|Home | About | Journals | Submit | Contact Us | Français|
The incorporation of histone H3 variants has been implicated in the epigenetic memory of cellular state. Using genome editing with zinc finger nucleases to tag endogenous H3.3, we report genome-wide profiles of H3 variants in mammalian embryonic stem (ES) cells and neuronal precursor cells. Genome-wide patterns of H3.3 are dependent on amino acid sequence, and change with cellular differentiation at developmentally regulated loci. The H3.3 chaperone Hira is required for H3.3 enrichment at active and repressed genes. Strikingly, Hira is not essential for localization of H3.3 at telomeres and many transcription factor binding sites. Immunoaffinity purification and mass spectrometry reveal that the proteins Atrx and Daxx associate with H3.3 in a Hira-independent manner. Atrx is required for Hira-independent localization of H3.3 at telomeres, and for the repression of telomeric RNA. Our data demonstrate that multiple and distinct factors are responsible for H3.3 localization at specific genomic locations in mammalian cells.
Genetic and biochemical evidence have recently converged to connect epigenetic mechanisms at the level of chromatin (Bernstein et al., 2007; Goldberg et al., 2007; Henikoff, 2008). In addition to nucleosome remodeling and covalent modifications, eukaryotic cells generate variation in chromatin by the introduction of variant histone proteins (Henikoff, 2008). Mammalian cells express three major types of non-centromeric histone H3 variants, H3.1, H3.2, and H3.3 (Hake and Allis, 2006; Hake et al., 2006). Although histone H3.3 differs from H3.2 and H3.1 at only 4 or 5 amino acids (Figure S1A), H3.3 is specifically enriched at transcriptionally active genes and regulatory elements in non-pluripotent cells (Ahmad and Henikoff, 2002; Jin et al., 2009; Mito et al., 2005, 2007).
Histone H3.3 is incorporated into chromatin in both a replication-coupled (RC) and replication-independent (RI) manner, while the incorporation of H3.2 is coupled to replication (Ahmad and Henikoff, 2002; De Koning et al., 2007). The histone chaperone CAF-1 is found in a complex with H3.1, and mediates RC nucleosome assembly (Smith and Stillman, 1989; Tagami et al., 2004). In contrast, the histone chaperone Hira has been found in a complex with H3.3, and mediates RI nucleosome assembly (Ray-Gallet et al., 2002; Tagami et al., 2004).
Hira has been implicated in H3.3-specific deposition and chromatin assembly (Loyola and Almouzni, 2007). Although Hira is required for chromatin assembly and H3.3 deposition in the male pronucleus of Drosophila, Hira is not required for global H3.3 deposition in Drosophila embryos or adult cells, suggesting that alternate pathways may mediate H3.3 nucleosome assembly (Bonnefoy et al., 2007; Loppin et al., 2005). Indeed, the chromatin remodeling factor CHD1 was shown to physically associate with Hira, and has been suggested to work with Hira to mediate H3.3 incorporation into chromatin in Drosophila (Konev et al., 2007).
In Drosophila, both Hira and H3.3 are required for fertility and for transcriptional regulation of specific genes, but not for developmental viability (Bonnefoy et al., 2007; Hodl and Basler, 2009; Nakayama et al., 2007; Sakai et al., 2009). However, in mice, targeted mutagenesis of Hira results in a more severe phenotype, with gastrulation defects leading to early embryonic lethality (Roberts et al., 2002). Given the conserved association between H3.3 and active chromatin, H3.3 has been speculated to play an important role in mammalian ES cells (Creyghton et al., 2008; Gaspar-Maia et al., 2009). However, no genome-wide studies in pluripotent cells distinguish between H3 variants, nor do they examine the genome-wide role of Hira or other histone chaperones in specifying H3.3 localization at specific genomic regions.
Here we report genome-wide profiles of histone H3 variant localization in mammalian ES cells and neuronal precursor cells (NPCs), and we establish the dependence and independence of these patterns on Hira. We find that Hira is required for genome-wide H3.3 enrichment at active and repressed genes in ES cells. Surprisingly, H3.3 enrichment at many transcription factor binding sites (TFBS) and telomeres is Hira-independent. To identify factors that might mediate specific Hira-independent localization of H3.3, we use immunoprecipitation and mass spectrometry. We identify Atrx and Daxx as proteins that specifically associate with H3.3 in both pluripotent and non-pluripotent cells, both in the presence and in the absence of Hira. Unlike Hira, Atrx is not required for H3.3 localization at genes or TFBS. However, Atrx is specifically required for enrichment of H3.3 at telomeres in ES cells, and for the repression of telomeric RNA.
