The histone variant macroH2A is an epigenetic regulator of key developmental genes

Abstract

The histone variants macroH2A1 and macroH2A2 are associated with X chromosome inactivation in female mammals. However, the physiological function of macroH2A proteins on autosomes is poorly understood. Microarray-based analysis in human male pluripotent cells uncovered occupancy of both macroH2A variants at many genes encoding key regulators of development and cell fate decisions. On these genes, the presence of macroH2A1+2 is a repressive mark that overlaps locally and functionally with Polycomb repressive complex 2. We demonstrate that macroH2A1+2 contribute to the fine-tuning of temporal activation of HOXA cluster genes during neuronal differentiation. Furthermore, elimination of macroH2A2 function in zebrafish embryos produced severe but specific phenotypes. Taken together, our data demonstrate that macroH2A variants constitute an important epigenetic mark involved in the concerted regulation of gene expression programs during cellular differentiation and vertebrate development.

Main

The structural unit of chromatin is the nucleosome. It consists of DNA wrapped 1.75 times around an octamer of histone proteins. In addition to their structural role, histones are integration sites for various signals and are highly modified on their N-terminal tails1. These modifications serve as binding platforms for other proteins and are crucially involved in the regulation of all chromatin functions, such as DNA repair and transcriptional regulation.

The term 'epigenetic' has been coined for information that does not rely on DNA sequence but can still be transmitted from mother to daughter cells2. Histone modifications and DNA methylation constitute a large part of the epigenetic memory of the cell. A common form of epigenetic modification is the complete exchange of a canonical histone for a variant. Among all known histone variants, macroH2A differs most from its canonical counterpart, but its function is poorly understood3. In addition to a homologous histone domain, macroH2A has a large C-terminal domain of unknown activity, called the macro domain. This domain is about twice the size of the histone domain and protrudes out of the compact structure of the nucleosome4. Thus, the macro domain is likely to serve as a binding interface for chromatin regulators. The vertebrate genome contains two closely related macroH2A variants, macroH2A1 and macroH2A2, which are enriched at the inactive X chromosome and centrosomes, suggesting their involvement in gene repression and heterochromatinization5. MacroH2A1 can interfere with transcription factor binding, SWI/SNF-induced nucleosome sliding and initiation of RNA polymerase II (RNAP II) transcription in vitro6,7. In Namalwa cells, a Burkitt lymphoma cell line, the presence of macroH2A1 in a positioned nucleosome on the proximal promoter rendered the interleukin-8 (IL8) gene inactive and insensitive to viral stress stimuli8. MacroH2A1 in complex with inactive poly (ADP-ribose) polymerase-1 (PARP1) contributes to the repression of heat shock–responsive hsp70 genes9. Upon heat shock, macroH2A1 is released from the promoter, thus activating PARP1, which in turn stimulates transcription through ADP-ribosylation. The inhibition of PARP1 by macroH2A1 also contributes to X chromosome inactivation10. Notably, the macro domain can bind ADP-ribose and some of its derivatives, but the consequence of such metabolite binding is not known11,12. Despite these recent advances, the physiological and cellular function of macroH2A remains elusive, largely because of our limited knowledge about the genes that are regulated by this atypical histone variant.

Using NTera2/D1 (NT2) cells, a testicular cancer cell line that retains some stem cell characteristics such as pluripotency, we conducted an extended analysis of macroH2A targets. Here we describe more than 800 target genes identified by analyzing chromatin immunoprecipations with human promoter arrays. Further analysis of these genes allowed us to describe the physiological function of macroH2A as an epigenetic regulator of development and cell fate decisions. This was validated in vivo by the generation of severe specific phenotypes upon disruption of macroH2A2 function in zebrafish embryos.

Results

Genomic analysis of macroH2A target genes

The genes encoding both macroH2A1 and macroH2A2 are expressed in NT2 cells (Supplementary Fig. 1a). To analyze their distribution in the genome, we first generated specific antibodies directed against the nonhistone regions of both forms (Supplementary Fig. 1a,b). Using these antibodies and a commercially available anti-macroH2A1 antibody, we performed a series of chromatin immunoprecipitations (ChIPs) in biological replicates (Fig. 1a). Although the commercial antibody had a weaker signal, it correlated positively with the other ChIPs (Spearman correlation coefficient 0.29). Input and pulled down DNA were labeled with different dyes and hybridized together to human promoter arrays (ChIP-on-chip). Data were analyzed with respect to nonspecific IgG controls. We scored genes as macroH2A target genes if they had at least four significantly enriched probes within 5.5 kb upstream and 2.5 kb downstream of the transcriptional start site (P < 0.05). To further validate our approach, we compared the list of macroH2A1 target genes calculated with nonspecific IgG samples as background controls to a list of target genes using macroH2A1 samples from cells treated with macroH2A-specific short hairpin RNA (shRNAs), denoted as shmacroH2A-treated, instead of with IgG (Supplementary Fig. 2a,b). As expected, when we used samples from shmacroH2A-treated cells, which retained residual macroH2A1 signal, for background subtraction, a smaller set of macroH2A1 target genes was yielded. Notably, this set of macroH2A1 target genes was almost fully represented in the set calculated with IgG (Supplementary Fig. 2c), thus validating the use of IgG as a negative control for our purposes.

