The Fission Yeast Jmj2 Reverses Histone H3 Lysine 4 Trimethylation*

Histone methylation regulates transcription, chromatin structure, and the epigenetic state of the cell. Recent studies identified the JmjC domain as a catalytic module for histone demethylation. Schizosaccharomyces pombe contains seven JmjC proteins, but it was unclear whether any of them functioned as histone demethylases. In this report, we show that the JmjC protein Jmj2, which is evolutionarily conserved from yeast to human, reversed trimethylated H3-Lys-4 to di- and mono-but not unmethylated products. Overexpression of Jmj2 but not a catalytically inactive mutant reduced H3-Lys-4 trimethylation levels in vivo and suppressed the toxicity caused by overexpression of the H3-Lys-4-me3-binding protein Yng1 in budding yeast. Genome-wide analysis showed that the loss of jmj2 was associated with an increase in the H3-Lys-4-me3 signal, which was enriched near the transcriptional start sites and the coding regions. At the mating-type locus, the loss of jmj2 or substitution of jmj2 with a catalytically inactive form is correlated with increased reporter gene transcription and H3-Lys-4-me3/2 levels, suggesting that Jmj2 and its demethylase activity may play a role in heterochromatin biology. Our findings identified a novel S. pombe histone demethylase with specificity toward di- and trimethylated histone H3-Lys-4 and a possible role in heterochromatin regulation.

Histone N-terminal tails are modified by phosphorylation, ubiquitination, acetylation, and methylation. Each modification can affect chromatin architecture, but the total sum of these modifications may be the ultimate determinant of the chromatin state, which governs gene transcription (1,2). Histone methylation has been implicated in multiple biological processes including heterochromatin formation, X-inactivation, genomic imprinting, and silencing of homeotic genes (3).
Methylation occurs on both lysine and arginine residues, including H3-Lys-4-, Lys-9, Lys-27, Lys-6, H4-Lys-20, as well as Lys-9 located within the core of histone H3 (4,5). Methylation at these sites has been linked to transcriptional activation and repression, as well as DNA damage response (5,6), demonstrating a widespread role for histone methylation in various aspects of chromatin biology (3). An added level of complexity results from the fact that lysine residues can be mono-, di-, or trimethylated, which may represent distinct interaction surfaces for the recruitment of different factors that differentially impact the functional outcomes.
JmjC domain is a recently identified catalytic module for histone demethylation (13)(14)(15)(16)(17)(18). Its chemistry allows for this type of enzymes to mediate demethylation of mono-, di-, and trimethylated lysine residues (reviewed in Refs. 19 and 20). Indeed, recent studies identified a subfamily of JmjC domain-containing proteins that mediate demethylation of trimethylated histone H3-Lys-9 and H3-Lys-36 (15)(16)(17)(18). In Schizosaccharomyces pombe, there are seven JmjC domain proteins, but it was unclear whether any of them functioned as demethylases. In this report, we provide both in vitro and in vivo evidence that Jmj2 is an H3-Lys-4 trimethyl histone demethylase that catalyzes demethylation of H3-Lys-4-me3 to H3-Lys-4-me1/2 but had no effect on methylation at other lysine residues. Consistent with these observations, overexpression of Jmj2 resulted in a specific reduction of H3-Lys-4-me3/2 in vivo. Genome-wide ChIP 5 -on-chip analysis of H3-Lys-4 trimethylation identified a correlation between the loss of jmj2 and increase in H3-Lys-4-me3 levels, further supporting the notion that Jmj2 is an H3-Lys-4 tridemethylase. Importantly, overexpression of Jmj2 as well as the Sac-* This work was supported by National Institutes of Health Grants GM 071004 and NCI118487 (to Y. S.) and Grant GM 46498 (to S. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains two supplemental figures. 1 Supported in part by a fellowship from Fundacion Ramon Areces. 2
Expression Vectors-jmj2 cDNAs (wild type and H468A mutant) were amplified from S. pombe genomic DNA using Platinum Pfx polymerase (Invitrogen) and introduced into the Invitrogen Gateway Entry System. After sequencing of the plasmids, the cDNAs were transferred to the FLAG-baculovirus vector.
