Identification and Characterization of an Epi-Allele of FIE1 Reveals a Regulatory Linkage between Two Epigenetic Marks in Rice^,

This work identifies an epi-allele of rice Fertilization Independent Endosperm1 (FIE1) with DNA hypomethylation, reduced H3 Lys 9 dimethylation, increased H3 Lys 4 trimethylation, ectopic FIE1 expression, and loss imprinting, plus altered H3 Lys 27 trimethylation and perturbed expression of hundreds of genes. This suggests a regulatory link among these epigenetic marks.

Abstract

DNA methylation and histone H3 Lys 9 dimethylation (H3K9me2) are important epigenetic repression marks for silencing transposons in heterochromatin and for regulating gene expression. However, the mechanistic relationship to other repressive marks, such as histone H3 Lys 27 trimethylation (H3K27me3) is unclear. FERTILIZATION-INDEPENDENT ENDOSPERM1 (FIE1) encodes an Esc-like core component of the Polycomb repressive complex 2, which is involved in H3K27me3-mediated gene repression. Here, we identify a gain-of-function epi-allele (Epi-df) of rice (Oryza sativa) FIE1; this allele causes a dwarf stature and various floral defects that are inherited in a dominant fashion. We found that Epi-df has no changes in nucleotide sequence but is hypomethylated in the 5′ region of FIE1 and has reduced H3K9me2 and increased H3K4me3. In Epi-df, FIE1 was ectopically expressed and its imprinting was disrupted. FIE1 interacted with rice Enhancer of Zeste homologs, consistent with its role in H3K27me3 repression. Ectopic expression of FIE1 in Epi-df resulted in alteration of H3K27me3 levels in hundreds of genes. In summary, this work identifies an epi-allele involved in H3K27me3-mediated gene repression that itself is highly regulated by DNA methylation and histone H3K9me2, thereby shedding light on the link between DNA methylation and histone methylation, the two important epigenetic marks regulating rice development.

INTRODUCTION

DNA methylation is a reversible epigenetic modification that plays important roles in transposon silencing, genomic imprinting, and regulation of gene expression (Martienssen and Colot, 2001; Law and Jacobsen, 2010). In Arabidopsis thaliana, methylation occurs in three sequence contexts: CG, CHG, and CHH (where H = A, T, or C). Heavily methylated DNA, and therefore silenced, is usually found in heterochromatic regions, where transposons and repeated sequences are enriched (Zhang et al., 2006). By contrast, only ∼5% of genes are methylated within their promoter regions, and these promoter-methylated genes often show tissue-specific expression patterns (Zhang et al., 2006).

A few naturally occurring epigenetic alleles harboring misregulated DNA methylation have been reported. The first described epi-allele was clark kent (clk) in Arabidopsis, which contains hypermethylated cytosines at the SUPERMAN locus. The clk epi-allele showed phenotypes similar to loss-of-function mutants of SUPERMAN, with increased numbers of stamens and carpels (Jacobsen and Meyerowitz, 1997). The clk epi-allele was not stable and reverted back to normal phenotypes with reduction of DNA methylation (Jacobsen and Meyerowitz, 1997). In Linaria vulgaris, hypermethylation at cycloidea led to conversion of bilateral flowers to radial flowers (Cubas et al., 1999). In tomato (Solanum lycopersicum), spontaneous hypermethylation at Colorless non-ripening resulted in inhibition of fruit ripening (Manning et al., 2006). In rice (Oryza sativa), the sole epi-allele discovered so far is Epi-d1, which showed DNA hypermethylation at DWARF1. Epi-d1 showed a metastable dwarf phenotype caused by the epigenetic modifications of Epi-d1 (Miura et al., 2009b).

In Arabidopsis, FLOWERING WAGENINGEN (FWA) is an imprinted gene in which only the maternal allele expresses in the endosperm in the wild type (Kinoshita et al., 2004). In contrast with hypermethylated epi-alleles, fwa is a gain-of-function epi-allele that showed a late flowering phenotype caused by DNA hypomethylation at two direct repeat elements upstream of the FWA locus (Soppe et al., 2000). Thus, hypermethylation usually causes repressed gene expression, but hypomethylation usually causes increased or ectopic expression of the affected gene.

Histone H3 Lys 9 dimethylation (H3K9me2) and H3 Lys 27 trimethylation (H3K27me3) are two crucial epigenetic marks involved in heterochromatin silencing and transcriptional repression, respectively (Fischle et al., 2003; Liu et al., 2010). H3K9me2 is often associated with DNA methylation. Polycomb group proteins (PcG) were first identified in Drosophila melanogaster and are required for normal development of embryo segmentation, acting by repressing homeotic genes (Struhl, 1981; Frei et al., 1985). PcG proteins repress gene expression by trimethylating H3K27 at target loci (Cao and Zhang, 2004). Unlike Drosophila, which contains a single Polycomb repressive complex 2 (PRC2), Arabidopsis contains at least three distinct PRC2-like complexes, including the Fertilization-Independent Seed (FIS), Embryonic Flower (EMF), and Vernalization (VRN) complexes. Fertilization-Independent Endosperm (FIE; the homolog of Drosophila Esc) serves as a core component of all three PRC2-like complexes in Arabidopsis (Chanvivattana et al., 2004; Pien and Grossniklaus, 2007; Köhler and Villar, 2008; Liu et al., 2010). A loss of function mutation of FIE resulted in development of diploid endosperm without fertilization and in embryo abortion; therefore, this fie allele could not be transmitted through the female gametophyte (Ohad et al., 1996, 1999; Chaudhury et al., 1997). Downregulation of FIE also produced dramatic morphological aberrations resulting from derepression of KNOTTED-like homeobox genes and MADS box genes (Katz et al., 2004).

Rice, one of the most important food crops in the world, has become the model monocot species for functional genomic analysis. Unlike Arabidopsis, which has heterochromatin located primarily around the centromeres, rice has heterochromatin that is less distinct and dispersed throughout the genome (Cheng et al., 2001; Yan and Jiang, 2007). Such discontinuous heterochromatin and euchromatin may make rice a unique species in which to characterize gene regulation. In rice, DNA methylation and H3K9 dimethylation are required for maintenance of silencing at heterochromatin. For example, RNA interference–mediated inactivation of rice DNA methyltransferase 1 (MET1) led to reactivation of a silenced transgene and reduction of CG methylation at the CentO repeat and CRR2 retrotransposon (Teerawanichpan et al., 2004; Miki and Shimamoto, 2008). Knockdown of SET Domain Group Protein 714 (SDG714), which encodes an H3K9 methyltransferase, reduced CG and CNG methylation and H3K9me2 and enhanced the transcription and transposition of the Tos17 retrotransposon (Ding et al., 2007). DNA Glycosylase 701 (DNG701) is rice Repressor of Silencing 1 homolog and functions as a DNA glycosylase/lyase (Gong et al., 2002; La et al., 2011). Downregulation of DNG701 resulted in hypermethylation, reduced expression of Tos17, and less transposition, but upregulation of DNG701 led to hypomethylation, enhanced expression of Tos17, and more transposition (La et al., 2011).