To distinguish H3 variants in our study without altering endogenous levels of H3 variant expression, we used designed zinc finger nucleases (ZFNs) (Carroll, 2008) both in gene addition (Moehle et al., 2007) and correction (Urnov et al., 2005) modes to engineer a panel of heterozygous ES lines carrying one allele of wild-type H3.3B, and another allele of H3.3B with a C-terminal EYFP tag (H3.3-EYFP), HA tag (H3.3-HA), or H3.3B with an epitope tag and simultaneous mutation towards H3.2 or H3.1 (H3.2-EYFP, H3.1-EYFP, H3.2-HA, H3.1S31-HA) (Figure S1). All heterozygously tagged ES cells retain three “wild-type” copies of H3.3 genes, including one copy of unmodified H3.3B (Figure S1G), and two copies of H3.3A. Western blots demonstrate equal protein expression in the epitope-tagged H3 variant ES lines (Figure S1H).
To assess the genome-wide localization of histone H3 variants at high resolution, we used chromatin immunoprecipitation followed by massively parallel sequencing (ChIP-seq) (Barski et al., 2007; Mikkelsen et al., 2007). We found 10,099 total genic and intergenic regions highly enriched for H3.3 in crosslinking ChIP-seq of ES cells, in contrast to 1,442 regions enriched for H3.2, and we observe that gene-rich regions of the genome show greater enrichment of H3.3 than H3.2 (Figure 1A, B). On a chromosomal scale map, H3.3 enrichment correlates with markers of transcription, including Ser-5 phosphorylated RNA polymerase II (RNAPII), H3K4me3, H3K36me3, and H3K4me1 (Figure 1A). Unbiased clustering analyses confirm the genome-wide correlation of H3.3 with active histone modifications, particularly H3K4me1 (Table S1). Despite different epitope tags, we found extremely similar profiles of H3.3-HA and H3.3-EYFP (Figure 1A, Table S1, Figure S2A–B). Genome-wide analysis also reveals specific enrichment of H3.3 at previously identified genic and intergenic ES TFBS (Chen et al., 2008), as well as peaks of H3.3 in specific unannotated intergenic regions (Figure 1B).
To analyze H3 variant enrichment in different classes of repeats, we determined the relative enrichment of specific repeat sequences from our ChIP-seq experiments and compared it to that from control input DNA. H3.3 was reproducibly depleted in satellite repeat sequences and in Y-chromosomal repeat DNA (Figure 1C). We found the most significant enrichment of H3.3 in the (TTAGGG)n repeat that is the conserved telomeric DNA sequence among vertebrates (Meyne et al., 1989) (Figure 1C). This telomeric enrichment of H3.3 is consistent with telomeric foci of incorporation visible on the largely heterochromatic Y-chromosome (Figure S1J–O), and with a recent report of telomeric localization of transfected epitope-tagged H3.3 in ES cells (Wong et al., 2009). Mutation of H3.3B to H3.2 or H3.1S31 alters genome-wide patterns of H3.3 enrichment, demonstrating that the amino acid sequence of an endogenous H3.3 gene determines its genomic localization in mammalian cells.
As there has been some disagreement regarding the patterns of H3.3 at active versus repressed genes in vertebrates (Jin and Felsenfeld, 2006; Jin et al., 2009; Sutcliffe et al., 2009; Tamura et al., 2009), we used ChIP-seq to address the genome-wide patterns of H3 variant enrichment around gene transcription start sites (TSS) in mouse ES cells. We find that H3.3 is not exclusively a marker of transcriptionally active genes in ES cells and NPCs.
Previous studies show that the majority of genes with high CpG content promoters (HCP genes) in both ES cells and differentiated cells are marked by histone H3K4me3 and the presence of RNAPII, regardless of whether the gene is active or repressed (Barski et al., 2007; Guenther et al., 2007; Mikkelsen et al., 2007). When we divide HCP genes into low, medium, and high expression in ES cells (see Methods), we find that H3.3, like H3K4me3 and RNAPII, is enriched around the TSS of both active and repressed genes (Figure 2A). In both native and crosslinking ChIP-seq, we find H3.3 less enriched at the TSS itself (Figure 2A, D, Figure S2A–D). This depletion of the −1 nucleosome at the TSS of active and inactive genes (Schones et al., 2008) has been attributed both to the instability of nucleosomes containing H3.3 and H2A.Z, and to the presence of specific sequences at CpG promoters that directly reduce nucleosome stability (Jin et al., 2009; Ramirez-Carrozzi et al., 2009).
