HP1β, but not HP1α, is essential to maintain pluripotency and cell proliferation in ESCs

In order to determine whether HP1α and/or HP1β isoforms have any role in stem cell pluripotency and early differentiation, we took advantage of the recently generated HP1α^−/− and HP1β^−/− knockout (KO) mice and of the derived pluripotent ESCs, the differentiated embryoid bodies (EBs), and the mouse embryonic fibroblast (MEF) cells from these KO strains [34, 52]. To explore whether HP1α or HP1β has a specific function in pluripotent/undifferentiated cells, we analyzed the morphology of HP1α^−/− and HP1β^−/− ESCs, their cell growth, and differentiation potential compared with their wild-type (WT) counterparts at identical passages under identical conditions. To validate the KO clones and the specificity of the HP1α and HP1β antibodies, we verified the absence of the specific HP1 protein in the appropriate cell line, using immunofluorescence (IF) and western blots (Figure S1a, b in Additional file 1). As we cultured the KO ESCs, we noticed unexpectedly that whereas WT and HP1α^−/− ESCs displayed normal colony morphology, most of the HP1β^−/− ESCs did not form the usual compact three-dimensional colonies. They tended instead either to differentiate spontaneously or to remain very small (Fig. 1a). This was observed both in the presence of leukemia inhibitory factor (LIF), which maintains ESCs in their undifferentiated state, and in its absence, where the effect was more pronounced. We also observed that the HP1β^−/− ESCs differentiated faster than WT and/or HP1α^−/− ESCs upon LIF depletion (Fig. 1a, lower panel). The same was true when differentiation was induced by retinoic acid (RA; data not shown). Finally and importantly, HP1β^−/− but not HP1α^−/− ESCs displayed significantly reduced growth rates (Fig. 1b), indicating a reduced capacity for self-renewal.

We next tested the differentiation potential of HP1α^−/− and HP1β^−/− ESCs. To this end, we performed a teratoma assay, which involves injecting HP1α^−/−, HP1β^−/−, and WT ESCs under the skin of SCID mice. Three weeks later the resulting teratomas were analyzed by histology. We detected increased neuroectoderm formation in teratomas derived from HP1β^−/− ESCs, although all three germ layers were present in all the teratomas of all cell lines tested (Fig. 1c). To validate this observation, we performed directed differentiation of WT and HP1β^−/− ESCs into neuroectoderm in vitro. HP1β^−/− ESCs displayed accelerated neuronal differentiation, as judged by morphology and increased Tuj1-positive cells (Fig. 1d). Together, these results argue that the absence of HP1β in ESCs compromises the maintenance of pluripotency and cell proliferation, and increases neuronal differentiation both in vitro and in vivo. This suggests that HP1β negatively regulates neuronal differentiation in pluripotent cells and is thereby required to maintain pluripotency. We confirmed the results for the KO ESCs by RNA interference for HP1β which, similarly, led to premature differentiation (Figure S2b in Additional file 2).

HP1β is 100 % conserved between mouse and human, and mouse HP1β and HP1α are 63 % identical (and 79 % similar). It was of interest, therefore, to examine the effects of HP1β loss on overall chromatin organization. First, visualizing pericentromeric heterochromatin by DAPI, we note that the absence of HP1β had no significant impact on the global structure of pericentromeric heterochromatin domains in ESCs (Figure S1c in Additional file 1), nor did loss of HP1α (Figure S1c in Additional file 1). In addition, H3K9me3 staining of pericentromeric heterochromatic foci, as shown by the overlap with the DAPI staining in MEFs and ESCs, was also not altered in HP1α^−/− and HP1β^−/− ESCs compared with their WT counterparts (Figure S1c in Additional file 1). This observation is in line with previous reports in differentiated 3T3 mouse fibroblasts [53].

