Control of transposon activity by a histone H3K4 demethylase in rice

Abstract

Transposable elements (TEs) are ubiquitously present in plant genomes and often account for significant fractions of the nuclear DNA. For example, roughly 40% of the rice genome consists of TEs, many of which are retrotransposons, including 14% LTR- and ∼1% non-LTR retrotransposons. Despite their wide distribution and abundance, very few TEs have been found to be transpositional, indicating that TE activities may be tightly controlled by the host genome to minimize the potentially mutagenic effects associated with active transposition. Consistent with this notion, a growing body of evidence suggests that epigenetic silencing pathways such as DNA methylation, RNA interference, and H3K9me2 function collectively to repress TE activity at the transcriptional and posttranscriptional levels. It is not yet clear, however, whether the removal of histone modifications associated with active transcription is also involved in TE silencing. Here, we show that the rice protein JMJ703 is an active H3K4-specific demethylase required for TEs silencing. Impaired JMJ703 activity led to elevated levels of H3K4me3, the misregulation of numerous endogenous genes, and the transpositional reactivation of two families of non-LTR retrotransposons. Interestingly, loss of JMJ703 did not affect TEs (such as Tos17) previously found to be silenced by other epigenetic pathways. These results indicate that the removal of active histone modifications is involved in TE silencing and that different subsets of TEs may be regulated by distinct epigenetic pathways.

Retrotransposons are RNA-mediated transposable elements (TEs), which are abundant in the genomes of both plants and animals (1, 2). Retrotransposons are classified into long terminal repeat (LTR) or non-LTR types, and they are mobilized in a “copy and paste” manner (3–5). The integration of a newly transposed copy might disrupt local gene structure and affect the expression of nearby genes. In humans, misregulation of retrotransposons causes numerous diseases (3, 6, 7).

Although transposition of TEs is a major driving force for genome evolution, host genomes have evolved diverse mechanisms to limit harmful mobilization (1, 7). DNA methylation and histone methylation are two reversible epigenetic modifications that control transposon activity (8–13). In plants, histone H3 that is dimethylated at residue K9 (H3K9me2) associates with methylated DNA sequences such as CpG, CHG, and CHH (where H = A, T, or C) and correlates with gene silencing, whereas H3 trimethylated at K4 (H3K4me3) is linked to active transcription (14–16). Histone methylation is highly dynamic. Lysine methylation is catalyzed by SET domain group (SDG) proteins and reversed by a family of Jumonji C (JmjC) domain-containing proteins, which use Fe (II) and α-ketoglutarate (αKG) as cofactors (17, 18). In Arabidopsis, JMJ25/IBM1 (increase in BONSAI methylation) encodes an H3K9 demethylase (19, 20). Mutation of IBM1 results in increased levels of H3K9me2 and DNA methylation in genes but not in transposons, indicating that IBM1 distinguishes genes from transposons to protect the active transcribed genes (19, 20). Moreover, mutations of JMJ14, which encodes an H3K4me3/2/1 demethylase, lead to decreased non-CG DNA methylation, compromising the maintenance of methylation but not de novo methylation activity (21–23). JMJ14 is also implicated in RNA-directed DNA methylation and the transcriptional silencing of endogenous TEs, although it is not fully clear whether such an activity directly limits transposition itself (24).