More than one-fifth of HCP genes in ES cells carry both H3K4me3 and H3K27me3 in their promoters, and these transcriptionally repressed ‘bivalent’ genes are proposed to represent genes that are poised for activation following cell differentiation (Bernstein et al., 2006; Mikkelsen et al., 2007). When we analyze the pattern of H3 variants at bivalent TSS by ChIP-seq, we find that H3.3 is enriched around the TSS of bivalent genes in ES cells, while mutation of H3.3 towards H3.2 or H3.1 abolishes this enrichment (Figure 2G, S3).
Although H3.3 is incorporated around the TSS of both active and repressed genes in ES cells, H3.3 is enriched in the body of active genes, but not that of repressed genes (Figure 2A, G, H, S2A–D). Mononucleosome resolution analysis indicates that H3.3 is incorporated into the +1 nucleosomes in both active and repressed genes, but up to +3 nucleosomes and further into the coding regions of active genes (Figure S2Q–W). The level of H3.3 in gene bodies is correlated with gene expression, particularly at highly expressed genes (Spearman’s rank correlation coefficient ρ = 0.54, p < 2.2e-16; see Figure 2A, S2). Upon mutation of H3.3 to H3.2 or H3.1S31, H3.3 specific patterns of enrichment around the TSS and gene body are lost, generating patterns similar to general H3 (Figure 2B, D).
In accordance with previous studies (Henikoff et al., 2009; Jin et al., 2009; Mito et al., 2005), H3.3 enrichment often extends beyond the gene body and past the transcriptional end site (TES) at highly expressed genes in ES cells (Figure 2A, H) and in differentiated NPCs (Figure S2CC). At highly expressed genes such as beta-actin, H3.3 enrichment increases immediately after the TES, and peaks at approximately +700 bp ± 200 bp after the TES (Figure 2H, S2A–D). We observe a similar pattern of H3.3 localization in the bodies and after the TES of transcribed non-coding RNA (Guttman et al., 2009) (Figure 2I). Interestingly, the distribution of post-TES peaks of serine-5 phosphorylated RNAPII are closely co-localized with peaks of H3.3 at active genes, with the phosphorylated RNAPII peak slightly downstream (~200 bp) from the H3.3 peak. We did not observe any significant post-TES enrichment of histone modifications that are associated with active genes, such as H3K4me3, H3K4me1, and H3K36me3 (Figure 2E, F, Figure S2). Post-TES enrichment of H3.3 is also dependent on amino acid sequence (Figure 2B, D, Figure S2). From these analyses, we conclude that H3.3 is localized around the TSS of both active and repressed HCP genes, and both H3.3 and phosphorylated RNAPII are significantly enriched beyond the TES of highly expressed genes in undifferentiated and differentiated mammalian cells.
To determine how genome-wide patterns of H3.3 change with cell differentiation, we differentiated both H3.3-HA and H3.2-HA ES cells to NPCs (Figure S3A) (Conti et al., 2005). In ES cells, H3.3 is enriched in the bodies of expressed pluripotency genes such as Esrrb, Nanog, and Oct4, correlated with H3K36me3 (Figure 3A, Table S2, S3B). Upon differentiation of ES cells to NPCs, the expression of most pluripotency genes is lost (Conti et al., 2005). Accordingly, the enrichment of H3.3 and H3K36me3 in the bodies of pluripotency genes largely disappears upon cell differentiation (Figure 3A, S3B).
Following differentiation of ES cells to NPCs, the profile of H3.3 changes with resolution of bivalent domains. In bivalent ES genes that resolve to H3K4me3 and become transcriptionally active in NPCs, H3.3 is maintained around the TSS and also incorporated into the gene body, in correlation with H3K36me3 and H3K4me1 (Figure 3B, S3C). For example, upon differentiation to NPCs, H3.3 extends into the gene body of the active epidermal growth factor receptor gene Egfr, correlating with H3K36me3 and H3K4me1 (Figure 3B). In contrast, for bivalent genes that remain transcriptionally repressed and resolve to either H3K27me3 or no mark in NPCs, H3.3 enrichment is reduced at the TSS upon differentiation (Figure S3C). As expected, H3.2 remains unenriched at the TSS in both ES cells and NPCs (Figure S3C).
While the pattern of H3.3 changes at cell-type-specific genes, housekeeping genes that remain highly expressed through differentiation retain similar enriched patterns of H3.3 incorporation. For example, H3.3 remains enriched around the TSS and within the gene body of the housekeeping gene lactate dehydrogenase A Ldha (Figure 3C, S3A). In both ES cells and NPCs, we find the greatest enrichment of H3.3 at highly expressed metabolic and housekeeping genes (Figure S3D). Our data show that the overall relationship between H3.3 incorporation and gene expression is retained in ES and NPCs, with H3.3 and H3K4 methylation marking the TSS of active and poised HCP genes, and H3.3 enriched in the body and TES of active HCP genes. At cell-type specific genes, H3.3 localization changes with cellular state.