Using a more quantitative approach, we monitored fluorescence recovery after photobleaching (FRAP) for H1-GFP, as an indicator for chromatin plasticity [5]. This is used to monitor the impact of HP1β depletion on general chromatin proteins, as previously reported for CHD1 in euchromatic regions [8]. However, H1 protein dynamics in WT and HP1β^−/− ESCs were not significantly different (Figure S1d in Additional file 1). Indeed, as described below, HP1β itself is relatively poorly associated with chromatin in ESCs (see Fig. 7). Finally, to test whether the reduced capacity for self-renewal of HP1β^−/− ESCs (Fig. 1b) could be explained by defects in chromosome segregation during mitosis, we monitored metaphase and anaphase cells in HP1α^−/−, HP1β^−/− and WT ESCs (Figure S2a in Additional file 2). No defects, such as DNA bridges, were detected in any of the anaphase ESCs, although H3K9me3 has been described to be important for chromosome segregation [29]. In addition, H3K9me3 staining in these cells was perfectly localized mainly at pericentromeric regions, as expected (Figure S2a in Additional file 2). This suggests that chromosome segregation can occur normally in HP1β^−/− ESCs.

HP1β regulates developmental genes and pluripotency factors in ESCs

Given the strong phenotypic effect of HP1β deletion on pluripotency, and the absence of change in chromatin organization, we next looked for effects on the level of gene expression. Using Affymetrix whole transcriptome microarrays (GSE65121), we analyzed transcription profiles of WT, HP1α^−/− and HP1β^−/− ESCs in duplicates, and after EB differentiation for 7 days. EBs are known to undergo non-directed differentiation and cell specification into the three germ lineages (endoderm, ectoderm, and mesoderm). To ensure that neither MEFs nor spontaneously differentiating cells contaminated our ESC preparations, we sorted the pluripotent SSEA1-positive cells from all ESC types using magnetic beads and a column-based method. This is particularly important in the case of HP1β^−/− ESCs, since, as noted above, these cells tend to spontaneously differentiate. Using a threshold of 1.5-fold change in mRNA level (corresponding to p < 0.05; Figure S3a in Additional file 3) comparing mutant and WT ESCs and EBs, we found that the loss of HP1β resulted in the misregulation of 495 and 1054 genes in ESCs and EBs, respectively. The loss of HP1α, on the other hand, had a more subtle effect in both ESCs and EBs, with 53 and 627 genes altered, respectively (Fig. 2a, right). When a stringent cutoff of 2.5-fold in transcription level was used (corresponding to p < 0.005; Figure S3a in Additional file 3), only one gene passed the threshold in the HP1α^−/− ESCs, and 97 genes did in the corresponding EBs. In contrast, the HP1β^−/− ESCs had 34 genes in the undifferentiated ESCs and 201 in the corresponding EBs that were at least 2.5-fold misregulated (Fig. 2a, left). Changes in gene expression were validated in both ESCs and EBs using quantitative RT-PCR (qRT-PCR) for several genes (r ²  > 0.8 between the two methods; Figure S3b, c in Additional file 3). We conclude that HP1β has a far more significant effect on gene expression in both ESCs and EBs than HP1α.

We next examined the misregulation of established lineage markers in ESCs and found that, once again, HP1α deficiency had a relatively mild effect, with none of the selected markers showing a significant change (Fig. 2b). In contrast, depletion of HP1β resulted in significant changes in the expression of genes from all lineages examined, including endoderm, mesoderm, ectoderm, and trophoectoderm (Eomes). The most pronounced effect was again in neuroectoderm lineage markers, where significant overexpression of a related set of genes was detected (Fig. 2b). This correlates well with the changes observed in protein levels of neuroectodermal markers and with the effect of HP1β deletion on teratoma formation (Fig. 1c, d). Consistently, Gene Ontology (GO) analysis for the genes upregulated >2.5-fold in the HP1β^−/− ESCs revealed a significant enrichment in categories reflecting neuronal differentiation and cell proliferation (Fig. 2c). In contrast, the effect of HP1α deletion was again insignificant, even when the more relaxed threshold of 1.5-fold was used. Importantly, loss of HP1β in ESCs also led to a significant downregulation of key pluripotency factors (Fig. 2d), a fact that may explain the partial loss of pluripotency characteristics of those cells (morphology, growth rate, etc.). This is unlike the loss of HP1α (Fig. 2d) and unlike depletion of HP1γ, which show normal expression of pluripotency markers [15, 51]. In summary, we find that the loss of HP1β in ESCs downregulates the expression of pluripotency factors and skews the expression of developmental genes. This correlates with premature ESC differentiation, particularly along the neuroectodermal lineage. Such effects are unique to HP1β.