In Vitro Demethylation Reactions-Demethylation reactions were carried out with 2-10 g of purified protein using either 1 M peptide or 5 g of calf thymus type II-A histones (Sigma) as substrates in a buffer containing 20 mM Tris-HCl, pH 7.3, 150 mM NaCl, 50 M (NH 4 ) 2 Fe(SO 4 ) 2 , 1 mM ␣-ketoglutarate, and 2 mM ascorbic acid for 3-5 h at 30°C. The reactions were inhibited omitting ␣-ketoglutarate in the reaction buffer or adding 250 M DFO (Sigma-Aldrich).
MALDI-TOF Mass Spectrometry-1 l of demethylation reaction was analyzed by MALDI-TOF mass spectrometry on an Applied Biosystems Voyager DE-STR instrument as shown previously (18).
Microarray Design and Hybridization-A whole-genome tiling array was implemented using a total of 191,581 distinct 55-mer tiles with 65-bp spacing, designed using the forward strand of the 2003 S. pombe genome assembly. No T m matching was performed. Each spot was replicated twice on the array, and random GC content probes were included for quality control purposes. The platform design can be found at NCBI GEO under accession number GPL4749.
Analysis of Microarray Data-Nimblegen-supplied data files were imported into the Bioconductor computing environment.Afterperformingquantilenormalization,genotypedependent ChIP enrichments were estimated for each probe by fitting a fixed linear model accounting for array, dye, and genotype effects to the data using the LIMMA package (46). Moderated t-statistics and the log-odds score for differential expression or ChIP enrichment were computed by empirical Bayes shrinkage of the standard errors with the false discovery rate controlled to 0.05. Results were loaded into a relational data base for positional analyses into a Generic Genome Browser. Files containing per-probe signal estimates are available (see supplemental data), and all microarray data have been deposited under the provisional NCBI GEO series GSE8175.
RT-PCR-For RT-PCR, 5 g of RNA was used in a RT reaction by using random N9 primer.
Ura4 ϩ Reporter Assay-5-fold serial dilutions of exponentially growing liquid cell cultures were spotted on plates starting with 2 ϫ 10 5 cells. The strains were grown on rich medium (YEA), Edinburgh minimal medium supplemented with amino acids without uracil, and Edinburgh minimal medium supplemented with amino acids and uracil and 1 mg/ml 5-fluoorotic acid. Cells were growing for 4 days at 32°C.

RESULTS
Jmj2 Is a Histone H3-Lys-4-me3-specific Demethylase-The fission yeast S. pombe has seven JmjC domain-containing proteins (supplemental Fig. 1). Among them, Epe1 has been shown to counteract H3-Lys-9 methylation and heterochromatin spreading (21,22). Although Epe1 is an ortholog of the human histone demethylase JHDM1 and requires the integrity of the JmjC domain to antagonize heterochromatin spreading, it does not appear to have the demethylase activity when assayed on histone substrates. Although it lacks some of the key residues required for histone demethylation (14), whether Epe1 has demethylase activity on other protein substrates remains to be determined. Another jmjC containing protein, Msc1, has been shown to be involved in the regulation of the histone acetylation (23), underlining the evolutionary conservation of these proteins as chromatin regulators.