Genome-scale studies have demonstrated that in addition to genes (Mito et al., 2005), H3.3 replacement marks the boundaries of regulatory elements (Jin et al., 2009; Mito et al., 2007). 36% (3,627) of H3.3 enriched regions in mouse ES cells are located outside of known annotated genes (Figure 1B, Figure 3D–F). We therefore compared our genome-wide distributions for histone H3 variants in ES cells and NPCs with established genome-wide maps for 13 distinct sequence-specific TFs in ES cells (Chen et al., 2008). We find that H3.3 is enriched genome-wide at genic and intergenic TFBS for all 13 characterized TFs in ES cells (Figure 3D, Figure S3E). Remarkably, 3,878 (38%) of the overall regions enriched for H3.3 in ES cells (Figure 1B) correspond to previously identified TFBS (Chen et al., 2008). Again, peaks of H3.3 at TFBS are dependent on H3.3 amino acid sequence (Figure 3D–G, S3E).
Many of the specific peaks of H3.3 at TFBS are cell-type specific, especially those located in intergenic regions (Figure 1B, 3D–G, S3B, data not shown). We find that H3.3 incorporation at TFBS bound by multiple TFs increases significantly with the number of bound TFs in ES cells, but these same elements do not show increased H3.3 in NPCs (Figure 3F, S3F). Genomic locations bound by more than four TFs are called multiple transcription factor-binding loci (MTL), and a subset of MTL bound by Oct4-Sox2-Nanog have been shown to serve as ES cell-specific enhanceosomes (Chen et al., 2008). For example, an intergenic region near the AP-2 transcription factor gene Tcfap2c is a target for multiple transcription factors in ES cells (Chen et al., 2008). We find H3.3 enriched at this region specifically in ES cells, with enrichment lost in NPCs (Figure 3F).
Following differentiation to NPCs, genome-wide enrichment of H3.3 at ES TFBS is reduced, but not eliminated (Figure 3E, S3E). At some specific TFBS (Figure 3F), H3.3 and H3K4me1 enrichment is lost upon differentiation to NPCs. However, at other TFBS, such as those 5′ and 3′ of Nanog, H3.3 and H3K4me1 enrichment is reduced, but partially maintained (Figure S3B). H3K4me1 and H3.3 have been separately shown to significantly co-localize with TFBS and other regulatory elements (Heintzman et al., 2007; Mito et al., 2007; Wang et al., 2008), but the patterns of H3.3 and H3K4me1 have never been directly compared. In our analysis, 26% of H3K4me1 peaks overlap with H3.3 peaks, and peaks of H3.3 are 5 times more likely to be associated with peaks of H3K4me1 than H3K4me3 (Table S1, p 0.00001). In addition to ES cell specific intergenic peaks of H3.3 and H3K4me1 (Figure 3F), we also found 724 intergenic peaks of H3.3 and H3K4me1 that are specific to differentiated NPCs and that do not correspond to any known TFBS (Figure 3G, data not shown). It is possible that these represent uncharacterized cell-type specific enhancers. Overall, our data demonstrate that the genome-wide localization of H3.3 changes with cell differentiation at cell-type specific genes and regulatory elements.
After establishing genome-wide patterns of H3.3 in undifferentiated ES cells and differentiated NPCs, we sought to determine whether these patterns were dependent on the H3.3 chaperone Hira. We therefore used ZFNs to tag one allele of the endogenous H3.3B gene in Hira −/− ES cells (Meshorer et al., 2006; Roberts et al., 2002) with a C-terminal EYFP tag (Figure 4A, Figure S4A), and compared the genome-wide localization of H3.3-EYFP in the presence and absence of Hira.
Using native ChIP-seq, we find that Hira is required for genome-wide H3.3 enrichment at active and repressed HCP genes (Figure 4B–G). In the absence of Hira, the pattern of H3.3 around active and repressed genes resembles H3.2, with no enrichment surrounding the TSS, and depletion of H3.3 at the TSS itself (Figure 4D, E). For example, in Hira −/− ES cells, the enrichment of H3.3 around the TSS of the repressed bivalent gene Pparg is abolished (Figure 4F). At highly expressed housekeeping genes such as the ribosomal protein-coding gene Rps19, the gene body and post-TES enrichment of H3.3 is similarly decreased in the absence of Hira (Figure 4G). Our data demonstrate that the vast majority of H3.3 enrichment at both active and repressed genes in ES cells is Hira-dependent.
