatory regions [19, 20]. Finally, non-allelic variants of canonical histones, named histone variants, are important elements in chromatin signalling pathways [21, 22]. Some variants are general players—expressed ubiquitously, contributing to various aspects of transcription and epigenetic regulations—while others are only expressed in certain cell types, such as germ cells [23]. Some of these variants are specifically expressed during sperm differentiation and are annotated TS for testis-specific [24,25,26]. Altogether, histone variants have been described for H3, H2A, H2B and H1; H4 is the only histone for which no variant has been identified in mammals, but some organisms, such as the urochordate Oikopleura dioica, ciliates and trypanosomes, have evolved H4 variants [10, 27,28,29].

Histone variants were initially discovered using classical biochemical approaches. Recently, the development of mass spectrometry (MS) techniques, with constant increases in sensitivity and speed of analysis, has facilitated their identification and functional characterization [14, 30,31,32,33,34]. In order to utilize these technologies, histones are first biochemically enriched taking advantage of their highly basic nature. Then, they are proteolyzed with proteases to form short peptides, which are then analysed by MS/MS. The acquired MS/MS spectra are interpreted and converted into amino acid sequences, from which the identity of the original histone protein and the possible presence of post-translational modifications on specific residues can be determined [35]. However, these analyses still remain restricted for a number of reasons. One of these is that the interpretation of MS/MS spectra relies on matching experimental data to theoretical peptide sequences obtained by an in silico proteolysis of a list of proteins. Therefore, the content of the theoretical protein sequence database conditions the interpretation of the experimental spectra and the subsequent identification of histones. Classical databases such as Swiss-Prot, trEMBL and NCBI are usually used with success. However, histones have not been precisely annotated in these resources. Manually curated databases such as Swiss-Prot lack several histones, while others, such as trEMBL or NCBI protein database, are more extensively populated with non-reviewed data. The latter contain more histone entries, but the degree of redundancy and the precision of the descriptions can make protein identification results difficult to interpret. Finally, naming of histones has been recently revisited with a new unified nomenclature [36]. The recent release of HistoneDB 2.0 consolidated the sequence information of a large variety of histones and their sequence variants in many organisms [37]. However, it has not yet been integrated in the above databases and the same variant can go by different names. For instance, the coding gene H2afb1 refers to proteins H2A.L.2 or H2A.Lap3 in the literature and to H2A-Bbd type 1 in the NCBI RefSeq and UniProtKB databases [38, 39]. In addition, a different protein coded by Gm14920 is also named H2A.Bbd.1 in other publications [40]. Here, a unified name is presented to identify uniquely each ambiguous entry and is also associated with its other names to facilitate its relationship with previously published work.

We have collected redundant histone entries from a number of public databases, gathering >700 entries for mouse and >1000 entries for human histones. We manually curated these lists to obtain a final count of 83 and 85 histone entries for mouse or human, respectively. Their annotations have been revisited to match the current histone nomenclature in accordance with the new resource “HistoneDB 2.0—with variants” [37]. About 30% of these entries have a fuzzy UniProtKB protein annotation, such as “predicted” or “inferred by homology”, and we performed MS analysis to clearly identify several of these imprecisely characterized entries (some of which had formerly been described to be detected by western blot).

Results

MS_HistoneDB, a resource containing unique and non-redundant histones

Our initial aim with this work was to generate an exhaustive and non-redundant resource that would facilitate histone analysis by MS. We identified and collected all the information available on human and mouse histones from the public databases of NCBI, Ensembl and UniProtKB (Table 1). This work was aided by the recent release of an updated version of the Histone Database, named “HistoneDB 2.0—with variants” [37]. This database contains 38,664 entries from 1624 species, with 761 and 1039 entries for mouse and human, respectively. In addition, several histones were also considered based on published articles [41,42,43,44,45].

