Xist-RNA dissociates from the inactive X chromosome at the metaphase to anaphase transition in mouse pre-B cells
Male and female pre-B cell clones were isolated from foetal liver following immortalisation with Abelson murine leukaemia virus, as described previously67. As anticipated, we detected Xist-RNA uniquely in female pre-B cells, coating the presumptive Xi and co-located with a H3K27me3-labelled, DAPI-intense, Barr body (Fig. 1a, top panels) in cells at interphase. Xist-RNA was not detected in male pre-B cells (Fig. 1a, lower panels), where a single X and Y chromosome were confirmed (Fig. 1.a, Supplementary Fig. 1a and Supplementary Fig. 1b). In female pre-B cells, upon entry into mitosis, Xist-RNA remains partially chromosome-associated at prophase and metaphase stages but is completely lost at anaphase/telophase (Fig. 1b). These data reaffirm mitotic displacement of Xist from chromosomes in female cells68 and that Xist is likely dispensable for XCI maintenance69, 70. The timing of displacement, however, is later than has been reported for human B cells and fibroblasts, where XIST is predominantly lost during prophase70.
Proteomic comparison of mitotic chromosomes from male or female pre-B cells
As a first approach to investigate the factors that may convey epigenetic memory of X chromosome states during cell division, we isolated individual X chromosomes from male and female pre-B cells and performed proteomic analysis. These cells were derived from littermates and have stable karyotypes and Xist expression. The rationale for this analysis is that in female cells 50% of X chromosomes are active and 50% inactive, while in male cells, 100% of X chromosomes are active. Consequently, the differences between female and male chromosome X therefore should be attributed to the inactive X. To identify factors that are uniquely enriched on Xi, we also performed proteomics on an autosome, chromosomes 19 (Fig. 1c). As previously described40, 66, chromosome suspensions were prepared from male and female metaphase-arrested pre-B cells, stained with the DNA dyes Hoechst 33258 and Chromomycin A3 to allow visualisation and isolation by flow cytometry of mitotic chromosomes en masse, or individually (according to GC:AT ratio) (Fig. 1d). One million individual flow-sorted metaphase chromosomes X and 19 were subjected to quantitative LC-MS/MS proteomics. We identified 2562 and 2375 protein hits differentially detected on female versus male chromosomes X and 19, respectively (Fig. 1e, Supplementary Datafile 1 and 2). Of these, 109 showed a significant enrichment on the X chromosome of female cells as compared to males, including histone macroH2A (H2afy) (red, Fig. 1e left and Supplementary Fig. 1c), a histone variant that is known to be preferentially incorporated at the Xi in interphase and maintained during cell division60, 71. As anticipated, an analysis of chromosome 19 showed no evidence of macroH2A enrichment in female versus male samples (Fig. 1e right).
Many proteins that are known to associate with, or to mediate Xist-RNA recruitment, including Rnf20, Trim28, Hnrnpc (highlighted in red, Fig. 1f) were significantly enriched in samples derived from females versus males. As enrichment of these factors was evident on both mitotic X chromosomes (red, left) and an autosome (chromosome 19 red, right), these data suggest that although many RNA-binding and potential Xist-interacting factors appear more abundant in female versus male samples, they are unlikely to be relevant for the specific maintenance of XCI in mitosis. Furthermore, chromosome repressor complexes that are known to be retained on mitotic chromosomes, such as DNMT1,3a, PRC2 and PRC140, 66, 72, did not show any significant increase in representation in female samples (containing a presumptive Xi) versus male samples (Fig. 1g). Despite this, a comprehensive analysis of proteins enriched on X chromosomes from female versus X chromosomes in male (Supplementary Fig. 1d and Supplementary Fig. 1e), revealed a small number of candidates (37 and 43) solely enriched on female and male chromosome X respectively, that showed no equivalent enrichment on female versus male chromosome 19 (Supplementary Fig. 1e and full lists provided in Supplementary Datafile 3). For example, significant enrichment of histone macroH2A was seen in female-derived mitotic X chromosomes, as anticipated. These data indicated that although the representation of chromosome-associated RNA-binding proteins at metaphase was different in female and male pre-B cells, there was no evidence implicating Xist-interacting factors in the preservation of XCI through mitosis.