To determine whether HP1β KO also affects later stages of differentiation, we compared the transcriptional profiles from 7-day-old EBs originating from WT, HP1α^−/− and HP1β^−/− ESCs. As in earlier stages, loss of HP1α had a mild effect on gene expression, and it subtly, but significantly, altered lineage markers of the three germ layers. Loss of HP1β, on the other hand, had a particularly robust effect on mesodermal lineage markers. For instance, loss of HP1β led to the downregulation of Bmp2, Bmp4, Des, and Fgf15 (Fig. 3a). GO analysis on the altered genes (using a threshold of 2.5-fold) in HP1β^−/− EBs indicated strong effects on heart and muscle development (Fig. 3b), consistent with mesodermal differentiation defects. This is consistent with the fact that modulation of the HP1β protein level has been found to impair MyoD target gene expression and muscle terminal differentiation [33]. Therefore, whereas the differentiation of HP1β^−/− ESCs was skewed towards neuroectoderm, HP1β^−/− EBs were skewed away from proper mesoderm formation. Interestingly, a relatively high number of actin, myosin, and related proteins, which we found as interacting partners of HP1β in differentiated cells (see below and Fig. 4), were found to be both up- and downregulated (GO category “actomyosin structural organization”) in the HP1β KO cells (Fig. 3b). Finally, several pluripotency genes, including Nanog, Oct4, Esrrb, Dppa2, Dppa5a, and Stat3 failed to be correctly downregulated in the differentiated HP1β^−/− EBs compared with WT EBs (Fig. 3c). It is important to point out that pluripotency factors are downregulated in HP1β^−/− ESCs but upregulated in the differentiating HP1β^−/− EBs. This result, together with the distinct effects that HP1β elimination has on ESCs and EBs, suggests that HP1β influences gene expression in opposite directions — or at the very least triggers distinct pathways of gene control — in pluripotent versus differentiated cells.

HP1β has different interacting partners in pluripotent and differentiated cells

Because HP1β has very distinct and contrasting effects on gene regulation in pluripotent versus differentiated cells, we checked whether HP1β is associated with different protein complexes in the two cell states. To examine HP1β’s interacting partners in pluripotent and differentiated cells, we immunoprecipitated the endogenous HP1β from both ESC and MEF extracts, and used liquid chromatography-tandem mass spectrometry (LC-MS/MS) to examine co-precipitating proteins. This allowed us to avoid potential artifacts due to overexpression or the addition of tags. Experiments were performed in two biological replicates and non-specific interactions were eliminated using anti-green fluorescent protein (anti-GFP) as a negative control. Several HP1β interaction partners were common to both ESCs and MEFs, including hnRNPH2, hnRNPA0, Rundc2a, Eif4enif1 and histone H2B (Fig. 4a; Additional file 4). However, the large majority of HP1β’s interaction partners differed between the two cell types (Fig. 4a), suggesting that the recovery is not a product of contamination. Moreover, the number of identified HP1β interacting partners overall was considerably lower in ESCs than in MEFs (30 versus 105 proteins; Additional file 4). Whereas it is impossible to infer function from simple immunoprecipitation, the fact that we recovered different sets of interacting partners is consistent with a distinct function for HP1β in differentiated cells.

