Repression of FLOWERING LOCUS C and FLOWERING LOCUS T by the Arabidopsis Polycomb repressive complex 2 components.

Polycomb group (PcG) proteins are evolutionarily conserved in animals and plants, and play critical roles in the regulation of developmental gene expression. Here we show that the Arabidopsis Polycomb repressive complex 2 (PRC2) subunits CURLY LEAF (CLF), EMBRYONIC FLOWER 2 (EMF2) and FERTILIZATION INDEPENDENT ENDOSPERM (FIE) repress the expression of FLOWERING LOCUS C (FLC), a central repressor of the floral transition in Arabidopsis and FLC relatives. In addition, CLF directly interacts with and mediates the deposition of repressive histone H3 lysine 27 trimethylation (H3K27me3) into FLC and FLC relatives, which suppresses active histone H3 lysine 4 trimethylation (H3K4me3) in these loci. Furthermore, we show that during vegetative development CLF and FIE strongly repress the expression of FLOWERING LOCUS T (FT), a key flowering-time integrator, and that CLF also directly interacts with and mediates the deposition of H3K27me3 into FT chromatin. Our results suggest that PRC2-like complexes containing CLF, EMF2 and FIE, directly interact with and deposit into FT, FLC and FLC relatives repressive trimethyl H3K27 leading to the suppression of active H3K4me3 in these loci, and thus repress the expression of these flowering genes. Given the central roles of FLC and FT in flowering-time regulation in Arabidopsis, these findings suggest that the CLF-containing PRC2-like complexes play a significant role in control of flowering in Arabidopsis.