Table 1 Histone entries in various publicly accessible databases
Full size table

The dataflow is presented in Fig. 1a. This curating process is exemplified with human H2A.Z on Fig. 1b. A total of 23 entries were collected from Swiss-Prot, Uniprot-trEMBL and NCBI databases. Fourteen were duplicated and removed to obtain nine unique entries, which were annotated as H2A.Z.1 and eight spliced isoforms of H2A.Z.2 using the release R86 of the Ensembl database (Fig. 1c, d). In summary, the following rules were applied. First, each entry is protein-centric and therefore defined by the final product, a unique mature protein. Second, it must be associated with gene, transcript and protein accession numbers in NCBI and/or Ensembl, unless published data document its existence. Third, histone names are not always consistent within the existing public databases. Some were renamed following the Talbert et al. nomenclature and in agreement with the HistoneDB 2.0 resource as detailed in the following sections [36, 37]. The final list of histone entries is presented as a phylogenetic tree in Fig. 2 and in Tables 2 and 3 for mouse and human species, respectively. We did not provide here an extensive review of the functional roles of each histone variant, which are already available elsewhere [4, 21, 23, 36, 46].

Fig. 1
figure 1

Methodology used to create a manually curated MS_HistoneDB resource. a Representation of the dataflow used to generate the MS_HistoneDB resource. b Example of the dataflow for human H2A.Z histone variants. The number of entries at each step is indicated. c Graphical representation of the exons of human H2A.Z.2 spliced isoforms. d Sequence alignment of human H2A.Z variant isoforms

Full size image
Fig. 2
figure 2

Phylogenetic trees for mouse and human histone entries in MS_HistoneDB. Please note that for clarity, some putative spliced isoforms of canonical histones were not included, as well as other very short spliced isoforms for some histone variants. The full lists are presented in Tables 2 and 3. Gene names are indicated in italic to identify histone isoforms grouped under a generic term (see Tables 2 and 3; Additional files 4, 5). Black dots highlight testis-specific variants

Full size image
Table 2 Manually curated list of mouse histones
Full size table
Table 3 Manually curated list of human histones
Full size table

Canonical histones

Canonical histones constitute the bulk of the proteins that organize DNA into chromatin. They are synthesized and incorporated into chromatin during replication [41]. Their expression is carefully regulated to provide enough proteins to be loaded onto newly synthesized DNA while preventing the accumulation of free histones [47, 48]. For this reason, they are denominated “replication-dependent” and their mRNA adopts a unique organization (for review, see [49]). They are the only RNA polymerase II transcripts which are not polyadenylated but instead possess a 3′ stem-loop, formed during the maturation of their mRNA and which is essential for their regulation [49]. However, polyadenylation events of replication-dependent histone mRNA have recently been identified in terminally differentiated cells and suggested to provide a replacement pool of canonical histones [50].

H2A and H2B canonical histones have minor sequence variations, and it is not clear yet whether these have a functional significance [51]. MS analysis can differentiate between these isoforms and their denomination had to be adapted for proteomic analysis. Here, we propose that canonical H2A and H2B isoforms can be regrouped under the generic term “canonical H2A” or “canonical H2B”, complemented by the gene name of each isoform (Tables 2, 3).

Histone variants are mostly replication-independent

In contrast to canonical histones, almost all histone variants are synthesized independently of the cell cycle and named “replication-independent” [49]. Their mRNA is polyadenylated and these histones are incorporated into chromatin at any time of the cell cycle. Two exceptions are the testis-specific (TS) histone variants TS H2A and TS H2B, which possess a 3′ stem-loop in their mRNA. For this reason, they have been classified as replication-dependent [49] even if expressed in differentiating germ cells which replicate their DNA only once before meiosis.

Spliced and putative isoforms

More than 40 spliced isoforms for all mouse and human histones are present in the Ensembl database. However, this information, mainly based on transcriptional data, remains questionable; notably whether the corresponding proteins are expressed and incorporated into chromatin is uncertain. Some spliced isoforms correspond to very short isoforms that lack the globular domain and are probably, if expressed, non-functional (mouse: cenH3-CENPA.s2, cenH3-CENPA.s3, cenH3-CENPA.s4, cenH3-CENPA.s5, OO H1.8.s.4, OO H1.8.s.5, H2A.Z.1.s3, H2A.Z.1.s4, H2A.Z.2.s2; human: H2A.J.s3, H2A.Z.2.s7, H2A.Z.2.s8, canonical H2B.s2, cenH3-CENPA.s2). Even though their expression remains highly uncertain, they have been included in MS_HistoneDB for their identification by MS to be possible. Observing the presence of a shorter non-functional sequence at the expense of the full-length histone would indeed constitute interesting information. Following the same rationale, several putative isoforms or pseudogenes have been included in this resource (Tables 2, 3). Their detection by MS will constitute an indispensable step to confirm the expression of their protein form.

