INTRODUCTION

The modulation of chromatin structure has emerged as a key level of regulation of gene expression in many tissues and stages of development (e.g. Stillman and Stewart, 2004). One level of chromatin modulation is via the covalent modification of histones, including methylation, acetylation and ubiquitination (reviewed by Fischle et al., 2003). Modified histone residues can serve as docking sites for downstream effector proteins: for instance, H3 methylated at Lys9 recruits HP1, and H3 methylated at Lys27 recruits the Polycomb repressive complex PRC1 (reviewed by Martin and Zhang, 2005). Certain histone modifications are known to lead to either the activation or repression of underlying genes (Fischle et al., 2003; Sims et al., 2003).

Within the germ line of Caenorhabditis elegans, both X chromosomes in XX hermaphrodites and the single X in XO males are silenced by global repression mechanisms involving modifications of histones (Kelly et al., 2002). The C. elegans proteins MES-2, MES-3, MES-4 and MES-6 have been implicated in this repression. Mutations in the mes genes result in maternal-effect sterility, due to defects in germ cell proliferation and necrotic degeneration of germ cells (Capowski et al., 1991; Garvin et al., 1998). Several findings suggest that this necrotic germline death is primarily a result of the aberrant expression of X-linked genes when silencing fails. First, mes mutant animals with two X chromosomes are more severely affected than mes mutants with only one X; in fact, single X animals are usually fertile (Garvin et al., 1998). Second, in germ cells of wild-type hermaphrodites the X chromosomes lack numerous marks of active chromatin (Kelly et al., 2002), whereas in germ cells of mes-2, mes-3 or mes-6 hermaphrodites the X chromosomes display those marks (Fong et al., 2002). Thus, the MES proteins are required for germ cell viability and probably function, at least in part, to silence the X chromosomes.

MES-2, MES-3 and MES-6 operate together in a complex (Ketel et al., 2005; Xu et al., 2001) and probably participate directly in X-chromosome silencing. MES-2 possesses a SET domain, a hallmark of histone methyltransferases (HMTs), and has been shown to have HMT activity on Lys27 of histone H3 (H3K27) (Bender et al., 2004), like its fly and vertebrate orthologs, E(Z) and EZH2, respectively (Cao et al., 2002; Czermin et al., 2002; Kuzmichev et al., 2002; Muller et al., 2002). The HMT activity of MES-2 requires the association of both MES-6, an ortholog of fly ESC and vertebrate EED, and MES-3, a novel protein (Ketel et al., 2005). Thus, the MES-2/MES-3/MES-6 complex resembles the Polycomb Repressive Complex PRC2 in its HMT activity, substrate specificity and certain partner requirements. In worms, the MES-2/MES-3/MES-6 complex is responsible for all detectable H3K27 methylation in most regions of the germ line and in early embryos (Bender et al., 2004). Notably, the MES-2/MES-3/MES-6 complex concentrates trimethylated H3K27 (H3K27me3) on the X chromosomes (Bender et al., 2004). This repressive mark is likely to contribute to the repressed state of the X chromosomes (Fong et al., 2002).

The function of MES-4 has until now been a mystery. MES-4 shows the unique property of associating with the five autosomes but not with the X chromosome (Fong et al., 2002). Here, we show that MES-4 is a histone H3 HMT that it is responsible for all detectable H3K36 dimethylation in most regions of the germ line and in early embryos, and that it concentrates H3K36me2 marks on the autosomes. In contrast to Set2-related H3K36 HMTs, which associate with elongating RNA polymerase II and methylate H3K36 within the coding regions of genes (e.g. Kizer et al., 2005; Morris et al., 2005), the binding of MES-4 to chromatin and its H3K36 HMT activity do not appear to depend on RNA polymerase II. This suggests that methylation of H3K36 can serve different roles in regulating chromatin function. Microarray analysis, performed on gonads dissected from wild type and mes-4 mutants, revealed that loss of MES-4(+) function results primarily in the upregulation of genes on the X chromosome. Our results suggest that in germline tissue MES-4 cooperates with MES-2/MES-3/MES-6 to achieve proper silencing of X-linked genes.

MATERIALS AND METHODS

Worm strains

C. elegans N2 variety Bristol was used as the wild type. The following mutations, balancers, and translocations were used.

LGII: rol-1(e91), mes-2(bn11), mnC1.

LGIV: mes-6(bn38), DnT1(IV;V), DnT1[qIs51](IV;V).

LGV: dpy-11(e224), mes-4(bn23, bn50, bn58, bn67, bn73, bn85, bn87), pgl-3(bn104), mnT12(IV;X).

RNAi analysis

RNAi was performed to deplete RNA Pol II/AMA-1, CDK-9 and TLK-1 as described by Kamath et al. (Kamath et al., 2003). Either wild-type or mes-4(bn73) L3 hermaphrodites were placed on plates containing dsRNA-expressing bacteria, at room temperature, and stained 36-40 hours later.