HP1β restricts reprogramming into iPSCs

The distinct effects on gene expression and the different interaction partners of HP1β in ESCs and MEFs prompted us to test its potential involvement in somatic cell reprogramming to iPSCs. To this end, we generated iPSC colonies from WT and HP1β KO MEFs by lentiviral infection expressing the four reprogramming factors Oct4, Sox2, Klf4, and cMyc. HP1β KO MEFs displayed increased reprogramming efficiency compared with WT MEFs as judged by the number of iPSC colonies generated after 12 days of reprogramming in identical conditions by alkaline phosphatase staining (Fig. 4b). This again suggested that, like HP1γ [15], HP1β helps maintain a proper differentiation state in WT differentiated cells by inhibiting efficient reprogramming. Indeed, heterochromatin reorganization was found to be one of the first steps in the rearrangement of chromatin from a somatic-like to a pluripotent-like state during the reprogramming process [14].

Importantly, and consistent with the phenotypes we observed in HP1β^−/− ESCs, fully reprogrammed HP1β KO iPSCs exhibit similar properties to those of HP1β^−/− ESCs. They tend to differentiate spontaneously and rapidly, especially in the absence of a feeder layer, losing their compact morphology after several passages (Fig. 4c). In contrast, iPSC colonies generated from HP1α KO MEFs were morphologically indistinguishable from WT iPSC colonies and HP1α KO ESCs (Fig. 4c). Taken together, our findings confirm that pluripotent cells such as ESCs and iPSCs that lack HP1β tend to differentiate spontaneously. On the other hand, HP1β^−/− differentiated cells could not maintain a proper differentiation state (EBs) and reprogrammed into iPSCs more easily than WT cells (MEFs) (Fig. 4d). This contrasting behavior argues that HP1β has distinct roles at different stages of differentiation. HP1β maintains pluripotency in ESCs, while in differentiated cells it helps maintain the differentiated state.

HP1β is highly expressed and diffuse in nuclei of pluripotent cells

We next asked how this can be achieved. Is there dissimilar expression and/or localization of HP1β in the different cell states? Indeed, by indirect immunofluorescence [14], we scored an approximately threefold higher expression level of HP1β in pluripotent nuclei of mouse Rr5 iPSCs [14] and R1 ESCs over that in MEFs (Fig. 5a, b). The Rr5 iPSC line contains both fully and partially reprogrammed iPSCs with otherwise similar properties (i.e., morphology, size, proliferation rate, nuclear volume) [14], conveniently enabling us to compare these two cell populations in the same field of view using Nanog staining as a marker of pluripotency. Only the “fully” reprogrammed and pluripotent Rr5 iPSCs showed high levels of HP1β, arguing that elevated HP1β levels are truly linked to the pluripotent state, and do not simply reflect proliferation rate or cell size. The MEF feeder layer (some of which are marked by asterisks in Fig. 5), which is used to maintain the undifferentiated state of pluripotent cells, provided us with an additional internal control, in the same image field for HP1β staining. We also confirmed that HP1β is present at higher levels in ESCs than in MEFs by western blotting extracts from mouse ESCs and MEFs (Figure S4a, b in Additional file 5 and Fig. 7c). Total levels of HP1β normalized to the amount of histone H3 shows an enrichment in ESCs of about threefold compared with MEFs (Figure S4b in Additional file 5), consistent with fluorescence intensity. Finally, we observed a slight, but reproducible, drop in HP1β levels after 7 days of ESC differentiation towards EBs (Figure S4c in Additional file 5).