Figure 1: Identification of macroH2A target genes by ChIP-on-chip analysis.
figure1

(a) Overview of ChIP-on-chip samples used. (b) Venn diagram showing the number and overlap of target genes calculated for macroH2A1 (mH2A1), macroH2A2 (mH2A2) and for the combined data set (*, P < 0.05; **, P < 0.005). (c) Enrichment plot of signals along four target genes. Significant signals are shown in black. Arrows indicate transcriptional start sites. (d) Standard ChIP analysis of identified target genes in control and shmacroH2A-treated cells (sh mH2A). Error bars indicate the s.d. of three independent experiments and asterisks indicate P-values of <0.05 with respect to shRNA-treated control samples. (e) Enrichment on the same target genes with respect to the average enrichment on control genes. Asterisks indicate P-values of <0.05 for both macroH2A1 and macroH2A2 compared to nontarget genes.

Among the about 17,000 genes that were represented in our analysis, we identified 250 high-confidence macroH2A1 and 424 high-confidence macroH2A2 target genes (Fig. 1b). Both sets of target genes strongly overlapped with almost two-thirds of all macroH2A1 target genes that also scored positive for macroH2A2. Next, we had a closer look at those genes that scored positive for one macroH2A form but not the other. We observed that the very same probes that were significantly enriched for one macroH2A form often showed enrichment of the respective other macroH2A protein just below threshold levels (P < 0.05). This suggested that our lists of target genes were an underestimation. Thus, we decided to perform a third analysis that treated all seven macroH2A variant samples as replicates. This approach yielded a set of 838 genes, which included more than 90% of both individual sets of target genes. This approach also led to the inclusion of 368 additional genes that showed a low but robust enrichment for both macroH2A forms. Detailed information on all three sets of macroH2A target genes can be found in Supplementary Table 1.

Enrichments of macroH2A1 and macroH2A2 were found at discrete gene regions, as indicated for E2F1, FBXL16, HOXA3 and HOXA10 (Fig. 1c). Signals replicated well in different biological samples (Supplementary Fig. 3) and were strongly reduced in cells treated with macroH2A shRNA (Fig. 1c, below). To validate these results, we first analyzed a set of target genes by conventional ChIP. Notably, macroH2A1 and macroH2A2 occupancy was confirmed in all cases, and the signal was strongly reduced in shmacroH2A-treated cells (Fig. 1d). Second, to exclude the effects of nonspecific co-precipitation, we confirmed that enrichment on target genes was several fold higher than on control genes (Fig. 1e). For the latter, we randomly selected a few nontarget genes and the two housekeeping genes GAPDH and PUM1.

MacroH2A variants are enriched at developmental genes

Analysis of the largest set of macroH2A1+2 targets (GEO GSE17531) revealed that they were enriched for genes that control developmental processes (Fig. 2a). Notably, the same held true whether we calculated the sets of macroH2A1 or macroH2A2 target genes individually or used samples from shmacroH2A-treated cells instead of IgG as a background control (Supplementary Fig. 4a,b). Genes encoding proteins with homeobox domains and those containing other DNA-binding domains were overrepresented among the set of macroH2A target genes (Fig. 2b). Among these homeobox genes were almost all genes of the HOXA, HOXB and HOXC clusters, but also others belonging, for instance, to the DLX, LHX and MEIS families (Fig. 2c).

Figure 2: Repressive macroH2A targets key developmental genes.
figure2

(a) Genes bound by macroH2A (mH2A) were compared to biological process gene ontologies. Only highly represented, nonredundant categories are shown. (b) mH2A target genes were analyzed for encoded protein domains. Significant categories including helix loop helix (HLH) proteins and cation transporter (C+) ATPases with a P-value < 0.001 are shown. (c) A selection of bound genes involved in developmental and transcriptional regulation. Small circles indicate number of genes. Genes were grouped according to pathways and families such as homeobox proteins (HOX), fibroblast growth factor (FGF) signaling and HLH transcription factors (TFs). The names of genes included in the representation are detailed in Supplementary Figure 2b. (d) mH2A occupancy correlates with low transcriptional activity. Genes were grouped into five expression categories according to average probe intensities on Agilent Human Expression arrays (0 indicates no signal/expression and 1–4 indicates very low to high expression, respectively). Distribution of targets has been corrected for the distribution of all genes analyzed by ChIP-on-chip. (e) Distribution of mH2A2 bound probes within a 8-kb window centered on the TSS of its target genes. The black line represents smoothing over of the actual data. (f) ChIP analysis of macroH2A2 occupancy on on TSS and upstream (UP) regions. Data have been normalized for H3 occupancy (*, P < 0.05). Results are expressed as the mean ± s.e.m.