We took a candidate approach and, in this study, focused on one of the members of this family, Jmj2 (SPAC1002.05c). Jmj2 contains the JmjN and JmjC domains, which have been shown to be important for histone demethylation (24). It also contains an AT rich interacting domain (ARID)/BRIGHT domain, which is a DNA binding motif (25,26), and a C5HC2 zinc finger whose function is unknown (supplemental Fig. 1). To determine whether Jmj2 was a histone demethylase, we isolated a full-length jmj2 cDNA from S. pombe genomic DNA by PCR, subcloned it into a Gateway entry vector, and subsequently transferred it to N-terminal FLAG tag baculovirus expression vector for expression and purification to near homogeneity (supplemental Fig. 2A). We used a MALDI-TOF mass spectrometric approach to determine whether Jmj2 had histone demethylase activity, using as substrates peptides corresponding to the methylated histone N-terminal tails. As shown in Fig. 1, Jmj2 mediated demethylation at H3-Lys-4 but not at H3-Lys-9, H3-Lys-36, or H4-Lys-20. Jmj2 appeared to be capable of demethylating mono-, di-, and trimethylated H3-Lys-4 in the peptide-based assay. However, when assayed on bulk histones, Jmj2 acted predominantly as an H3-Lys-4-me3/2-specific demethylase, with no detectable effect on H3-Lys-4-me1/2 ( Fig. 2A). Like other JmjC domain demethylases, Jmj2 requires Fe(II) and ␣-ketoglutarate as cofactors for the reactions (14,18). Omission of ␣-ketoglutarate, as well as the point mutation of one of the conserved histidine residues important for iron chelation (H468A) (18,19,24,27), resulted in a robust inhibition of the demethylation reaction mediated by Jmj2 (supplemental Fig. 2, B and C).
Regulation of H3-Lys-4 Methylation by Jmj2 in Vivo-We investigated the ability of Jmj2 to regulate H3-Lys-4 methylation as a demethylase in vivo. We generated S. pombe strains in which the wild type Jmj2 or the catalytically inactive Jmj2 mutant (H468A) were under the control of the thiamine-repressible promoter nmt1, which in the absence of thiamine allows for high levels of expression of the gene of interest (28). As shown in Fig. 2B, overexpression of Jmj2 resulted in a signif-icant decrease in the H3-Lys-4-me3 level, as well as an appreciable reduction of H3-Lys-4-me2, but had no significant effect on H3-Lys-4-me1/2. This is consistent with the in vitro data on bulk histones ( Fig. 2A), suggesting that Jmj2 is predominantly an H3-Lys-4-me2/3-specificic demethylase. Importantly, overexpression of the catalytically inactive Jmj2 (H468A) had no effect on either H3-Lys-4-me3 or H3-Lys-4-me2 (Fig. 2B, compare lane 4 with lane 2, top and middle panel), indicating that the changes in H3-Lys-4-me2/3 levels upon overexpression of Jmj2 were likely to be due to its histone demethylase activity. However, deletion of Jmj2 had no discernable effects on global H3-Lys-4 methylation (Fig. 2B, lane 2, top panel). Yeast cells deleted of Jmj2 were viable, suggesting that Jmj2 is not an essential gene. The lack of change of the H3-Lys-4-me3 level could be due to either the presence of additional H3-Lys-4 trimethyl demethylases or the possibility that Jmj2 regulates only a small percentage of the genome; therefore, its ablation would not be expected to cause a global increase of H3-Lys-4-me levels. As a further control, mono-, di-, and trimethylation of H4-Lys-20 and trimethylation of H3-Lys-36 were shown to be unaffected by either the overexpression or the loss of Jmj2 (Fig. 2C), supporting the idea that Jmj2 is an H3-Lys-4-me-specific demethylase.