We compared our results with previous reports and with publicly available gene expression datasets [54, 55] to ensure that this variation is broadly observed, even at the transcriptional level. Consistent with our findings, the Amazonia dataset [56] shows higher HP1β expression levels in human pluripotent cells compared with all other differentiated cell types (Figure S4d in Additional file 5). In previous datasets, HP1β had a threefold higher level in undifferentiated mouse ESCs over 7-day-old neuronal progenitor cells (NPCs) derived from those ESCs by in vitro differentiation [18]. HP1γ also displayed approximately threefold higher levels in ESCs compared with NPCs, in contrast to HP1α, which was only ~1.5 fold higher in the undifferentiated cells. The fact that HP1β and HP1γ levels decrease more sharply than HP1α levels upon differentiation supports the results we obtained by immunofluorescence on pluripotent and differentiated cells (Figs. 5d-e and 7e for HP1γ; and [14] for HP1α).

Akin to other chromatin proteins, the localization of the HP1 isoforms may be more important than their absolute levels. In support of this, we found that HP1β has a diffuse nucleoplasmic staining pattern in the nuclei of iPSCs and ESCs, in stark contrast to the characteristic heterochromatic foci found in the nuclei of differentiated MEFs (Fig. 5a, c; see Figure S6a in Additional file 6 for shorter exposure). This phenomenon was also true for HP1γ (Fig. 5d, e), but was not the case for HP1α. Whereas HP1α is somewhat diffuse in the nuclei of pluripotent cells, it also clearly labels heterochromatic foci [6, 14]. We quantified these differences by counting the average number of HP1β-positive foci in each cell type. We scored, on average, 12.2 ± 2.4 HP1β foci per nucleus in MEFs and 0.1 ± 0.4 in either fully reprogrammed iPSCs or ESCs (Fig. 5c). These observations were reproducible under different conditions, and are consistent with previous studies which showed fewer HP1β heterochromatic foci in E14 mouse ESC line (4 foci per ESC) than in a more differentiated state (11 HP1β foci per cell [57]). We note that the E14 ESCs displayed a lower level of histone acetylation and a diminished ability to reprogram MEFs by cell fusion than the R1 ESCs used above [58]. Consistent with the stronger pluripotency character of our R1 ESCs over E14 cells, we see that HP1β assumes a completely diffuse pattern in the nucleoplasm of R1 ESCs, while it was partially accumulated at heterochromatin foci in E14 ESCs (Figure S6b in Additional file 6).

A final confirmation that HP1β changes localization during differentiation came from the use of an endogenously tagged fluorescent protein library (our own unpublished resource), in which HP1β is endogenously fused with the mCherry fluorescent protein. By scoring HP1β localization in living cells we can avoid potential artifacts of fixation or overexpression. Spinning disk time lapse imaging of ESC differentiation showed that HP1β has a diffuse pattern in undifferentiated cells, which transitions to HP1β focus accumulation. This occurred within 24–36 hours, at which point all cells displayed some degree of HP1β foci (≥1–2 foci per cell; Fig. 5f; Additional file 7). Taken together, we conclude that HP1β is more highly expressed and has a diffuse subnuclear localization in pluripotent stem cells, whereas it becomes heterochromatin-enriched in differentiated cells, consistent with the different roles it has in the two cell states.

HP1β is enriched within genes in pluripotent cells

In order to confirm these imaging results, we investigated the distribution of HP1β genome-wide using ChIP-Seq. ChIP-Seq analysis in ESCs showed that HP1β is significantly enriched in genes, especially within exons (p < 10⁻⁴, hypergeometric test; Fig. 6a; [GEO:GSE64946]). Moreover, HP1β is largely depleted from intergenic regions in ESCs, which would normally show enrichment for heterochromatin. Moreover, HP1β was largely depleted from proximal promoters (Fig. 6a) and transcription start sites (Fig. 6b), yet showed a clear enrichment gradient that increased from introns to exons: indeed, HP1β is more strongly enriched on exons than on introns (Fig. 6c–e). This preferential association of HP1β with exons is consistent with a unique role in pluripotent cells, and suggests a potential role in exon recognition, that may coincide with histones bearing H3K36me3 [59]. Interestingly, ‘alternative splicing’ was the most highly enriched category in GO analysis performed for the HP1β-bound genes (Figure S5a in Additional file 8). These correlations suggest a potential role for HP1β in exon recognition and/or pre-mRNA processing in ESCs. This observation is in line with a recent study that showed that HP1β regulates the alternative splicing of a subset of genes in a DNA methylation-dependent manner [60], which is thought to be achieved by the recruitment of splicing factors to DNA methylated genes through HP1β [60].