The role of macroH2A proteins in transcriptional regulation is not fully understood. The presence of both macroH2A forms on the inactive X chromosome suggested early on that they could function in repression. However, in some cases macroH2A1 was also found to be enriched upstream of highly transcribed genes such as Albumin1 (ref. 13). Hence, we asked how the presence of macroH2A correlates with the transcriptional activity of its target genes. For this purpose, we grouped genes into five categories based on their expression levels and then analyzed the relative distribution of the set of macroH2A target genes. MacroH2A target genes were overrepresented in the group of genes with no detectable transcription and progressively underrepresented among groups of genes with increasing expression (Fig. 2d). Although this shows that macroH2A correlates with gene repression, its presence is not a strict indicator for transcriptional inactivity. We next aligned all target promoters at their transcriptional start site (TSS) and calculated the average distribution of macroH2A. Our results indicate that the presence of macroH2A is minimal in the direct vicinity of the transcription start site, but peaks at 2–3 kb upstream and, to a lesser extent, 2–3 kb downstream of the TSS (Fig. 2e). The reduction of the macroH2A signal at the TSS was confirmed for several target genes and was sustained after normalization for nucleosome density monitored by histone H3 ChIPs (Fig. 2f). These results suggest that macroH2A is a repressive mark that peaks at TSS distal regions, which in some cases allows transcription to occur.

MacroH2A regulates the timing of HOXA activation

In mammals, the HOXA genes are organized into one of four HOX clusters. They have an important function in embryonic patterning and cell differentiation, and their deregulation is linked to cancer14. During the early phase of retinoic acid (RA)–induced neuronal differentiation of NT2 cells, the genes of the HOXA cluster follow a strict activation pattern15, with genes become progressively activated in a temporospatial manner, starting with the more anterior genes. In NT2 cells, both macroH2A forms scored positive at regions along the complete 120-kb HOXA locus (Fig. 3a). Upon addition of RA, activation of anterior HOXA2 and HOXA3 genes was detectable early on, whereas mild activation of the posterior HOXA10 gene was visible only after 10 d (Fig. 3b). MacroH2A2 occupancy was reduced to less than half of its starting level on the HOXA2 promoter after 1 d of RA treatment, whereas its occupancy on HOXA10 was not affected until 10 d. Reduction of macroH2A2 was accompanied by a loss of histone H3 Lys27 trimethylation (H3K27me3) and an increase in the presence of H3K4me2. We observed a similar effect at only one of the three alternative promoters of HOXA3 that responded to RA stimulation (Supplementary Fig. 5). Furthermore, shRNA-mediated knockdown of both macroH2A forms resulted in increased transcription of all three HOXA genes upon induction of differentiation by RA (Fig. 3b), demonstrating that macroH2A variants represent one of the layers regulating temporal transcriptional activation.

Figure 3: Occupancy by macroH2A inversely correlates with HOXA cluster activation.
figure3

(a) Enrichment plot of both macroH2A (mH2A) variants over the 120-kb HOXA cluster. Significant probes P < 0.05 are sketched in black. The exon-intron structure of all genes is represented schematically below. (b) Analysis of the anterior (early-expressed) genes HOXA2 and HOXA3 and the posterior (late-expressed) HOXA10 gene. Left, relative mRNA levels after treatment with 0.1 μM RA for different periods of time. Middle, ChIP analysis of mH2A2, H3K27me3 and H3K4me2 occupancy on promoter regions. Right, mRNA levels in macroH2A-depleted and control cells after treatment with 0.1 μM RA for 24 h. (c) Recovery of HOXA2 repression and mH2A2 occupancy after transient RA treatment, as demonstrated by quantitative RT-PCR and ChIP. Results are expressed as the mean ± s.e.m.

Focusing on the RA-responsive HOXA2 promoter, we analyzed the dynamics of macroH2A2 occupancy in more detail. We washed the cells after 1 d of RA treatment and over the following days analyzed levels of HOXA2 mRNA as well as the presence of macroH2A2 on its promoter. After an initial increase, mRNA levels returned to almost basal levels within 3 d (Fig. 3c). Concomitantly, macroH2A2 occupancy on the HOXA2 promoter also recovered from its initial drop and returned to its basal level, thus demonstrating that macroH2A is a regulated and dynamic chromatin modification.

MacroH2A overlaps with Polycomb repressive complex 2

Similar to our results for macroH2A1+2, data from others have shown that Polycomb proteins target key developmental genes16,17,18,19,20. The core Polycomb repressive complex 2 (PRC2) consists of the proteins EZH2, EED and SUZ12 (ref. 21) and catalyzes H3 Lys27 di- and trimethylation (H3K27me2 and H3K27me3, respectively)22. Focusing on the PRC2 complex, we explored the possibility that macroH2A proteins and Polycomb could functionally overlap in the regulation of these genes. By ChIP analysis we found that both EZH2 and H3K27me3 are enriched at the same sites as macroH2A1+2, albeit with varying patterns (Fig. 4a). To prove that both marks actually occur on the same allele, we used sequential ChIP (Re-ChIP) to confirm that anti-H3K27me3–precipitated material also scored positive for macroH2A2 (Fig. 4b). shRNA-mediated knockdown of macroH2A1+2 expression resulted in a reduction of both EZH2 and H3K27me3 signals, with a concomitant increase in histone H4 acetylation at the HoxA2 promoter (Fig. 4c). These alterations on the chromatin could explain the observed increased sensitivity to RA-induced activation.