PRC2-mediated gene repression is a highly conserved mechanism in both plants and animals. However, the exact composition of each PRC2-like complex varies among different species. The genomes of monocot plants, such as rice (Luo et al., 2009), barley (Hordeum vulgare) (Kapazoglou et al., 2010), and maize (Zea mays) (Springer et al., 2002), lack homologs of MEDEA (MEA) and FIS2, which encode FIS complex components involved in central cell repression and endosperm development in Arabidopsis (Kinoshita et al., 1999; Kiyosue et al., 1999; Luo et al., 1999, 2000). In addition, monocot plants have duplicated homologs of Extra Sex Combs (Esc), Enhancer of Zeste [E(z)], and Suppressor of Zeste 12 [Su(z)12], which encode components of the PRC2-like complex (Spillane et al., 2007; Luo et al., 2009; Rodrigues et al., 2010). For example, there are two Esc homologs in maize and rice. Zm FIE1 is exclusively expressed in developing endosperm and shows maternal-specific expression; Zm FIE2 is expressed throughout the plant (Danilevskaya et al., 2003). Similar to the maize Esc homologs, Os FIE1 is imprinted, with the maternal allele specifically expressed in endosperm, and Os FIE2 is expressed in a wide range of tissues (Luo et al., 2009). Unlike Arabidopsis FIE, loss-of-function Os FIE1 mutants did not show autonomous endosperm development (Luo et al., 2009).

Previous studies of H3K9me2 have focused on repression of transposons; the role of H3K9me2 in regulation of protein-coding genes has not been extensively studied. Furthermore, PcG proteins and their repression functions are not yet fully understood in monocot plants, especially in rice. In this study, we characterized a rice dominant dwarf and flower aberrant mutant, Epi-df, and found that its mutant phenotype was due to ectopic expression of FIE1. FIE1 displays maternal origin-specific expression in endosperm, and this imprinting pattern is regulated by DNA methylation and histone H3K9me2. The ectopic expression in Epi-df resulted from DNA hypomethylation, reduced H3K9me2, and increased H3K4me3 without alteration of the DNA sequence. Ectopic expression of FIE1 in Epi-df caused misregulation of hundreds of genes, associated with altered H3K27me3 levels. Our studies identify a DNA hypomethylation epi-allele in rice and demonstrate that repression of FIE1 via DNA methylation and H3K9me2 is essential for normal H3K27me3 distribution and plant development.

RESULTS

Rice Epi-df Is a Reversible Dominant Mutant

A dominant dwarf and flower aberrant mutant (df) was identified from anther-cultured rice plants (Qin et al., 2005). Compared with the wild type, df showed reduced plant height from the seedling to the mature stage, reaching about half the height of wild-type plants, and had more than twofold more tillers at the mature stage (Figures 1A and 1B). In df, each internode was evenly shortened relative to the wild type (see Supplemental Figure 1A online). When df was crossed with the wild type, the F1 plants showed a phenotype resembling df, indicating that the mutation acts in a dominant manner (Figure 1A). Also, the mutant phenotype was mitotically and meiotically heritable.

Figure 1.

Characterization of Epi-df, an Epigenetic Mutant Involved in Dwarfism and Floral Defects.

(A) Three-week-old seedlings of the wild type (left) and Epi-df (right).

(B) A wild-type (left), hybrid (middle), and Epi-df (right) plant at reproductive phase.

(C) The normal spikelet of the wild type.

(D) The mutated spikelets of Epi-df with loss of one empty glume (left), one elongated empty glume (middle), and two elongated empty glumes (right).

(E) The elongated lodicules of Epi-df.

(F) The normal spikelet of the wild type with removed palea and lemma.

(G) The mutated spikelets of Epi-df with removed palea and lemma showing five stamens (left), seven stamens (middle), and nine stamens (right). The pistils are also abnormal.

(H) The normal pistil with one carpel and two stigmas.

(I) to (M) The aberrant pistils of Epi-df. One carpel with four stigmas (I), two carpels with many stigmas (J), three carpels with many stigmas (K), and the aberrant carpels with amorphous tissues ([L] and [M]).

ca, carpel; eg, extra glume; le, lemma; lo, lodicule; pa, palea; st, stamen; stg, stigma; WT, the wild type. Bars = 10 cm in (A) and (B), 2 mm in (C) to (G), and 1 mm in (H) to (M).

A wild-type rice flower has one lemma, one palea, two lodicules, six stamens, and one central pistil with one carpel and two stigmas. Two degenerated flowers, known as empty glumes, are located under the normal flower (Figures 1C, 1F, and 1H). We noted that in df, only 6% of flowers developed normally, out of 200 flowers observed (see Supplemental Table 1 online). The mutant flowers showed a set of defects, including formation of extra palea or lemma (Figure 1D), elongated lodicules (Figure 1E; see Supplemental Table 1 online), and altered numbers of stamens and pistils. For example, 30.5% of flowers had fewer than six stamens, and 8.5% of flowers had more than six stamens (Figure 1G; see Supplemental Table 1 online). In addition, a large percentage of flowers harbored altered numbers of carpels or stigmas (Figures 1I to 1M; see Supplemental Table 1 online). As a result, df displayed a notable reduction in seed set (13.5% in df in contrast with 93.1% in the wild type), although it had normal pollen viability (see Supplemental Figures 1B to 1D online).

Although df showed heritable dwarfism and floral defects, normal plants (revertants) spontaneously occurred in the df-selfed population (see Supplemental Figures 2A and 2B online). For example, in 2006, a total of 67 revertants, a frequency of 1.97%, were observed among 3397 df progeny; in the 2008 and 2009 growing season, the frequencies of revertants were 0.31% (57 from 18,408) and 0.20% (35 from 17,555) respectively. Among the 67 revertants found in 2006, 38 still produced both mutant and wild-type plants in later generations (see Supplemental Table 2 online). These results suggest that df might be an epigenetic mutant that is heritable but unstable with a low frequency of spontaneous reversion back to a normal phenotype (Jacobsen and Meyerowitz, 1997; Cubas et al., 1999; Manning et al., 2006). We thus refer to df as Epi-df hereafter.

Cloning and Characterization of Epi-df

To understand the molecular mechanism responsible for the Epi-df phenotype, we isolated the mutant gene by map-based cloning. Using an F2 mapping population generated from a cross between Epi-df and PA64 (indica), Epi-df was mapped to a 49-kb DNA region containing seven open reading frames (ORFs) on the short arm of chromosome 8 (Figure 2A). No DNA sequence alteration was found in this region; however, examination of gene expression showed an upregulation of ORF5 (Os08g04290) in Epi-df but not in the wild type (Figure 2B). We also found that the ORF5 expression level was low in revertants (see Supplemental Figure 2C online). In their segregating progeny, normal plants had low ORF5 expression and the dwarf plants showed much higher ORF5 expression (see Supplemental Figure 2D online). The close association between the mutant phenotype and increased expression of ORF5 suggested that Os08g04290 corresponds to Epi-df. To confirm that the upregulation of ORF5 caused the Epi-df phenotype, we overexpressed ORF5 in Kasalath (indica) under the control of the maize ubiquitin promoter. All the 27 positive transformants showed high expression levels of ORF5 and displayed dwarfism and abnormal floral development (Figures 2C and 2D). A similar result was also observed when ORF5 was overexpressed in Nipponbare (japonica). Thus, we concluded that ectopic expression of Os08g04290 is responsible for the Epi-df phenotype.

Figure 2.

Molecular Cloning of Epi-df.

(A) Epi-df was first mapped to the short arm of chromosome 8 between markers SS-1 and SS-2 covering three BACs, AP005620, AP003896, and AP003876, and was further fine-mapped to a 49-kb region between markers In-4 and In-6 containing seven open reading frames (ORFs). The markers and numbers of recombinants (Rec) are indicated.

(B) RT-PCR analyses of the seven predicted ORFs in the 49-kb region. ORF3 and ORF5 corresponding to FIE2 and FIE1, respectively, are shown. ORF1 and ORF2, whose expression was not detected in the wild type or Epi-df, are not shown. Actin1 was used as a control. WT, the wild type.