ChIP-on-Chip Identifies Genomic Regions That Are Regulated by Jmj2-Since deletion of the Jmj2 gene does not have a significant impact on the global H3-Lys-4 methylation levels, we next determined the effect of Jmj2 deletion by surveying the entire genome using ChIP-chip analysis. We noticed misregulation of H3-Lys-4 trimethylation where H3-Lys-4-me3 was significantly increased at diverse genomic loci in Jmj2⌬ (Fig.  3A). Specifically, average gene analysis, which includes all the open reading frames in the genome, showed that, in Jmj2⌬, H3-Lys-4-me3 was maximally enriched relative to wild type   JULY 27, 2007 • VOLUME 282 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 21665
cells around the transcriptional start site, as well as in the 5Ј half of the coding region and throughout the promoter (Fig. 3B). Significantly, this analysis also showed substantial decreases in H3-Lys-4-me3 levels in other genomic regions in Jmj2⌬ cells (Fig. 3A), including the 3Ј end of coding sequences (Fig. 3B). To confirm the ChIP-chip findings, we carried out independent chromatin IP experiments. As shown in Fig. 3C, both H3-Lys-4-me3 and H3-Lys-4-me2 were enriched relative to WT near the transcriptional start site at a number of gene loci upon deletion of jmj2. These included the genes spcc285.03, cgs2, and cox18 (right panels). This provided further in vivo evidence that Jmj2 functions as a demethylase that reverses H3-Lys-4 di-and trimethylation.
Jmj2 and Yjr119c Antagonize Yng1, Which Binds H3-Lys-4-me3 and Is Important for H3-Lys-4-me-directed Transcription-In budding yeast, the catalytic subunit of transcription factor II H (TFIIH), Kin28, phosphorylates serine 5 of C-terminal domain of polymerase II (CTD) to recruit Set1 containing COMPASS complex (29). Therefore, serine 5 phosphorylation of CTD and H3-Lys-4-me3 are enriched at the 5Ј end of genes (29,30). Recently, it was shown that NuA3 histone acetyl transferase (HAT) complex preferentially binds to H3-Lys-4-me3 through the plant homeo domain (PHD) finger of Yng1 and promotes H3-Lys-14 acetylation and transcription of a subset of genes (31,32). Yng1 overexpression results in a high toxicity that is strongly dependent on H3-Lys-4 methylation (32). It is possible that removal of Lys-4 methylation by overexpression of an H3-Lys-4 histone demethylase might suppress this toxicity. Therefore, Yng1-overexpressing cells were used to isolate a budding yeast-specific H3-K4 histone demethylase in vivo. High copy plasmids expressing each JmjC domain protein were introduced in cells expressing Yng1 from GAL1 promoter, and the resulting transformants were tested for the ability to grow on galactose medium. Confirming the previously published data, Yng1 overexpression strongly inhibited the cell growth, and the toxicity was completely abolished by deletion of the H3-Lys-4 methyltransferase SET1 (Fig. 4A). Interestingly, overexpression of Yjr119c, the ortholog of Jmj2 in budding yeast, also conferred resistance to Yng1 overexpression, suggesting that Yjr119c may function as a trimethyl-specific H3-Lys-4 histone demethylase in budding yeast (Fig. 4B). Importantly, four other JmjC domain proteins, Jhd1, Rph1, Gis1, and Ecm5, did not rescue the Yng1 toxicity (data not shown). We next examined the effect of Yjr119c on H3-Lys-4 methylation by Western blot analysis. As shown in Fig. 4D, H3-Lys-4-me3 was significantly reduced in Yjr119c-overexpressing cells but not in cells overexpressing a catalytic mutant yjr119c (H427A), which has a mutation in the Fe(II) binding site. Interestingly, overexpression of the mutant Yjr119c did not rescue the Yng1 toxicity, although the mutant protein was expressed at levels comparable with wild type Yjr119c (Fig. 4, B and D), indicating that Yjr119c rescues the Yng1 toxicity by directly regulating H3-Lys-4 methylation in vivo. Importantly, overexpression of Jmj2, but not the catalytically inactive mutant H468A, also rescued the Yng1 toxicity in the budding yeast (Fig. 4C). Taken together, these findings suggest that Jmj2 and Yjr119C are likely to play an important role in regulating transcription via demethylation of H3-Lys-4-me2/3.