Figure 4: Functional overlap of macroH2A and PRC2.
figure4

(a) ChIP enrichment profile of macroH2A1 (mH2A1) and macroH2A2 (mH2A2), EZH2 and H3K27me3 on different HOXA genes. Results are expressed as the mean ± s.e.m. (b) Co-occupancy on HOXA2 demonstrated by Re-ChIP using H3K27me3-precipitated material as input material. (c) ChIP analysis comparing control cells and shmacroH2A1+2-treated cells. (d) Global overlap of mH2A1+2 and SUZ12 target genes identified previously17. (e) Affinity-purified Flag-tagged mH2A1 and H2A complexes were analyzed by anti-HDAC2 and SUZ12 immunoblotting and colloidal Coomassie staining (Coo.). (f) Anti-EZH2 immunoprecipitates and input material were analyzed by western blotting using antibodies for the proteins indicated. The abundant nuclear SAF-A and heterogenous nuclear ribonucleoprotein C1 (hnRNP C1) proteins were included as negative controls.

Comparing our data with the SUZ12 ChIP-on-chip data generated by the Farnham laboratory17 showed that a highly significant amount of overlap between the gene sets (P < 2.2 × 10−16; Fig. 4d and Supplementary Table 2). If we take into account that we used different microarray platforms and that ChIP experiments were performed by slightly different methods, this is likely to be an underestimate of the true number of co-regulated genes. This overlap is further exemplified by the enrichment of SUZ12 and EZH2 in macroH2A1 precipitates compared to control precipitates with canonical H2A (Fig. 4e and data not shown). The enrichment of SUZ12 even exceeds that of HDAC2, a known interactor of macroH2A1 (ref. 4), which was included as a positive control. In the converse experiment, both macroH2A forms were enriched upon immunoprecipitation of EZH2 (Fig. 4f). This binding was sensitive to treatment with DNA-intercalating agents (data not shown), suggesting that these co-precipitations reflect co-occupancy of the same chromatin regions rather than direct binding of macroH2A1+2 to PRC2. Taken together, these data suggest that macroH2A1+2 and Polycomb proteins such as those of the PRC2 indeed co-operate in the regulation of developmental genes.

MacroH2A is required for proper zebrafish embryogenesis

To test the role of macroH2A in vivo, we decided to use zebrafish embryogenesis as a model system. The zebrafish genome possesses both a macroH2A1-encoding and a macroH2A2-encoding gene. The proteins are 70% identical to their human counterparts (Supplementary Fig. 6a). In contrast to many human cells, we found that zebrafish embryos predominantly expressed macroH2A2, whereas macroH2A1 was hardly detectable (Fig. 5a). MacroH2A2 was readily detectable by western blot (Supplementary Fig. 6b), and mRNA in situ hybridization indicated expression in various tissues, with particularly high levels in the developing brain (Fig. 5b).

Figure 5: MacroH2A is essential for normal zebrafish embryogenesis.
figure5

(a) MacroH2A (mH2A) expression in early zebrafish embryos. Quantitative RT-PCR analysis of both mH2A mRNAs in early zebrafish embryos at different time points (hours post fertilization (hpf)). Expression data were normalized to GAPDH expression, and data for different mH2A forms were normalized with identical amounts of corresponding plasmidic DNA. (b) Spatial expression of mH2A2 and negative control staining in the 24-hpf embryo is shown by in situ RNA hybridization. (c) Loss-of-function phenotype of a zebrafish embryo 48 h after injection of morpholinos (MO) directed against mH2A2 and control. Scale bars show 500 μm. (d) Characterization of morphants and control embryos at the 11 somite stage by in situ RNA hybridization. Arrowheads indicate alterations in regions corresponding to rhombomere 4 and the MHB. Scale bars show 100 μm. (e) Rescues were generated by injection of embryos with mH2A2 morpholino in combination with human (h.s.) mH2A2 mRNA. Dorsal views of embryos at the 11 somite stage are shown after in situ RNA hybridization for the MHB marker pax2a. Scale bars show 100 μm. Below, further-magnified sections of the MHB. (f) Chromatin was prepared from MO control and MO macroH2A2-injected embryos 24 hpf. ChIP demonstrates that macroH2A2 is specifically enriched on several hox genes (*, P < 0.05 with respect to IgG and MO macroH2A2-injected controls). Results are expressed as the mean ± s.e.m.

We used specific morpholinos targeting the translational start site of the macroH2A2 mRNA to efficiently knock down macroH2A2 protein expression (Supplementary Fig. 6c,d). Injection of these inhibitors resulted in several developmental defects, the most obvious being severe deformations of the body structure (Fig. 5c and Supplementary Fig. 6e). As in situ hybridizations showed that macroH2A2 is highly expressed in the head region, we decided to analyze the developing brain in more detail using expression of different genes as markers.