H1 histones

H1 histones (or linker histones) are different from core histones with respect to their structure, function and evolution. Therefore, it is not possible to single out one of its isoforms as canonical. H1 variants are known to encompass isoforms named H1.0–H1.10. H1.1–H1.5 from histone gene cluster 1 and orphan genes H1.0 (H1°) and H1.10 are usually referred to as somatic variants [36]. The linker variant H1.0 has been described to be involved in cell differentiation (for review [2, 52]). H1.10 has been identified in human and plays an essential role for mitotic progression [53]. H1 variants also include the TS proteins TS H1.6 [54], TS H1.7 [55, 56], TS H1.9 [57, 58] and the oocyte-specific OO H1.8 variant [59, 60]. Finally, a new mouse entry, H1.11, was identified here while performing an in silico search using sequence alignments.

H2A variants

H2A variants comprise H2A.X, H2A.Z, macro-H2A and a number of TS variants, TS H2A, H2A.L/H2A.P and H2A.B.

Only one H2A.X protein has been described; this variant is involved in double-strand break repair, genome stability and chromatin remodelling and silencing in male meiosis [61,62,63,64]. H2A.Z is involved in transcription regulation and is encoded by two different genes, H2afz and H2afv [43]. In mouse, four putative spliced isoforms may be expected in addition to the two original sequences, while in human eight H2A.Z.2 isoforms have been suggested, of which two have been demonstrated to be stable at the protein level [65]. The specific functional roles of these isoforms are not well understood yet, but in some specific tissues, such as in the brain, some H2A.Z spliced isoforms could provide context-specific signalling information [65].

Macro-H2A is the largest histone variant with a long C-terminal domain [66]. This histone variant is associated with transcription repression, although recent evidence suggests that in some conditions it may also promote transcription (reviewed in [51]). Macro-H2A is known to be encoded by two or three different genes, for human and mouse, respectively, some of which are differentially spliced. These variable forms allow differential binding of NAD [67].

Finally, many H2A variants are specifically expressed in the testis. First, TS H2A.1 was originally identified in 1982 in the testis, where it plays an important role and was later detected in the ovary [41, 68,69,70,71].

Fourteen other mouse TS H2A variants have been grouped into three main classes, H2A.L, H2A.B and H2A.P (Fig. 3). This class also regroups human variants, with two H2A.B and one H2A.P proteins (Fig. 3). They are involved in transcription regulation and the final chromatin reorganization during post-meiotic differentiation of sperm cells [26, 38, 39]. The mouse variants have been described by different research groups [38, 39], and a denomination used here follows previous publications [36, 37]. When potential protein products of different genes have only minor sequence variations and no functional difference has been characterized, we grouped them under the name with the same number suffix (e.g. H2A.L.1); however, the gene name is provided in the name of the entry as the second qualifier. Future studies might warrant splitting of such groups of proteins if functional differences between the members are detected. Currently, H2A.B.2 and H2A.B.3 are proposed to be numbered following their gene name, i.e. H2AFB2 and H2AFB3, respectively. In 2007, new TS H2A variants were identified and named H2AL1 and H2AL2. [38]. A few years later, these histones were independently identified and named H2A.Lap2 and H2A.Lap3, respectively [39]. This latter work also reported the identification and functional characterization of a third member baptized H2A.Lap1 which falls into H2A.B group and is proposed to be regrouped with the highly similar protein H2A.B.3. H2A.L.3 was originally identified by S. Khochbin’s group [38] and is the same as H2A.Lap4, also identified by D. Tremethick’s group [39]. However, it forms a separate phylogenetic clade in placental mammals and is named H2A.P here according to [36].

Fig. 3
figure 3

Phylogenetic tree of the mouse and human H2A.L, H2A.B and H2A.P histone variants

Full size image

H2B variants

Variants TS H2B.1, H2B.L and H2B.W were first identified as TS. In the testes, these proteins are involved in the chromatin-to-protamine transition [38, 69, 72, 73]. Then, TS H2B.1 and TS H2A.1 were also identified in human oocytes, where they favour the generation of induced pluripotent cells [70, 71]. In human, some genes (e.g. H2BFM, H2BFS) still await characterization and have been denoted as putative variants in this work (Table 3).