Jmj2 Demethylase Activity Is Required for Regulation of Heterochromatin at the Mating-type Region-Fission yeast heterochromatin is characterized by H3-Lys-9 hypermethylation and H3-Lys-4 hypomethylation (33), raising the possibility that H3-Lys-4 demethylases may play a role in the establishment and/or maintenance of the heterochromatic regions. The ChIP-chip experiments did not reveal a significant misregulation of H3-Lys-4-me3 at the centromeres or subtelomeres, where H3-Lys-4-me3 is present at a very low level in both WT and jmj2⌬. The strain that we used for ChIP-chip does not contain the K-region, which is required for the efficient establishment of the silenced chromatin state at the mating-type region (34), precluding our analysis of any potential effects of Jmj2 in this heterochromatin region. To determine whether the heterochromatin in this region is affected by jmj2 deletion, we used a reporter strain that contains the ura4 ϩ gene inserted in the mating-type region, adjacent to mat3M (Fig. 5A). In this strain, due to the position effect variegation, ura4 ϩ remains silenced, and the cells are unable to grow in the absence of uracil. As shown in Fig. 5B, jmj2⌬ and the catalytically inactive point mutant H468A strains were able to grow slightly better in medium lacking uracil, as compared with cells with WT or overexpressed levels of Jmj2. This improved growth was most likely due to the ability to express the inserted ura4 ϩ gene since the mutant cells are sensitive to FOA (Fig. 5B). Chromatin IP identified significantly higher levels of H3-Lys-4 di-and trimethylation of the ura4 ϩ reporter gene in both the jmj2⌬ and H468A strains (Fig. 5C). This increase of H3-Lys-4 methylation was also correlated with a higher level of the ura4 ϩ mRNA detected by reverse transcription-PCR (Fig. 5D). These experiments suggest that Jmj2 might be involved in regulating H3-Lys-4 methylation and heterochromatin at the mating-type locus as a histone demethylase, either directly or indirectly. The exact role and mechanism of action of Jmj2 in heterochromatin regulation will require further investigation.
When histone peptides were used as substrates, Jmj2 demethylated all three methylation states at H3-Lys-4 in vitro (Fig. 2). However, its activity became restricted to H3-Lys-4-me3 and H3-Lys-4-me2 when bulk histone proteins were used as substrates in vitro (Fig. 3). Consistently, analysis of endogenous histone H3-Lys-4 methylation states upon overexpression of Jmj2 also showed restricted activity of Jmj2 toward tri-and dimethylated H3-Lys-4 (Fig. 4). Thus, it is most likely that the major histone substrates for Jmj2 are H3-Lys-4-me2/3. It is interesting to note that deletion of jmj2 or replacement of the wild type jmj2 with a catalytically inactive jmj2 had no effect on the global histone H3-Lys-4 methylation states in vivo (Fig. 4A). This suggests possible functional redundancy and the existence of additional H3-Lys-4 demethylases in S. pombe. Supporting this view, S. pombe has two other Jmj2-related JmjC domain proteins, Msc1 and Lid2, that share sequence and structural similarities to the human JARID1 family of proteins, composed of four members, JARID1A (RBP2), JARID1B (PLU-1), JARID1C (SMCX), and JARID1D (SMCY) (reviewed in Refs. 19 and 20), all of which have recently been shown to be H3-Lys-4 demethylases (35). Collectively, we have identified Jmj2 as a histone H3-Lys-4 tri-and dimethyl-specific demethylase, and this finding suggests that Msc1 and Lid2 may be H3-Lys-4 demethylases as well.
H3-Lys-4 trimethylation has been linked to active transcription (7)(8)(9)(10)(11). Consistent with this view, genome-wide ChIP-chip analysis identified H3-Lys-4 trimethylation to be enriched around transcriptional start sites (Fig. 3B). Importantly, deletion of jmj2 resulted in an increase in H3-Lys-4 trimethylation at these genomic loci, consistent with the idea that Jmj2 is an H3-Lys-4-me3-specific demethylase. Interestingly, jmj2 deletion showed lower than wild type levels of H3-Lys-4 trimethylation toward the 3Ј end of genes, which may contribute to our failure to detect significant global H3-Lys-4 methylation changes in the jmj2 deletion mutant by Western blot using modification-specific antibodies (Fig. 2B). Despite our efforts, we were unsuccessful at ChIP-chipping or ChIPping Jmj2. Although we consider it likely that Jmj2 directly demethylates histones at the target genomic loci, the lack of the ChIP result formally leaves open the question of whether the effect is direct or indirect.
