OsVIL2 functions with PRC2 to induce flowering by repressing OsLFL1 in rice


For correspondence (e-mail genean@khu.ac.kr).


Flowering is exquisitely regulated by both promotive and inhibitory factors. Molecular genetic studies with Arabidopsis have verified several epigenetic repressors that regulate flowering time. However, the roles of chromatin remodeling factors in developmental processes have not been well explored in Oryza sativa (rice). We identified a chromatin remodeling factor OsVIL2 (O. sativa VIN3-LIKE 2) that promotes flowering. OsVIL2 contains a plant homeodomain (PHD) finger, which is a conserved motif of histone binding proteins. Insertion mutations in OsVIL2 caused late flowering under both long and short days. In osvil2 mutants OsLFL1 expression was increased, but that of Ehd1, Hd3a and RFT1 was reduced. We demonstrated that OsVIL2 is bound to native histone H3 in vitro. Chromatin immunoprecipitation analyses showed that OsVIL2 was directly associated with OsLFL1 chromatin. We also observed that H3K27me3 was significantly enriched by OsLFL1 chromatin in the wild type, but that this enrichment was diminished in the osvil2 mutants. These results indicated that OsVIL2 epigenetically represses OsLFL1 expression. We showed that OsVIL2 physically interacts with OsEMF2b, a component of polycomb repression complex 2. As observed from osvil2, a null mutation of OsEMF2b caused late flowering by increasing OsLFL1 expression and decreasing Ehd1 expression. Thus, we conclude that OsVIL2 functions together with PRC2 to induce flowering by repressing OsLFL1.


Spatial and temporal gene expression is controlled through a balance between activation and repression, as well as through the stable maintenance of gene silencing. Polycomb group (PcG) proteins play a crucial role in these mechanisms as key transcriptional regulators (Schuettengruber et al., 2007; Morey and Helin, 2010). These proteins function by forming chromatin-associated multiprotein complexes. Drosophila contains three different PcG protein complexes: polycomb repressive complex 1 (PRC1), PRC2 and Pho repressive complex (PhoRC) (Muller and Verrijzer, 2009). PRC1 catalyzes histone H2A lysine 119 ubiquitylation (H2AK199u; Francis et al., 2004; Napoles et al., 2004), whereas PRC2 has histone methyltransferase activity that acts upon lysine 27 of histone H3 (H3K27; Cao et al., 2002). PhoRC silences genes via nucleosome compaction (Klymenko et al., 2006). All of these PcG complexes are recruited to target genes by cis-acting polycomb response elements (Schwartz and Pirrotta, 2007).

PRC2 in Drosophila has four core proteins: ENHANCER OF ZESTE [E(z)], SUPPRESSOR OF ZESTE 12 [Su(z)12], EXTRA SEX COMBS (ESC) and P55 (Schuettengruber and Cavalli, 2009). The E(z) protein exhibits methyltransferase activity on lysines 9 and 27 of histone H3 (H3K9 and H3K27; Cao et al., 2002; Czermin et al., 2002). Most PRC subunits have plant homologs. For example, Arabidopsis contains three E(z) genes [CURLY LEAF (CLF), SWINGER (SWN) and MEDEA] plus three Su(z)12 genes [FERTILIZATION INDEPENDENT SEED 2 (FIS2), VERNALIZATION 2 (VRN2) and EMBRYONIC FLOWER 2 (EMF2)] (Luo et al., 1999; Gendall et al., 2001; Yoshida et al., 2001). ESC is present as a single-copy gene, FERTILIZATION INDEPENDENT ENDOSPERM (FIE), whereas P55 has five homologues: MULTICOPY SUPPRESSOR OF IRA 1–IRA 5 (MSI1–MSI5; Ach et al., 1997; Ohad et al., 1999; Hennig et al., 2003). In Oryza sativa (rice), E(z) and ESC are present in two copies: OsiEZi and OsCLF for the former, and OsFIE1 and OsFIE2 for the latter (Luo et al., 2009). Rice also has two Su(Z)12-like genes: O. sativa EMBRYONIC FLOWER 2a (OsEMF2a) and OsEMF2b (Hennig and Derkacheva, 2009; Luo et al., 2009).

Three PRC2-like complexes have been identified in Arabidopsis. One is the FIS complex, which contains MEA/SWN, FIS2, FIE and MSI1, and which functions during seed development and gametogenesis. A second, the EMBRYONIC FLOWER (EMF) complex, is comprised of CLF/SWN, EMF2, FIE and MSI1; this complex is involved in the suppression of early flowering.

The third, the VERNALIZATION (VRN) complex, is active in vernalization (Hennig and Derkacheva, 2009). This VRN complex associates with VERNALIZATION INSENSITIVE 3 (VIN3), VIL1/VRN5 and VIL2/VERNALIZATION-LIKE 1 (VEL1) to form a PHD-PRC2 complex (Wood et al., 2006; De Lucia et al., 2008). The VIN3 protein enhances the trimethylation of H3K27 (H3K27me3) throughout the target loci to a level sufficient for stable silencing. For example, it increases the level of H3K37me3 in FLOWERING LOCUS C (FLC) and some FLC clade members (Sung and Amasino, 2004; Sung et al., 2006; Sheldon et al., 2009). Four VIN3-like (VIL) genes exist: VIL1, VIL2, VIL3 and VIL4 (Sung et al., 2006; Greb et al., 2007). They code for proteins that contain conserved motifs of the PHD finger domain, the fibronectin type-III (FNIII) domain and the VIN3-interacting domain (VID) (Sung et al., 2006; Greb et al., 2007). VIL1 acts with VIN3 to repress FLC and FLOWERING LOCUS M (FLM) through histone modification and H3K27me3 (Sung et al., 2006; Greb et al., 2007). VIL2 associates more broadly over FLC and also represses MAF5 by H3K27me3 (Napoles et al., 2004; Kim and Sung, 2010).