Since HP1β is not known to bind methylated H3K36, we next compared the HP1β ChIP-Seq data with other existing genome-wide datasets in ESCs (Figure S5b in Additional file 8). We found significant correlation of HP1β-bound loci (p < < 10⁻¹⁶) with H3K36me2/me3, which is also enriched within exons [59, 61], as well as with H3K9me3 (p < < 10⁻¹⁶). This suggests that HP1β, while largely euchromatic and exonic in ESCs, may also be associated in some regions with H3K9me3.

To understand if the changes in gene expression in the HP1β^−/− ESCs resulted from transcriptional regulation by HP1β or from post-transcriptional regulation through HP1β, we tested the correlation between HP1β binding to the genome and the changes in expression level of the corresponding genes or promoter regions. Comparing the list of the misregulated genes (>1.5-fold) in the HP1β^−/− ESCs with the list of the promoters or gene bodies directly bound by HP1β (ChIP-Seq data), we found that promoter regions bound by HP1β do not correlate significantly with misregulation of the adjacent genes (hypergeometric p value > 0.9; Figure S5c in Additional file 8). The HP1β-bound exons/gene bodies selected with a mild threshold (p < 0.01) also had no significant correlation with upregulated transcripts in the HP1β KO ESC samples, whereas a slight correlation was found with downregulation. When a more stringent threshold was used for the HP1β-bound genes (p < 0.001), a higher significance level was observed for a group of 15 genes that were clearly downregulated in HP1β^−/− ESCs (Figure S5c–e in Additional file 8), suggesting that HP1β could potentially upregulate the transcription of this subset of genes in WT ESCs. Nonetheless, since the majority (>97 %) of HP1β-bound genes in ESCs had no change in their expression level in HP1β^−/− ESCs, it appears that, in pluripotent ESCs, HP1β by itself probably does not act principally by modulating transcription. Supporting this view, we found that the genes that are misregulated in HP1β KO ESCs and that are included in biological process categories such as “regulation of cell proliferation” or “regulation of cell development” (Fig. 2c; e.g., Inpp5D, Ifitm3, Nefl, Nefm, Tnfrsf12a) are not genes or promoter regions bound by HP1β in ESCs. Nor are pluripotency factors such as Nanog or Klf4 downregulated in HP1β KO ESCs (see below). In addition, none of the genes (listed in Figure S5e in Additional file 8) bound by HP1β and misregulated in HP1β KO ESCs seem a priori able to explain all the phenotypes observed in HP1β KO ESCs. Alternatively, HP1β may work by modulating mRNA processing or export or may serve to maintain a chromatin state that only affects gene expression at a later point in development.

HP1β binds chromatin in a distinct manner in pluripotent and differentiated cells

We next asked whether the more diffuse distribution of HP1β found in ESCs versus differentiated cells reflects a different mode of binding to chromatin. To that end, we first co-stained MEFs and ESCs with the heterochromatin markers H3K9me3 and HP1β. Whereas HP1β almost completely overlapped with H3K9me3 in MEFs, consistent with recognition of H3K9me3 by its chromodomain, it did not co-localize with bright H3K9me3 foci in ESCs (Fig. 7a). In the case of HP1α, a major overlap with the H3K9me3 foci was scored in both ESCs and differentiated cells [6]. Therefore, we suggest that the correlation of HP1β with H3K9me3 by ChIP-Seq in ESCs probably does not represent HP1β association with H3K9me3-containing chromocenters, but rather recognition of this modification at other loci. On the other hand, in the somewhat less pluripotent E14 ESCs, the few HP1β foci that we observed did co-localize largely with H3K9me3 heterochromatin (Figure S6c in Additional file 6).