Rice (Oryza sativa L.) is a worldwide crop species and a model organism for monocotyledons. More than 40% of the rice genome consists of repetitive sequences or TEs (25–27), including 14% LTR retrotransposons and ∼1% non-LTR retrotransposons (28). Non-LTR retrotransposons are comprised of long interspersed elements (LINEs) and short interspersed elements (SINEs) (29–31). Several LTR retrotransposons, namely Tos2, Tos17, and Tos19 (32), and a non-LTR retrotransposon LINE element Karma (33), have been identified as mobile TEs in rice. Tos17, a Ty1-copia type retrotransposon was originally found to transpose under prolonged tissue culture conditions (32). During normal growth conditions, histone methylation and DNA methylation prevent Tos17 from mobilization (34). Knockdown of the rice gene SDG714, which encodes an H3K9 methyltransferase, reduces H3K9me2 and DNA methylation, resulting in transposition of Tos17 (8). Knockout or knockdown of the DNA glycosylase/lyase DNG701 leads to DNA hypermethylation and reduced expression of Tos17. Moreover, overexpression of DNG701 increases retrotransposition of Tos17 (9). In contrast to LTR retrotransposons, the factors that regulate the activity of non-LTR retrotransposons remain to be elucidated. Active LINE elements have been identified in rice and LINE elements constitute approximately 40% of the mammalian genome (3), but how they are controlled is largely unknown.

Here, we show that JMJ703 is a histone H3K4-specific demethylase in rice. Impaired JMJ703 leads to increased H3K4me3, enhanced genome-wide transcription, and pleiotropic developmental defects. In addition, two LINE elements, Karma and its N-terminal truncation, were identified as direct targets of JMJ703. These two elements display increased transposition frequency in jmj703 mutants, whereas the LTR retrotransposon Tos17 is not affected. Therefore, our work uncovers histone demethylation as a unique mechanism to control retrotransposon activity, which further strengthens the link between epigenetic silencing and genome stability.

Results

JMJ703 Specifically Demethylates Histone H3K4.

We previously identified 20 JmjC domain containing proteins in the rice genome and predicted JMJ703 as one of 13 potentially active histone demethylases (35). JMJ703 is the homolog of the Arabidopsis H3K4 demethylase JMJ14, which is involved in flowering time regulation and gene silencing (21–24). To determine whether JMJ703 is an active demethylase, we performed enzymatic activity assays in vivo and in vitro as described (22) (Fig. 1A). When JMJ703-YFP-HA was overexpressed, no differences in the levels at H3K9, H3K36, and H3K27 were observed between cells with or without YFP signal (Fig. S1A); in contrast, H3K4me1, H3K4me2, and H3K4me3 (H3K4me3/2/1) were dramatically reduced (Fig. 1B). Mutation of His394, a conserved iron-binding amino acid, to Ala completely abolished the demethylation activity (Fig. S1B). Statistical analysis also showed that JMJ703-YFP-HA, but not the mutated version, specifically demethylated H3K4 in vivo (Fig. 1B and Fig. S1B). In addition, immunoaffinity-purified JMJ703-YFP-HA but not JMJ703H394A-YFP-HA specifically decreased levels of H3K4me3/2/1 (Fig. 1C). Taken together, these results demonstrate that JMJ703 is an H3K4 demethylase both in vivo and in vitro.

Fig. 1.

JMJ703 is an H3K4 demethylase. (A) Schematic representation of JMJ703-YFP-HA construct. (B Left) Overexpression of JMJ703-YFP-HA reduces the levels of H3K4me3/2/1 in vivo. A JMJ703-YFP-HA fusion protein was transiently expressed in tobacco cells, and the nuclei were isolated for immunolabeling. Nuclei transfected with JMJ703-YFP-HA can be visualized by YFP signal (Left). Immunofluorescence with methylation-specific histone antibodies (Center) was used to analyze the substrate specificity of JMJ703. DAPI staining indicates the location of nuclei in each field (Right). Cells exhibiting expression of JMJ703 are marked by arrows. (Scale bars: 2 µm.) More than 50 pairs of nontransfected nuclei versus transfected nuclei in the same field of view were observed. Statistical analyses are shown at B Right. (C) JMJ703-YFP-HA reduces H3K4me3/2/1 in vitro. The methylation-specific antibodies are shown at Right. Anti-H3 was used as a loading control.

JMJ703 Loss-of-Function Mutant Displays Pleiotropic Phenotypes.

To investigate the biological function of JMJ703 in rice, we identified jmj703, a T-DNA insertion mutant from our mutant collection (36, 37). The T-DNA was inserted into the eighth exon of JMJ703 (LOC_Os05g10770) (Fig. 2A) and no full-length cDNA was detected (Fig. S2A).