The rice genome contains four genes that are related to Arabidopsis VIN3: O. sativa VIN3-LIKE 1–4 (OsVIL1–4; Fu et al., 2007). Similar to Arabidopsis, all OsVIL proteins carry three conserved domains: a Plant Homeo Domain (PHD) finger, a fibronectin type III (FNIII) domain, and a VIN3-interacting domain (VID) (Fu et al., 2007). The PHD finger domain exists widely in eukaryotic nuclear proteins, and is characterized by a conserved Cys4-His-Cys3 (C4HC3) zinc finger motif (Aasland et al., 1995; Bienz, 2006). Increasing evidence indicates that PHD fingers can bind to histone with post-translational modifications to mediate gene transcription and chromatin dynamics (Mellor, 2006; Kouzarides, 2007; Taverna et al., 2007; Liu et al., 2010; Li and Li, 2012). For example, in humans and mice, the PHD finger protein 2 (PHF2) modulates histone demethylation through the recognition of histone H3 trimethylation on lysine 4 with its PHD finger (Chan et al., 2009; Wen et al., 2010; Horton et al., 2011).

Although plants have relatively fewer studied PHD-finger-containing proteins compared with animals (Liu et al., 2010), they are the focus of much research. For example, the Arabidopsis origin recognition complex 1 (ORC1) preferentially binds to H3K4me3 with its PHD finger to regulate the transcription of genes that help define the origins of DNA replication (Sanchez and Gutierrez, 2009). The Arabidopsis inhibitor of growth 1 (ING1) binds to H3K4me3 and H3K4me2 via PHD fingers (Lee et al., 2009). In addition, H3K9me2 is recognized by the PHD finger of Arabidopsis VIL2 protein (Kim and Sung, 2010). Although their protein structures are highly conserved between rice and Arabidopsis, it is largely unknown how they function within plant developmental processes. The only report has been that mutation in OsVIL2 causes a change in leaf angle because of altered cell division in the lamina joint (Zhao et al., 2010).

In rice, flowering is promoted by the transcription of Heading date 3a (Hd3a) and Rice FT-like 1 (RFT1) (Yano et al., 2000; Hayama et al., 2002; Kojima et al., 2002). Expression of both of them is regulated by Early heading date 1 (Ehd1), which acts as a flowering activator under both short days (SDs) and long days (LDs) (Doi et al., 2004; Izawa, 2007). Several genes regulate Ehd1, including O. sativa LEAFY COTYLEDON 2 and FUSCA 3-LIKE 1 (OsLFL1), which attenuates its expression via binding to the Ehd1 promoter (Peng et al., 2007, 2008). OsId1/Ehd2/RID1 promotes flowering by upregulating Ehd1 (Matsubara et al., 2008; Park et al., 2008; Wu et al., 2008). OsCOL4, controlled by OsphyB, functions as a flowering repressor upstream of Ehd1 (Lee et al., 2010). OsMADS50 and Ghd7 are LD-preferential flowering repressors, functioning upstream of Ehd1 (Lee et al., 2004; Xue et al., 2007). OsMADS51 induces Ehd1 preferentially under SD (Kim et al., 2007). In addition, Hd3a and RFT1 are regulated by heading date 1 (Hd1), which is controlled by OsGI (Yano et al., 2000; Hayama et al., 2002).

Here, we studied the functional role of OsVIL2 in controlling flowering time in rice. We showed that OsVIL2 associates with OsEMF2b, a component of the PRC2-like complex, to regulate the expression of relevant genes.


Mutations in OsVIL2 cause delayed flowering

To determine whether any OsVIL genes are involved in controlling flowering time, we used T-DNA insertion mutant lines in japonica rice that had been generated previously (Jeon et al., 2000; Jeong et al., 2002; An et al., 2003; Ryu et al., 2004). We identified two independent mutants in OsVIL2 (Os02g05840.1 in TIGR and Os02g0152500 in RAP-DB) that were the most closely related to Arabidopsis VIN3 among rice and Arabidopsis OsVIN3-like proteins (Figure S1). This gene contains four exons and three introns. T-DNAs were inserted into the second intron and third exon of OsVIL2 in osvil2-1 (line 3A-10852) and osvil2-2 (line 1B-21530), respectively (Figure 1a). Transcripts of OsVIL2 were not detected in either mutant, indicating that they were null alleles (Figure 1b).

Figure 1.

OsVIL2 genomic structure and flowering time of T-DNA insertional mutants. (a) Gene structure of OsVIL2. Block boxes indicate exons in coding region; lines connecting boxes are introns. Triangles are T-DNA insertions of osvil2-1 (line 3A-10852) and osvil2-2 (line 1B-21530). Arrows F1, R1 and RB are the primers used for genotyping osvil2-1; F2, R2 and RB are the primers used for genotyping osvil2-2. Arrows F1 and R2 are primers for RT-PCR analyses of OsVIL2 transcript levels. Scale bar: 100 bp (b) Measurement of OsVIL2 transcript levels by RT-PCR. Rice Ubi served as the control. (c) Phenotypes of wild type (WT), osvil2-1 and osvil2-2 at the heading stage in a paddy field. (d) Days to heading in WT, osvil2-1 and osvil2-2 plants under SDs (10 h of light/14 h of dark) or LDs (14 h of light/10 h of dark). The number of days to heading was scored when the first panicle bolted. Error bars indicate standard deviations; n = 15 plants.