H3 variants

H3 has several isoforms: H3.1 and H3.2 are replication-dependent; H3.3 is considered to be a replication-independent histone variant, while TS H3.4 and H3.5 are TS [74, 75]. Several new isoforms of H3.3 were included in the database developed here along with two other human H3 histone variants, H3.X and H3.Y [44].

CenH3/CENPA is a well-known centromeric H3 variant with many spliced isoforms. Its name has been the subject of heated discussion, which is out of the scope of our work [36, 76, 77]. We therefore propose to use both names, cenH3-CENPA, until a consensus has been reached by the community.

Generation of MS-based databases

De novo MS data interpretation methods are naive and do not rely on pre-existing databases. However, MS data acquired on histones are generally matched to a database containing all the protein sequences that could theoretically be found in the sample. Using this approach, a given histone protein cannot be identified if its sequence is not present in the database explored by the MS/MS data interpretation software. We used MS_HistoneDB to create a new search space dedicated to the analysis of histones. Basically, mouse or human non-redundant and well-annotated Swiss-Prot FASTA files were cleared of their histone sequences and then repopulated using MS_HistoneDB. This resource is included as Additional files 1 and 2, providing resources to study histones in mouse and human samples, respectively.

Identification of new histones in mouse

About 30% of the proteins in MS_HistoneDB have imprecise protein annotations in UniProtKB and are presented in Tables 4 and 5. These tables regroup histones that are annotated in the UniProtKB and NCBI databases as “inferred from homology” or “predicted”. Even though a certain number of these histones have already been described in publications, which provide clear evidence of their existence at mRNA and protein levels, they may not have been identified by MS yet. This could explain their poor annotation status in UniProtKB.

Table 4 Mouse histone variants with poor annotation status in the UniProtKB database
Full size table
Table 5 Human histone variants with poor annotation status in the UniProtKB database
Full size table

At the RNA level, almost all histone variants have been detected in the testis [38,39,40, 55, 57, 68, 78,79,80,81]. Moreover, the expression profile of the mouse H2A.L.1 isoforms which are described in this study has been explored. RNA-seq data from nine mouse tissues have been obtained from a recently published dataset [82]. Expression data were available for 7 out of the 8 H2A.L.1 mouse histone entries and confirm that all of them are mainly expressed in the testis, similarly to H2A.L.2 and H2A.P (Fig. 4a) [38, 39]. Gene expression profiles during spermatogenesis have been obtained from Ref. [83]. It also confirms that H2A.L.1 isoforms are expressed in the post-meiotic stage in spermatids, similarly to H2A.L.2 and H2A.P (Figs. 3c, 4b) [38, 39].

Fig. 4
figure 4

Mouse H2A.L, H2A.B and H2A.P isoforms are expressed in the post-meiotic stages of spermatogenesis. a Expression profile of a selection of H2A.L, H2A.B and H2A.P genes across mouse tissues. All of them are testis-specific. Data were extracted using gene names provided in Table 2 and Additional file 4 from a dataset downloaded from Expression Atlas [87] and published in Ref. [82]. b Expression profiles of a selection of H2A.L, H2A.B and H2A.P genes during mouse spermatogenesis, revealing a maximum expression in the post-meiotic stage. Data have been obtained from Ref. [83]. Lepto leptotene, Zygo zygotene. c Coomassie-stained SDS-PAGE gel loaded with histones extracted from whole testis and elongating spermatids. d Peptides of the mouse H2A.L histone variants identified by mass spectrometry analysis (identified peptides are highlighted in red boxes)

Full size image

We next decided to test whether MS_HistoneDB would allow the identification by MS of histone entries with imprecise protein annotation using mouse testis. Histones were purified from whole testis or from elongating spermatids (Fig. 4c). Mass spectrometry analysis combined with MS_HistoneDB allowed identification of nine of these poorly annotated proteins (Table 4; Additional file 3). Each newly MS-identified variant was detected by 1–10 specific peptide sequences. The current guidelines for the identification of previously undetected human proteins (“missing proteins”) require the identification of two different peptide sequences of at least nine amino acids in length [84]. To stringently apply the same rules to validate new histone variants would be demanding, given the very high level of sequence homology between some variants. However, out of the nine histone variants detected here for the first time at the protein level by MS-based approaches, six were identified with at least two non-overlapping peptides of length ≥9 amino acids. Almost all the newly identified variants are TS. This analysis thus confirmed the existence of H2A.L.1 encoded by H2al1b in mouse testes (Fig. 4d). In addition, this analysis confirmed the existence of the histone variant H3.3 encoded by the gene Gm10257, for which a specific peptide has been identified, even if its corresponding NCBI protein record has been recently removed (XP_003084990.1).