(C) The morphologies of transgenic plants overexpressing FIE1 and the control plant (Kasalath, an indica variety).

(D) The flower morphologies of the transformed Kasalath T0 plants.

Bars = 10 cm in (C) and 2 mm in (D).

ORF5 (Os08g04290) encodes FIE1, a WD40 domain–containing protein. FIE1 is a close homolog of Arabidopsis FIE, a core component of PRC2 (Ohad et al., 1996, 1999; Luo et al., 1999) (see Supplemental Figure 3 online). The rice genome also contains a close homolog of FIE1, Os08g04270 (FIE2, corresponding to ORF3), which is located 13 kb upstream of Os08g04290. FIE1 has 15 exons and 14 introns and encodes a protein of 466 amino acids. The amino acid sequences of FIE1 and FIE2 are 75% identical, and they share 66 and 70% identity with Arabidopsis FIE, respectively. The high similarity among these Esc homologs suggests that FIE1 and FIE2 may also participate in PcG-mediated repression.

Hypomethylation of FIE1 in Epi-df

For FIE1, no nucleotide acid sequence difference was found between the wild type and Epi-df, and spontaneous revertants occurred in the mutant population, suggesting that Epi-df may be regulated in an epigenetic manner. We therefore investigated the DNA methylation status of the FIE1 locus. McrBC is an endonuclease that cuts methylated but not unmethylated DNA. After McrBC treatment, methylated DNA will be digested and therefore will not be amplified by PCR. We found that the 5′ region of FIE1 corresponding to fragments “a” (−72 to +408) and “b” (+246 to +817) was hypermethylated in the wild type but hypomethylated in Epi-df (Figure 3B). By contrast, the middle and 3′ regions of FIE1 corresponding to fragments “c” (+1833 to +2378) and “d” (+3650 to +4433) were not methylated in the wild type and Epi-df (Figure 3B).

Figure 3.

DNA Hypomethylation in the 5′ Region of FIE1 in Epi-df.

(A) Schematic representation of FIE1. Exons are shown as black boxes, introns and intergenic regions are shown as lines, and 5′ and 3′ untranslated regions are shown as gray boxes. Different regions for the McrBC-PCR analysis (a to d), H/M-PCR analysis (e and f), and DNA methylation-sensitive DNA gel blot probes (1 and 2) are shown. H/M indicates HpaII and MspI restriction enzymes, and E and P indicate EcoRI and PstI restriction enzymes.

(B) McrBC-PCR analysis of DNA methylation at FIE1. Equal amounts of genomic DNAs from the wild type (WT) and Epi-df (Mu) were digested with McrBC for 0, 0.5, 3, and 8 h, followed by PCR amplification of regions a to d shown in (A).

(C) HpaII-PCR and MspI-PCR analyses of DNA methylation at FIE1. Equal amounts of genomic DNAs from the wild type and Epi-df (Mu) were digested with HpaII or MspI for 3 h, followed by PCR amplification of regions e to g shown in (A); region g is used as a control that contains no HpaII-MspI site.

(D) DNA gel blot analyses of DNA methylation in the wild type and Epi-df (Mu). Genomic DNAs of the wild type and Epi-df (Mu) were double digested with E/H (EcoRI and HpaII), E/M (EcoRI and MspI), P/H (PstI and HpaII), and P/M (PstI and MspI), respectively, and the blots were hybridized with the probes shown in (A). The positions of the methylated (M) and unmethylated (U) bands are shown.

HpaII and MspI were also employed to detect the methylation status of the two CCGG sites within fragments “e” (+50 to +734) and “f” (+2931 to +3679). PCR results suggested that the outer C of the first CCGG lost its methylation in Epi-df compared with the wild type (Figure 3C). Furthermore, DNA methylation-sensitive DNA gel blot analysis verified the endonuclease-based PCR results (Figure 3D). When the CentO repeat, 5S rDNA, and Tos17 were used as probes, no differences were observed between the wild type and Epi-df (see Supplemental Figure 4 online), suggesting that the DNA methylation changes are specific to the 5′ region of FIE1.

Methylation Patterns of FIE1 in the Wild Type and Epi-df

Unlike animals, in which cytosine methylation predominantly occurs in the symmetric CG context, in higher plants cytosine methylation is found in both CG and non-CG contexts (Ehrlich et al., 1982; Cokus et al., 2008). To explore in detail the methylation status of FIE1, we used bisulfite sequencing to analyze a 6966-bp genomic DNA region consisting of a 1763-bp upstream sequence, the entire FIE1 gene, and a 242-bp downstream region (Figure 4A). All the altered methylation sites were found in a contiguous 1.8-kb (−802 to +1024) region in the promoter and 5′ region of FIE1 (Figure 4A). Compared with the wild type, Epi-df showed a substantial reduction in methylation at CG (from 94.78 to 5.86%), CHG (from 58.13 to 0.74%), and CHH (from 3.82 to 0.48%) sites across this 1.8-kb region (Figure 4B). We also analyzed the methylation patterns of this 1.8-kb region in three wild-type rice varieties, Asominori (japonica), Kasalath (indica), and PA64 (indica). All three varieties showed similar DNA methylation levels to the wild type used in this study (see Supplemental Figure 5 online). Thus, silencing of FIE1 by DNA methylation is conserved in rice varieties.

Figure 4.

Detailed Methylation Patterns of FIE1 in the Wild Type, Epi-df, and Six Revertants.

(A) Analyses of DNA methylation at FIE1 in the wild type (WT), Epi-df, and six revertants (Rev1 to 6) by bisulfite sequencing. Sequencing data were analyzed using Kismeth software (Gruntman et al., 2008). The region from −2006 to +4960 bp represents the entire sequenced DNA fragment; the region from −802 to +1024 bp represents the fragment that shows different methylation levels between the wild type and Epi-df. The translation start site was designed as +1. The colored lines indicate the percentage of methylation at individual cytosine sites. Two black bars under the Rev6 represent the common regions with high methylation level in six revertants.

(B) Histograms represent the percentage of CpG, CpHpG, and CpHpH methylation in the wild type, Epi-df, and six revertants.

We also analyzed the 1.8-kb (−802 to +1024) region in six spontaneous revertants of Epi-df (Rev1-Rev6). Surprisingly, even though FIE1 was silenced, DNA methylation was not restored to wild-type levels (Figure 4A). CG methylation in the six revertants ranged from 23.40 to 41.27%, which is much higher than that of Epi-df (5.86%), but still not up to wild-type levels (94.78%). CHG methylation was 58.13% in the wild type, but that of the revertants ranged from 5.62 to 8.33%. Cytosine methylation of CHH sites was also only partially restored (Figure 4B). The cytosine sites where methylation was restored in the revertants are enriched in two CG-rich regions, the first region is around the transcription start site and the second region is within the third exon (Figure 4A). These results suggest that methylation levels of CG rich regions are important for silencing of FIE1.

Alteration of Histone Methylation of FIE1 in Epi-df

Non-CG DNA methylation is usually associated with H3K9me2, a hallmark of transcriptional silencing (Bernatavichute et al., 2008). Given that there is high methylation level at CHG sites, we further compared H3K9me2 and H3K4me3 (associated with gene activation) levels at the FIE1 locus between the wild type and Epi-df by chromatin immunoprecipitation (ChIP) assays. The precipitated DNA was analyzed by real-time PCR using primer sets corresponding to six regions distributed across the 1.8-kb (−802 to +1024) hypomethylated region in Epi-df (Figure 5A). Compared with the wild type, H3K9me2 levels were reduced and H3K4me3 levels were increased within the 1.8-kb region in Epi-df (Figure 5B). These results are consistent with the expression pattern and DNA methylation pattern of Epi-df, suggesting that silencing of FIE1 in the wild type is regulated by DNA methylation and H3K9me2, and the ectopic expression of FIE1 is due to DNA hypomethylation and altered histone methylation.