One possible explanation for the loss of H3.3 localization at active and repressed genes in Hira −/− ES cells is a significant alteration in patterns of gene expression. Remarkably, microarray analysis reveals that global patterns of ES cell gene expression are maintained in Hira −/− ES cells (Figure S4B-D). Moreover, genome-wide patterns of H3K36me3 are extremely similar in wild-type and Hira −/− ES cells (Figure S4E-H). This result is similar to the recent observation that global patterns of ES cell gene expression are maintained following knock-down of CHD1 (Gaspar-Maia et al., 2009). We conclude that genome-wide enrichment of H3.3 at active and repressed genes is Hira-dependent, but is not required for maintenance of the undifferentiated ES cell transcriptome.
We next examined the Hira-dependence of H3.3 enrichment at genic and intergenic TFBS. Surprisingly, global H3.3 profiles are largely maintained at ES cell TFBS in Hira −/− ES cells (Figure 5A, S5, Table S3), indicating that H3.3 enrichment at most known regulatory elements is Hira-independent. For example, H3.3 enrichment at an intergenic MTL approximately 30 kb from the Mmp17 gene is maintained in Hira −/− ES cells (Figure 5B), as is H3.3 enrichment at an intergenic MTL near the repressed Foxi3 gene (Figure 5C). Indeed, levels of H3.3 are increased at some TFBS in Hira −/− ES cells (Figure S5A). However, levels of H3.3 are also reduced at other TFBS in Hira −/− ES cells (Table S3, Figure S5B). Of all previously identified ES TFBS (Chen et al., 2008), 34% show greater than 2-fold more H3.3 tags in Hira +/+ than in Hira −/−, while 12% show greater than 2-fold more H3.3 tags in Hira −/−, indicating that targeting of H3.3 to the majority (54%) of known ES TFBS is Hira-independent (Table S3, Figure S5). Global profiles of H3K4me1 are very similar in wild-type and Hira −/− ES cells (Figure S4E–F, S5D), indicating that Hira is also not required to maintain the localization of H3K4me1. Overall, these data demonstrate that Hira is not essential for the localization of H3.3 at many TFBS in mammalian ES cells.
To identify candidates that might mediate Hira-independent localization of H3.3, we used immunoaffinity purification and mass spectrometry (Cristea et al., 2005) (Figure 6A, S6). We found many interacting proteins common to all H3 variants in wild-type and Hira −/− ES cells, including core histones and previously described members common to both RC and RI chromatin assembly complexes, such as Nasp, Asf1a, Asf1b, and Rbap48 (see Table S4). Notably, the previously described H3.3 chaperone Hira (Tagami et al., 2004) was identified specifically in proteins isolated with H3.3 (Figure 6A–B).
In addition to Hira, we also identified Atrx and Daxx as proteins that specifically associate with H3.3 (Figure 6A–B). Atrx is a member of the SNF2 family of chromatin remodeling factors (Picketts et al., 1996). Mutations of human ATRX give rise to the ATR-X syndrome, a disorder characterized by a form of X-linked mental retardation that is frequently associated with alpha thalassemia (Gibbons et al., 2008). ATRX co-exists in a chromatin-remodeling complex with the death domain-associated protein Daxx, and these proteins localize to heterochromatin and promyelocytic leukemia (PML) nuclear bodies in human and mouse cells (Tang et al., 2004; Xue et al., 2003). As Atrx and Daxx specifically associate with H3.3 in both wild-type and Hira −/− ES cells, we conclude that this association is Hira-independent.
To determine if the association between H3.3, Atrx, and Daxx was conserved in differentiated human cells, we isolated oligonucleosomes and chromatin-associated proteins from HeLa cells stably expressing FLAG-HA tagged H3.3 or H3.1 (Tagami et al., 2004) (Figure 6C–D). Following FLAG affinity purification, immunoblots of H3.3 and H3.1-associated proteins reveal that both Daxx and Atrx are specifically associated with H3.3 but not H3.1 oligonucleosomes in human cells (Figure 6E).
To determine if Atrx is required for H3.3 localization, we again used ZFNs to knock-in an epitope tag into the endogenous allele of H3.3B, generating heterozygous H3.3B/H3.3B-EYFP in Atrxflox and Atrxnull mouse ES cells (Garrick et al., 2006) (Figure 7A). We then used native ChIP-seq to generate genome-wide profiles of H3.3 in the presence and absence of Atrx. We find that Atrx is not required for H3.3 incorporation at active or repressed genes, or at regulatory elements (Figure S5A, S7A-E), as genome-wide profiles of H3.3 are similar at genes and TFBS in Atrxflox and Atrxnull ES cells. Strikingly, both ChIP-seq (Figure 7B) and cell imaging analysis (Figure 7C) demonstrate that Atrx, but not Hira, is specifically required for H3.3 enrichment at telomeres. In accordance with the requirement of Atrx for telomeric localization of H3.3, ChIP analysis shows that Atrx itself is physically associated with telomeres in Atrxflox ES cells (Figure 7D).