Immunofluorescence staining

Samples were fixed using methanol/paraformaldehyde (Han et al., 2003) or methanol/acetone (Strome and Wood, 1983). Rabbit antibodies to MES-4 were raised against the C-terminal 19 amino acids+Cys, or against amino acids 530-898, then affinity purified and used at 1:100 to 1:500 dilution. Other primary antibodies used were affinity-purified rabbit anti-H3K36me2 (Tsukada et al., 2006) at 1:200, mouse monoclonal antibody H5 to RNA Pol II CTD pSer2 (Covance) at 1:200, rabbit anti-H4K20me2 [Upstate and a gift from Yi Zhang (Fang et al., 2002)] at 1:100, chicken anti-H3K27me2 (Upstate) at 1:25, rabbit anti-H3K27me3 (Upstate) at 1:200, rabbit anti-acetylated histone H4 (Upstate) at 1:5000, mouse monoclonal antibodies PA3 at 1:1000 and PL4-2 at 1:2000 [a gift from M. Monestier (Monestier et al., 1994)], mouse monoclonal antibody OIC1D4 (Hird et al., 1996) at 1:5, and rat anti-PGL-3 (Kawasaki et al., 2004) at 1:10,000. Secondary antibodies from Jackson Immunologicals (TRITC-conjugated anti-rabbit IgG and anti-mouse IgM) and Molecular Probes (Alexa 488-conjugated anti-rat and anti-rabbit IgG) were used at 1:200. Images were acquired with an UltraVIEW LCI spinning-disk confocal laser and Nikon Eclipse TE200 microscope with UltraVIEW software (Perkin-Elmer), assembled with Adobe Photoshop, and displayed as projections of images taken at 0.5 μm intervals through the sample.

Bacterial expression of MES-4 and HMT assays

Full-length MES-4 cDNA was subcloned into a pET28b bacterial expression vector (Novagen). Transformed BL21-Gold (DE3) E. coli were grown to an OD₆₀₀ of 0.6 and induced with 0.1 mM isopropyl-D-thiogalactoside for 24 hours at 20°C. His-tagged MES-4 was detected in western blots using mouse monoclonal SC-8036 anti-His antibodies (Santa Cruz Biotechnology). HMT assays were performed by incubating 4 μl of bacterial lysate with 16 μg of chicken oligonucleosomes and 1.0 μCi of S-adenosyl-L-[methyl-³H]methionine (Amersham Pharmacia Biotech) in methyltransferase buffer (25 mM Tris-HCl pH 8.0, 5% glycerol) for 30 minutes at 20°C in a total volume of 20 μl. Half of the reaction was analyzed by SDS-PAGE, followed by Coomassie staining and fluorography.

Microarray analysis of RNA from dissected gonads

Dissection of gonad arms from hermaphrodites, isolation of RNA, and linear amplification were performed as described by Chi and Reinke (Chi and Reinke, 2006). Gonad arms were dissected from wild-type and mes-4(bn85) M⁺Z^- young adult hermaphrodites containing one to two fertilized embryos. During each of four dissection sessions, 100 gonad arms were collected from each genotype. Fluorescently labeled cDNA samples were prepared and hybridized to microarrays as previously described (Reinke et al., 2000). DNA microarrays were prepared as described elsewhere (Jiang et al., 2001; Reinke et al., 2004). Two hybridization experiments were performed with Cy3-labeled wild-type cDNA and Cy5-labeled mes-4 cDNA, and two with the dyes swapped. For every gene in each microarray hybridization experiment, the ratio of wild type (wt)/mes-4 was transformed into a log₂ value and the mean log₂ ratio calculated. Confidence levels were determined using a two-tailed paired t-test. Genes were considered to be significantly altered in the level of mRNA accumulation in wild type versus mes-4 if they displayed a mean fold-difference ratio of 1.8 or higher, and a confidence level of greater than 95% (P<0.05) (Whetstine et al., 2005). The GEO accession number for microarray data is GSE5454.

Real-time PCR analysis of RNA from dissected gonads

Fifty gonad arms were dissected from wild-type and mes-4(bn85) M⁺Z^- young adult hermaphrodites and total RNA isolated as described by Chi and Reinke (Chi and Reinke, 2006). Poly-adenylated cDNA was prepared using the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). Real-time PCR was performed in triplicate using iQ SYBR Green Supermix (Bio-Rad) and the iCycler iQ Multi-Color Real-Time PCR Detection System (Bio-Rad). The Autoprime program (www.autoprime.de) was used to design primers to span an exon-exon junction. All data were normalized to him-3 and F14B4.2, and the Pfaffl method (Pfaffl, 2001) was used to calculate relative fold changes.