Fig. 2.

jmj703 mutants display pleiotropic defective phenotypes. (A) Structure of JMJ703 and the T-DNA insertion mutation in exon 8. F1/R1 and F2/R2 are primers used for genotyping and RT-PCR, respectively. (B) jmj703 mutant plants show pleiotropic defective phenotypes. The mutant displays a semidwarf phenotype (B1). Gross morphology of single tillers (B2). Elongation pattern (B3) and length statistics of the uppermost six internodes (I to VI) (n = 30) (B4). Lamina joint angle phenotype (B5). Panicle morphology (B6), panicle structure (B7) and statistical analysis of rachis branches (n = 30) (B8). Grain morphology (B9) and statistical analysis of grain length, width and thickness (n = 100) (B10). In each graph, the wild type is on the left and the mutant is on the right. (Scale bars: B1–B3, B5, B6, 10 cm; B7, 5 cm; B9, 5 mm.) **P < 0.01 with Welch's t test. Error bars correspond to the SD of biological repeats.

jmj703 mutants displayed pleiotropic phenotypes, the most obvious of which was dwarfism. At mature stages, mutant plants were only 70% as tall as wild type (WT) (Fig. 2 B, 1 and 2). The top three internodes (I, II and III) were significantly shortened in the mutant (I, P = 1.9 × 10⁻¹³; II, P = 6.0 × 10⁻¹⁶; III, P = 3.6 × 10⁻¹⁵, Welch's t test, n = 30), whereas the lengths of other internodes (IV, V, and VI) were marginally affected (Fig. 2 B, 3 and 4). Moreover, jmj703 had erect leaves (Fig. 2 B, 5), and significantly decreased secondary panicle branches (P = 2.3 × 10⁻¹⁰, t test, n = 30) (Fig. 2 B, 6–8). In addition, the grains of mutants showed abnormal phenotypes, including reduced length, width, and thickness (P = 2.2 × 10⁻¹⁶, t test, n = 100) (Fig. 2 B, 9 and 10).

Fig. 3.

jmj703 activates retrotransposition of Karma. (A) H3K4me3 status of the genomic regions 50 Kb upstream and downstream of Karma. (B) Structure of Karma. Vertical lines above ORF1 show the EcoRV cleavage sites. The regions for ChIP-qPCR assay (regions 1–4), qPCR analysis (regions 1 and 2), and DNA gel hybridization probes (blue lines) are shown. B Lower shows anti-H3K4me3 ChIP-seq and RNA-seq data for Karma locus in WT and jmj703 mutants. (C) qPCR validation of ChIP- and RNA-seq for Karma. Anti-H3 was used as an internal reference for ChIP-qPCR. Os04g22450 was used as a negative control (53). eEF1α was used as an internal reference for qPCR. Error bars correspond to the SD. (D) DNA-methylation profiles of a 400-bp region containing Karma (as shown in B). The cytosine methylation status in WT and mutant are shown. The numbers on the x axis indicate the cytosine positions, and the y axis indicates the methylation levels. (E) The percentage of methylated cytosines in CpG, CHG, and CHH contexts in WT and mutant. (F) Retrotransposition analysis of Karma in cultivated calli and seedling leaves of WT and jmj703. Tubulin8 was used as a control for complete digestion of DNA by EcoRV. Black asterisks on top of the images represent the independent individuals. Red asterisks denote the positions of newly transposed Karma elements. DNA ladders are shown on the right.

Fig. 4.

Model of the distinct mechanisms controlling TE silencing. (A Upper) A TE located in a euchromatic region containing active genes (associated with H3K4me3, marked by green triangles) is silenced by additional repressive marks like H3K9me2. (A Lower) If H3K9me2 methyltransferase is impaired, the repressive marks are removed and the TE is activated. (B Upper) A TE located in a heterochromatic region is silenced by active H3K4me3 demethylase (with no H3K4me3, marked by gray triangles). (A Lower) If H3K4me3 demethylase is impaired, the active marks are not removed and the TE is activated.