Other variants may not be detectable by MS in our analysis. For example, a trypsin digestion does not generate any peptides distinguishing mouse H2al1k protein from highly homologous H2A.L.1 variants. Its specific detection would require a more extensive analytical work, e.g. using an alternative protease for protein sample processing, which is beyond the scope of the current work.

Conclusions

MS is a powerful technique to identify histones, their variants and their post-translational modifications but relies on databases with contradictory naming and excessive redundancy. Here we exhaustively collected histone sequences for mouse and human and used manual curation to establish a protein-centric list, MS_HistoneDB, dedicated to the proteomic study of mouse and human histones. Histone variants whose protein status is uncertain in UniProtKB and NCBI but whose protein existence has been established by experimental evidence described in the literature have been included. This work confirmed the expression of isoforms of previously identified TS histone variants and allowed the detection of one H3.3 isoform whose status was so uncertain that its record had been deleted from the NCBI protein database. We hope that this resource will facilitate the study of histone variants, especially by MS, and their functional roles in physiological and pathological contexts.

Methods

Phylogenetic tree representation

Multiple sequence alignments of mouse and human histones were performed using Clustal Omega [85]. Tree data were downloaded in aln format and displayed with iTOLv3 [86].

Analysis of RNA-seq data

Tissue-specific expression data were obtained from Huntley et al. [82] through the Expression Atlas Repository [87]. RNA-seq data at different stages of spermatogenesis were obtained from da Cruz et al. [83]. Data were imported and treated in R using the pheatmap library (https://CRAN.R-project.org/package=pheatmap).

Purification of histones from mouse testis

Histones were extracted from two types of biological samples, namely whole testis and elongated spermatids, to maximize the number of histone variants identified at specific maturation stages of male germ cells. Pure fractions of spermatid nuclei were obtained by sonicating mouse testes, as previously described [38]. Histones were isolated from testis cells and spermatids using sulfuric acid [38] or saline extraction [88]. They were then separated by SDS-PAGE, and proteins were visualized by Coomassie staining.

Sample preparation and analysis by MS

Histones were reduced and alkylated as described previously [89]. Histones were either derivatized with propionic anhydride before and after in-gel trypsin digestion [90], or only submitted to trypsin digestion [89]. The dried extracted peptides were resuspended in 2.5% acetonitrile and 0.05% trifluoroacetic acid and analysed via online nano-LC–MS/MS using an Ultimate 3000 LC system coupled to an LTQ-Orbitrap instrument (CID fragmentation mode) or a Q Exactive Plus instrument (HCD fragmentation mode) (Thermo Fisher Scientific).

Protein sequence database search and manual verification

MS RAW files produced by LC–MS/MS analysis of proteolyzed histones were processed as follows. All MS/MS spectra were submitted to the Mascot program (version 2.5.1) for searching against the MS_HistoneDB protein sequence database. The parse rules for MS_HistoneDB Fasta files in Mascot are using the accession rule >\([^]*\) and the description rule \(.*\). In addition, the taxonomy and sequence report sources are indicated as “Swiss-Prot FASTA” and “FASTA file”, respectively. No taxonomy was specified when using MS_HistoneDB with Mascot Daemon.

Classical histone modifications were included in the variable modifications: N-terminal protein acetylation; Lys acetylation; and Lys and Arg mono- or di-methylation. For all Mascot searches, the tolerance on mass measurement was set to 5 ppm for peptides and to 0.6 or 0.025 Da for fragment ions when considering LTQ-Orbitrap or Q Exactive acquisitions, respectively. Up to four tryptic missed cleavages were allowed for samples that were not propionylated in vitro, as trypsin does not cleave acetylated lysine, a frequent modification. The enzyme ArgC and up to two missed cleavages were specified for the interpretation of data acquired on propionylated samples. All MS/MS spectra leading to the identification of tryptic peptides specific to newly described variants were carefully manually examined: all major intensity fragment peaks had to be interpreted in terms of y/b ions; a continuous sequence of at least five amino acids had to be read in all cases for validation. Proteomics data are available from ProteomeXchange (PXD005489).