RESULTS

MES-4 is a histone H3 lysine 36 methyltransferase

The presence in MES-4 of a SET domain (Fong et al., 2002) predicted that it has histone methyltransferase (HMT) activity. The mouse SET-domain protein bearing the closest similarity to MES-4 (Fong et al., 2002), NSD1, has been reported to methylate histones H3 and H4 in vitro (Rayasam et al., 2003). As shown in Fig. 1A, full-length 6his::MES-4 expressed in bacteria and incubated with chicken oligonucleosomes methylates H3. No other histones are detectably methylated.

We used antibodies specific for H3 and H4 peptides containing methylated lysines to investigate the residue specificity of MES-4. Mouse NSD1 has been reported to dimethylate H3K36 and H4K20 (Rayasam et al., 2003). Based on the staining of dissected C. elegans germ lines and early embryos with an antibody specific for H3 dimethylated at K36 (H3K36me2) (Tsukada et al., 2006), this mark is abundant on chromatin in wild-type worms but is absent from mes-4 mutants (Fig. 1B-I). Thus, MES-4 is required for H3K36 dimethylation in vivo, at least in germline tissue (see below). Given its H3 HMT activity in vitro, we propose that MES-4 functions as an H3K36 HMT in vivo. Based on the staining of wild type and mes-4 mutants, MES-4 is not required in vivo for any of the other H3 methyl marks we tested (H3K4me2 or me3, H3K9me2, H3K27me2 or me3, H3K79me2) or for H4K20me2 (Fong et al., 2002) (see Fig. S2E,F in the supplementary material; data not shown). The latter result agrees with the lack of H4 HMT activity of MES-4 in vitro. Furthermore, we were unable to detect H4 methylation activity for recombinant human NSD1 (amino acids 1556-1950) in vitro under conditions where strong H3 methylation was seen (data not shown). Therefore, we propose that both MES-4 and human NSD1 are specific for histone H3K36.

MES-4 is required for H3K36 dimethylation in most regions of the germ line and in early embryos

We used the antibody specific for H3K36me2 to investigate the tissue distribution of H3K36 dimethylation. H3K36me2 is abundant on the chromatin of most or all nuclei in wild-type worms (Fig. 1B-I; see also Fig. S1A in the supplementary material). To learn which tissues and stages require MES-4 function, we compared the H3K36me2 patterns in wild type to those in mes-4 adult hermaphrodites, specifically in the fertile F1 progeny of mes-4/+ heterozygotes. These F1 progeny have a maternal load of mes-4 products but no expression from the zygotic genome (we refer to them as M⁺Z^-). The embryos produced by M⁺Z^- hermaphrodites are M^-Z^- and develop into sterile adults. In mes-4 M⁺Z^- germ lines, H3K36me2 is undetectable in nuclei from the mitotically dividing distal region of the gonad (Fig. 1C) through the pachytene region (Fig. 1E). Some H3K36me2 signal is visible in late-pachytene/diplotene nuclei and in oocytes (Fig. 1G). In addition, some H3K36me2 staining is detectable in the polar body products of oocyte meiosis, although H3K36me2 staining is undetectable in the blastomere nuclei of mes-4 embryos through the∼ 35-cell stage (Fig. 1I, see Fig. 4B). Thus, MES-4 is responsible for all detectable H3K36 dimethylation in most regions of the adult germ line and in early stages of embryogenesis. A different H3K36 HMT(s) apparently is active in adult somatic cells (arrowheads in Fig. 1C,E), in the diplotene/diakinesis region of the oogenic germ line, and in >40-cell-stage embryos (Fig. 4D,F,H). These findings are consistent with the accumulation of MES-4 in the germ line and in early embryos (Fong et al., 2002), and with mes-4 mutants displaying defects primarily in the germ line (Capowski et al., 1991).

MES-4 and H3K36me2 are restricted to the autosomes and the left tip of the X chromosome

A striking feature of the wild-type H3K36me2 pattern in germline and early embryo nuclei is that a pair of chromosomes lacks this mark (Fig. 1D,H) and one oocyte bivalent has low levels of this mark (Fig. 1F). Because MES-4 is dramatically enriched on autosomes (Fong et al., 2002), we predicted that the chromosomes lacking H3K36me2 are the X chromosomes. H3K36me2 staining of X:autosome translocations in embryos confirmed that prediction: H3K36me2 is absent from the presumed X-chromosome portion of each translocation chromosome (Fig. 2B). Interestingly, in embryos containing an X:autosome translocation with a free left end of the X chromosome, we noticed a pinpoint spot of MES-4 (or of H3K36me2) staining on the exposed tip of the X portion (Fig. 2A,B). A `dot' of MES-4 was detected on 16 out of 31 X-chromosome left ends examined. A dot of MES-4 was never found on the right end (18 X-chromosome right ends examined; data not shown). Clues as to the significance of the dot of X-chromosome staining are considered in the Discussion.