Figure 5.

ChIP Analyses of FIE1 in the Wild Type and Epi-df.

(A) Schematic representation of the FIE1 gene, with exons shown as black boxes, introns and intergenic regions shown as lines, and 5′ and 3′ untranslated regions shown as gray boxes. The regions amplified to detect the chromatin status are indicated as black bars.

(B) ChIP with anti-H3K9me2 and anti-H3K4me3 antibodies to determine the levels of H3K9me2 and H3K4me3 at FIE1 in the wild type (WT) and Epi-df. Values are means ± sd of three biological replicates.

Expression and Imprinting Patterns of FIE1 Are Perturbed in Epi-df

FIE1 was reported to be an imprinted gene that is exclusively expressed in the endosperm from the maternal allele (Luo et al., 2009). The ectopic expression of FIE1 prompted us to test whether its overall expression and imprinting patterns were altered in Epi-df. In contrast with the wild type, which had no detectable expression of FIE1 in various vegetative tissues, FIE1 expression in Epi-df was detected in all tissues tested, including leaf, leaf sheath, culm, young panicle, and root (Figure 6A). We also dissected the embryo and endosperm and confirmed that FIE1 expression was only detected in the endosperm of the wild type (Figure 6B). In addition, FIE1 was highly expressed and reached its peak at 9 d after pollination (DAP), declining after this stage. After 15 DAP, almost no expression of FIE1 could be detected (Figure 6C). Consistent with a previous study, we also found that FIE2 was ubiquitously expressed in all tissues (see Supplemental Figures 6A to 6C online) and was not affected by the ectopic expression of FIE1 or revertants of Epi-df (see Supplemental Figure 6D online). Imprinting analysis showed that only the maternal transcript of FIE1 could be detected in hybrid endosperm; however, both parental transcripts could be detected when Epi-df was used as the paternal strain (Figure 7A), suggesting that imprinting is lost in Epi-df due to hypomethylation.

Figure 6.

Expression Pattern and Methylation Analyses of FIE1.

(A) Expression pattern of FIE1 in various organs of the wild type and Epi-df. C, culm; L, leaf; LS, leaf sheath; P, young panicle; R, root; S, seedling; WT, the wild type. Expression in the wild type was undetectably low in all tissues.

(B) Expression analysis of FIE1 in embryo (EM) and endosperm (EN) of the wild type at 1 week after pollination.

(C) Expression analysis of FIE1 in endosperm of the wild type at different stages (4, 6, 9, 12, 15, 18, and 21 DAP).

(D) Methylation patterns of the 1.8-kb 5′ region of FIE1 in various organs of the wild type.

Relative expression levels of FIE1 in various tissues were compared with that in seedling of Epi-df in (A), endosperm in (B), and 15 DAP in (C). Values are means ± sd of three biological replicates in (A) to (C).

Figure 7.

Imprinting Analysis of FIE1 in Endosperm.

(A) Allele-specific RT-PCR shows maternal allele-specific expression of FIE1 in endosperm at 1 week after pollination, and binary expression of FIE1 could be detected when using Epi-df (Mu) as the paternal strain. 93-11 (indica) was used to intercross with the wild type (WT) or Epi-df.

(B) Schematic representation of FIE1 showing the promoter (line before the first box), the first three introns (lines), the first four exons (boxes), and the 5′ untranslated region (gray boxes). The DNA methylation analyzed regions 1 and 2 under the gene model are shown.

(C) Percentage of methylation of two regions of maternal and paternal alleles, respectively, in endosperm at 1 week after pollination. These two regions could be distinguished by two single nucleotide polymorphisms (T/G in intron 2 and G/A in exon 3) between the wild type and 93-11.

To clarify the mechanism of the expression and imprinting of FIE1, we analyzed the methylation patterns of the 1.8-kb region (−802 to +1024) in different tissues. DNA methylation levels in leaf, culm, and young panicle were much higher than in endosperm at 6, 9, and 12 DAP (Figure 6D). Using two single nucleotide polymorphisms (T/G in intron 2 and G/A in exon 3) within the 1.8-kb region between the wild type and 93-11(an indica variety), we also analyzed methylation patterns of each parental FIE1 allele in hybrid endosperm (Figure 7B). The methylation level of the maternal allele was much lower than that of the paternal allele; when Epi-df was used as the paternal strain, both the parental alleles had lower methylation levels (Figure 7C). These results suggest that DNA methylation is essential for imprinting in endosperm and repression in other tissues.

FIE1 Interacts with Rice E(z) Homologs

FIE has been reported to interact with E(z) proteins to form the core of the PRC2 complex, which deposits H3K27me3 repressive marks on targets in Arabidopsis (Yadegari et al., 2000; Katz et al., 2004; Liu et al., 2010). FIE1 shares high amino acid sequence identity with Arabidopsis FIE, suggesting that FIE1 likely also participates in PRC2-mediated gene repression. To test this hypothesis, we performed yeast two-hybrid assays to test the interaction between FIE1 and rice E(z) homologs. There are two E(z) homologs in rice: iEZ1 (for indica Enhancer of Zeste 1; Os03g19480) and CLF (for Curly Leaf; Os06g16390) (Luo et al., 2009). The full-length FIE1 CDS and iEZ1 CDS (or CLF CDS) were cloned into pGADT7 and pGBKT7, respectively. Yeast cells containing pGADT7-FIE1 and pGBKT7-iEZ1 or pGBKT7-CLF could grow well on Trp-, Leu-dropout and Trp-, Leu-, His-dropout media (Figure 8A), indicating an interaction between FIE1 and iEZ1 and between FIE1 and CLF. We also examined the interaction between FIE2 and E(z) homologs. Yeast two-hybrid results showed that FIE2 could interact with iEZ1 but not CLF (Figure 8A). The interaction between FIE2 and CLF may be too weak to be detected, or FIE2 may interact with CLF indirectly via another component. Together, these results support the hypothesis that FIE1 and its homolog FIE2 participate in PRC2-mediated gene repression.

Figure 8.

Interaction between Rice Esc and E(z) Homologs and Increased H3K27me3 at Target Gene.

(A) Examination of the interaction between FIE1 and rice E(z)-like proteins (left), and FIE2 and rice E(z)-like proteins (right). The interactions were verified by growing the yeast on selective medium [SD/-Leu-Trp-His with 3-amino-1, 2, 4-triazole (3-AT)].

(B) Expression analysis and H3K27me3 status of Os11g31770 gene. The expression of Os11g31770 gene was validated by RT-qPCR (left). H3K27me3 ChIP-seq data for Os11g31770 gene are shown (right), and three subregions (1, 2, and 3) were validated by ChIP-qPCR analysis (middle). The wild type (WT) is shown in red in the top part, and Epi-df is shown in cyan in the middle part. Gene models are shown at the bottom. The two DNA stands are indicated by “+” and “−.” Values are means ± sd of three biological replicates for transcription analysis and two biological replicates for ChIP analysis.