The osvil2 mutants displayed several phenotypic alterations in their leaf angle, tiller number, floret development and flowering time when grown in a paddy field. Leaf angles were increased in the mutants, as previously observed (Zhao et al., 2010). The mutants also had fewer tillers, with the number decreasing from an average of 17 (in the wild type, WT) to seven or eight (in osvil2). In addition, the mutant displayed an abnormal floret structure: the palea was depressed and the number of empty glumes was reduced (Figure S2). Each mutant spikelet also had an extra lodicule (Figure S2).

Because flowering time is an important trait for yield potential, we focused on this phenotype. Both alleles flowered late, by about 3 weeks, compared with the WT in a paddy field (Figure 1c,d). The osvil2 mutants flowered late under both SD (10 h of light/14 h of dark) and LD (14 h of light/10 h of dark) conditions (Figure 1d). Compared with our WT control, this delay was about 2 weeks under SDs and about 3 weeks under LDs (Figure 1d). Plants heterozygous for the T-DNA insertions were normal, indicating that the osvil2 mutation is recessive. The introduction of full-length cDNA under the control of the Zea mays (maize) ubiquitin promoter complemented the mutant phenotypes. Because both alleles displayed similar phenotypes, we selected only one, osvil2-2, for further characterization.

OsLFL1 expression is increased in the osvil2 mutant

We studied the relationship between OsVIL2 and previously identified flowering time regulators by monitoring changes in gene expression. Flowering signals are generated in the leaves and transmitted to the shoot apex. In rice, the florigen Hd3a is produced in the leaves and moves to the shoot apical meristems (Tamaki et al., 2007). Therefore, we prepared RNA samples from the leaf blades of 6-week-old SD plants and 8-week-old LD plants. We chose those particular stages of development for study because floral transition occurred around those times. Transcript levels were estimated by quantitative real-time RT-PCR using Ubiquitin (Ubi) as a standard. Under LD conditions, mRNA levels of Hd3a, RFT1 and Ehd1 were significantly reduced in the osvil2 mutant (Figure 2b–d), suggesting that OsVIL2 functions upstream of these three genes.

Figure 2.

Expression pattern of floral regulators in wild-type (WT) and osvil2-2 plants under long-day (LD) conditions. Quantitative RT-PCR analyses of OsVIL2 (a), Hd3a (b), RFT1 (c), Ehd1 (d), OsLFL1 (e), OsId1 (f), OsCOL4 (g), OsGI (h), Hd1 (i), Ghd7 (j), OsMADS50 (k) and OsMADS51 (l). Leaf blades were sampled from WT and osvil2-2 plants grown for 56 days under LDs; closed circles, WT; open circles, osvil2-2; y-axis, relative transcript level of each gene compared with that of rice Ubi. Error bars indicate standard deviations; n = 6 or more.

Because Ehd1 is a convergence point for various flowering signals, we examined the regulatory genes that function upstream of it. Under LDs, the expression of OsId1 and OsCOL4 was not affected in the osvil2-2 mutant (Figure 2f,g), whereas that of OsLFL1 was markedly elevated (Figure 2e). We also observed that transcript levels for OsGI, Hd1, Ghd7, OsMADS50 and OsMADS51 were unchanged in the mutant (Figure 2h–l). This suggested that OsVIL2 is an upstream repressor of OsLFL1. Similar results were obtained under SD conditions, implying that OsVIL2 functions in the same manner, regardless of day length (Figure S3).

The OsVIL2-mediated flowering pathway is independent of Hd1, OsId1 and OsCOL4

To elucidate further the relationship between OsVIL2 and the other flowering genes, we examined its transcript level in mutant plants that had altered expression of the signaling genes Ehd1, Hd1, OsID1 and OsCOL4. Transcripts of OsVIL2 were not significantly changed in Ehd1 RNAi plants (Figure 3a,g). Likewise, expression was not altered in hd1 mutants or Hd1-overexpressing plants (Figure 3b,c,h,i). We also tested OsId1 RNAi plants because that gene is a major regulator of Ehd1 independent of OsLFL1 (Park et al., 2008). As expected, transcripts of OsVIL2 were not changed in osid1 knock-down plants (Figure 3d,j), nor were they altered in OsCOL4 knock-out or overexpression plants (Figure 3e,f,k,l). OsCOL4 is a repressor of Ehd1 that functions independently from OsLFL1 and OsId1. All of these results supported our conclusion that OsVIL2 functions independently from Hd1, OsId1 and OsCOL4.

Figure 3.

Expression of OsVIL2 in mutant plants with altered expression of Ehd1, Hd1, OsId1 and OsCOL4. Plants were grown for 8 weeks under LDs. (a, g) Transcript levels of Ehd1 (a) and OsVIL2 (g) in Ehd1 RNAi plants. (b, h) Transcript levels of Hd1 (b) and OsVIL2 (h) in hd1 mutants. (c, i) Transcript levels of Hd1 (c) and OsVIL2 (i) in Hd1-overexpressing (OX) plants. (d, j) Transcript levels of OsId1 (d) and OsVIL2 (j) in OsId1 RNAi plants. (e, k) Transcript levels of OsCOL4 (e) and OsVIL2 (k) in oscol4-2 mutants. (f, l) Transcript levels of OsCOL4 (f) and OsVIL2 (l) in OsCOL4 activation-tagging plants. WT, wild-type siblings of each mutant. Error bars indicate standard deviations; n = 15 plants.