To investigate the functional consequences of Atrx deletion and the loss of Atrx-dependent telomeric enrichment of H3.3, we examined the chromatin state and transcriptional output of ES cell telomeres. The chromatin of ES and induced pluripotent cell telomeres has previously been shown to have lower levels of the heterochromatin marker H3K9me3 and increased transcription of telomeric repeat-containing RNA (TERRA) in comparison to differentiated cells (Marion et al., 2009). In particular, TERRA has recently been identified as a component of telomeric heterochromatin, and levels of TERRA have been shown to be regulated by chromatin modifying enzymes (Azzalin et al., 2007; Luke and Lingner, 2009; Schoeftner and Blasco, 2008). ChIP of H3K4me3 and H3K9me3 does not show a significant difference in telomeric enrichment between Atrxflox and Atrxnull ES cells (Figure S7F). However, northern blots from Atrxflox, Atrxnull, and Atrxflox ES cells 4 days after treatment with Cre reveal reproducible (~1.7-fold) upregulation of TERRA in the absence of Atrx (Figure 7E–G). Our data demonstrate that Atrx is required for Hira-independent localization of H3.3 at telomeres and for repression of TERRA.
In this study, we examine H3 variant localization in mammalian ES cells and differentiated NPCs. Genome-wide patterns of H3.3 are dependent on H3.3-specific amino acid sequence; the enrichment of H3.3 at cell-type specific genes and TFBS is dependent on cellular state. We describe three general categories of H3.3 enrichment in mammalian cells: 1) genes and other transcribed non-repetitive sequences, 2) TFBS, and 3) telomeres. Remarkably, we find that each of these general categories of H3.3 enrichment in ES cells is mediated by distinct mechanisms. As expected, Hira is required for genic enrichment of H3.3. Unexpectedly, localization of H3.3 at specific TFBS and telomeres is Hira-independent, and we have identified Atrx as required for H3.3 localization at telomeres. Our results demonstrate that distinct factors control H3.3 localization at specific genomic locations in mammalian cells.
We find that H3.3 is constitutively enriched around the TSS of active and repressed HCP genes in mammalian ES cells and NPCs, including the TSS of repressed bivalent genes in ES cells. Although a recent genome-wide study found that H3.3 is unenriched at the TSS of repressed genes in HeLa cells (Jin et al., 2009), these results are not necessarily in conflict with our findings. Low CpG content promoter (LCP) and HCP genes have been described to display distinct modes of regulation (Mikkelsen et al., 2007; Ramirez-Carrozzi et al., 2009; Saxonov et al., 2006). Most HCP genes show evidence of transcriptional initiation, assemble unstable nucleosomes, and do not require SWI/SNF nucleosome remodeling complexes for gene induction, while LCP genes assemble stable nucleosomes and require SWI/SNF (Guenther et al., 2007; Ramirez-Carrozzi et al., 2009). Less differentiated cells such as ES cells contain large numbers of HCP genes with characteristics of transcriptional initiation (Guenther et al., 2007; Mikkelsen et al., 2007). Indeed, nearly all (99%) of HCP genes are marked by H3K4me3 in mouse ES cells, whether they are transcriptionally active or repressed (Mikkelsen et al., 2007). In contrast to HCP genes, we do not observe any significant pattern of H3.3 enrichment at LCP genes in ES cells and NPCs (data not shown). Our results are therefore consistent with a model in which H3K4 methylation and H3.3 localization at HCP TSS are coupled to transcriptional initiation.
We find that enrichment of H3.3 in the gene body and after the TES is proportional to transcriptional activity. As with previous studies in Drosophila and human cells (Henikoff et al., 2009; Jin et al., 2009; Mito et al., 2005), we find peaks of H3.3 after the TES of highly active genes, and we observe that these peaks are closely paralleled by peaks of Ser-5 phosphorylated RNAPII itself. We demonstrate that chromatin-based “transcriptional punctuation” (Siegel et al., 2009; Talbert and Henikoff, 2009) by H3.3 and phosphorylated RNAPII marks the boundaries of highly expressed genes in both undifferentiated and differentiated mammalian cells, calling attention to a potentially more universal mechanism for histone variant utilization as a genomic “boundary marker.”