Ectopic Expression of FIE1 Leads to Alternation in H3K27me3 Patterns in Epi-df

The pleiotropic defects of Epi-df suggest that ectopic expression of FIE1 may affect the expression of other genes required for normal rice development. Microarray analyses of Epi-df and wild-type seedlings revealed that 83 and 222 genes were up- and downregulated, respectively, by greater than twofold (q-value ≤ 0.05) (see Supplemental Date Set 1 online). In particular, FIE1 was the most strongly upregulated gene in Epi-df, while FIE2 expression showed no change (1.04-fold) compared with the wild type. This is in consistent with our RT-PCR analysis (Figure 2B). The most strongly downregulated gene (Os11g31770, which has five homologs in the rice genome) encodes an unknown protein and has an expression level in Epi-df of only 0.02-fold compared with that in the wild type. One of its homologs (Os11g31470) also showed reduced expression (0.33-fold compared with the wild type). Real-time PCR confirmed the microarray analysis and found that another two homologs (Os11g31430 and Os11g31705) were also downregulated (Figure 8B; see Supplemental Figure 7 online). The genes with altered expression in Epi-df are involved in a wide range of biological processes, including response to stress, lipid metabolism, signal transduction, and biosynthesis. This result suggests that silencing of FIE1 via DNA methylation and H3K9me2 is essential for regulation of numerous genes and therefore is required for normal rice development.

The interaction between FIE1 and rice E(z) homologs suggests that FIE1 participates in PRC2 repression and ectopic expression of FIE1 may result in misregulated H3K27me3. To test this hypothesis, we examined the global histone methylation status in Epi-df. Immunoblot analysis showed that global histone modification of H3K27me3 and others, including H3K27me2, H3K9me2, and K3K4me3, were not altered in Epi-df (see Supplemental Figure 8 online). We further examined the global H3K27me3 levels by ChIP sequencing (ChIP-seq) assays. As a control, we did not detect H3K27me3 at Tos17 (see Supplemental Figure 9A online), which was usually silenced and enriched in H3K9me2 (Ding et al., 2007), indicating a low background in our ChIP-seq assay. H3K27me3 levels were not detectable in FIE1 and FIE2 (see Supplemental Figure 9B online), which in consistent with their gene expression pattern. Compared with the wild type, 486 genes showed increased H3K27me3 by greater than twofold in Epi-df, and 14 of them overlapped with the 222 downregulated genes (P = 7.3 × 10⁻⁹ by hypergeometic distribution), while no upregulated genes showed increased H3K27me3 in Epi-df mutants (P = 0.5 by hypergeometic distribution), suggesting that H3K27me3 is positively linked to downregulated gene expression. It is worth noting that most of the H3K27me3 enrichment genes showed low or no expression, which was out of the detection range of microarray technology. All the four downregulated genes (Os11g31770, Os11g31470, Os11g31430, and Os11g31705) mentioned in the above paragraph showed increased H3K27me3, which were viewed by the snapshots of H3K27me3 (Figure 8B; see Supplemental Figure 7 online) and were further confirmed by ChIP-quantitative PCR (qPCR) assays, suggesting that ectopic expression of FIE1 enhanced the repressive function of PRC2. Reduced H3K27me3 levels were also found at three upregulated MADS box genes (MADS17, MADS27, and MADS16), which were viewed by the snapshots of H3K27me3 and further confirmed by real-time PCR and ChIP-qPCR assays (see Supplemental Figure 10 online). This reduced H3K27me3 levels associated with upregulation in gene expression may be caused by a secondary effect of enhanced PRC2 function mediated by Epi-df. Alternatively, ectopic expression of FIE1 may have a dominant effect to compete with FIE2 and therefore leads to PcG complex disruption and reduced H3K27me3. Based on genetic analysis, Epi-df displayed distinct phenotypes compared with loss-of-function mutant emf2b (Luo et al., 2000), a key component of PRC2 complex, which makes the later situation less likely. Taken together, these results indicate that ectopic expression of FIE1 results in directly and indirectly misregulated H3K27me3 and suggest that normal rice development requires silencing of FIE1 via epigenetic repressive marks.

DISCUSSION

In this work, we report the identification of an epi-allele of rice FIE1, which exhibits hypomethylation in the promoter and 5′ region, reduced H3K9me2 and increased H3K4me3, causing ectopic FIE1 expression, loss of imprinting in endosperm, dwarfism, and various floral defects. We found that FIE1 expression was tightly regulated by DNA methylation. Epi-df plants were not stable and all revertants were remethylated at CpG to a certain level and reduced gene expression, resulting in a wild-type phenotype. Remarkably, although FIE1 is the only detected site showing altered DNA methylation here, since FIE1 encodes a core component for PRC2 complex and, therefore, the Epi-df mutant shows altered histone H3 Lys 27 trimethylation and perturbed expression of hundreds of genes, suggesting a regulatory linkage between DNA methylation, H3K9me2 and H3K27me3 in coordinating genome-wide gene expression and plant development (Figure 9).

Figure 9.

Summary of the Epi-df Allele in Regulation of H3K27me3 by DNA Methylation and H3K9me2.

FIE1 is an endosperm-specific expressing gene, which is silenced in other tissues via epigenetic repressive modifications. In Epi-df, DNA hypomethylation, reduced H3K9me2, and increased H3K4me3 lead to ectopic expression of FIE1 and loss of imprinting in endosperm, whereas revertants were remethylated, leading to silence of FIE1 and normal phenotype. Ectopic expression of FIE1 resulted in hypermethylated H3K27me3 at PRC2 target genes.

Epi-df Is a DNA Hypomethylated Rice Mutant with Distinct Characteristics

Although epigenetic silencing of transposons has been broadly studied in plants, very few epigenetic mutants have been discovered in plants, especially in the important crops. Epi-d1, exhibiting a metastable dwarf phenotype, is the first epigenetic mutant found in rice; this epi-allele results from metastable silencing of DWARF1 by DNA hypermethylation and histone modification variations (Miura et al., 2009b). fwa is the first identified DNA hypomethylated mutant discovered in plants (Soppe et al., 2000). Hypomethylation at two SINE-related direct repeats upstream of FWA results in its ectopic expression and late flowering phenotype (Soppe et al., 2000). Although both fwa and Epi-df are hypomethylated mutants, and both FWA and FIE1 are imprinted, Epi-df shows distinct features. First, unlike FWA, which is fully methylated at two direct repeats upstream of the translation start site (Soppe et al., 2000), the methylation sites of FIE1 are located in the coding region, especially the third exon, suggesting that methylation within the gene body affects silencing of FIE1. Second, the adjacent genes up- or downstream of FIE1 exhibit high expression levels (Figure 2B), indicating that FIE1 locus is surrounded by euchromatin. Thus, a mechanism likely exists by which the heterochromatic status of FIE1 can be distinguished from the surrounding euchromatic regions. Finally, like most hypermethylated mutants, Epi-df could produce rare revertants; however, fwa is rather stable and no revertant was discovered from fwa populations (Soppe et al., 2000). In addition, we also observed that even though FIE1 is silent in revertants, only some regions have returned to high methylation levels at CG sites, suggesting that high levels of CG methylation are crucial for silencing FIE1.

Recently, studies revealed that rice Domains Rearranged Methylase 2 (DRM2) is responsible not only for de novo and non-CG methylation, but also for CG methylation (Moritoh et al., 2012); this is different from Arabidopsis DRM1/2 (Cao and Jacobsen, 2002; Cao et al., 2003). Previously, we showed that downregulation of rice H3K9 methyltransferase SDG714 affected both CG and CHG methylation (Ding et al., 2007), whereas loss of function of Arabidopsis H3K9 methyltransferase KRYPTONITE (KYP)/SUVH4 mainly affected CHG methylation (Jackson et al., 2002; Malagnac et al., 2002). Bisulfite sequencing results showed high levels of CG (94.78%) and CHG (58.13%) methylation but much lower levels of CHH (3.82%) methylation within FIE1, indicating that silencing of FIE1 primarily depends on CG and CHG methylation. We speculate that silencing of FIE1 via DNA methylation may be synergistically regulated by MET1, DRM2, and SDG714. Epi-df is a DNA hypomethylated mutant discovered in rice, and its distinct characteristics provide an intriguing opportunity to discover epigenetic modifications for developmental regulation in important crops.