MacroH2A2 morpholino-treated embryos displayed several malformations in the brain (Fig. 5d). For instance, rhombomere 4 of the hindbrain was reduced in size compared with the other rhombomeres, as indicated by staining of the neighboring rhombomeres with a krx20 probe. Staining for pax2a further indicated a defect in the formation of the midbrain-hindbrain boundary (MHB). Noteably, a combination of similar rhombomere 4 and MHB malformations has been observed upon miR-9 RNA overexpression23, suggesting developmental co-regulation of these regions. As further shown by otx2 and hoxa3a staining, the extension of the most anterior head region was also reduced in macroH2A2-deficient embryos compared with normal embryos (Fig. 5d).

Next, we decided to test whether human macroH2A2 can rescue the phenotype of the morphants. Therefore, we again turned to pax2a staining of the MHB: whereas control animals displayed the circular disc of an intact boundary in a dorsal view, macroH2A knockdown animals showed a cleft separating the structure into two halves (Fig. 5e). Notably, co-injection of human macroH2A2 mRNA fully rescued this phenotype, thus revealing the functional conservation of macroH2A function across species. Finally, we found that macroH2A2 was specifically enriched on the promoters of several zebrafish hox genes, suggesting that target genes of macroH2A are also shared between different species (Fig. 5f). Taken together, these results demonstrate that macroH2A function is required for proper zebrafish development and is enriched at hox genes in this species too.

Discussion

The data presented here and elsewhere9,24 suggest that macroH2A1+2 occupies genes that need to be maintained in a chromatin state that is repressed but sensitive to signal-mediated activation. This would hold true for the many developmental regulators that we found to be macroH2A1+2 targets, which eventually need to be activated during cell fate switches in pluripotent cells. In NT2 cells, for instance, macroH2A1+2 target genes of the HOXA cluster were readily activated by administration of RA, which initiates the neuronal differentiation program. The fact that knockdown of macroH2A1+2 increased the sensitivity of HOXA genes to RA suggests that the macroH2A variants act as repressors in the fine-tuning of such developmental expression programs. In vitro data suggested that the presence of macroH2A in nucleosomes at the proximal promoter could interfere with transcription initiation7. However, our results indicate that macroH2A occupancy is minimal in the direct vicinity of the transcription start site, pointing toward an alternative mechanism of repression.

The spatial expression of macroH2A was not uniform in the zebrafish embryo, suggesting that macroH2A has specific functions for a subset of developmental processes or particular tissues. The defects observed upon elimination of macroH2A could be caused by the deregulation of a single pivotal key regulatory gene. However, taking into account that macroH2A targets many developmental genes, the loss-of-function phenotype is probably the consequence of alterations in the expression of several genes at once. In macroH2A-deficient animal, alterations in temporospatial gene expression patterns may not need to be strong to make them, in their sum, capable of perturbing the complex process of embryonic development. The generation of macroH2A-deficient embryos was technically facilitated by the fact that zebrafish embryos express only one of the two macroH2A forms. Otherwise, loss of a single macroH2A form might have been functionally compensated by the continuing presence of the other at the same chromatin regions. Knockout mice lacking only macroH2A1 developed normally24; the slower development of mice compared to zebrafish might further allow for additional repressive pathways to compensate for loss of macroH2A function. It remains to be seen how a double knockout of both macroH2A forms will affect mammalian development, but on the basis of our results in zebrafish we would anticipate specific embryonic defects.

The presence of the histone variant macroH2A is a dynamic nucleosome modification. On the HOXA cluster, macroH2A is removed upon activation with RA and is re-incorporated following removal of RA. It is conceivable that several protein complexes are involved in these loading and unloading processes. Likely candidates are complexes formed around histone chaperones and chromatin remodelers, which have both been shown to be able to catalyze similar reactions with other histone proteins. The histone chaperone HIRA, for instance, was shown to specifically mediate H3.3-containing nucleosome assembly25. Conversely, SWI/SNF-related remodeling complexes can mediate the removal of H2A–H2B dimers from a promoter upon hormonal activation26.

To fully understand the function of macroH2A, future work will have to thoroughly address both the regulation of macroH2A incorporation in determined loci and the consequence of its presence for the function of the surrounding chromatin.

Methods

Antibodies, reagents and plasmid.

We generated specific antibodies against macroH2A1 and macroH2A2 by immunization of rabbits with bacterially expressed and purified His6- or glutathione S-transferase (GST)-tagged macro domains. After terminal bleeding, we affinity purified antibodies using corresponding GST-tagged macro domains. ChIP-on-chip data sets obtained with our macroH2A1-specific antibody gave stronger signals when compared to the commercial antibody from Upstate. We also used the following antibodies: tubulin and Flag M2 (Sigma-Aldrich); SUZ12, H3 C-terminal, H3K27me3, H3K4me2, macroH2A1.2 and H4 Ac (Upstate); H3K4me3 (Diagenode), EZH2 BD43 and AC22 (ref. 27), hnRNP U 4G5 (provided by G. Dreyfuss, Howard Hughes Medical Institute), and hnRNP C1 (Santa Cruz). We generated CMV-promoter driven expression constructs and pRetroSUPER constructs by standard PCR and cloning techniques. Sequences of all oligonucleotides used are listed in Supplementary Table 3.