To our knowledge, our report is the first genome-wide study to compare chromatin in the presence and absence of a mammalian histone chaperone. We find that H3.3 enrichment at active and repressed genes is dependent on the histone chaperone Hira. Previous studies suggest that H3.3 deposition in actively transcribed gene bodies may be coupled to transcription, potentially mediated by factors associated with elongating polymerase (Daury et al., 2006; Janicki et al., 2004; Schwartz and Ahmad, 2005). Our data are consistent with Hira-dependent transcription-coupled deposition of H3.3 at transcribed non-repetitive sequences.
Intriguingly, we do not observe significant abnormalities in Hira −/− ES cells, despite a global lack of H3.3 enrichment at active and repressed genes, and despite the requirement of Hira for early embryonic development (Roberts et al., 2002). We speculate that Hira −/− ES cells may be rescued by the replication-coupled deposition of histones during the frequent S-phases of rapidly dividing ES cells (Burdon et al., 2002). Hira −/− ES cells divide as rapidly as wild-type ES cells, and show a similar preponderance of cells in S phase (A.C. and P.J.S., unpublished data). Overall, our data are consistent with a role for Hira in genic deposition of H3.3.
H3.3 enrichment has recently been shown at TFBS in Drosophila and human cells (Jin et al., 2009; Mito et al., 2007). Deposition of H3.3 at TFBS may serve as a mechanism for the maintenance of regulatory elements in a more accessible chromatin conformation (Henikoff, 2008). Close comparison of our data to a recent dataset of 13 different TFs in mouse ES cells (Chen et al., 2008) shows H3.3 enriched in ES cells at all known types of TFBS genome-wide, whether in gene bodies, promoters, or intergenic regions. We also find a strong positive correlation between MTL and H3.3 localization, indicating particular enrichment of H3.3 at enhancer elements. Our data demonstrate that Hira is involved in H3.3 localization at some genic and intergenic TFBS. However, we also find that genome-wide H3.3 enrichment at many regulatory elements is Hira-independent and Atrx-independent. Our data therefore suggest that H3.3 localization at TFBS may be mediated by multiple and distinct factors, including Hira, with as yet unidentified factors mediating H3.3 localization at specific regulatory elements.
We find that H3.3 is specifically enriched in the canonical (TTAGGG)n repeat that is the hallmark of telomeres in vertebrates (Meyne et al., 1989). Previous immunofluorescence studies localizing GFP-tagged Hira to telomeres suggested that Hira facilitates H3.3 deposition at this location (Wong et al., 2009). However, using genome-wide ChIP-seq and cell imaging analyses in Hira −/− ES cells, we show that localization of H3.3 at telomeres in ES cells is Hira-independent. Further, we identify Atrx and Daxx as proteins that associate with H3.3 nucleosomes in the presence and absence of Hira.
Recent studies have shown that the Drosophila homolog of Atrx, XNP, co-localizes with H3.3 at sites of nucleosome replacement on polytene chromosomes, but is not required for H3.3 localization at these sites (Schneiderman et al., 2009). We find that Atrx is required for enrichment of H3.3 at mammalian ES cell telomeres, suggesting a divergence of homolog function. Moreover, we demonstrate that in the absence of Atrx, ES cells show upregulation of TERRA. Could ATRX and Daxx serve as specific H3.3 variant deposition machinery for specialized regions of heterochromatin? The ATRX/Daxx complex has previously been shown to have chromatin remodeling activity (Tang et al., 2004; Xue et al., 2003), and we show that Atrx is physically associated with ES cell telomeres. In addition to telomeres, our preliminary studies indicate that Atrx is also required for H3.3 enrichment at ribosomal DNA (data not shown), another transcribed repetitive element with characteristics of heterochromatin (McStay and Grummt, 2008).
Our findings raise multiple questions. What is the function of H3.3 at genes, TFBS, and telomeres? Do cellular requirements for H3.3 differ in dividing versus post-mitotic cells, where replication-independent deposition might play a larger role? Is Atrx-mediated localization of H3.3 also replication-independent, like Hira, or does it occur during replication? Broadly, our study raises the prospect that distinct, region-specific chaperone and remodeling complexes may mediate the localization of a single histone variant (H3.3) to particular genomic regions. Although key factors required for region-specific H3.3 localization have now been identified, the exact deposition mechanisms at play remain an important challenge for future work.
Mouse ES cells were cultured under standard conditions and differentiated to NPCs as previously described (Conti et al., 2005). Hira −/− ES cells, Atrxflox, and Atrxnull ES cells have been described previously (Garrick et al., 2006; Meshorer et al., 2006). For more detail, see Supplemental Experimental Procedures.