Biphasic distribution of H3K27me3 on mitotic X chromosomes enables isolation of Xa and Xi chromosomes and reveals a differential compaction at metaphase
Since female-derived samples contain both active and inactive X mitotic chromosomes, it is possible that heterogeneity may mask proteomic differences between Xi and Xa. To resolve this, we took advantage of elevated H3K27me3 levels on the Xi to separately purify Xa and Xi mitotic X chromosomes by flow cytometry directly from female pre-B cells. We refined mitotic chromosome flow-sorting strategy by adding one additional step that consists of staining mitotic chromosome suspension with a directly coupled antibody (AF647) that recognises trimethylated H3K27 prior to staining with Hoechst 33258 and Chromomycin A3 (outlined in Fig. 2a), in order to be able to distinguish two populations (high-H3K27me3 and low-H3K27me3) among female X chromosomes. As shown in Fig. 2b, H3K27me3 distribution was bimodal in female pre-B cells (right histogram), whereas H3K27me3 distribution was unimodal in male-derived X chromosomes (right histogram).
Native mitotic chromosomes corresponding to the presumptive inactive (high-H3K27me3) and active (low-H3K27me3) X chromosomes were separately isolated by flow cytometry. Parallel analyses of autosomes derived from female cells (chromosomes 3 and 19, Supplementary Fig. 2a) confirmed that bimodal H3K27me3 distribution was a unique feature of female X chromosomes. To validate the epigenetic status of purified X chromosomes, we analysed DNA methylation at selected CpG islands of X linked genes. In the high-H3K27me3 population we observed extensive CpG methylation at two X-linked genes, Hprt and Pdk3 (Fig. 2c) while the Xist locus was unmethylated (Supplementary Fig. 2b). Conversely, in the low-H3K27me3 chromosome X population, Hprt and Pdk3 were unmethylated, while Xist showed extensive CpG methylation. Eif2s3x, an X-linked locus that escapes inactivation73, was shown to be similarly unmethylated in high- and low-H3K27me3 samples, as anticipated (Supplementary Fig. 2b, right panel). These results confirmed that female-derived metaphase X chromosomes isolated as high-H3K27me3 represent the Xi, while those isolated as low-H3K27me3 comprise the Xa.
To investigate whether the Xi and Xa display different structures or sizes at metaphase, DNA-FISH, immunolabelling and microscopy was used to examine isolated Xi and Xa chromosomes. As predicted, metaphase Xi and Xa chromosomes were differentially labelled with antibody to macroH2A (Fig. 2d) and H3K27me3 (Fig. 2e), with more intense label intensity at the female Xi (Fig. 2e, quantified in lower dot plot). In addition, inactive X chromosomes were significantly smaller and more compact at metaphase than their active counterparts (Fig. 2e, upper graph) whether isolated from female or from male cells. In comparison, the chromosome size (volume) of mitotic autosomes such as metaphase chromosome 19, was unchanged in samples derived from male- versus female-derived pre-B cells (Fig. 2f). Taken together these results indicate that some of the distinctive differences in chromatin state and compaction of Xa and Xi chromosomes that have been described in interphase31–33, are retained as chromosomes condense at metaphase.
Proteomic comparison of Xa and Xi metaphase chromosomes
Proteomic comparisons were performed on an equivalent number (one million) of metaphase Xa, Xi chromosomes from females, as well as male X chromosomes and revealed that of 2418 candidate proteins identified, 29 showed significant enrichment on the Xa (as compared to the Xi), and 43 were significantly enriched on the Xi (Fig. 2g, Supplementary Datafile 4 and 5). Principal component analysis (PCA) analysis comparing the Xa, Xi and male X mitotic chromosome samples, showed that each sample category clusters separately (Supplementary Fig. 2c), and as anticipated macroH2A showed a significant enrichment selectively on the Xi (Fig. 2g). No significant difference of macroH2A abundance was observed between female Xa and male X (Supplementary Fig. 2d). Although chromatin repressors remain bound to mitotic chromosome as compared to mitotic lysate of pre-B cells (Supplementary Fig. 2e and Supplementary Datafile 6), consistently with other reports40, 66, 72, we didn’t observe any enrichment of repressive complexes (such as DNMT’s, PRC1, PRC2) on mitotic Xi as compared to Xa (Fig. 2h). We also did not detect any significant enrichment of Xist-associated proteins on mitotic Xi as compared to Xa (Fig. 2i), including Trim28 (Supplementary Fig. 2f), a B-cell specific Xist-interacting factor recently implicated in XCI maintenance74. While acknowledging that there is still conflicting literature about Xist-associated proteins and XCI maintenance6, 27, 74–77, here we saw no evidence of Xist-interacting proteins differentially marking the Xi in mitosis. GO-term analysis of Xi and Xa mitotic proteomes did however reveal that factors that catalyse histone H3 and H4 lysine acetylation were significantly enriched on Xa chromosome (Supplementary Fig. 2g).