Cell culture and gene transfer.

We cultivated NT2 and HEK293T cells in DMEM supplemented with 10% (v/v) FCS. We performed infections as described28. For knockdown of macroH2A1 and macroH2A2, we sequentially infected cells with retroviral vectors expressing specific or control shRNAs and selected with 1 μg ml−1 puromycin and 100 μg ml−1 hygromycin.

Protein analysis.

Lysis, immunoprecipitation and western blot analysis were as described29. To solubilize chromatin, we lysed nuclei from cells expressing Flag-tagged macroH2A or H2A by sequential and progressive sonication of insoluble material. After binding to anti-Flag M2 agarose (Sigma-Aldrich), we washed complexes with lysis buffer and eluted with 200 ng ml−1 Flag peptide (Sigma-Aldrich). For western blot analysis of nuclear zebrafish proteins, we disaggregated whole embryos in hypotonic lysis buffer (10 mM Tris, 85 mM KCl, 0.5% (w/v) Triton X100, 1 mM PMSF) and collected nuclei by mild centrifugation.

Gene expression analysis and chromatin immunoprecipitation.

Following the supplier's instruction, we purified RNA from 2 × 106 cells using the Qiagen RNeasy Mini Kit, including DNase 1 digestion to avoid potential contaminations of DNA. We reverse transcribed 1 μg of total RNA using cDNA Synthesis Kit (Roche) and oligo-dT primers. We quantified the expression of target genes of interest by real-time quantitative PCR (Roche LightCycler) using the LightCycler FastStart SYBR Green 1 Kit from Roche. We generally normalized values to the expression of housekeeping genes. For the expression analysis, we labeled and hybridized cDNA to Agilent Human Expression Arrays following the supplier's instructions. We performed and analyzed ChIPs essentially as described30. We quantified immunoprecipitated DNA by real-time quantitative PCR. We listed all PCR primers in Supplementary Table 3. For ChIP-on-chip, we amplified samples following the modified WGA Sigma protocol described previously31. We carried out labeling and hybridization to Agilent Human Promoter Arrays following the supplier's guidelines. We performed Re-ChIP was as described32.

Data analysis and statistics.

We normalized microarray data using LIMMA software in Bioconductor33. For each slide, we corrected background with the 'normexp' method34, resulting in strictly positive values and reducing variability in the log ratios for genes with low levels of hybridization signal. We normalized each slide with the 'global loess' method and, finally, we adjusted the slide values for batch effect and scaled them by their median absolute deviation to make the arrays comparable. We performed analysis of macroH2A target genes from ChIP-on-chip data using an adapted version of the rank product nonparametric method proposed previously35,36 for gene expression analysis. First, we carried out the identification of probes with a significantly enriched signal (100 permutations per test) (P < 0.05). Multiple testing was taken into account by setting a false discovery rate (FDR) threshold. We statistically assigned significant probes to genes if they were within −5.5 kb upstream to +2.5 kb downstream of its annotated TSS (by GeneSymbol and RefSeq). We considered target genes if they had at least four significant probes among their associated probe set, and not fewer than five probes. We analyzed the macroH2A data sets against an unspecific antibody IgG used as a background control (threshold set at 0.01). For the gene ontology analysis, we calculated the P-values of Biological Process Gene Ontology terms by Fisher's exact tests using EASE37. To account for multiple testing, we corrected P-values with the Bonferroni method. We used DAVID38 to analyze enriched genes coding for proteins sharing PFAM protein domains. For the comparison of macroH2A and Polycomb target genes, we processed SUZ12 array data (GEO accession: GSM109884, generated by Squazzo et al.17) according to the method described above (100 permutations and an FDR threshold of 0.05). We assessed the significance of the set of overlapping target genes with Chi-squared test.

For the correlation of ChIP data with gene expression, we grouped transcripts in categories 0 (no signal), 1 (very low) to 4 (high) according to average spot intensity of corresponding probes on three Agilent human expression arrays. Delimitators of the groups were set arbitrarily to result in groups of similar size covering a similar range of values for categories 1–3 as far as possible. We normalized the distribution of macroH2A target genes in these categories for the distribution of all genes included in the ChIP-on-chip analysis.

To calculate the average profile of macroH2A, we computed the distribution of the binding signals on target genes with respect to the TSSs as follows: first, for each TSSs annotated in Ensembl, we built Agilent probe sets by collecting those probes lying in the neighboring −4 kb to +4 kb. Each probe was then extended to 500 bp, equally toward its 3′ and 5′ ends, to fit the average length of the sonicated ChIP fragments. We divided the number of bound probes at a given position by the total number of probes, bound or unbound, at that particular position.

Embryo injections and analysis.