OsVIL2 acts as a repressor of OsLFL1

Peng et al. (2007, 2008) have proposed that OsLFL1 negatively regulates Ehd1 by binding to the promoter. Thus, if OsVIL2 negatively controls OsLFL1, their levels of expression should be opposite to each other. We examined their transcripts during developmental stages prior to floral transition. Under LD conditions, flowering signals were undetectable during the first 40 days of vegetative growth after germination (Ryu et al., 2009). This was based on sampling at 2-day intervals that began at 22 days after germination (DAG) and continued until 42 DAG. OsVIL2 expression gradually increased until 28 DAG, before rapidly rising at 30 DAG to a level that was maintained thereafter (Figure 4a). In contrast, OsLFL1 expression remained high until 28 DAG, then started to decline at 30 DAG before reaching its lowest level at 38 DAG (Figure 4a). In osvil2-2 mutants, the OsLFL1 transcript level did not drop but was continuously increased (Figure 4b). Consequently, Ehd1 transcript levels remained very low in the osvil2 mutant compared with the WT (Figure 4c). Our results, therefore, supported that OsVIL2 is a negative regulator of OsLFL1.

Figure 4.

Temporal expression patterns of OsVIL2, OsLFL1 and Ehd1 at various developmental stages, and transcription levels of OsLFL1 homologous genes. (a–c) Temporal expression patterns of OsVIL2 and OsLFL1 in WT (a), OsLFL1 in WT and osvil2-2 (b), and Ehd1 in WT and osvil2-2 (c). At 2-day intervals from day 22 to 42, samples were harvested at Zeitgeber time zero (ZT0) from second leaf blades from the top of plants grown under LDs. Y-axis, relative transcript level of each gene compared with that of rice Ubi. (d) Transcript levels of OsLFL1 homologous genes in WT and osvil2 plants at 56 days after germination.

Three genes are homologous to OsLFL1: OsLFL2 (LOC_Os08g01090), OsLFL3 (LOC_Os04g58010) and OsLFL4 (LOC_Os01g68370) (Figure S4). To examine whether they are also affected by the osvil2 mutation, we prepared mRNAs from leaf blades of the WT and the osvil2-2 mutant at Zeitgeber time zero (ZT0), when OsVIL2 transcripts were abundant. Whereas the OsLFL1 mRNA level was significantly increased in the mutants, no significant change was observed in transcripts of OsLFL2, OsLFL3 and OsLFL4 (Figure 4d).

OsVIL2 binds to native histone H3

The apparent structural similarity between the OsVIL2 PHD finger and others implies that the former may function as a module to interact with histone (Figure S5). To investigate this, we first examined whether the OsVIL2 PHD finger could bind to any specific histone. Our GST pull-down assay results showed that the GST-fused OsVIL2 PHD finger did bind with histone H3 from a calf thymus histone mixture, and that its binding affinity was relatively weaker than that of the GST-fused AtING1-PHD and GST-fused PHF2-PHD (Figure 5a). Thus, using native H3 histone pull-down assays, we confirmed that the GST-fused OsVIL2 binds to native H3 histone (Figure 5b).

Figure 5.

OsVIL2 binds with native histone H3 and associates with OsLFL1 chromatin. (a) GST pull-down assays of the GST-fused OsVIL2 canonical PHD finger (GST-OsVIL2cPHD aa 163–232) with the calf thymus histone mixture. The GST-PHF2PHD and GST-AtING1PHD are positive controls. The mock (only beads of equal volume) and GST are negative controls. The protein samples were separated in 15% SDS-PAGE gel and stained with Commassie blue. (b) Western-blot analyses. GST pull-down assay of the GST-fused OsVIL2 canonical PHD finger (GST-OsVIL2cPHD aa 163–232) with the native histone H3. The GST-fused PHD fingers of the Arabidopsis AtING1 and the human PHF2 are positive controls, whereas the mock (only beads of equal volume) and GST are negative controls. (c) Genomic structure of OsLFL1 and Ehd1, and regions tested in ChIP assay. (d) ChIP analysis of OsLFL1 chromatin. OsVIL2-Myc epitope-tagged transgenic lines were used to detect enrichment of OsVIL2 at OsLFL1 chromatin. Actin chromatin was used as control. (e) ChIP analysis of Ehd1 chromatin. OsVIL2-Myc epitope-tagged transgenic lines were used to detect enrichment of OsVIL2 at Ehd1 chromatin. Actin chromatin was used as a control. (f) Trimethylation of lysine 27 of histone H3 (H3K27me3) analysis of OsLFL1 chromatin. Chromatin signatures at OsLFL1 chromatin were determined in the wild type (WT) and the osvil2 mutant using antibodies against H3K27me3. Enrichments of controls and OsLFL1 regions were calculated based on fractions of precipitates compared with each input. Actin chromatin was used as a control.

OsVIL2 associates with OsLFL1 chromatin and mediates the enrichment of H3K27me3 at OsLFL1 chromatin

Our findings suggested that OsLFL1 is a potential target of OsVIL2 (Figure 2). To examine this possibility, we generated transgenic plants expressing Myc-tagged OsVIL2. As a control, we also generated transgenics that expressed only Myc. Chromatin immunoprecipitation (ChIP) assays using the anti-Myc antibody revealed an enrichment of OsVIL2 at OsLFL1 chromatin, especially in the promoter region and the transcript start region (Figure 5c,d). However, no such enrichment was observed with Myc alone (Figure 5d). As a control, we examined Ehd1, which functions downstream of OsLFL1. We observed that Ehd1 chromatin was not enriched in the transgenic plants expressing Myc-tagged OsVIL2 (Figure 5e). We also showed that Actin chromatin was not enriched by the anti-Myc antibody (Figure 5d).