When jmj703 was crossed with wild-type plants, all F1 plants showed a wild-type phenotype, and the F2 population segregated into WT and mutants, the latter of which uniformly displayed jmj703 phenotypes. The ratio of WT to mutants was 245:76 (X² = 0.30, χ² test), indicating that jmj703 is a recessive mutation. We generated JMJ703 RNAi transgenic plants and all progeny (n = 55) in the T1 generation with reduced JMJ703 mRNA levels resembled the jmj703 phenotype (Fig. S2 B–D). This observation further suggests that these pleiotropic phenotypes are caused by JMJ703 mutation.

Effects of JMJ703 on H3K4me3 and Gene Expression.

JMJ703 was expressed at relatively high levels in leaves of 7 d-after-germination (DAG) seedlings compared with all of the other tissues tested (Fig. S3). To further elucidate the function of JMJ703 in an unbiased manner, we performed chromatin immunoprecipitation sequencing (ChIP-seq) by using anti-H3K4me3 antibody. In the WT, 89% (12,814 of 14,422) of peaks overlapped with previously identified H3K4me3 peaks with similar patterns, in which 5′ transcriptional start sites are enriched (38) (Fig. S4 A and B). We found that 57 H3K4me3 putative binding sites covering 40 genes showed at least 1.5-fold increase in H3K4me3 in jmj703 compared with WT (Dataset S1). We also performed RNA-seq analysis and found that the T-DNA insertion disrupted JMJ703 transcription (Fig. S5A). From wild-type and jmj703 sequencing libraries, 14.3 million (86.5%) and 12.0 million (84.0%) contiguous reads were aligned to the rice genome, respectively (Fig. S5B). Approximately 51% of total matched reads were mapped to unique loci, representing 11,970 and 11,397 genes in WT and jmj703 mutants, respectively [reads per kilobase of exons per million mapped reads (RPKM) > 3] (Fig. S5B). Gene Ontology analysis revealed that 1,718 up-regulated genes (Dataset S2) or 2,172 down-regulated genes (Dataset S3) in jmj703 mutants are enriched in chromatin assembly functions (P = 1.0 × 10⁻⁷) (Fig. S6). Nine of 40 H3K4me3-hypermethylated genes overlapped with the up-regulated genes (P = 2.06 × 10⁻⁶) (Fig. S4C), whereas only two down-regulated genes showed increased H3K4me3 in jmj703 mutants (P = 0.27), suggesting that elevated H3K4me3 levels in the jmj703 mutant were preferentially associated with transcriptional up-regulation of corresponding genes. Several protein coding or hypothetical genes with increased H3K4me3 and up-regulated gene expression in jmj703 mutants were further confirmed by anti-H3K4me3 chromatin immunoprecipitation (ChIP)-quantitative PCR (qPCR) and qPCR using samples from independent biological replicates (Fig. S4B). Additionally, we found that four of these validated target genes showed the decreased CpG methylation at their 5′ regions with the increased H3K4me3 in the mutant compared with WT (Fig. S4B).

Karma Is Directly Associated with JMJ703.

Intriguingly, Karma, a non-LTR LINE-type retrotransposon (LOC_Os11g44750), showed up-regulated gene expression and significantly increased association with H3K4me3 in jmj703 (Fig. 3 A and B). LINE elements are composed of two ORFs, ORF1 and ORF2, which encode different proteins required for retrotransposition. In jmj703 mutants there was a sixfold increase of H3K4me3 at the 5′ region of ORF1 in Karma and a twofold increase of H3K4me3 in the other coding regions (Fig. 3C).