Methylation Is Essential for Normal Endosperm-Specific Expression and Imprinting of FIE1

FIE1 is silent in vegetative tissues and young panicle and is highly expressed in early development of the endosperm. It has been reported that impairment of FIE1 by T-DNA insertion showed no morphological changes (Luo et al., 2009), suggesting that FIE1 may function redundantly with FIE2 to regulate early endosperm development. Consistent with its expression pattern, the 5′ region of FIE1 is subject to high DNA methylation levels in vegetative tissues and young panicle but has decreased DNA methylation in the endosperm. Consistent with previous studies finding that FIE1 showed maternal-specific expression (Luo et al., 2009), bisulfite sequencing results showed lower methylation levels of the maternal allele of FIE1 than that of paternal allele, and hypomethylation of FIE1 resulted in disturbance of the imprinting pattern. These results indicate that the expression and imprinting pattern of FIE1 is regulated by DNA methylation. It has been reported that the common ancestor of the grass family underwent genome duplication 50 to 70 million years ago (Wang et al., 2005; Messing and Bennetzen, 2008), which may have resulted in the duplication of Esc in rice. Like Zm FIE1 (Danilevskaya et al., 2003), Os FIE1 might have evolved an endosperm-specific expression pattern and inherited the imprinting feature; epigenetic marks, such as DNA methylation, may play an important role during this evolution process.

PRC2 (FIE1)-Mediated H3K27me3 Repression of Gene Expression Is Regulated by H3K9me2

H3K27me3 and H3K9me2 are two crucial repressive marks found in both animals and higher plants (Fischle et al., 2003; Liu et al., 2010); however, the relationship between these two repressive marks has not been extensively studied. Unlike in animals, where H3K27me3 spans large genomic regions and covers multiple genes (Schwartz et al., 2006; Tolhuis et al., 2006), H3K27me3 in Arabidopsis was enriched in genic regions and restricted to transcribed regions of single genes (Zhang et al., 2007). In Drosophila and mammals, H3K27me3 is deposited by PRC2, but in Arabidopsis, at least three PRC2-like complexes have been found, and each contains one FIE and different E(z) and Su(z)12 forms (Chanvivattana et al., 2004; Pien and Grossniklaus, 2007; Köhler and Villar, 2008; Liu et al., 2010). Phylogenetic analysis of PcG proteins showed that the components of PRC2-like complexes were different between Arabidopsis and rice (Luo et al., 2009). For example, the E(z) gene family in Arabidopsis has three members, MEA, CLF, and SWINGER, whereas only two E(z) genes were found in rice, Os CLF and Os iEZ1. Also, FIE is the sole Esc gene in Arabidopsis, whereas rice has two Esc homologs, Os FIE1 and Os FIE2, and three Su(z)12 homologs, FIS2, EMF2, and VRN2, were found in Arabidopsis, whereas only two Su(z)12 homologs, EMF2a and EMF2b, were found in rice. Thus, the rice PRC2-like complexes and its targets may be to some extent different from those of Arabidopsis. In rice, the sole loss-of-function PcG mutant identified was a T-DNA insertion in EMF2b (Luo et al., 2009). emf2b caused a recessive phenotype including flowering earlier in long-day conditions and arrest of floret development; whereas the Epi-df showed a dwarfism and floral organ defects in a dominant manner, but no effects on flowering time. Thus, the phenotypes of emf2b and Epi-df are distinct. Therefore, it is less likely that FIE1 competes with FIE2 and reduced H3K27me3. To further investigate the mechanism of H3K27me3-mediated gene repression in rice, a genome-wide scale analysis of the H3K27me3 distribution and identification of more PRC2 targets are necessary in the future.

H3K9me2 is known as a hallmark for inactivation of transposons and maintenance of genome integrity (Liu et al., 2010). In Arabidopsis, H3K9me2 is mainly enriched in heterochromatin, such as centromeric and pericentromeric regions, where repetitive elements are prevalent (Bernatavichute et al., 2008). Previous genome-wide analyses of histone and DNA methylation patterns in Arabidopsis revealed a very high coincidence between H3K9me2 and non-CG methylation (Bernatavichute et al., 2008). Arabidopsis has two distinct mechanisms to distinguish active and inactive genes; the first mechanism involves plant-specific CHROMOMETHYLASE3 (containing a chromo domain that can bind to H3K9me2) and histone H3K9 methyltransferase KYP/SUVH4 (containing a SRA domain that can bind to methylated CHG). These two methyltransferases form a self-reinforcing feedback loop responsible for maintenance of CHG methylation and H3K9me2 in the inactive genes (Bartee et al., 2001; Lindroth et al., 2001; Jackson et al., 2002; Malagnac et al., 2002). In the second mechanism, the H3K9 demethylase Increase in Bonsai Methylation 1 (IBM1) removes H3K9 methylation in the active genes (Saze et al., 2008). The loss-of-function mutants of IBM1 showed ectopic CHG methylation and H3K9me2 in thousands of genes (Miura et al., 2009a; Inagaki et al., 2010). Recent studies discovered that endosperm H3K27me3 profile identified some transposable elements and protein coding genes that contained a lower level of DNA methylation and were subjected to H3K27me3, whereas these sequences contain dense DNA methylation and are absent of H3K27me3 in vegetative tissues, suggesting that DNA methylation prevents the establishment of H3K27me3 at targets in vegetative tissues (Weinhofer et al., 2010).

Corresponding to its more complex genome compared with Arabidopsis, rice may have distinct epigenetic regulation mechanisms. In rice, heterochromatin is less distinct and dispersed throughout the genome (Cheng et al., 2001; Yan and Jiang, 2007). Such discontinuous heterochromatin and euchromatin may make rice a unique species in epigenetic gene regulation. In rice, genome-wide analysis showed that DNA methylation is not restricted in the centromeric/pericentromeric regions but is distributed evenly along the chromosome arms (Li et al., 2008). Moreover, DNA methylation preferentially occurred in the gene body but not the promoter region in rice (Li et al., 2008). We found that there was a high level of H3K9me2 within the 5′ region of FIE1, which coincided with the densely methylated CHG sites within this region. Reduction of DNA methylation and H3K9me2 resulted in ectopic expression of FIE1 and pleiotropic defects in Epi-df, suggesting that silencing of FIE1 by H3K9me2 is essential for normal rice development.

METHODS

Plant Materials and Growth Conditions

Epi-df spontaneously emerged from F5 offspring of a self-crossed diploid plant that is derived from pollen culture of autotetraploid rice (Qin et al., 2005). The mutant phenotype of Epi-df was not exhibited until the fifth generation from cell culture. The normal offspring of pollen culture plants, with a nearly identical genetic background to Epi-df, were used as the wild type. Rice materials were grown in the paddy field at the Chinese Academy of Agricultural Sciences (Beijing, China) in the natural growing season or grown in pots in a greenhouse at 25 to 28°C with 16 h light and 8 h dark.

Investigation of Pollen Fertility

Pollen grains sampled from mature flowers were stained with 1% iodine potassium iodide solution and observed under a microscope. Pollen grains were classified as sterile or fertile according to their staining behavior. All round and dark-brown stained pollen grains were scored as fertile, and irregular-shaped, yellowish, or unstained pollen grains were scored as sterile.

