Histone H3 lysine-to-methionine mutants as a paradigm to study chromatin signaling

See allHide authors and affiliations

Science  29 Aug 2014:
Vol. 345, Issue 6200, pp. 1065-1070
DOI: 10.1126/science.1255104

Chromatin mutations disrupt development

Histone proteins form the core packaging material for our genomic DNA, and covalent modifications to amino acid residues in their structure play an important role in the epigenetic control of gene expression. Herz et al. show that specific mutations in the residues that are normally modified to regulate expression cause severe disruption of normal development in the fruit fly. Similar mutations are known to be involved in a subtype of aggressive pediatric brain cancers. Insights into the epigenetic regulatory pathways disrupted by these mutations in Drosophila may suggest possible treatments for human cancers.

Science, this issue p. 1065


Histone H3 lysine27-to-methionine (H3K27M) gain-of-function mutations occur in highly aggressive pediatric gliomas. We established a Drosophila animal model for the pathogenic histone H3K27M mutation and show that its overexpression resembles polycomb repressive complex 2 (PRC2) loss-of-function phenotypes, causing derepression of PRC2 target genes and developmental perturbations. Similarly, an H3K9M mutant depletes H3K9 methylation levels and suppresses position-effect variegation in various Drosophila tissues. The histone H3K9 demethylase KDM3B/JHDM2 associates with H3K9M-containing nucleosomes, and its misregulation in Drosophila results in changes of H3K9 methylation levels and heterochromatic silencing defects. We have established histone lysine-to-methionine mutants as robust in vivo tools for inhibiting methylation pathways that also function as biochemical reagents for capturing site-specific histone-modifying enzymes, thus providing molecular insight into chromatin signaling pathways.

Histone proteins constitute the core of eukaryotic chromatin (1). SET domain–containing histone methyltransferase complexes such as complex of proteins associated with Set1 (COMPASS) and polycomb repressive complex 2 (PRC2) methylate lysine residues within the histone H3 amino-terminal tail and are essential for normal development (1). Establishing direct functions for modified lysine residues in histones is difficult because there are multiple histone gene copies in metazoans (2). Moreover, histone methyltransferase enzymes occur in multimember families with potential redundant activities and histone methylation–independent functions (3). In Drosophila, replacing all copies of histone H3 with H3 Lys27-to-Arg27 (H3K27R) in a clonal substitution recapitulates the phenotype of mutating E(z), the PRC2 H3 Lys27 methyltransferase gene, suggesting this mark is indeed required for PRC2-mediated repression (4). Single-allele mutations of histone H3.3 Lys27-to-Met27 (H3.3K27M) occur in a subtype of aggressive pediatric brain cancers (5, 6) and act in a dominant manner to deplete H3K27 methylation by inhibiting PRC2 methyltransferase activity (7, 8). Other histone H3 lysine-to-methionine mutants also possess dominant gain-of-function activities (8), making them attractive tools for in vivo functional studies of histone lysine modifications. Trimethylation of histone H3 Lys27 (H3K27me3) and Lys9 (H3K9me3) are associated with distinct forms of transcriptionally silenced chromatin. Histone H3K27me3 catalyzed by PRC2 is enriched at so-called facultative heterochromatin and is implicated in the silencing of key developmental genes, in particular the homoeotic gene clusters (9). By contrast, H3K9me3 is associated with “constitutive” heterochromatin at telomeres and centromeres (10).

We established wild-type histone H3.3, H3.3K27M, and H3.3K9M constructs with a C-terminal FLAG–hemagglutinin (HA) tag for tissue-specific overexpression in Drosophila. Overexpression of H3.3K27M in the posterior compartment of wing imaginal discs driven by engrailed-GAL4 caused a strong reduction in all three H3K27 methylation states and derepression of the PRC2 target gene Ultrabithorax (Ubx) (Fig. 1, A to H′, and fig. S1), thus phenocopying knockdown of the catalytic PRC2 subunit E(z) (fig. S2). Also, increased H3K27 acetylation was observed for H3.3K27M overexpression in Drosophila and mammalian cells and E(z)-RNAi in Drosophila (fig. S3). Genome-wide RNA sequencing (RNA-seq) analysis of H3.3K27M-overexpressing wing imaginal discs revealed up-regulated RNA transcripts for known polycomb target genes, including Ubx, wingless (wg), and the PRC1 subunits Posterior sex combs (Psc) and Suppressor of zeste 2 [Su(z)2] (Fig. 1, I and J; fig. S4, A and B; and table S1). Other Homeobox (Hox)–containing genes such as engrailed (en) and invected (in), and signaling pathway components such as cubitus interruptus (ci), were down-regulated upon H3.3K27M overexpression (fig. S4, C and D, and table S1). Moreover, flies expressing H3.3K27M under a tissue-specific Distal-less–GAL4 driver exhibit gross morphological defects—such as severe leg malformations and fusion phenotypes and malformed, reduced, or missing proboscis—and die around eclosion, phenocopying E(z)-RNAi under the same conditions (fig. S4, E to G).