We obtained macroH2A2 and control morpholinos from Gene Tools (listed in Supplementary Table 3). We generated capped, poly(A) tail mRNA as well as labeled RNA by in vitro transcription using the mMessage mMachine Kit (Ambion) and the Dig/Fluor RNA labeling Mix (Roche). We crossed wild-type zebrafish of the Wik strain, and we injected fertilized eggs with 5 ng of morpholino or 100 ng of mRNA. We staged the embryos as described39 and fixed them in 4% (v/v) paraformaldehyde in PBS overnight at 4 °C. We carried out in situ hybrizations essentially as described40.

Accession Codes.

Gene Expression Omnibus: Data from the ChIP-on-chip analyses have been deposited with accession code GSE17531.

Note: Supplementary information is available on the Nature Structural & Molecular Biology website.

Accession codes

Accessions

Gene Expression Omnibus

References

  1. 1

    Kouzarides, T. Chromatin modifications and their function. Cell 128, 693–705 (2007).

  2. 2

    Wolffe, A.P. & Matzke, M.A. Epigenetics: regulation through repression. Science 286, 481–486 (1999).

  3. 3

    Sarma, K. & Reinberg, D. Histone variants meet their match. Nat. Rev. Mol. Cell Biol. 6, 139–149 (2005).

  4. 4

    Chakravarthy, S. et al. Structural characterization of the histone variant macroH2A. Mol. Cell. Biol. 25, 7616–7624 (2005).

  5. 5

    Costanzi, C. & Pehrson, J.R. Histone macroH2A1 is concentrated in the inactive X chromosome of female mammals. Nature 393, 599–601 (1998).

  6. 6

    Angelov, D. et al. The histone variant macroH2A interferes with transcription factor binding and SWI/SNF nucleosome remodeling. Mol. Cell 11, 1033–1041 (2003).

  7. 7

    Doyen, C.M. et al. Mechanism of polymerase II transcription repression by the histone variant macroH2A. Mol. Cell. Biol. 26, 1156–1164 (2006).

  8. 8

    Agelopoulos, M. & Thanos, D. Epigenetic determination of a cell-specific gene expression program by ATF-2 and the histone variant macroH2A. EMBO J. 25, 4843–4853 (2006).

  9. 9

    Ouararhni, K. et al. The histone variant mH2A1.1 interferes with transcription by down-regulating PARP-1 enzymatic activity. Genes Dev. 20, 3324–3336 (2006).

  10. 10

    Nusinow, D.A. et al. Poly(ADP-ribose) polymerase 1 is inhibited by a histone H2A variant, MacroH2A, and contributes to silencing of the inactive X chromosome. J. Biol. Chem. 282, 12851–12859 (2007).

  11. 11

    Karras, G.I. et al. The macro domain is an ADP-ribose binding module. EMBO J. 24, 1911–1920 (2005).

  12. 12

    Kustatscher, G., Hothorn, M., Pugieux, C., Scheffzek, K. & Ladurner, A.G. Splicing regulates NAD metabolite binding to histone macroH2A. Nat. Struct. Mol. Biol. 12, 624–625 (2005).

  13. 13

    Changolkar, L.N. & Pehrson, J.R. macroH2A1 histone variants are depleted on active genes but concentrated on the inactive X chromosome. Mol. Cell. Biol. 26, 4410–4420 (2006).

  14. 14

    Eklund, E.A. The role of HOX genes in malignant myeloid disease. Curr. Opin. Hematol. 14, 85–89 (2007).

  15. 15

    Houldsworth, J., Heath, S.C., Bosl, G.J., Studer, L. & Chaganti, R.S. Expression profiling of lineage differentiation in pluripotential human embryonal carcinoma cells. Cell Growth Differ. 13, 257–264 (2002).

  16. 16

    Jørgensen, H.F. et al. Stem cells primed for action: Polycomb repressive complexes restrain the expression of lineage-specific regulators in embryonic stem cells. Cell Cycle 5, 1411–1414 (2006).

  17. 17

    Squazzo, S.L. et al. SUZ12 binds to silenced regions of the genome in a cell-type-specific manner. Genome Res. 16, 890–900 (2006).

  18. 18

    Boyer, L.A. et al. Polycomb complexes repress developmental regulators in murine embryonic stem cells. Nature 441, 349–353 (2006).

  19. 19

    Lee, T.I. et al. Control of developmental regulators by Polycomb in human embryonic stem cells. Cell 125, 301–313 (2006).

  20. 20

    Bracken, A.P., Dietrich, N., Pasini, D., Hansen, K.H. & Helin, K. Genome-wide mapping of Polycomb target genes unravels their roles in cell fate transitions. Genes Dev. 20, 1123–1136 (2006).

  21. 21

    Kuzmichev, A., Nishioka, K., Erdjument-Bromage, H., Tempst, P. & Reinberg, D. Histone methyltransferase activity associated with a human multiprotein complex containing the Enhancer of Zeste protein. Genes Dev. 16, 2893–2905 (2002).

  22. 22

    Kirmizis, A. et al. Silencing of human Polycomb target genes is associated with methylation of histone H3 Lys 27. Genes Dev. 18, 1592–1605 (2004).