Selective enrichment of HAT complexes and acetyl-binding proteins on Xa metaphase chromosomes
Several studies have reported that histone acetylation is dramatically and globally reduced as cells enter mitosis, while histone methylation is preserved, and H3S10p increases70, 78–80. In addition, histone deacetylase activity has been reported to be critically important for correct chromosome compaction and segregation at cell division79, 81–84. Against this backdrop we asked whether histone acetylation was a distinguishing feature of mitotic X chromosomes. Specifically, we examined HAT components, acetyl-binding proteins, and proteins associated with active transcription among the LC-MS/MS data derived from native mitotic Xa or Xi chromosomes. As shown in Fig. 3a (upper left-hand panel) several HAT components were significantly enriched in female-derived Xa samples as compared to female-derived Xi samples, as highlighted in green. This included Kat7, a component of the Hbo1 HAT complex that is enriched on mitotic chromosomes (as compared to mitotic lysates) (Supplementary Fig. 3a) and showed significant enrichment on the Xa as compared to the Xi (Supplementary Fig. 3b). Similarly, we also observed enrichment of histone acetyl binding proteins and factors associated with active transcription (Fig. 3b,c top panels). For example, Atad2b, an acetyl binding protein and several proteins associated with active transcription were also significantly and selectively co-enriched in female-derived Xa samples. Importantly, control comparisons made between mitotic Xa samples and male-derived X chromosomes (Fig. 3a-c lower panels) showed no enrichment of HAT components except for Msl1 that was enriched on female Xa compared to male (active) X chromosomes.
The increased representation of multiple components of three HAT complexes (Hbo1, Moz/kat6, Mof/msl, illustrated in Fig. 3d) prompted us to quantify further the relative abundance of these components on mitotic active and inactive X chromosomes, and an autosome (chromosome 19). As shown in Fig. 3e, a significant increase in Meaf6, Msl1 and Kat7 was seen on the Xa (green) compared to other chromosomes analysed. As a control, macroH2A levels were significantly higher only on the inactive X (red). Since Hbo1, Moz/kat6 and Mof-Msl complexes target histone H3 (at lysine 9, 14 and 23) and histone H4 (at lysine 5, 8, 12 and 16), we compared the levels of acetylated peptides in our proteomic dataset (Supplementary Datafile 7). We detected increases in the levels of several acetylated histone peptides in Xa versus Xi metaphase samples (Fig. 3f). In comparison, acetylated H2B (K5, K16) remained unchanged between Xa and Xi samples. Importantly, total histone quantification of H2B, H3 and H4 was also comparable (Fig. 3d right). Taken together, these results highlight significant differences in histone H3 and H4 acetylation level between the Xa and Xi.
Among factors that showed a significant enrichment on the mitotic Xi relative to the Xa, we detected a non-canonical (histone deacetylase (HDAC)-independent) inhibitor of histone acetyltransferase activity, Noc2l85 (Supplementary Fig. 3c) (validated by immunostaining in supplementary Fig. 3d), and a sequence specific DNA binding protein Mybbp1a, that functions as a repressor in part through HDAC activity86, 87 (Supplementary Fig. 3c). Noc2l and Mybbp1 were also more abundant on the Xi than other chromosomes analysed (Supplementary Fig. 3e). Collectively, these data show that differential histone acetylation pathways are a feature of Xi and Xa chromosomes in mitosis.
Relative size of the mitotic Xa versus Xi is dependent on acetylation of histone H3K14
To verify the histone acetylation modifications that distinguish X chromosomes in mitosis, we performed immunolabelling using antibodies that recognise pan-acetyl H4, acetyl H3K14 and acetyl H4K16, of female Xa and Xi metaphase chromosomes, isolated by flow cytometry (Fig. 4a). Figure 4b shows that Xa chromosomes could be defined as being low-H3K27me3 and high-acetyl H3/H4, relative to Xi chromosomes. Bimodal distribution of acetyl H4K16 (H4K16ac), acetyl H3K14 (H3K14ac) and pan-H4 acetylation (H4ac), displayed by mitotic X chromosomes, was not seen in female-derived autosomes such as chromosomes 3 and 19 (Supplementary Fig. 4a,b). To investigate the functional relevance of enhanced H3K14 acetylation on Xa, we treated female pre-B cells with a Kat7-specific inhibitor, WM-383588 (Fig. 4c). Cells exposed to WM-3835 for 48 hours showed a global reduction in histone H3K14ac levels as compared to control cells (vehicle), while acetylation of H4K16 and pan acetylation of H4 was unaffected (Fig. 4d). To discover the impact of reduced H3K14ac on mitotic chromosomes and, in particular at the active X, we purified chromosomes 3, 19, Xi and Xa from WM-3835 treated and control cells (Fig. 4e, and supplementary Fig. 4a-c). Although WM-3835 treatment did not significantly affect H3K27me3 distribution on any chromosome (Fig. 4e and supplementary Fig. 4c), and did not affect the size of the Xi (Fig. 4f), we observed a significant reduction in size of metaphase Xa chromosomes in response to experimentally induced H3K14ac reduction (Fig. 4f). The impact of WM-3835 on autosome size was variable (Supplementary Fig. 4d). Xist-RNA expression in interphase pre-B cells did not change following 48h of Hbo1 inhibition (Supplementary Fig. 4e). These results show that loss of Hbo1-mediated H3K14ac enhanced Xa chromosome compaction, so that it more closely resembles the size of the Xi in mitosis.