To investigate whether JMJ703 affects Karma DNA methylation, bisulfite genomic sequencing was performed in WT and jmj703. CpG methylation of the 5′ region of Karma was dramatically reduced from 75.89% in WT to 30.55% in jmj703, whereas non-CG methylation was marginally decreased (CHG from 5.27 to 1.55%; CHH from 3.34 to 1.57%) (Fig. 3 D and E). Therefore, our results indicate that JMJ703 deficiency leads to increased H3K4me3 that is inversely correlated with DNA methylation, suggesting that Karma is directly associated with JMJ703.

Prolonged tissue culture can promote Tos17 transposition (32). In contrast to Tos17, transposition of Karma only occurs in regenerated plants, but not in cultured cells (33). We found that Karma ORF1 but not ORF2 transposed in jmj703 mutants and integrated into extra sites in the rice genome (Fig. 3F). We also demonstrated that the retrotransposition of Karma occurred in mutant seedlings but not WT seedlings (Fig. 3F). Moreover, inheritance of retrotransposition were observed from self-pollinated mutant progenies indicated as the detection of de novo insertion site (Fig. S7). Therefore, we conclude that JMJ703 mediates H3K4me3 dynamics in vivo and reinforces the repressed state of Karma, thus preventing ectopic retrotransposition.

Effects of JMJ703 on a LINE Element.

We also looked for loci in addition to Karma that are potential direct targets of JMJ703. LINE1 (LOC_Os05g23140) has high similarity with Karma except for a 5′ deletion and several single nucleotide polymorphisms (33) (Fig. S8A). In jmj703, LINE1 transcription was up-regulated and the genomic DNA was associated with increased levels of H3K4me3 (Fig. S8 B–D). By genomic blotting, we also found a newly transposed band corresponding specifically to LINE1 but not Karma (Fig. S8E), indicating that JMJ703 targets more retrotansposons rather than an effect on a single target.

Karma Is in a Chromosome Niche Depleted in H3K4me3.

We further tested whether the JMJ703 mutation affected Tos17. Neither the transcription level nor histone and DNA methylation of Tos17 was changed in jmj703 mutants (Fig. S9 A–E). In addition, no transposition occurred in jmj703 mutants (Fig. S9F), suggesting that Tos17 is not a direct target of JMJ703. Previous studies indicated that Tos17 is heavily methylated by the histone H3K9 methyltransferase SDG714 (8). Tos17 transposition was observed in SDG714IR transgenic plants (Fig. S10A). However, Karma transposition was not observed in the same SDG714IR transgenic plants (Fig. S10B).

We analyzed the chromosomal regions surrounding Karma and Tos17 based on the genome annotation (http://rice.plantbiology.msu.edu/cgi-bin/gbrowse/rice/) (39). We found that the regions 50 Kb upstream and downstream of Karma and LINE1 are mainly composed of TEs and are heterochromatic in nature. Consistent with our H3K4me3 ChIP-seq dataset, these regions are depleted in H3K4me3 modifications (Fig. 3A and Fig. S8B). However, Tos17 is mainly flanked by protein-coding genes, resembling euchromatin and enriched in H3K4me3 (Fig. S9A). These results indicate that Karma and Tos17 are controlled by different epigenetic modifications on the basis of their chromosomal localization in rice.

Discussion

Using both in vivo and in vitro assays, we have identified JMJ703 as a unique functional histone H3K4 demethylase in rice (Fig. 1). Genome-wide RNA-seq analysis has revealed that the loss-of-function mutant jmj703 derepresses thousands of genes, especially those involved in chromatin assembly. jmj703 displays pleiotropic defects that could be due to ectopic expression or up-regulation of direct targets of JMJ703 (Fig. 2). Using a combination of ChIP-seq and RNA-seq analysis, we have identified the non-LTR retrotransposable elements Karma (Fig. 3) and LINE1 (Fig. S8) as direct targets of JMJ703. In jmj703, Karma displays increased levels of histone H3K4me3; reduced DNA methylation at CpG and, to some extent, CHG; and up-regulated gene expression (Fig. 3). These changes enhanced the movement of transposons that are normally silenced and immobile (33). Our findings show that plants use H3K4me3 demethylase to constitutively remove active chromatin marks and maintain the silent status of a subset of retrotransposons to preserve genome integrity.