Arabidopsis VIN3 protein increases the H3K27me3 to suppress the target loci (Sung and Amasino, 2004; Sung et al., 2006). Therefore, we examined the H3K27me3 level of OsLFL1 chromatin and found that it was significantly enriched at the promoter and translation start regions of OsLFL1 chromatin in the WT, but that this enrichment was diminished in the osvil2-2 mutant (Figure 5f). These results suggest that OsVIL2 mediates the H3K27me3 of OsLFL1.

OsVIL2 interacts with OsEMF2b, the absence of which causes delayed flowering

Because Arabidopsis VIN3 is a component of PRC2 (De Lucia et al., 2008), we speculated that OsVIL2 might also function in a PRC2 complex. If so, then mutations in another element of rice PRC2 would show an alteration in their flowering time. We screened late-flowering T-DNA insertion mutants for lesions in PRC2 element genes. This allowed us to identify a T-DNA mutant in OsEMF2b. OsEMF2b is homologous to Su(z)12, which is a component in Drosophila PRC2 (Hennig and Derkacheva, 2009; Luo et al., 2009). Compared with the WT, mutant plants of OsEMF2b flowered about 3 weeks later under SDs, and about 4 weeks later under LDs, or when grown in a paddy field (Figure 6a). This phenotype was similar to that of osvil2, suggesting that OsEMF2b and OsVIL2 function together in controlling flowering time. It was previously reported that the osemf2b mutant shows the phenotype of early flowering under LD conditions (Luo et al., 2009). However, we were unable to repeat this observation. As Luo et al. (2009) had obtained the mutant line from our laboratory this discrepancy in results was not likely to result from a difference in genetic background. Instead, because the plants described in the earlier report had been cultured under extreme LD conditions (16 h of light/8 h of dark), this might have caused the unusual response. By contrast, under our experimental conditions, the osemf2b mutant repeatedly exhibited a late-flowering phenotype.

Figure 6.

Heading dates for the wild type (WT) and osemf2b, and co-immunoprecipitation analyses between OsVIL2 and OsEMF2b.(a) Number of days to heading for WT and osemf2b plants under short days (SDs) or long days (LDs); error bars indicate standard deviations; n = 15 plants. (b) Co-immunoprecipitation assay using OsVIL2-myc transgenic plants. Protoplasts from transgenic plants expressing OsVIL2-Myc were transfected with OsEMF2b-HA and co-immunoprecipitation was performed with anti-HA antibody. Non-transgenic plants (NT) were used as a control for the specificity of precipitation. (c) Co-immunoprecipitation assay using OsVIL2-HA transgenic plants. Protoplasts from transgenic plants expressing OsVIL2-HA were transfected with OsEMF2b-Myc and co-immunoprecipitation was performed with anti-Myc antibody. Non-transgenic plants (NT) were used as a control for the specificity of precipitation.

To examine whether OsVIL2 and OsEMF2b bind to each other, we performed co-immunoprecipitation assays using OsVIL2::Myc transgenic plants. The HA-tagged OsEMF2b plasmid was introduced into mesophyll protoplasts, and OsEMF2b-HA protein was detected by the precipitates using an antibody against c-Myc (Figure 6b). We obtained similar results when OsVIL2::HA transgenic lines were transformed with the Myc-tagged OsEMF2b plasmid (Figure 6c). These findings suggested that OsVIL2 binds to OsEMF2b and functions as a member of the rice PRC2 complex.

The osvil2 osemf2b double mutant was seedling lethal. The occurrence of this severe phenotype implied that OsVIL2 and OsEMF2b function together with other elements, e.g. OsEMF2a and OsVIL2-related proteins (Figure S7).

OsEMF2b is a negative regulator of OsLFL1

In addition to late flowering, the osemf2b mutant displayed other phenotypes similar to osvil2, including fewer tillers and abnormal floret development. These osemf2b plants were also semi-dwarfed (Figure S8). If OsEMF2b indeed functions together with OsVIL2 in controlling flowering time, then osemf2b would affect the genes in the OsLFL1-Ehd1-Hd3a pathway. Therefore, we measured the expression of such regulators in the mutant. Under LDs, transcript levels of Ehd1, Hd3a and RFT1 were reduced in the mutant (Figure 7b–d). By contrast, OsLFL1 transcripts were upregulated, as had also been observed with our osvil2 mutants (Figure 7e). However, transcript levels of OsId1, OsCOL4, OsGI, Hd1, Ghd7 and OsMADS51 were not altered in the osemf2b mutants (Figure 7f–k). Similar results were obtained under SDs (Figure S6). Because of this, we could conclude that both OsEMF2b and OsVIL2 repress the expression of OsLFL1, thereby functioning together to control flowering time (Figure 8).

Figure 7.

Expression patterns of floral regulators in osemf2b under long-day (LD) conditions. Quantitative RT-PCR analyses of OsEMF2b (a), Hd3a (b), RFT1 (c), Ehd1 (d), OsLFL1 (e), OsId1 (f), OsCOL4 (g), OsGI (h), Hd1 (i), Ghd7 (j), OsMADS50 (k) and OsMADS51 (l) in osemf2b (open circles) compared with wild type (WT, closed circles). Y-axis, relative transcript level of each gene compared with that of rice Ubi. Error bars indicate standard deviations; n = 6 or more.