Mof/Msl-mediated H4K16ac is required to sustain differential compaction and H3K27me3 profiles of X chromosomes in female fibroblasts
Our results indicated that components of Mof/Msl complex and histone H4K16 acetylation were enriched in female Xa- versus Xi-derived metaphase chromosomes. To investigate the role of Mof/Msl-mediated H4K16 acetylation, MEFs were derived from E13.5 embryos Moffl/fl Cre-ERT2T/+ inducible knockout (KO) and Msl2 constitutive KO. Pre-B cell lines could not be derived from either Msl2 KO embryos or following induction of Mof knockdown suggesting a possible role for Mof/Msl complex in B cell development or survival. As we have previously shown that native metaphase chromosomes can be isolated from a range of cell types including fibroblasts40, we purified chromosomes X, 3 and 19 from Msl2 knockout mouse embryonic fibroblasts (MEFs) (Fig. 5a,b and Supplementary Fig. 5a,b). WT and Msl2 KO female MEFs were arrested in metaphase and released chromosomes were stained with Hoechst 33258 and chromomycin A3 (Fig. 5a). X chromosomes were purified by gating (as indicated) and stained with antibody to H3K27me3 (Fig. 5b). In X chromosomes of WT MEFs, a clear bimodal distribution for H3K27me3 was observed. In the Msl2 KO X chromosome samples, however, bimodal H3K27me3 distribution was lost in favour of a distribution that clustered around intermediate values (Fig. 5b). Interestingly, this change in H3K27me3 levels was also accompanied by a significant reduction in X chromosome size (Fig. 5c). No or very little changes of H3K27me3, and variable chromosome size changes was observed for mitotic autosomes 3 and 19 (Supplementary Fig. 5b and c).
To strengthen evidence of a causal link between reduced H4K16ac and alterations in mitotic Xa chromatin, we also purified chromosomes from female MEFs Moffl/fl Cre-ERT2T/+ (Fig. 5d and Supplementary Fig. 6a) in which Mof activity could be conditionally deleted upon the addition of tamoxifen (4-OHT). As anticipated, 4-OHT treatment resulted in a dramatic global reduction in Mof expression in MEFs Moffl/fl Cre-ERT2T/+ cells (Supplementary Fig. 6b,c), and a significant decline in H4K16ac mean intensity (Supplementary Fig. 6c). Flow-isolated chromosomes X, 3 and 19 were prepared from metaphase- arrested Moffl/fl Cre-ERT2T/+ cells treated with 4-OHT (or vehicle) (Fig. 5d,e and Supplementary Fig. 6d), and purity of X chromosome sorting was verified by chromosome paint DNA-FISH as being > 80% (Fig. 5f). Conditional depletion of Mof resulted in loss of H3K27me3 distribution on X chromosomes, with a shift in the profile from bimodal (in vehicle treated) to unimodal in 4-OHT induced samples (Fig. 5e). In addition, conditional Mof deletion and reduced histone H4K16ac resulted in reduced metaphase X chromosome size (Fig. 5g), with no or very modest changes to chromosomes 3 and 19 (Supplementary Fig. 6d,e). To exclude the possibility that 4-OHT treatment caused X-chromosome loss, rather than altering the properties of Xi and Xa chromosomes, we performed DNA FISH using X-paints (and examined Xist-RNA expression) on metaphase spreads of MEFs treated with 4-OHT or vehicle (Fig. 5h). Xist-RNA in Mof KO fibroblasts was not significantly perturbed (Fig. 5i) and the majority of fibroblasts retained similar number of X chromosomes in both 4-OHT or vehicle conditions. This exclude that loss of bimodal H3K27me3 distribution was the result of a selective loss of Xi in Mof-deficient cells. These data therefore show that Mof deletion interferes with both the differential condensation and differential H3K27me3 marking of Xa and Xi metaphase chromosomes.