Targeting of MES-4 to chromosomes requires the first PHD finger

MES-4 contains three CysHis fingers previously classified as PHD (plant homeodomain) motifs, plus an AWS-like domain and a post-SET domain flanking the SET domain (Fong et al., 2002) (Fig. 3A). We sequenced seven EMS- or gamma radiation-induced mes-4 mutations (Capowski et al., 1991; Fong et al., 2002) (Fig. 3A; see also Table S1 in the supplementary material). Some of the mes-4 lesions shed light on features of MES-4 required for its association with chromatin, in particular the first PHD domain.

Two point mutations in the first PHD finger, bn50 and bn67, lead to an apparently complete dissociation of MES-4 from chromosomes (Fig. 3B, row 4; data not shown). PHD domains are predicted zinc-binding fingers and are considered to be a signature of chromatin-associated proteins (Bienz, 2006). Studies of ISWI and p300 implicate their PHD domains in nucleosome binding (Eberharter et al., 2004; Ragvin et al., 2004), and more recently the PHD domains of ING2 and of BPTF have been shown to bind H3K4me3 (Li et al., 2006; Pena et al., 2006). Although the first PHD domain of MES-4 is fairly atypical (Bienz, 2006), it appears to be crucial for MES-4 to associate with chromatin.

The in-frame deletion allele bn85, which disrupts the SET domain, results in partial dissociation of MES-4 from chromosomes (Fig. 3B, row 3; see Table S1 in the supplementary material). The small amount of bn85 mutant protein bound to chromosomes does not lead to detectable H3K36me2 signal (data not shown). The partial dissociation of bn85 protein from chromosomes may implicate the SET and/or post-SET domains in having a minor role in the localization of MES-4 to chromatin.

The point mutation bn58, within the region between the third PHD domain and the AWS-like domain, also results in partial dissociation of MES-4 from chromosomes (Fig. 3B, row 2). bn58 is the only mes-4 allele that displays detectable H3K36me2 signal (Table S1 in the supplementary material), suggesting that the fraction of mutant protein that is associated with chromatin has at least partial HMT activity. This residual activity may confer upon bn58 animals the ability to produce more germ nuclei than other mes-4 mutants (Capowski et al., 1991).

Exclusion of MES-4 and H3K36me2 from the X requires MES-2, MES-3 and MES-6, and also depends on gamete history

Previously we showed that the SET-domain protein MES-2, in a complex with MES-3 and MES-6, is required for H3K27 di- and trimethylation in the C. elegans germ line (Bender et al., 2004), and that MES-4 patterns are altered in mes-2, mes-3 and mes-6 mutants. Specifically, in mes-2, mes-3 and mes-6 M⁺Z^- germ lines, MES-4 appears ectopically on X chromosomes in late oogenesis (Fong et al., 2002). Here, we have examined both MES-4 and H3K36me2 patterns in mes-2, mes-3 and mes-6 M^-Z^- early embryos (Fig. 2C,D). Consistent with our previous study, in early embryos MES-4, and also H3K36me2, spread to the oocyte-derived X chromosome. Thus, the activity of the MES-2/MES-3/MES-6 complex participates in repelling MES-4 and H3K36me2 from the X chromosomes at late stages of oocyte differentiation and in early embryos.

Curiously, in mes-2, mes-3 and mes-6 embryos, in contrast to the oocyte-derived X, the sperm-contributed X remains unstained (or is occasionally very faintly stained) by MES-4 and H3K36me2. After sperm chromatin is apparently stripped of histone modifications during spermatogenesis, an unidentified `imprint' on the sperm X chromosome causes it to reacquire H3K4me and H3K9/K14ac (acetyl) marks several cell cycles later than the autosomes, which acquire those modifications shortly after fertilization (Bean et al., 2004). This imprint is likely to explain the failure of MES-4 and H3K36me2 to decorate the sperm X in mes-2, mes-3 and mes-6 mutants.

The MES-2/MES-3/MES-6 complex may repel MES-4 and its HMT activity not only from the oocyte-derived X chromosome but also from particular regions of the autosomes. The pattern of H3K36me2 on autosomes is patchy or banded in appearance in wild type (Fig. 1D, see also Fig. S2 in the supplementary material), and appears more uniform and intense in mes-2, mes-3 and mes-6 nuclei (Fig. 2, compare rows B and D; Fig. S2, compare rows A and B, and rows C and D). An attractive mechanistic model to explain this finding is that MES-2-catalyzed methylation of H3K27 prevents MES-4 from binding to and methylating a nearby residue of the H3 tail, K36.