Although mobile genetic elements in plants have been studied extensively, previous work has largely focused on the identification of active transposons and characterization of how they contribute to genome evolution and adaptation. The detailed regulatory mechanisms controlling transposition are not fully defined. Epigenetic regulation has been implicated in playing a major role in controlling transposon silencing. Genome-wide profiling in Arabidopsis revealed that methylation of H3K9me2 is required for the transcriptional silencing of transposons and repetitive DNA, whereas methylation of H3K4 is exclusively associated with gene expression (40). CHG methylation is linked to H3K9me2 in Arabidopsis (41, 42). So far, only a few mutants with defects in transposition of different types of transposable elements have been identified in Arabidopsis and rice (8–13). Among those mutants, inactivation of DNA methylation has been demonstrated to cause robust transposition in plants.

In Arabidopsis, DDM1 (decrease in DNA methylation 1) and DDM2 (also known as MET1), encode a putative chromatin remodeling protein and CpG maintenance methyltransferase, respectively (43). Mutations in ddm1 and met1 result in the loss of ∼70% of total DNA methylation especially at CpG sites. Chromomethyltransferase 3 (CMT3) is important for non-CpG methylation (43). In Arabidopsis cmt3 met1 double mutants, CACTA DNA transposable elements were transcriptionally activated and then mobilized (44). The transposition of CACTA and several other TEs was also observed in ddm1 mutants (13). Transcriptional activation is necessary but not sufficient for mobilization of retrotransposons because a posttranscriptional mechanism exists to prevent retrotransposition (45–47). This observation became evident when it was found that the mobility of Évadé (EVD), a copia-type retrotransposon, occurred when both RNA-directed DNA methylation and H3K9me2 were compromised (11). In addition, the siRNA pathway has been shown to prevent transgenerational retrotransposition of a copia-type retrotransposon ONSEN under heat stress (48). Recently, it has been demonstrated that there is a link between RNA-mediated silencing of transposable elements and demethylation of histone H3 by JmjC-domain containing proteins in Arabidopsis, although no direct mobilization of transposable elements was detected (19–21, 24).

In rice, activation of transcription and transposition of Tos17 was shown to associate with decreased DNA methylation (34). Additionally, decreased H3K9me2 in SDG714 RNAi transformants or overexpression of DNA glycosyase/lyase DNG701 also induced hypomethylation and activated transposition of Tos17 (8, 9). Here, we found that impaired function of the H3K4 demethylase JMJ703 did not affect transcription or transposition of Tos17. Instead, two LINE elements, Karma and its 5′ truncation LINE1, were transposed in jmj703 mutants.

Heterochromatin in rice is discontinuous and less distinct compared with that in Arabidopsis (49, 50). This scattered pattern of TEs throughout the rice genome may influence genes nearby, and differential epigenetic regulation might be required to specifically silence TEs but not regular genes. Our analysis indicated that Karma and LINE1 are embedded in heterochromatic genomic regions that are enriched in transposons and have low levels of H3K4me3 (Fig. 3A and Fig. S8B). In contrast, Tos17 is flanked by protein-coding genes with relatively high levels of H3K4me3 deposition (Fig. S9A). In addition, we confirmed that Tos17 but not Karma was mobile in SDG714IR transformants (Fig. S10). Therefore, we propose that specific epigenetic modifications are co-opted to distinguish TEs from genes. When a TE localizes in a region flanked by genes, additional repressive marks such as H3K9me2 might be required to reinforce its silencing status (Fig. 4A); however, if the TEs cluster in the host genome, H3K4me3 demethylase may be required to persistently remove the active marks (Fig. 4B). If H3K9me2 methyltransferase or H3K4me3 demethylase are impaired, TE silencing will no longer be maintained, resulting in reduced DNA methylation, and increased gene expression that may cause transposition (Fig. 4). These findings imply that appropriate epigenetic modifications are discriminatively used in the rice genome to specifically control TEs based on their local chromosomal niche.