Fig. 1 Overexpression of histone H3.3K27M results in loss of H3K27 methylation and derepression of polycomb target genes.

(A to H) Overexpression of H3.3WT-FLAG-HA or H3.3K27M-FLAG-HA in the posterior compartment of the wing imaginal disc. Green fluorescent protein (GFP) expression in green marks the posterior domain where the respective histones are overexpressed. Nonapostrophed panels represent merged images that include the GFP signal in green and the respective antibody signal in red. Apostrophed panels only contain the respective antibody signal in red with a white arrow pointing toward the posterior compartment. Overexpression of H3.3WT-FLAG-HA does not result in bulk changes in H3K27 trimethylation (A and A’), H3K27 dimethylation (C and C’), or H3K27 monomethylation (E and E’). Decreased levels of H3K27 trimethylation (B and B’), H3K27 dimethylation (D and D’), and H3K27 monomethylation (F and F’) can be observed when H3.3K27M-FLAG-HA is overexpressed in the posterior compartment of the wing imaginal disc. Ubx remains silenced in wing imaginal discs overexpressing H3.3WT-FLAG-HA (G and G’). However, Ubx becomes derepressed when H3.3K27M-FLAG-HA is overexpressed (H and H’). RNA-seq analysis of wing imaginal discs expressing either H3.3WT-FLAG-HA or H3.3K27M-FLAG-HA with a T80-GAL4 driver and ChIP-chip analysis for H3K27 trimethylation of wild-type wing imaginal discs. RNA-seq tracks shown are an average of two biological replicates. Polycomb target genes that are highly enriched for H3K27 trimethylation in wild-type wing imaginal discs, such as Ubx (I) and wg (J), are derepressed when H3.3K27M-FLAG-HA is overexpressed in wing imaginal discs.

Trimethylation of histone H3K9me3 by supressor of variegation 3-9 [Su(var)3-9] proteins is a hallmark of constitutive heterochromatin (11). Histone H3K9me3 serves as a binding substrate for heterochromatin protein 1 α (HP1α, also known as CBX5) (1215) and establishes a transcriptionally repressed state (1618). Euchromatic genes that become abnormally juxtaposed to heterochromatic regions are subject to transcriptional silencing through position-effect variegation (PEV) (16). Less is known about the direct role of H3K9 methylation in the regulation of gene expression. Indeed, studies in fission yeast point to H3K9 methylation–independent functions for the Su(var)3-9 homolog Clr4 in chromatin silencing (19). To test a direct role for H3K9 methylation in the regulation of gene expression in metazoans, we overexpressed H3.3K9M in Drosophila wing imaginal discs and mammalian cells and observed a global depletion of H3K9 methylation levels (Fig. 2, A to F, and fig. S5) but no effect on H3K27 methylation (fig. S6, A to C′). In contrast, H3K9 mono- and dimethylation were slightly reduced when H3.3K27M was overexpressed (fig. S6, D to F′). We purified mononucleosomes from wild-type H3.3-, H3.3K9M-, and H3.3K27M-overexpressing human embryonic kidney (HEK) 293 cells and subjected these samples to multidimensional protein identification technology (MudPIT) mass spectrometry (Fig. 2, F and G). The bindings of HP1α (CBX5), HP1β (CBX1), and HP1γ (CBX3) were substantially reduced for H3.3K9M-containing mononucleosomes, as were the interactions of the HP1-associated proteins chromatin assembly factor 1a (CHAF1A/p150) (20, 21) and CHAF1B/p60 (Fig. 2G). We also found substantially increased association of the H3K9 demethylase KDM3B and the H3K9/K56 deacetylase SIRT6 with H3K9M-containing mononucleosomes (Fig. 2G).

Fig. 2 Histone H3.3K9M overexpression results in depletion of H3K9 methylation levels and alters recruitment of HP1 family members and other H3K9-modifying enzymes.