We next performed ChIP-qPCR to test the association of HP1β with major satellite repeats in ESCs. The major satellite is the main sequence element in heterochromatic pericentromeric regions and these generally map to the chromocenters where HP1β binds in differentiated cells [53, 62]. Unlike the situation in MEFs, HP1β was not highly enriched on major satellite repeats in pluripotent ESCs (Fig. 7b). These results are consistent with a recent study in which HP1β was shown to be only moderately enriched at pericentromeric regions in ESCs, while HP1α was strongly enriched at these sites, as monitored by a quantitative locus purification method [63]. The large absence of HP1β on major satellites in pluripotent ESCs compared with MEFs is consistent and reinforces the almost complete absence of pericentromeric foci enriched with HP1β in ESCs.

In order to measure the association of HP1β with chromatin in differentiated and undifferentiated cells biochemically, we fractionated MEFs and ESCs into cytoplasmic (S1), nucleoplasmic/chromatin unbound (S3) and chromatin-bound (P3) fractions, and analyzed HP1β levels in each fraction using immunoblots. Interestingly, HP1β was highly enriched in the nucleoplasmic fraction of ESCs, and was only weakly associated with the chromatin fraction, whereas in the differentiated MEFs, HP1β was more enriched in the chromatin-bound fraction (Fig. 7c, d).

We obtained similar results for HP1γ (Fig. 7e), which also displayed a diffuse nuclear localization in pluripotent ESCs (Fig. 5d, e). This is in contrast to HP1α distribution, which largely overlaps with pericentromeric heterochromatic foci at all stages of differentiation (data not shown and [6]). Finally, to test whether HP1β and HP1γ have redundant functions in ESCs, we knocked down over 70 % of the level of HP1γ by small interfering RNA (siRNA) in the HP1β^−/− ESCs (Figure S6d in Additional file 6), and found that depletion of HP1γ led to a slight (~18 %, p = 0.01) reduction in the proliferation rate of WT cells (Figure S6e in Additional file 6) [15, 51], yet there were no additive effects on cell growth and survival in the HP1β KO/HP1γ knock-down (Figure S6f in Additional file 6).

Taken together, we conclude that, unlike the situation in differentiated cells, HP1β does not associate predominantly with chromatin in ESCs, does not localize to pericentromeric H3K9me3 foci, and is not enriched on major satellite repeats. Importantly, we show by ChIP-Seq that HP1β in ESCs is enriched on exons over the genome, even though this may represent a minor fraction of total HP1β in ESCs, given that most HP1β is not chromatin-bound. The distribution and expression levels of HP1β and HP1γ are similar, yet loss of HP1β in ESCs resulted in precocious differentiation in cultured ESCs, and HP1β^−/− embryos died perinatally [34], while depletion of HP1γ affected cell growth and differentiation only under certain conditions [51]. Thus, this dual and opposing function in pluripotent and differentiated cells appears to be unique to HP1β and is not shared redundantly with HP1γ or HP1α.

Here we have reported unique characteristics and an unexpected role for HP1β in mouse ESCs. Functionally, we found that HP1β is required to maintain the undifferentiated/pluripotent ESC state, given that HP1β depletion in both ESCs and iPSCs resulted in precocious differentiation. The differentiation was mostly towards neuronal cell types. This is in line with the aberrant cerebral cortex development phenotype observed in vivo in the HP1β^−/− mutant mice [34], which die around birth with defective cerebral corticogenesis and reduced proliferation of neuronal precursors. Whereas HP1β^−/− MEFs proliferate at a similar rate to that of WT MEFs, HP1β^−/− ESCs display slower proliferation rates than WT or HP1α^−/− ESCs, in conjunction with other observations [64].