In budding yeast, H3-Lys-4 methylation recruits NuA3 HAT complex to promote H3-Lys-14 acetylation and transcription of genes including YML062C and YGR157W (31). Yng1, a component of NuA3, binds to H3-Lys-4 methylation through its PHD finger, and genome-wide analysis showed that Yng1 is localized at the 5Ј end of genes enriched with H3-Lys-4-me3 (31,36). Consistent with this observation, we found that H3-Lys-14 acetylation together with H3-Lys-4-me3 was increased in jmj2 deletion mutant (data not shown). A, the Yng1 toxicity is dependent on H3-Lys-4 methylation. pRS316 and pGAL-YNG1 were introduced into wild type and set1⌬ mutant. The resulting transformants were grown in synthetic complete (SC) medium containing 2% glucose, and then 10-fold serial dilutions of yeast cells were plated on SC medium lacking uracil and tryptophan containing either 2% glucose or 2% galactose as a carbon source. SC plates containing glucose were photographed after 48 h, and plates containing galactose were photographed after 72 h. B and C, overexpressions of Yjr119c and Jmj2 confer resistance to Yng1 overexpression. The indicated plasmids were introduced into a wild type strain containing either pRS316 or pGAL-YNG1. Cells were spotted onto SC medium lacking uracil and tryptophan (B) or uracil and leucine (C) containing either 2% glucose or 2% galactose as a carbon source. SC plates containing glucose were photographed after 48 h, and plates containing galactose were photographed after 72 h. D, Yjr119c is a histone demethylase for H3-Lys-4-Me3 in budding yeast. The indicated plasmids were transformed into wild type or set1⌬ mutant strain, and the transformants were grown in medium containing 2% glucose as a carbon source. Equal amounts of chromatin fractions from the indicated cells were analyzed by Western blot analysis with anti-H3, anti-H3-Lys-4-me3, anti-H3-Lys-4-me2, anti-H3-Lys-36-me3, anti-hemagglutinin (HA), or anti-Rpb3 antibody. E, expression of Jmj2 in budding yeast. The indicated plasmids were transformed into wild type, and the resulting transformants were grown in medium containing 2% galactose as a carbon source. Equal amounts of total cell extracts from the indicated cells were analyzed by Western blot analysis with anti-hemagglutinin or anti-Rpb3 antibody.
Furthermore, we showed that both Jmj2 and Yjr119c rescue the Yng1 toxicity by directly antagonizing H3-Lys-4-me3 in vivo (Fig. 4). These findings suggest a negative role of these histone demethylases in transcriptional regulation. Demethylation of H3-Lys-4 by Jmj2 or Yjr119c might inhibit transcription by disrupting the Yng1-H3-Lys-4-me3 interaction, resulting in the loss of H3-Lys-14 acetylation at the 5Ј end of genes. Although the findings support the idea that Jmj2 and Yji119c regulate transcription via their demethylase activity, much still remains to be understood. For instance, it is unclear why we did not observe transcriptional changes of those genes where deletion of jmj2 resulted in an increase in H3-Lys-4-me3 (Fig. 3) (data not shown). Lastly, the biological function of Jmj2 is only beginning to be understood. Our current data suggest that Jmj2 likely plays a role in heterochromatin biology by regulating H3-Lys-4 methylation and transcription at the mating-type locus. Although ChIP-chip failed to identify H3-Lys-4-me3 alteration at the centromeric and subtelomeric regions, it is possible that a lack of an effect reflects functional redundancy. Thus, whether Jmj2 plays a role at other heterochromatin regions still remains to be determined.