Map-Based Cloning of Epi-df

To clone the Epi-df gene, an F2 mapping population was generated from a cross between Epi-df and PA64 (indica). In the F2 population, 1978 recessive normal plants were selected for gene mapping. The Epi-df gene was first mapped to an interval between simple sequence repeat markers SS-2 and SS-1 on the short arm of chromosome 8 using 521 F2 normal plants, then the location of Epi-df was further narrowed down to a 49-kb region between InDel markers In-6 and In-4 using 1457 F2 normal plants. This region corresponds to BAC clones AP005620, AP003896, and AP003876. Sequencing of the 49-kb region showed no sequence alternation between Epi-df and the wild type. cDNAs of the five ORFs were amplified from both the wild type and Epi-df at the seedling stage, and the PCR products were confirmed by sequencing. The primer sequences are listed in Supplemental Table 3 online.

To confirm that FIE1 corresponds to the Epi-df gene, the 1.6-kb cDNA sequence containing the whole coding sequence of FIE1 was amplified and inserted into vector pCUbi1390 under the control of the maize (Zea mays) ubiquitin promoter. The primer sequences used for the construction of this vector are listed in Supplemental Table 3 online. Because it is difficult to get regenerated plants from wild-type or Epi-df calli, this construct was introduced into a normal indica variety Kasalath and japonica variety Nipponbare using Agrobacterium tumefaciens–mediated transformation method as described before (Hiei and Komari, 2008).

RT-PCR and Real-Time PCR

Total RNA was extracted from tissues of the wild type and Epi-df, respectively, using the RNA Prep Pure plant kit (Tiangen), and treated with DNase I following the manufacturer’s instructions. The embryos were isolated using a stereomicroscope and washed with RNase-free water. The endosperm was isolated without the seed coat. The first-strand cDNA was synthesized using oligo(dT) as the primer. RT-PCR was performed to amplify the FIE1 transcripts with 29 PCR cycles, and Actin1 transcripts were also amplified with 26 PCR cycles as a control. The RT-PCR analysis was repeated three times with similar results. Real-time PCR analysis was conducted using ABI7900HT Fast real-time PCR system with the SYBR Premix Ex Taq (TaKaRa; RR041A). The rice (Oryza sativa) Ubiquitin gene was used as the endogenous control. The relative expressions of specific genes were quantitated using the 2^−ΔΔCt (cycle threshold) method. ΔCt values were calculated by first normalizing Ct values to the endogenous control and subsequently calculating ΔΔCt values using the ΔCt value of one of the examined samples as a reference (for example, the wild-type seedling was used as a reference and set at 1.0 in Figure 8B). Expression levels were calculated from three biological replicates, each consisting of three technical replicates. Gene-specific primers are shown in Supplemental Table 3 online.

McrBC-, MspI-, and HpaII-PCR

Genomic DNA was isolated from seedlings using the cetyltrimethylammonium bromide (CTAB) method and the quantity was measured by the Quant-i dsDNA assay kit (Q33120). McrBC is an endonuclease that cuts DNA between methylated cytosine residues but not unmethylated DNA. After McrBC treatment, a methylated DNA will be digested and therefore will not be amplified by PCR. For McrBC-PCR analysis of FIE1 gene, 1 μg of genomic DNA was digested with 20 units of McrBC restriction endonuclease (New England Biolabs) in 50-μL reaction mixes for 0, 0.5, 3, and 8 h. Equal amounts of McrBC-digested DNAs were used for PCR amplification, and the products were separated on 1% agarose gels. The MspI- and HpaII-PCR procedures were similar to that of McrBC-PCR except the reaction time was 3 h.

DNA Gel Blotting

Genomic DNA was extracted using the CTAB method, and 10 μg of genomic DNA was used for DNA gel blot analysis. DNA was digested with appropriate restriction endonucleases (New England Biolabs) for 3 h. The digested DNA was loaded onto 1% agarose gels, electrophoresed at 30 V for 10 h, and blotted on Hybond N⁺ membranes (Amersham). The probes used for the detection of DNA methylation of FIE1 and Tos17 were DNA fragments amplified using specific primers (see Supplemental Table 3 online) and labeled with [³²P]dCTP. The probe used for the detection of CentO was a [³²P]dCTP-labeled mixture of CentO sequence DNA oligomers (see Supplemental Table 3 online).

Bisulfite Sequencing

Genomic DNAs were isolated from 3-week-old rice seedlings using the CTAB method. Briefly, 1 μg of genomic DNA was treated with sodium bisulfite using the EZ DNA methylation kit (Zymo Research) according to the manufacturer’s instructions. The treated DNA was dissolved in 15 μL distilled water, and 3 μL of this solution was then used as the template in a 50 μL PCR. The PCR products were purified and cloned into PMD18-T vectors. For each region, at least 12 individual clones were sequenced, and the sequencing data were analyzed with Kismeth software (Gruntman et al., 2008). The primer sequences used for bisulfite sequencing are listed in Supplemental Table 3 online.

Immunoblot Analysis

Immunoblot assays were performed as described previously (Ding et al., 2007). Histones were extracted from the wild type and Epi-df, respectively, following by segregating on 15% SDS-PAGE gel. H3 Lys methylation status in wild-type and Epi-df plants was determined by immunoblot with the methylation-specific antibodies listed below. Anti-H3 immunoblot was used as a loading control. Antibodies were as follows: anti-H3K27me3, 07-449 (Millipore); anti-H3K27me2, 07-452 (Millipore); anti-H3K9me2, ab1220 (Abcam); anti-H3K4me3, 07-473 (Millipore); and anti-H3, ab1791 (Abcam).

ChIP and ChIP-Seq Analyses

ChIP assays were performed as described previously (Nagaki et al., 2004; Li et al., 2008; Lu et al., 2011) with minor modifications. Approximately 2 g leaves of the 3-week-old rice seedlings was ground to fine powder with liquid nitrogen, resuspended in 37 mL of 1% formaldehyde in extraction buffer I (0.4 M Suc, 10 mM Tris-HCl, pH 8, 10 mM MgCl₂, 5 mM α-mercaptoethanol [α-ME], 0.1 mM phenylmethylsulfonyl fluoride [PMSF], and proteinase inhibitors [Complete; Roche Molecular Biochemicals]), inverted gently and thoroughly to release the nuclei, and then stopped the cross-linking by adding Gly to a final concentration of 0.125 M. The solution was filtered through Miracloth (Calbiochem; catalog no. 475855) into a fresh 50-mL falcon tube. The filtered solution was spun to remove supernatant, and the pellet was resuspended in 1 mL of extraction buffer 2 (0.25 M Suc, 10 mM Tris-HCl, pH 8, 10 mM MgCl₂, 1% Triton X-100, 5 mM α-ME, 0.1 mM PMSF, and proteinase inhibitors). The solution was transferred to a 1.5-mL Eppendorf tube and centrifuged to remove supernatant, and the pellet was resuspended in 300 μL of extraction buffer 3 (1.7 M Suc, 10 mM Tris-HCl, pH 8, 2 mM MgCl₂, 0.15% Triton X-100, 5 mM α-ME, 0.1 mM PMSF, and proteinase inhibitors), layered on top of another clean 300 μL of extraction buffer 3, and spun for 1 h at 16,000 g at 4°C. The supernatant was removed and the chromatin pellet resuspended in 200 μL of nuclei lysis buffer (50 mM Tris-HCl, pH 8, 10 mM EDTA, 1% SDS, and proteinase inhibitors), and the chromatin solution was sonicated into 0.2 to 1.0 kb. The sonicated chromatin can be stored at –80°C or used immediately for immunoprecipitation.