Karyotype analysis of human chromosomes has shown that condensed heterochromatic regions are dispersed, similar to rice (51). LINE elements constitute approximately 40% of the mammalian genome (3). It has been shown that LINE-1 is active in humans and more than 90 diseases, including haemophilia, neurofibromatosis, and breast cancer, are caused by LINE-1 insertions (52). It is plausible that the human genome may adopt the same mechanisms as rice to control TE activity.

In conclusion, we have demonstrated a role of histone H3K4me3 demethylase in silencing retrotransposons in higher plant. This work uncovers a link between removal of an active epigenetic mark and retrotransposon silencing, which may shed light on tumorigenesis in humans.

Materials and Methods

Plant Materials and Growth Conditions.

Rice plants used in the study were Oryza sativa spp. japonica cv Dongjin. Primers for genotyping jmj703 are listed in Dataset S4. All of the transgenic plants were regenerated from transformed calli. Plants used for phenotypic investigation were grown in the field until maturity. Plant materials used for ChIP-seq or RNA-seq analysis were 7 DAG seedlings cultivated in the incubator with cycles of 14 h 30 min of light at 27 °C followed by 9 h 30 min of dark at 23 °C.

Transient Expression in Tobacco Leaves.

The constructs for JMJ703 or JMJ703H394A fused with YFP-HA were transformed into Agrobacterium tumefaciens cells (strain EHA105). These cells were then injected into Nicotiana benthamiana leaves, which were harvested for nuclear isolation and immunostaining or immunoprecipitation after 48 h (22).

In Vivo Histone Demethylation Assays.

Nuclear isolation and immunostaining were performed as described (22). Detailed information is available in SI Materials and Methods.

In Vitro Demethylation Assays.

JMJ703-YFP-HA and JMJ703H394A-YFP-HA were immunoaffinity purified with HA antibody from tobacco leaves then incubated with calf thymus histones (Sigma H9250) in demethylation buffer (20 mM Hepes⋅NaOH, 150 mM NaCl, 50 M Fe(NH₄)₂(SO₄)₂, 1 mM α-ketoglutarate, and 2 mM ascorbate at pH 8.0) for 3 h at 37 °C. The reaction products were subjected to Western blot to determine the enzymatic activities by using histone modification-specific antibodies.

ChIP and Data Analysis.

ChIP was performed as described with minor modification (53). Os04g22450 was used as a negative control for the rice anti-H3K4me4 ChIP assay as described (53). Primers used for ChIP-qPCR analysis are listed in Dataset S4. The detailed ChIP-seq data analysis is available in SI Materials and Methods.

RNA Sequencing and Data Analysis.

Total RNAs were extracted from 7 DAG seedling leaves of WT and jmj703 by using TRIzol Reagent (15596–026; Invitrogen). Polyadenylated RNAs were isolated by using a Dynabeads mRNA Purification Kit (610-06; Invitrogen). RNA-seq libraries were prepared by using the Illumina Directional mRNA-Seq Library Prep. v1.0 protocol and sequenced on an Illumina GAII to generate high-quality single-end reads of 80 nt in length. The detailed RNA-seq data analysis is available in SI Materials and Methods.

Bisulfite Sequencing.

Bisulfite genomic sequencing for Karma and Tos17 were performed as described (8). Primers used for bisulfite sequencing are listed in Dataset S4.

DNA Gel Blot Assay.

Rice genomic DNA (10 µg) was isolated from WT and jmj703 calli or seedling leaves, respectively, followed by digestion with 30 units of the appropriate restriction enzymes (New England Biolabs) for 6 h. The digested DNAs were separated on 1.0% agarose gels and transferred onto Hybond XL (Amersham). Blots were prepared and hybridized as described (54). Primers used for amplification of probes are listed in Dataset S4.