ZFNs directed against the mouse H3.3B gene were designed using an archive of validated two-finger modules (Doyon et al., 2008; Urnov et al., 2005). ES cells were transfected with ZFNs and donor using Amaxa nucleofection. Fluorescent ES colonies were picked and screened by genomic PCR of H3.3B alleles, sequencing of PCR products, flow cytometry, and Southern blot. For more detail, see Supplemental Experimental Procedures.
Whole cell extracts were prepared by resuspending cell pellets in SDS-Laemmli sample buffer, followed by brief sonication and boiling.
Crosslinking and native ChIP were performed as described (Barski et al., 2007; Lee et al., 2006), with minor modifications detailed in Supplemental Experimental Procedures. ChIP DNA was validated by real-time PCR, prepared for Illumina/Solexa sequencing, and sequenced using the Illumina Genome Analyzer II. The CTD4H8 antibody was raised against a chemically synthesized phospho-Ser 5 peptide sequence from the CTD of the largest RPB1 subunit of RNAPII, and has been extensively characterized previously (Stock et al., 2007). Other antibodies and more detailed ChIP-seq methods are described in Supplemental Experimental Procedures. ChIP-seq assays performed are listed in Table S5. To confirm telomere enrichment, ChIP DNA from Atrxflox and Atrxnull ES cells was probed with a TTAGGG repeat probe as described (Sfeir et al., 2009).
ChIP-seq or input reads were mapped to the mouse genome (build 37, or mm9) using the ELAND alignment software within the Illumina Analysis Pipeline. Profiles in specific genomic regions were displayed in the Affymetrix Integrated Genome Browser. For analysis of repetitive elements, reads were aligned directly to a library of mouse consensus repetitive sequences (http://www.girinst.org) and enrichments were computed for ChIP against input samples. For TSS/TES profiling, we segregated reference genes (refSeq) into low, medium and high expression based on a previous microarray analysis (Mikkelsen et al., 2007). For TFBS profiling, binding sites for 13 TFs in mouse ES cells were obtained from a previous ChIP-seq analysis (Chen et al., 2008). To generate density profiles ± 5kb around TSS, TES, bivalent genes, and TFBS, we used a sliding window method to count the number of ChIP-seq reads in each 200 bp window, and the resulting counts were then normalized by the totals of genes or TFBS and the total mapped reads. For details of data analysis, see Supplemental Experimental Procedures.
RNA expression data for W9.5 and Hira −/− ES cells was generated from polyA RNA and random primers using the GeneChip Mouse Gene 1.0 ST Array kit (Affymetrix).
Immunoaffinity purifications of EYFP-tagged H3.3, H3.2, H3.1, HIRA−/− H3.3, and HIRA −/− H3.2 were performed as described (Cristea et al., 2005). MALDI MS and MS/MS analyses were performed as described (Luo et al., 2009). For more detail, see Supplemental Experimental Procedures.
Telomere FISH of ES cells was performed using a peptide nucleic acid TAMRA-TelG telomere probe (Sfeir et al., 2009), and immunofluoresence was performed using a previously described anti-GFP antibody (Cristea et al., 2005). For more detail, see Supplemental Experimental Procedures.
Our ChIP-seq and microarray datasets have been deposited in the GEO database with accession numbers GSE16893 and GSE19542.
We thank L Baker, EM Duncan, and GG Wang for critical reading of the manuscript, P Wu, A Sfeir, and T de Lange for telomere reagents and helpful discussion, PD Adams for Hira antibodies, R Jaenisch for F1 hybrid male 129SVJae x M. m. castaneus ES cells, G Almouzni for H3.1 and H3.3-FLAG-HA HeLa cells, K Zhao for sharing his native ChIP-seq protocol, E Moehle for drawing the gene editing schemes, N Jina of the UCL Genomics Core, KR Molloy for assistance with mass spectrometric analysis, and S Mazel and A North of the RU Flow Cytometry and Bio-Imaging Resource Centers. ADG is supported by NIH MSTP grant GM07739. LAB is a Damon Runyon CRF fellow. This work was funded by institutional support from The Rockefeller University and grants from the Tri-Institutional Stem Cell Initiative (funded by the Starr Foundation), Empire State Stem Cell fund through NYSDOH Contract #C023046, HHMI (SR), BBRC UK and BHF (PJS), startup funds from AECOM of Yeshiva University (DZ), and NIH Grants RR00862, RR022220, DP1DA026192 (IMC), GM53122 and GM53512 (CDA).
While this manuscript was in press, Wong et al. reported that Atrx associates with histone H3.3 and localizes to telomeres in ES cells (Wong et al., 2010). These data further support our conclusion that distinct factors control H3.3 localization at specific genomic regions.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.