Another H3K36 HMT becomes active by the ∼40-cell stage of embryogenesis

Prior to the ∼35-cell stage of embryogenesis in mes-4 M^-Z^- embryos, no H3K36me2 is visible on chromosomes (Fig. 4B). As mes-4 embryos reach the ∼40-cell stage, several nuclei show a faint H3K36me2 signal (Fig. 4D). The intensity of staining increases as the embryos develop (Fig. 4F). These results demonstrate that at least one other H3K36 HMT, in addition to MES-4, becomes active in embryos. Notably, this non-MES-4 H3K36 HMT does not appear to be active in the primordial germ cell P4 or its newly formed daughters Z2 and Z3 (Fig. 4D,F), where MES-4 is active (Fig. 4C,E). Later, during the comma stage of embryogenesis, H3K36me2 appears in Z2 and Z3 of mes-4 embryos (Fig. 4H). H3K36me2 levels are somewhat higher in all nuclei of older wild-type embryos and L1 larvae when compared with mes-4 mutants (Fig. 4G,H), probably because of the contribution of MES-4 activity to overall levels of H3K36me2.

MES-4 HMT activity does not appear to be linked to transcription elongation

Because MES-4 is concentrated on autosomes, and because autosomes, but not X chromosomes, are actively transcribed in the germ line, we investigated whether MES-4 H3K36 HMT activity is directly associated with the progression of transcription. Studies in Saccharomyces cerevisiae showed that Set2, an H3K36 HMT, associates with the elongating form of RNA Pol II (Krogan et al., 2003; Li et al., 2002; Li et al., 2003; Schaft et al., 2003; Xiao et al., 2003). The emerging view is that Pol II phosphorylated on Ser2 of its C-terminal domain (CTD) recruits Set2, which in turns methylates nearby nucleosomes. Transcription elongation-coupled methylation of H3K36 appears to be widely conserved across eukaryotic species (Adhvaryu et al., 2005; Kizer et al., 2005; Morris et al., 2005; Sun et al., 2005).

To test whether MES-4-mediated H3K36 methylation is linked to transcription, we examined MES-4 localization and H3K36 methylation after RNAi depletion of the large subunit of Pol II, AMA-1, which contains the CTD. We focused on young embryos (<40 cells) in which all detectable H3K36 methylation requires MES-4 (see above). ama-1(RNAi) embryos lacked detectable CTD Ser2 phosphorylation (and in fact lacked detectable Pol II; not shown), but displayed normal-appearing MES-4 localization to chromosomes, and to autosomes in particular, and normal levels of H3K36me2 (Fig. 5A-F). Conversely, mes-4 mutant embryos displayed apparently normal levels of CTD Ser2 phosphorylation (data not shown). These findings suggest that MES-4 association with chromatin and HMT activity are not directly linked to transcription elongation, and that MES-4-mediated H3K36 methyl marks serve a different role in C. elegans than Set2-mediated H3K36me marks.

To investigate whether the non-MES-4 H3K36 HMT activity that becomes active in >40-cell C. elegans embryos is linked to transcription elongation, we examined whether the residual H3K36me2 mark in >40-cell mes-4 embryos is affected by ama-1(RNAi). H3K36me2 was undetectable in mes-4; ama-1(RNAi) ∼100-cell embryos (Fig. 6A-D). Depletion of CTD Ser2 kinases (CDK-9 or TLK-1) (Han et al., 2003; Shim et al., 2002) in a mes-4 background also eliminated H3K36me2 signal (see Fig. S3 in the supplementary material), revealing that the residual H3K36me2 signal in mes-4 embryos requires the phosphorylated form of CTD Ser2 in embryos. These results suggest that: (1) non-MES-4-mediated dimethylation of H3K36 is directly linked with transcription; and/or (2) the non-MES-4 H3K36 HMT must be transcribed from the embryonic genome.

Loss of MES-4 results in desilencing of X-linked genes

To investigate directly the impact of MES-4 chromatin regulation on gene expression, microarray analysis was used to compare the profile of mRNA accumulation in wild-type and mes-4 mutant germ lines. Ideally, we would compare the nascent germ lines in young wild-type and M^-Z^- mutant larvae, prior to the onset of germline degeneration in mutants. Because those germ lines contain very few cells, we instead analyzed isolated gonads dissected from wild-type hermaphrodites and from fertile mes-4 M⁺Z^-hermaphrodites. Several observations justified analyzing the M⁺Z^- generation. First, MES-4 levels are below detection in M⁺Z^- adults (see Fig. S1C in the supplementary material). Second, as described above, MES-4-mediated H3K36me2 is also undetectable in M⁺Z^- germ lines. Third, mes-4 M⁺Z^- germ lines are compromised: transgenes are desilenced, brood size is reduced, and RNAi depletion of other chromatin regulators renders mes-4 M⁺Z^- worms, but not wild-type worms, sterile (Capowski et al., 1991; Kelly and Fire, 1998; Xu and Strome, 2001). We reasoned that altered patterns of gene expression are likely to underlie these M⁺Z^- germline phenotypes, and that elucidating the alterations would provide insights into MES-4 function.