Figure 8.

Model for regulatory network controlling flowering time by OsVIL2 and OsEMF2b in rice.


We have obtained evidence that OsVIL2 controls flowering time in rice by the chromatin remodeling of a target gene. Several PHD-containing proteins have been reported to have this function. Rice Early heading date 3 (Ehd3), with a PHD finger motif, preferentially promotes flowering under LDs by repressing Ghd7 (Matsubara et al., 2011). In Arabidopsis, a PHD-containing protein EARLY BOLTING SHORT DAYS (EBS) is involved in chromatin remodeling: mutations in EBS cause late flowering under SDs by repressing FT (Gómez-Mena et al., 2001; Piñeiro et al., 2003). VIN3 delays flowering by repressing FLC during vernalization (Sung et al., 2006; Greb et al., 2007). VIL1 and VIL2 are also involved in the vernalization-mediated repression of flowering, and are required for the repression of MAF1 and MAF5, respectively, especially under SDs (Napoles et al., 2004; Sung et al., 2006; Kim and Sung, 2010).

OsVIL2 directly regulates OsLFL1 expression

In rice, Ehd1 is an important integrator of floral transition, with numerous other genes controlling its expression (Komiya et al., 2008). Here, we showed that OsVIL2 regulates Ehd1 through OsLFL1. Transcripts of OsLFL1 were increased in osvil2 mutants, and temporal expression patterns of OsVIL2 and OsLFL1 were opposite to each other. That is, when levels of OsVIL2 were low, before 28 DAG, expression of OsLFL1 was high. Between 28 and 30 DAG, when OsVIL2 expression was markedly increased, that of OsLFL1 started to decrease rapidly. Finally, our ChIP assays demonstrated that OsVIL2 directly binds to OsLFL1 chromatin. Thus, OsVIL2 appears to directly mediate OsLFL1 expression through epigenetic regulation.

OsVIL2 is a component of the PRC2 complex

We found that OsVIL2 interacts in vitro with OsEMF2b, a homologue of Drosophila Su(z)12 (Hennig and Derkacheva, 2009). As observed from osvil2, the OsLFL1 mRNA level was elevated and Ehd1 transcripts decreased in the osemf2b mutant, which also displayed a delayed-flowering phenotype. This indicated that OsVIL2 and OsEMFb2 are both components of the complex that represses OsLFL1. Therefore, we might speculate that the OsVIL2OsEMF2b complex functions similar to PHD-PRC2 in Arabidopsis (Cao et al., 2008).

In rice, two genes (EMF2a and EMF2b) belong to the Su(z)12 family. Both are homologous to Arabidopsis EMF2. The VRN2 and FIS2 homologous genes are not present in rice. Because vernalization does not control flowering time in rice and an FLC-homologous gene is not present in rice, OsEMF2b may not function like Arabidopsis VRN2. Our osemf2b mutants showed phenotype changes in flowering time and floral organs. Therefore, it is likely that OsEMF2b independently regulates both of those. Although plants of osvil2 and osemf2b displayed similar phenotypes, the phenotypes of osemf2b were more severe (Figures 6,S1,S8). The mutation in OsEMF2b displayed semi-dwarfing and flowering that was delayed by 1 week compared with the osvil2 mutants. This was probably because OsEMF2b is a core component of the PRC2 complex. OsVIL2-like proteins could have partially compensated for the lack of OsVIL2. Thus, we propose that OsEMF2b interacts with OsVIL2 and regulates the expression of OsLFL1 in the flowering pathway. Further study will be needed to determine whether OsEMF2b is indeed a component of rice PRC2 and what is the role of OsEMF2a in controlling flowering time.

We also showed that the recombinant protein of OsVIL2cPHD was bound to native H3 histone in vitro. It is possible that the PHD domain binds to the H3 histone tail of target genes, and that the FNIII domain interacts with catalytic subunits such as the PRC2 complex. However, the mechanism by which effector proteins recognize and transduce different histone modification signals into specific cellular functions and biological outputs is largely unknown, especially in plants (Liu et al., 2010). OsVIL2 is a component of the OsEMF2b-PRC2-like complex that adds methylation marks on histone H3 at lysine 27. A dynamic change in the H3K27me3 pattern reflects an integrative balance between histone H3K27 methylation and de-methylation. In Arabidopsis, RELATIVE OF EARLY FLOWERING 6 (REF6), also known as Jumonji domain-containing protein 12 (JMJ12), specifically demethylates H3K27me3 and H3K27me2 (Lu et al., 2008, 2011). REF6 belongs to a large family of lysine demethylase 4 (KDM4)/JmjC domain-containing histone demethylation 3 (JHDM3). This family also contains five members in rice: OsJMJ701, OsJMJ702, OsJMJ705, OsJMJ706 and OsJMJ707 (Lu et al., 2008). Until now, there have been no reports on H3K27me3 demethylation activity and the biological importance of those members in rice. Therefore, more efforts are needed to deepen our knowledge about the mechanism by which OsVIL2 localizes to specific chromatin regions. We must also continue to explore its recruitment of the PRC2 complex to facilitate further histone modification and chromatin remodeling in order to regulate gene expression during rice development.