A meta-analysis of all available ChIP-Seq datasets in ESCs [65] revealed that the HP1β promoter is bound by Oct4, Nanog, Klf4, Esrrb, Nr5a2, and Sall4, which are all factors of the pluripotency network. This may well account for the high levels of HP1β in ESCs. Indeed, a knockdown of Oct4 in ESCs downregulated HP1β, while knockdown of Nanog or Klf4 did not [66]. However, we have made the intriguing finding that the depletion of HP1β in ESCs leads to the downregulation of most of the key pluripotency factors, including Nanog, Klf4, and Esrrb, but not of Oct4 (Fig. 5d). This suggests that Oct4 acts upstream of HP1β, and may be responsible for the high expression level of HP1β in ESCs. This in turn appears to contribute by regulating the other pluripotency factors. Nonetheless, the effect of HP1β on the global pluripotency gene expression signature does not appear to be through direct transcriptional control. One possible mode of action is that the nucleoplasmic fraction of HP1β stabilizes or potentiates selected long intergenic non-coding RNAs (lincRNAs) that were shown to associate with HP1β in ESCs and to regulate pluripotency [67]. While this is possible, further studies are needed to examine the effects of HP1β loss on lincRNAs in ESCs and the role of potential HP1β–RNA complexes on pluripotency.

The diffuse localization of HP1β in undifferentiated ESCs remains particularly intriguing, especially since H3K9me3 and HP1α foci are clearly visible [2, 13]. This rules out the possibility that the diffuse localization of HP1β is due to the absence of pericentromeric foci in ESCs, and suggests that HP1β has a differential affinity for H3K9me3 in ESCs versus differentiated cells [68]. This may reflect the preferential binding of HP1β to another histone modification that prevents or competes for its binding to H3K9me3, or else, possibly, competition for HP1β between RNA and H3K9me3-containing nucleosomes. We can rule out a role for H3S10 phosphorylation in this phenomenon, as we see no differences in H3S10P in ESCs and MEFs (data not shown). We do not rule out, however, that other histone modifications that are differentially abundant in pluripotent and differentiated cells might impact HP1β localization [69–71]. HP1β in vivo undergoes multiple post-translational modifications, including acetylation, phosphorylation, methylation, and many more [72], and several of these modifications have been correlated with the different functions of HP1 [21, 73–75]. Thus, HP1β itself could be differentially modified in pluripotent and differentiated cells. Alternatively, in order to explain HP1β diffuse localization in ESCs, HP1β may be targeted to sites of action by binding differentially to KAP1/TRIM28/TIF1β [76] in pluripotent versus differentiated cells, although this interaction was not detected under our LC-MS/MS experimental conditions.

Our findings suggest that HP1β has distinct interaction partners in ESCs compared with differentiated MEFs. In MEFs, HP1β interacting partners could be classified into the following categories: ‘cell structure and motility’, including actin, myosin, lamin, and other filaments; ‘protein biosynthesis’, including mostly ribosomal proteins; ‘chromatin and nucleotide’; and ‘RNA processing’ (Additional file 4). Based on these findings, we speculate that HP1β association with nuclear filaments such as lamin, myosin, and/or tubulin may contribute to its association with stable heterochromatic foci in differentiated cells (MEFs). The interaction of HP1β with an RNA-processing protein category also led us to wonder whether this category of proteins could be involved in the silencing function of HP1β in differentiated cells. In addition to the conventional mechanism of transcriptional repression by heterochromatin, we propose that HP1β and RNA-processing proteins could serve to recognize RNA transcribed from heterochromatin, leading to its sequestration and/or degradation. Such a role has been reported for the HP1^Swi6 protein in fission yeast [77]. In addition, association between Drosophila HP1a and a broad set of repetitive RNAs has been recently reported [78], and interactions between HP1a, RNA transcripts, and some RNA-processing heterogeneous nuclear ribonucleoproteins (hnRNPs) were also shown to be involved in regulation of gene expression and heterochromatin formation [79].