Gonads from wild-type and mes-4 M⁺Z^- gravid hermaphrodites were dissected, RNA linearly amplified, and cDNA prepared and labeled with Cy3 and Cy5. Hybridizations were performed in quadruplicate to DNA microarrays representing ∼16,400 of the ∼20,000 predicted C. elegans genes (Reinke et al., 2004). Seventy-one genes displayed significantly different accumulation in mes-4 relative to wild type (>1.8 average fold difference, P<0.05, two-tailed paired t-test, see Materials and methods; Fig. 7A,C; see also Table S2 in the supplementary material). The microarray results were validated by real-time PCR for 15 out of 15 genes (P<0.05) selected at random from the upregulated, downregulated, and non-regulated classes (see Table S3 in the supplementary material).

Two aspects of the altered gene expression profile in mes-4 are particularly striking and informative. First, 67 of the 71 affected genes displayed higher expression (upregulated) in mes-4 mutants relative to wild type, and four genes displayed lower expression (downregulated). Second, 61 of the 67 genes upregulated in mes-4 mutants are located on the X chromosome. Given the relative sizes of the six chromosomes and gene representation on our microarrays, random chance alone would predict that 10 of the 67 upregulated genes would map to the X. Thus, the primary effect of loss of MES-4 function on gene accumulation patterns in M⁺Z^- germ lines is upregulation of genes on the X chromosome.

Given the concentration of MES-4 on autosomes, we considered the possibility that an apparent upregulation of genes on the X chromosome in fact reflects widespread downregulation of autosomal genes in mes-4 mutants. In this scenario, MES-4 would serve as an activator of autosomal genes; loss of MES-4 would cause a reduced accumulation of autosomal transcripts and therefore of total mRNA, and our use of equivalent input mRNA to prepare microarray probes would result in artificially elevated levels of X-linked mRNAs. This scenario is highly unlikely for several reasons. (1) Absolute hybridization intensities for all ∼16,400 genes represented on the microarrays showed similar profiles in wild type and mes-4 mutants (Fig. 7B). (2) Given the banded appearance of MES-4 and H3K36me2 on the autosomes, downregulation of autosomal genes in mes-4 mutants would be expected to show gene-to-gene variation. Only 10 autosomal genes of the ∼14,000 represented on the microarrays showed a >1.8-fold difference in accumulation in mes-4 gonads relative to in wild-type gonads, and only four of those were downregulated in mes-4 mutants (Fig. 7A; see also Table S2 in the supplementary material). (3) If the downregulation of autosomal genes led to an apparent upregulation of X-linked genes in mes-4 gonads, then those genes on the X chromosome would be expected to be fairly uniformly affected. Instead, only 61 of the ∼2400 X-linked genes represented on the microarrays showed a >1.8-fold difference in accumulation in mes-4 gonads relative to wild type. (4) Real-time PCR analysis using non-amplified RNA from 50 dissected mes-4 and wild-type gonads verified the up- or downregulation of 10 genes, and the approximately equivalent accumulation of 5 genes (see Table S3 in the supplementary material).

Taken together, our results lead to the surprising conclusion that MES-4, a chromatin regulator concentrated on the five autosomes, functions to repress genes on the X chromosome. How X-chromosome silencing may be achieved by MES-4, in collaboration with the MES-2/MES-3/MES-6 complex, is discussed below.

DISCUSSION

MES-4 is an H3K36 methyltransferase in the germ line and in early embryos

Our in vitro HMT assays and in vivo staining results demonstrate that MES-4 methylates histone H3 on K36. The MES-4-related HMT, mouse NSD1, was reported to methylate H3K36 and, additionally, H4K20 (Rayasam et al., 2003). We do not see evidence that MES-4 participates in H4K20 methylation. Mouse NSD1 was further reported to methylate oligonucleosomes, core histones, and recombinant H3 and H4. In our assays, MES-4 and human NSD1 methylated oligonucleosomes, but not core histones or recombinant H3 or H4 (not shown). Our results suggest that both MES-4 and human NSD1 prefer nucleosomal substrates and are specific for H3K36. Mouse NSD1 may have a broader substrate range.