Role of OsVIL2 in rice development

The spikelets of our osvil2 mutants showed the phenotypes of 40% empty glumes and 60% defective palea. The mutants also had more lodicules, stamens and carpels than normal. Zhao et al. (2010) have found that a mutation in Rice leaf inclination 2 (LC2), another name for OsVIL2, has enlarged lamina joint angles caused by increased cell division. Patterns of expression for cell division-related genes and hormone-related genes are altered in those mutants. Furthermore, microarray analysis of emf mutants (emf1-1, emf1-2 and emf2-1) in Arabidopsis have shown that EMF genes affect flowering and the development of floral organs, and that they function in response to stress and cold signaling (Kim et al., 2010).

Our double mutant osvil2 osemf2b was seedling lethal. Although seeds developed normally, roots and shoots were extremely short and the seedlings eventually died. Such additive phenotypes have also been reported with Arabidopsis bli clf double mutants between PcG proteins and their interacting protein (Schatlowski et al., 2010). BLISTER (BLI) binds to PRC2 histone methyltransferase CLF protein: its loss results in small cotyledons and fewer trichomes in bli mutants. Although bli clf double mutants are very early flowering and have upwardly curled leaves, neither is a phenotypic character of the single mutant. BLI and CLF appear to act in both common and separate pathways during Arabidopsis development. Isolation of OsVIL2-containing complexes will reveal whether OsVIL2 indeed interacts with the PRC2-like complex or another PcG complex.

Experimental Procedures

Plant materials and growing conditions

T-DNA mutant lines were isolated from a T-DNA tagging line (O. sativa japonica cv. Dongjin and Hwayoung; Jeon et al., 2000; Jeong et al., 2002). Their flanking sequences were determined by inverted PCR (An et al., 2003; Ryu et al., 2004; Jeong et al., 2006). Previous reports have described: the T-DNA insertional mutants oscol4, OsCOL4-D and hd1; the RNAi-suppressed plants of Ehd1 RNAi and OsID1 RNAi; and the Hd1-overexpression plants (Park et al., 2008; Ryu et al., 2009; Lee et al., 2010). Our rice plants were raised either in a paddy field or in controlled growth rooms under SDs (10 h of light at 30°C/14 h of dark at 25°C) or LDs (14 h of light at 30°C/10 h of dark at 25°C).

Vector construction and plant transformation

The OsVIL2 cDNA clones were isolated by PCR, using two primers, 5′-CAATTCGCCATG GATCCACC-3′ and 5′-ATGCCAAAGTTCCATGCA-3′. This resulted in the removal of the termination codon. An amplified fragment was digested with HindIII and SpeI, and inserted into pGA3428 and pGA3438 vectors to generate the OsVIL2:HA3 and OsVIL2:Myc4 fusions, respectively (Kim et al., 2009). These constructs were transferred into Agrobacterium tumefaciens LBA4404 by the freeze–thaw method (An et al., 1988). The procedure for transforming rice via Agrobacterium-mediated co-cultivation has been described previously (Jeon et al., 1999; Zhang et al., 2011; Saeng-ngam et al., 2012).

RNA extraction and RT-PCR analyses

Total RNA was isolated using RNA iso (Takara, http://www.takara-bio.com). First-strand cDNA was synthesized with 2 μg of total RNA and Moloney murine leukemia virus reverse transcriptase (Promega, http://www.promega.com). The primers were as follows: for OsVIL2, 5′-TTAATGCCAAAGTTCCAT-3′ and 5′-TTAATGCCAAAGTTCCAT-3′; and for Ubi1, 5′-AACCAGCTGAGGCCCAAGA-3′ and 5′-ACGATTGATTTAACCAGTCCATGA-3′. Synthesized cDNA was prepared from leaf blades, and quantitative real-time RT-PCR was performed with a Rotor-Gene 6000 (Corbett Research, http://www.corbettlifescience.com). Rice Ubi was used as an internal control. Primers for studying gene expression are listed in Table S1.

In vitro binding assay

Calf thymus histone binding assays were performed as described previously (Matthews et al., 2007). The OsVIL2 canonical PHD finger, corresponding to amino acids 163–232, was predicted from a protein search at the SMART website (http://smart.embl-heidelberg.de). OsVIL2 cPHD cDNA was amplified with two primers, 5′-TACTTCCAATCCAATGCGATGGATCCACCCTACGCAGG-3′ and 5′-TTATCCACTTCCAATGCGCTA TTAATGCCAAAGTTCCATGC-3′. It was then cloned into ligation-independent cloning (LIC) vector pMCSG7 (Stols et al., 2002). The recombinant GST-OsVIL2cPHD protein was expressed in Escherichia coli strain BL21 (RIL) and purified using Glutathione Sepharose beads (Amersham Biosciences, now GE Healthcare Life Sciences, http://www.gelifesciences.com). A 10 μg of the target protein was first incubated for 1 h with 50 μl of a Glutathione Sepharose bead slurry (Amersham) at 4°C with continuous gentle shaking. After the solution was removed, the beads were blocked with 5% BSA protein in buffer A (50 mm Tris-HCl, pH 7.5, 0.3 m NaCl and 0.1% IGEPAL CA-630), and the mixture was incubated at 4°C for 2 h to prevent non-specific binding. The beads were then washed three times with buffer A before 10 μg of either core histone mixture from calf thymus (10223565001; Roche, http://www.roche.com) or a native histone H3 from calf thymus (11034758001; Roche, http://www.roche.com) were added. This was followed by incubation for 1 h at 4°C with continuous gentle shaking. After low-speed centrifugation, the beads were washed four times with buffer B (50 mm Tris-HCl, pH 7.5, 1 m NaCl and 0.1% IGEPAL CA-630). The bound protein was eluted with loading buffer and analyzed by SDS–PAGE with Coomassie blue staining. The mock (beads only) and GST protein were used as negative controls, whereas GST-AtING1PHD (Lee et al., 2009) and GST-PHF2PHD (Chan et al., 2009) were the positive controls. To identify which specific histone the GST-OsVIL2cPHD binds to, we electrophoresed the eluted samples and transferred them to a PVDF membrane (Bio-Rad, http://www.bio-rad.com), where they were probed with specific histone antibodies: anti-H3 (abcam), self-made anti-glutathione S-transferase antibody or HRP-conjugated secondary antibody (GE Healthcare). SuperSignal West Dura Chemiluminescent Substrate (34076; Thermo Fisher Scientific, http://www.thermofisher.com) was used for exposure detection. The GST-AtING1PHD and GST-PHF2PHD were kindly provided by Dr Chang Seob Kwon of the University of KAIST in Korea and Dr Jiemin Wong of East China Normal University in China, respectively.