(A to D) Histone H3.3WT-FLAG-HA and H3.3K9M-FLAG-HA were overexpressed in wing imaginal discs as described in Fig. 1. Overexpression of H3.3WT-FLAG-HA does not result in significant bulk changes of H3K9 trimethylation (A and A’), but leads to a marginal decrease in H3K9 dimethylation (C and C’). A substantial decrease in the levels of H3K9 trimethylation (B and B’) and H3K9 dimethylation (D and D’) can be observed when H3.3K9M-FLAG-HA is overexpressed. (E) Western blotting of whole-cell extracts from human HEK 293 cells expressing H3.3WT-FLAG-HA or H3.3K9M-FLAG-HA. Bulk histone H3K9 trimethylation levels and, to some extent, H3K9 dimethylation levels are reduced in cells expressing H3.3K9M-FLAG-HA. Dox, doxycycline-induced for 9 days. (F) Western blot of purified mononucleosomes from human HEK293 cells expressing H3.3WT-FLAG-HA or H3.3K9M-FLAG-HA. Histone H3K9 dimethylation and H3K9 trimethylation are strongly reduced on H3.3K9M-FLAG-HA–containing nucleosomes. (G) MudPIT analysis of mononucleosome-interacting proteins. dNSAF, distributed normalized spectral abundance factor; averaged of 9, 5, and 12 replicates for H3.3WT-FLAG-HA, H3.3K9M-FLAG-HA, and H3.3K27M-FLAG-HA, respectively. Green color indicates proteins significantly enriched in lysine-to-methionine mutants containing nucleosomes; red color indicates proteins that are depleted, based on power law global error model signal-to-noise ratios and P values; and gray indicates factors that are not significantly changed between wild type (WT) and lysine-to-methionine mutants.

Reduced dosage of Drosophila HP1 [α] [also known as Su(var)205] and Su(var)3-9 results in suppression of PEV (2224). By using a heat shock-inducible lacZ construct inserted within Y-chromosomal heterochromatin (25, 26), we found that overexpression of H3.3K9M results in suppression of PEV in both Drosophila salivary glands and eye-antenna imaginal discs (Fig. 3, A to D). Bulk histone H3K9 methylation levels were decreased in H3.3K9M-overexpressing salivary glands (fig. S7). We also assessed the effects of H3.3K9M overexpression on heterochromatic silencing in Drosophila ovaries. The gypsy-lacZ construct is normally silenced in almost all follicle cells but is up-regulated upon loss of heterochromatin function (27). Overexpression of H3.3K9M results in derepression of lacZ (Fig. 3, E to F’). Thus, the H3.3K9M mutation disrupts heterochromatic silencing of retroelements.

Fig. 3 Overexpression of H3.3K9M suppresses heterochromatic silencing.

Hsp70-lacZ variegating salivary glands overexpressing H3.3WT-FLAG-HA (A) and H3.3K9M-FLAG-HA (B), stained for lacZ expression by using X-gal. Hsp70-lacZ variegating eye-antenna discs overexpressing H3.3WT-FLAG-HA (C) and H3.3K9M-FLAG-HA (D). Arrows in (C) and (D) indicate the position of the morphogenetic furrow. gypsy-lacZ containing ovaries overexpressing H3.3WT-FLAG-HA (E and E’) and H3.3K9M-FLAG-HA (F and F’), stained for lacZ expression by using X-gal.

KDM3B is a JumonjiC domain-containing histone demethylase that shows specificity toward H3K9 and is involved in gene activation in leukemia cells (28, 29). Because KDM3B specifically interacts with H3.3K9M-containing nucleosomes (Fig. 2, G), we wanted to test whether changes in KDM3B levels would alter H3K9 methylation by knocking down or overexpressing its Drosophila homolog, JHDM2, in wing imaginal discs. Depletion of JHDM2 results in increased H3K9 mono- and dimethylation (Fig. 4, A to C’). Conversely, the overexpression of JHDM2 in wing imaginal discs results in depletion in H3K9 dimethylation levels and, to a lesser extent, H3K9 trimethylation (Fig. 4, D to F’) and suppresses PEV in both Drosophila salivary glands and eye-antenna imaginal discs (Fig. 4, G to L). JHDM2 and SIRT6 also globally affect H3K9 acetylation to a similar degree as H3.3K9M overexpression (figs. S8 and S9). Sirt6 is not a major regulator of PEV in eye-antenna imaginal discs and salivary glands, but Sirt6-RNAi results in a somewhat modest derepression of the gypsy-lacZ reporter (fig. S10).

Fig. 4 The histone H3K9 demethylase KDM3B/JHDM2 interacts with H3K9M and regulates H3K9 methylation levels and heterochromatic silencing.