MES-4 is apparently the sole active H3K36me2 HMT in the regions of the germ line extending from the distal mitotic stem cells through the meiotic pachytene region. In those regions, MES-4 and H3K36me2 are excluded from all of the X chromosome except the leftmost tip (Fong et al., 2002) (this report), and the X chromosomes are silenced (Kelly et al., 2002). During oogenesis, the X chromosomes become at least partially activated late in pachytene (Kelly et al., 2002). This turn-on of the X chromosomes is not accompanied by a detectable appearance of MES-4 on them, which argues, as do our microarray results, against a model in which MES-4 is required to activate gene expression in the germ line. However, concomitant with X activation, an H3K36me2 HMT distinct from MES-4 becomes active on the X chromosomes as well as the autosomes, leading to methylation of H3K36 on all chromosomes. This non-MES-4 HMT may serve a Set2-like role during transcription elongation (e.g. Kizer et al., 2005).

In embryos, MES-4 remains autosomally concentrated until at least the 100-cell stage (data not shown), and is responsible for all detectable H3K36me2 until about the 40-cell stage, at which time another H3K36 HMT(s) also becomes active. The current view is that C. elegans early embryos inherit a large stockpile of maternal transcripts, initiate embryonic transcription of at least some genes by the four-cell stage, and undergo a `mid-blastula transition' from maternal to embryonic control of development at about the 40-cell stage (Baugh et al., 2003; Edgar et al., 1994; Seydoux and Fire, 1994). Methylation of H3K36 catalyzed by the non-MES-4 HMT, which becomes detectable by the 40-cell stage, is temporally correlated with activation of the embryonic genome and may serve an essential role in that process. This non-MES-4 HMT(s) is dependent on Pol II and thus may function similarly to yeast Set2.

H3K36me2 marks may serve diverse roles

In the yeast S. cerevisiae, Set2 catalyzes all H3K36 methylation and requires association with Pol II for this activity (Strahl et al., 2002; Kizer et al., 2005). H3K36me2 in S. cerevisiae is recognized by the Rpd3S complex, which deacetylates nucleosomes within gene coding regions, to aid in suppressing aberrant intragenic transcription initiation (Carrozza et al., 2005; Joshi and Struhl, 2005; Keogh et al., 2005). H3K36me2 marks also may serve to distinguish actively transcribed sequences from inactive genes and from regulatory sequences in yeast and higher eukaryotes (Bannister et al., 2005; Rao et al., 2005; Sun et al., 2005).

Our studies suggest that MES-4 associates with and methylates chromatin independently of Pol II. This in turn suggests that MES-4-catalyzed methylation of H3K36 serves a role distinct from those described above. The same may be true of the MES-4-related proteins in humans: NSD1, NSD2/MMSET and NSD3. Significantly, all three have been implicated in causing or promoting human cancers when mutated or overexpressed (Schneider et al., 2002). Our findings on C. elegans MES-4 invite speculation that the NSD family HMTs may also operate independently of Pol II.

MES-4 and H3K36me2 on the left tip of the X chromosome

MES-4 and H3K36me2 both decorate autosomes in a banded pattern and are excluded from all regions of the X chromosome except the left tip. The left end of the X chromosome has emerged as being different from the remainder of the X in several respects. The meiotic pairing center for the X chromosome is located less than 2 Mb from the left end (MacQueen et al., 2005). This region binds to the zinc-finger protein HIM-8 and associates with the nuclear envelope during meiotic prophase (Phillips et al., 2005). The left end of the X is also enriched relative to the rest of the X chromosome for AA/TT dinucleotides that are periodically spaced along one face of the DNA helix, and that may influence DNA bending and chromatin structure (Fire et al., 2006). The X-chromosome left end also contains three copies of a 14 base pair perfect repeat that is distributed abundantly over the autosomes (696 copies total), but found nowhere else on the X chromosome (I. Korf and J. Bedell, personal communication). Tests to date have not revealed an involvement of this 14-mer in MES-4 binding specificity (P. Poole, C.R.C. and S.S., unpublished).

In the germ line, MES-4 participates in silencing X-linked genes

Earlier observations that MES-4 is concentrated on the autosomes, and that in the germ line many autosomal genes are expressed, while the X chromosome is globally silenced, led to the expectation that MES-4 serves an activating role on autosomes. Our microarray results do not support that view and instead lead to the surprising conclusion that autosomally concentrated MES-4 somehow participates in repressing genes on the X chromosome. Although relatively few X-linked genes (61 of 2423 genes tested) were upregulated >1.8-fold in a mes-4 background, this small number is likely to be a consequence of performing the analysis in the M⁺Z^- generation. Such M⁺Z^- mes-4 mutant hermaphrodites are fertile, and so, although germline health is compromised in these individuals (see Results), they were not expected to display dramatic alterations in mRNA profiles. We envision that the maternal load of wild-type mes-4 gene product in young M⁺Z^- worms and its epigenetic influences are sufficient to enable the nascent germ line to launch the normal pattern of gene expression, but that as the maternal load declines and development proceeds, transcription patterns become altered. We predict that the M^-Z^- offspring of M⁺Z^- mes-4 mothers will show earlier and more dramatic alterations in germline gene expression patterns. Given the small size of the nascent germ line (i.e. two cells in a newly hatched 550-cell larva), testing that prediction is challenging, but is a high priority.