In vivo co-immunoprecipitation assays

The OsEMF2b cDNA clone, without its stop codon, was generated by PCR with primer set 5′-CTGATACCAGATGTGCCGCCA-3′ and 5′-ATTTTTCTTTGGATCCGAGC-3′. The amplified fragment was digested with SpeI and HpaI, and then inserted into pGA3697 to generate the fusion molecule OsEMF2b-Myc (Kim et al., 2009). This construct was introduced into protoplasts prepared from transgenic plants that constitutively expressed OsVIL2-HA, as described previously (Park et al., 2011). Similarly, the cDNA clone was inserted into pGA3698, generating OsEMF2b-HA. Protoplasts from transgenic plants expressing OsVIL2-Myc were transfected with OsEMF2b-HA molecules. After 12 h of incubation, the protoplasts were harvested and used for co-immunoprecipitation analysis, as previously reported (Ryu et al., 2009), but with the following modification. Protoplasts were re-suspended in nuclei extraction buffer [250 mm sucrose, 10 mm Tris, pH 8.0, 10 mm MgCl2, 1% Triton X-100, 1 mm phenylmethylsulfonyl fluoride (PMSF) and one tablet of proteinase inhibitor mixture (Roche)] and incubated on ice for 15 min. The samples were then centrifuged at 13400 g for 10 min at 4°C and the pellets were re-suspended with 1 ml of lysis buffer [50 mm Tris, pH 8.0, 10 mm EDTA, pH 8.0, 1% Triton X-100, 0.1% SDS, 1 mm PMSF and one tablet of proteinase inhibitor mixture (Roche)]. After sonication, the resuspended samples were centrifuged at 4°C for 5 min at 13400 g.

Before the protoplasts were harvested, 1 μl of anti-c-Myc antibody (monoclonal, #2276; Cell Signaling Technology, http://www.cellsignal.com) and 10 μl each of protein A and G agarose beads (Upstate, Temecula, CA, USA; www.millipore.com) were incubated for 3 h with 1 ml of binding buffer (50 mm Tris-HCl, pH 7.5, 75 mm NaCl, 5 mm EDTA, 1 mm DTT, 0.1 mm PMSF and 1% Triton X-100). Total protein had been pre-cleared for 1 h with 10 μl of protein A/G plus agarose beads at 4°C and continuous gentle shaking. Afterwards, 10% of this pre-cleared protein extract was saved and the remaining 90% was precipitated with anti-c-Myc coated beads. The precipitated proteins were washed three times with a buffer of 50 mm Tris (pH 7.5), 75 mm NaCl, 5 mm EDTA, 1 mm DTT, 0.1 mm PMSF and 0.1% Triton X-100, before the precipitated protein complex was eluted with 20 ml of immunoprecipitation buffer. Samples were then resolved in SDS–PAGE gels and transferred for western blot analysis with peroxidase-conjugated high-affinity anti-HA antibodies (Roche). We used enhanced Chemiluminescence Plus Western Blotting Detection Reagents (RPN2132; Amersham, now GE Healthcare) and LAS-3000 film (Fuji, http://www.fujifilm.com). Next, we precipitated with anti-HA (monoclonal, 12CA5; Roche) and performed western blot analysis with peroxidase-conjugated high-affinity anti-Myc antibodies (monoclonal, #2040; Cell Signaling Technology).

Chromatin immunoprecipitation (ChIP) analysis

Transgenic plants expressing OsVIL2-Myc4 were used for ChIP analysis. The c-Myc antibodies (monoclonal; #2276) were purchased from Cell Signaling. The anti-trimethyl-Histone H3 (Lys27) was from Millipore (#CS200603; http://www.millipore.com). ChIP assays were performed following the previously reported method (Haring et al., 2007). Primer sequences are listed in Table S2. All assays were performed at least three times from two biological replicates.


We thank Youngsook Lee for valuable discussion, Kyungsook An for generating the transgenic lines and handling the seed stock, and Priscilla Licht for editing the English of the article. This work was supported in part by grants from: the Next-Generation BioGreen 21 Program (Plant Molecular Breeding Center, No. PJ008128), Rural Development Administration, Republic of Korea; the Basic Research Promotion Fund, Republic of Korea (KRF-2007-341-C00028); and Kyung Hee University (20120227) to G. A. and also from: the National Basic Research Program of China (grant no. 2009CB941500); and the National Natural Science Foundation of China (grant no. 30930048 and 30921061) to X. C.