(A to F) RNA interference (RNAi)–mediated knockdown of JHDM2 or JHDM2 overexpression were carried out in wing imaginal discs as described in Fig. 1. RNAi-mediated knockdown of JHDM2 results in increased levels of H3K9 monomethylation (A and A’) and H3K9 dimethylation (B and B’) but not H3K9 trimethylation (C and C’). JHDM2 overexpression does not affect H3K9 monomethylation (D and D’) but leads to decreased H3K9 dimethylation (E and E’) and a very weak reduction in H3K9 trimethylation (F and F’). (G to I) Hsp70-lacZ variegating salivary glands with reporter construct (Hsp70-lacZ) only (G); reporter construct and UAS-JHDM2 (H); and reporter construct, UAS-JHDM2, and daughterless-GAL4 driver (I) stained for lacZ expression by using X-gal. Compared with salivary glands containing only the reporter construct (G), salivary glands with the reporter construct and UAS-JHDM2 display a weak enhancement in PEV (H), possibly because of leakiness of the UAS-JHDM2 construct. PEV is further enhanced when UAS-JHDM2 is expressed under the control of daughterless-GAL4 (I). (J to L) Hsp70-lacZ variegating eye-antenna imaginal discs with reporter construct only (J); reporter construct and UAS-JHDM2 (K); and reporter construct, UAS-JHDM2, and daughterless-GAL4 driver (L) stained for lacZ expression by using X-gal. No X-gal staining is observed with the reporter construct only (J). A very weak enhancement is detected with the reporter construct and UAS-JHDM2 combined (K), possibly because of leakiness of the UAS-JHDM2 construct. PEV is further enhanced when UAS-JHDM2 is expressed under the control of daughterless-GAL4 (L), particularly in differentiating areas posterior to the morphogenetic furrow. (M) Model describing the activation of KDM3B/JHDM2 on H3K9M-containing nucleosomes and its possible role in heterochromatic silencing.

We used histone lysine-to-methionine mutants to globally modulate histone methylation in vivo. We established a Drosophila animal model of the H3K27M mutation, which may help elucidate the molecular pathogenesis of pediatric gliomas (5, 6). To gain mechanistic insight into the molecular function of these mutants, we used an unbiased proteomic strategy to identify histone lysine-to-methionine-interacting partners. Our biochemical studies do not identify PRC2 components, such as EZH2, SUZ12, and EED, as significantly enriched on H3.3K27M-containing nucleosomes (Fig. 2G) as previously suggested (8). However, we detected an increase in H3K27 acetylation levels (fig. S3) and association of bromodomain-containing protein 1 (BRD1) and BRD4 to H3.3K27M-containing nucleosomes (Fig. 2G). These findings suggest that inhibitors of H3K27 acetylation or BRD4 inhibitors, such as JQ1 and iBET, could be useful for the treatment of the H3.3K27M-mutated subtype of aggressive pediatric glioblastomas.

We also demonstrated that H3K9M globally depletes H3K9 methylation levels in vivo, disrupts interaction of HP1 proteins, and thus suppresses PEV. Via our unbiased proteomic strategy, we identified KDM3B/JHDM2 and Sirt6 as regulators of H3K9 methylation–dependent heterochromatic silencing (Fig. 4M). Indeed, JHDM2 acts as a suppressor of variegation in multiple tissues in our assays, whereas Sirt6 function seems to be restricted to retroelement silencing. Mutations of histone H3.3K36M were recently discovered in a subtype of bone cancer (30). Thus, histone lysine-to-methionine mutations are associated with highly tissue-specific cancer types. Given the importance of heterochromatin in maintaining genomic stability (17, 18), it is plausible that as-yet-uncharacterized H3K9M mutations might occur in some cancers. The system that we established will provide a powerful tool to inhibit histone lysine modifications at specific residues in vivo and allow to biochemically capture the molecular players involved in chromatin signaling pathways.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

References (3137)

References and Notes

  1. Acknowledgments: We thank T. Jenuwein, J. Müller, R. White, and the Bloomington Stock Center for providing antibodies and fly stocks; the Stowers Molecular Biology core for generating histone point mutants; the Stowers Tissue Culture Core for generating histone mutant cell lines; and S. Marshall and W. Hodges for technical assistance. Histone lysine-to-methionine studies in the Shilatifard laboratory were supported in part by NIH grant CA R01CA089455. Data generated for this study are deposited under Gene Expression Omnibus (GEO) accession number GSE59891. Chromatin immunoprecipitation (ChIP)–chip data of H3K27me3 are from GEO accession number GSE42106. Antibodies toward H3K27M were provided by Cell Signaling Technology. A.S. is a paid member of the scientific advisory board of Cell Signaling Technology.
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article