Among several models that might be postulated, we consider two to explain the derepression of some X-linked genes in mes-4 mutants. (1) MES-4 normally activates the expression of an autosomally encoded repressor that selectively represses genes on the X chromosome (Fig. 8, upper right). The expression of four autosomal genes is, in fact, downregulated >1.8-fold in mes-4 mutants. However, these genes do not possess motifs or show homologies that make them good candidates for serving as transcriptional repressors. Furthermore, RNAi depletion of each gene did not result in sterility (Gonczy et al., 2000; Kamath et al., 2003; Rual et al., 2004; Sonnichsen et al., 2005) (C.R.C. and S.S., unpublished). (2) An autosomal concentration of MES-4 or its H3K36me2 mark repels a global repressor, thereby concentrating repressor action on the X chromosomes (Fig. 8, lower right). We hypothesize that MES-2/MES-3/MES-6-catalyzed H3K27me3 concentrated on the X chromosome acts to repel MES-4 (Fig. 8, left), as suggested by the observation that MES-4 spreads onto the X chromosome in mes-2, mes-3 and mes-6 mutants (Fong et al., 2002) (this study), and that MES-4 and/or H3K36me2 concentrated on the autosomes repel an unidentified repressor (`R' in the figure). The possibility that this unidentified repressor is the MES-2/MES-3/MES-6 complex itself is unlikely, as the pattern of H3K27me3 catalyzed by that complex is not visibly altered in mes-4 mutants (see Fig. S2E,F in the supplementary material).

We postulate two types of X repression: a direct mechanism mediated by MES-2/MES-3/MES-6, and a novel, indirect mechanism mediated by MES-4. Loss of MES-2/MES-3/MES-6 would lead to the loss of repressive histone modifications and to at least some desilencing of the X chromosome. Loss of MES-4 would lead to insufficient levels of global repressor (model 1, Fig. 8) or promiscuous binding of the global repressor to autosomal regions, and, if the repressor is in limiting supply, to titration of repressor away from the X chromosome, causing at least some desilencing of the X (model 2, Fig. 8). One might expect promiscuous binding of the repressor to also cause widespread repression of autosomal loci, which we did not observe. If there is a limited supply of repressor and it is distributed over potentially six times more chromatin (10 autosomes in addition to the 2 Xs) in mes-4 mutants than in wild type, then the critical concentration of repressor needed for repression may not be achieved at most autosomal loci.

Our notion that MES-4 may have a role in repelling a global repressor from the five autosomes, thereby focusing repressor action on the X chromosome, has a precedent in van Leeuwen and Gottschling's proposed `gaining specificity by preventing promiscuity' model for the function of the S. cerevisiae H3K79 HMT Dot1 (van Leeuwen and Gottschling, 2002). Dot1 methylates ∼90% of H3K79 residues in the genome; notably, silent chromatin, which comprises ∼10% of the genome, is hypomethylated at that residue. Loss of Dot1 function and H3K79 methylation causes the SIR silencing proteins to spread from normally silent chromatin into euchromatin, and causes loss of silencing. This illustrates how a globally distributed histone modification can reduce nonspecific binding of silencers, and focus their binding and silencing effects to discrete domains.

MES-4 in the soma

Although MES-4 is not essential for the health and viability of somatic tissue (Capowski et al., 1991), recent studies of synMuv (for synthetic multivulva) genes have revealed a role for MES-4 in somatic cells (Unhavaithaya et al., 2002; Wang et al., 2005; Cui et al., 2006). Several synMuv class B mutants show a remarkable phenotype: somatic cells display germline traits, including expression of the germline marker PGL-1 and enhanced RNAi. Concomitant loss of mes-4 function suppresses the `ectopic germline traits' and other synMuv phenotypes, and in the case of mep-1 suppresses its larval lethality. This has led to a model in which MES-4 participates in conferring germline identity on cells, and the synMuv B regulators suppress or antagonize that function in somatic cells, thus protecting their somatic fates (Unhavaithaya et al., 2002; Strome, 2005). MES-4 is not alone in serving that proposed role; other mes genes and genes encoding additional chromatin regulators show similar genetic interactions with synMuv B mutants (Unhavaithaya et al., 2002; Wang et al., 2005; Cui et al., 2006). The targets of MES-4 regulation in somatic cells, and the mechanism by which MES-4 and synMuv B chromatin regulators antagonize each other, remain to be determined.

Supplementary material

Supplementary material for this article is available at http://dev.biologists.org/cgi/content/full/133/19/3907/DC1