  23. 23

    Leucht, C. et al. MicroRNA-9 directs late organizer activity of the midbrain-hindbrain boundary. Nat. Neurosci. 11, 641–648 (2008).

  24. 24

    Changolkar, L.N. et al. Developmental changes in histone macroH2A1-mediated gene regulation. Mol. Cell. Biol. 27, 2758–2764 (2007).

  25. 25

    Tagami, H., Ray-Gallet, D., Almouzni, G. & Nakatani, Y. Histone H3.1 and H3.3 complexes mediate nucleosome assembly pathways dependent or independent of DNA synthesis. Cell 116, 51–61 (2004).

  26. 26

    Vicent, G.P. et al. DNA instructed displacement of histones H2A and H2B at an inducible promoter. Mol. Cell 16, 439–452 (2004).

  27. 27

    Pasini, D., Bracken, A.P., Jensen, M.R., Lazzerini Denchi, E. & Helin, K. Suz12 is essential for mouse development and for EZH2 histone methyltransferase activity. EMBO J. 23, 4061–4071 (2004).

  28. 28

    Buschbeck, M., Hofbauer, S., Croce, L.D., Keri, G. & Ullrich, A. Abl-kinase-sensitive levels of ERK5 and its intrinsic basal activity contribute to leukaemia cell survival. EMBO Rep. 6, 63–69 (2005).

  29. 29

    Villa, R. et al. The methyl-CpG binding protein MBD1 is required for PML-RARα function. Proc. Natl. Acad. Sci. USA 103, 1400–1405 (2006).

  30. 30

    Frank, S.R., Schroeder, M., Fernandez, P., Taubert, S. & Amati, B. Binding of c-Myc to chromatin mediates mitogen-induced acetylation of histone H4 and gene activation. Genes Dev. 15, 2069–2082 (2001).

  31. 31

    O'Geen, H., Nicolet, C.M., Blahnik, K., Green, R. & Farnham, P.J. Comparison of sample preparation methods for ChIP-chip assays. Biotechniques 41, 577–580 (2006).

  32. 32

    Métivier, R. et al. Estrogen receptor-α directs ordered, cyclical, and combinatorial recruitment of cofactors on a natural target promoter. Cell 115, 751–763 (2003).

  33. 33

    Smyth, G.K. Limma: linear models for microarray data. in Bioinformatics and Computational Biology Solutions Using R and Bioconductor (Springer, New York, 2005).

  34. 34

    Ritchie, M.E. et al. A comparison of background correction methods for two-colour microarrays. Bioinformatics 23, 2700–2707 (2007).

  35. 35

    Breitling, R., Armengaud, P., Amtmann, A. & Herzyk, P. Rank products: a simple, yet powerful, new method to detect differentially regulated genes in replicated microarray experiments. FEBS Lett. 573, 83–92 (2004).

  36. 36

    Hong, F. et al. RankProd: a bioconductor package for detecting differentially expressed genes in meta-analysis. Bioinformatics 22, 2825–2827 (2006).

  37. 37

    Hosack, D.A., Dennis, G., Jr., Sherman, B.T., Lane, H.C. & Lempicki, R.A. Identifying biological themes within lists of genes with EASE. Genome Biol. 4, R70 (2003).

  38. 38

    Dennis, G. Jr. et al. DAVID: database for annotation, visualization, and integrated discovery. Genome Biol. 4, 3 (2003).

  39. 39

    Kimmel, C.B., Ballard, W.W., Kimmel, S.R., Ullmann, B. & Schilling, T.F. Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253–310 (1995).

  40. 40

    Thisse, C. & Thisse, B. High-resolution in situ hybridization to whole-mount zebrafish embryos. Nat. Protoc. 3, 59–69 (2008).

Download references

Acknowledgements

We are indebted to T. Rasmussen, G. Dreyfuss (Howard Hughes Medical Institute) for the hnRNP U 4G5 antibody, S. Dimitrov and C. Pujades for reagents and members of the Di Croce laboratory for helpful discussions. This work was supported by grants from the Spanish “Ministerio de Educación y Ciencia” and from “La Marató TV3” and Consolider. M.B. was supported by a Fellowship from Deutsche Forschungsgesellschaft (DFG) and I.U. by a predoctoral fellowship from Spanish “Ministerio de Educación y Ciencia”.

Author information

Affiliations

Authors

Contributions

M.B., I.U., I.W., A.G. and L.M. performed the experiments; M.B., P.R., D.M., R.G., H.L.-S. and L.D.C. analyzed the data; M.B. and L.D.C. designed the project and wrote the manuscript.

Corresponding author

Correspondence to Luciano Di Croce.

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–6 and Supplementary Tables 1–3 (PDF 4174 kb)

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Buschbeck, M., Uribesalgo, I., Wibowo, I. et al. The histone variant macroH2A is an epigenetic regulator of key developmental genes. Nat Struct Mol Biol 16, 1074–1079 (2009). https://doi.org/10.1038/nsmb.1665

Download citation

Further reading