The chromatin sample was diluted 10 times with ChIP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.1, 167 mM NaCl, and proteinase inhibitors), and then 1 mL of diluted chromatin sample was precleared with 40 μL of Protein A Agarose beads (Sigma-Aldrich; P3476) for 1 h at 4°C with rotation. Then 2 μg of antibodies was added and incubated for 4 h or overnight at 4°C with gentle agitation. A total of 10 μL of 100 mg/μL yeast tRNA (Invitrogen 15401-029) was added to final 1 mg/μL as a blocking agent. Immune complexes were collected with 50 μL of Protein A Agarose beads for 1 h at 4°C with gentle agitation. At the end of the incubation, the protein A-Sepharose was washed stepwise in low-salt wash buffer (150 mM NaCl, 0.1% SDS, 1% Triton X-100, 2 mM EDTA, and 20 mM Tris-HCl, pH 8.1), high-salt wash buffer (500 mM NaCl, 0.1% SDS, 1% Triton X-100, 2 mM EDTA, and 20 mM Tris-HCl, pH 8.1), LiCl wash buffer (0.25 M LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA, and 10 mM Tris-HCl, pH 8.1), and TE buffer (10 mM Tris-HCl, pH 8, and 1 mM EDTA).

Immune complexes were eluted by adding 250 μL of elution buffer (1% SDS and 0.1 M NaHCO₃) to the pelleted beads and incubated at 65°C for 15 min with gentle agitation. The beads were spun and the supernatant fraction was carefully transferred to another tube to repeat elution. The two eluates were combined. Twenty microliters of 5 M NaCl and 1 μL of RNase A (Qiagen; catalog no. 19101) was added to the eluate and reverse cross-linked at 65°C for at least 6 h. Ten microliters of 0.5 M EDTA, 20 μL of 1 M Tris-HCl, pH 6.5, and 2.5 μL of 10 mg/mL proteinase K were added to the eluate and incubated for 1 h at 45°C. The immunoprecipitated DNA was extracted with phenol/chloroform.

Antibodies used for ChIP were anti-H3K9me2 (Abcam, ab1220), anti-H3K4me3 (Abcam, ab8580), anti-H3K27me3 (Millipore, 07-449) and anti-H3 (Abcam, ab1791). The ChIPed DNA was then used for real-time PCR. The primers used for real-time PCR are listed in Supplemental Table 3 online.

For ChIP-seq analysis, ChIPed DNA was used for library construction and sequencing by Beijing Genomics Institute. Sequence reads were mapped to the reference genome of rice (ssp japonica cv Nipponbare 6.1) using Bowtie (Langmead et al., 2009) allowing up to two mismatches. Reads with unique locations were retained for further analysis. The ChIPDiff program (Xu et al., 2008) was used for quantitative comparison of H3K27me3 levels in the wild type and Epi-df. Regions with more than twofold change were kept for downstream analysis. The MACS program was used to shift the reads and convert the data to WIG format (Zhang et al., 2008). WIG files were visualized with the Integrated Genome Browser (Nicol et al., 2009).

Imprinting Analysis

The hybrid seeds between Epi-df (or the wild type) and 93-11 (an indica variety) were harvested 1 week after pollination. Endosperm was dissected from embryo and was squeezed out of the seed coat. Approximately 20 seeds were dissected and RNAs were extracted as above. The first-strand cDNAs were synthesized using oligo(dT) as the primer. Allele-specific PCR was performed using FIE1- and VPE1-specific primers listed in Supplemental Table 3 online. The specific PCR products of FIE1 and VPE1 were digested with Fnu4HI and Csp6I (New England Biolabs), respectively, and then separated on 1% agarose gels.

Microarray Analysis

Total RNA was isolated from seedlings 3 weeks after germination using TRIzol reagent and was further purified using the Qiagen RNeasy Mini Kit according to the manufacturer’s instructions. Three biological replicates were performed for the wild type and Epi-df, respectively. Hybridization of Affrymetrix rice GeneChips and initial data collection were conducted at CapitalBio. The hybridization data were analyzed using GeneChip Operating software (GCOS 1.4). The scanned images were first assessed by visual inspection then analyzed to generate raw data files using the default setting of GCOS 1.4. An invariant set normalization procedure was performed to normalize the different arrays using DNA-chip analyzer (dChip). In a comparison analysis, we applied a two-class unpaired method in the Significant Analysis of Microarray software (Tusher et al., 2001) to identify significantly differentially expressed genes between Epi-df and wild-type groups. A total of 305 genes were determined to be significantly differentially expressed with a selection threshold of false discovery rate ≤ 5% and fold change > 2.0 in the Significant Analysis of Microarray software output result.

Yeast Two-Hybrid Experiments

The Gal4 system (Clontech) was used for yeast two-hybrid analysis of protein interactions. The coding sequences of FIE1, FIE2, iEZ1, and CLF were amplified using gene-specific primers (see Supplemental Table 3 online). FIE1 and FIE2 were inserted into the pGADT7 vector, and iEZ1 and CLF were inserted into the pGBKT7 vector, respectively. Test constructs were transformed into the yeast strain AH109, and interactions were monitored by growth on SD plates lacking His, Trp, and Leu.

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL databases under the following accession numbers: FIE1, AK242200; FIE2, AK111761; Actin1, AK100267; Ubiquitin, AK059011; C-kinase, AK071174; VPE1, AK103573; iEZ1, AK243493, Os03g19480; CLF, AK111743, Os06g16390; MADS6, AK069103; MADS16, AK069317; MADS17, AK070540; MADS27, CT835409; Os11g31430, AY224562; Os11g31470, AK101049; Os11g31705, NM_001189633; Os11g31770, AK102677; 5S rRNA, D26370; and Tos 17, D85876. The microarray and ChIP-seq data were deposited in the Gene Expression Omnibus database (accession numbers GSE39298 and GSE40674).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure 1. Comparison of Plant Height and Fertility between the Wild Type and Epi-df.

• Supplemental Figure 2. The Morphology of Spontaneous Revertants of Epi-df and the Expression of FIE1 in Revertants and Their Progenies.

• Supplemental Figure 3. Alignment of FIE1 and Its Homologs.

• Supplemental Figure 4. Methylation Status of CentO, 5S rDNA, and Tos17 in the Wild Type and Epi-df Plants.

• Supplemental Figure 5. Histograms Representing the Percentage of CpG, CpHpG, and CpHpH Methylation at FIE1 in Three Normal Varieties.

• Supplemental Figure 6. Expression Pattern of FIE2.

• Supplemental Figure 7. Downregulated Expression and Increased H3K27me3 at Specific Loci.

• Supplemental Figure 8. Immunoblot Analysis of in Vivo Histone Methylation Status in the Wild Type and Epi-df.

• Supplemental Figure 9. H3K27me3 ChIP-seq at Tos 17 and FIE Genes.

• Supplemental Figure 10. Upregulated Expression and Reduced H3K27me3 at MADS Box Genes.

• Supplemental Table 1. Flower Organ Morphology of Epi-df.

• Supplemental Table 2. Phenotype Segregation in Progenies of Revertants.

• Supplemental Table 3. Oligonucleotides Used in This Study.

• Supplemental Data Set 1. The 305 Misregulated Genes with Expression Changes of at Least Twofold and Q-Value (%) ≤ 5 in Epi-df.

Glossary

H3K9me2

H3 Lys 9 dimethylation

H3K27me3

H3 Lys 27 trimethylation

ChIP

chromatin immunoprecipitation

DAP

days after pollination

ChIP-seq

ChIP sequencing

qPCR

quantitative PCR

CTAB

cetyltrimethylammonium bromide

PMSF

phenylmethylsulfonyl fluoride

α-ME

α-mercaptoethanol