Introduction
The transition from a vegetative to a reproductive phase (i.e., flowering) is a major developmental switch in the plant life cycle that must be properly timed to ensure maximal reproductive success. In Arabidopsis thaliana, this transition is genetically controlled by several pathways, including the autonomous pathway, the photoperiod pathway and the vernalization pathway, which form a regulatory network [1,2]. This network integrates the endogenous developmental state of the plant with environmental cues (e.g., day length and temperature) to precisely control the timing of the floral transition [1,2].
A key component in this regulatory network in Arabidopsis is FLC, a MADS box transcription factor that quantitatively inhibits the floral transition [3,4]. FLC expression is delicately controlled by various activators and repressors. The autonomous pathway, which includes FVE [5,6], FCA [7] and FLOWERING LOCUS D (FLD) [8], constitutively represses FLC expression to promote flowering, whereas FRIGIDA (FRI) activates FLC expression to delay flowering [9]. The vernalization pathway also represses FLC expression in response to a prolonged cold exposure (a typical winter) to accelerate flowering in Arabidopsis [10,11]. Besides FLC, in the Arabidopsis genome there are five close FLC relatives including FLOWERING LOCUS M (FLM), MADS AFFECTING FLOWERING 2 (MAF2), MADS AFFECTING FLOWERING 3 (MAF3), MADS AFFECTING FLOWERING 4 (MAF4) and MADS AFFECTING FLOWERING 5 (MAF5); these FLC relatives also appear to repress the floral transition [12,13].
Chromatin modification plays an important role in the regulation of FLC expression. Activation of FLC expression in the presence of FRI is associated with the H3K4 trimethylation and also requires deposition of the histone variant H2A.Z in FLC chromatin [14,15,16]. The autonomous-pathway represses FLC expression partly through generating repressive histone modifications in FLC chromatin. FLD is involved in the H3K4 demethylation and deacetylation of FLC chromatin [8,17,18]; FCA functions closely with FLD and is involved in H3K4 demethylation in FLC chromatin [18]; FVE is partly involved in the histone deacetylation of FLC chromatin [5,8]. In addition, histone H4 dimethylation at arginine 3 (H4R3) in FLC chromatin by Type I and Type II arginine methyltransferases is also associated with FLC repression [19,20,21]. Furthermore, small RNA-mediated repressive histone modifications are also linked to FLC repression [22,23]. Recent studies also reveal that vernalization leads to repressive histone modifications in FLC chromatin such as increased trimethylation of histone H3 at lysine 9 and H3K27, and H4R3 dimethylation [24,25,26,27].
FLC inhibits the floral transition partly by reducing expression of a key flowering-time integrator, FT [28]. FT was first identified as a component of the photoperiod pathway, which promotes flowering in response to increased day length [29,30,31]. In the presence of light, FT expression is activated by CONSTANS (CO), another component in the photoperiod pathway [31]. FT is expressed in the vasculature [32], and subsequently, FT proteins are translocated from veins to the shoot apex to promote flowering [33,34,35]. FLC binds to the FT locus and represses its expression, and thus antagonizes the activation by CO [28]. Hence, FT acts as a flowering-time integrator that integrates signals from the photoperiod pathway and the FLC-mediated flowering pathways to promote the Arabidopsis flowering. Recent studies indicate that chromatin modification may play a role in the regulation of FT expression. It has been shown that LIKE HETEROCHROMA-TIN PROTEIN 1 (LHP1) directly interacts with FT chromatin and represses FT expression [36,37,38]; in addition, recent wholegenome analysis of H3K27 trimethylation in Arabidopsis has revealed that this repressive mark is associated with FT chromatin [39]. However, how H3K27me3 is deposited in FT chromatin and its role in FT regulation remain elusive.
Repressive H3K27me3 is deposited by the PRC2 complex in Drosophila. PRC2 is composed of four core proteins including Enhancer of zeste (E(z); an H3K27 methyltransferase), Extra sex comb (Esc), Suppressor of zeste 12 (Su(z)12) and p55, and deposits trimethyl H3K27 to silence the expression of homeotic genes in Drosophila (reviewed in [40]). Homologs of Drosophila PRC2 components have also been identified in Arabidopsis, and play important roles in the control of plant developmental processes such as floral induction, flower organogenesis, seed development and sporophyte development (reviewed in [41,42]). To date, a PRC2-like complex composed of MEDEA (MEA), FIE, FERTIL-IZATION INDEPENDENT SEED 2 and MULTICOPY SUPPRESSOR OF IRA1 (MSI1), which are relatives of E(z), Esc, Su(z)12 and p55 respectively, has been biochemically characterized [43,44]. This complex represses the MADS box gene PHERES1 during seed development and thus controls this developmental process [45,46].
Recent studies have also shown that CLF, an Arabidopsis homolog of E(z), directly mediates the repression of AGAMOUS (AG) via H3K27 trimethylation and thus controls floral organogenesis [47,48]. CLF plays multiple roles in plant development, and also directly represses the expression of SHOOTMERISTEM-LESS (STM) and a flowering gene, AGAMOUS LIKE 19 (AGL19), during vegetative development [48,49]. Recent studies also reveal that VERNALIZATION 2 (VRN2), a homolog of Su(z)12, plays an important role in the vernalization-mediated FLC repression [50]. VRN2 is required for FLC repression by vernalization treatment [50]; VRN2 forms a complex with CLF, SWINGER (SWN; another homolog of E(z)), FIE and VERNALIZATION INSENSITIVE 3 to repress FLC expression in response to vernalization treatment [51]. In addition, EMF2, a relative of VRN2 and Su(z)12, also plays an important role in sporophyte development, and maintains vegetative development by repressing the floral induction [52,53,54]. However, the underlying mechanisms of the EMF2-mediated floral repression are unclear [54].
Here we report that Arabidopsis PRC2-like complex subunits CLF, EMF2 and FIE repress the expression of FLC and FLC relatives including MAF4 and MAF5, and that CLF directly binds to and mediates the deposition of H3K27me3 in FLC, MAF4 and MAF5 chromatin. Furthermore, we show that during vegetative development CLF and FIE strongly repress FT expression, and that CLF also directly interacts with and mediates the deposition of H3K27me3 in FT chromatin. Theses results imply that PRC2like complexes containing CLF, EMF2 and FIE deposit repressive H3K27me3 in and directly repress the expression of these flowering genes, and thus control the flowering program in Arabidopsis.

PRC2 Subunits CLF, EMF2 and FIE Repress the Expression of FLC, MAF4 and MAF5 in Vegetative Development
Arabidopsis PRC2-like complex components including VRN2, FIE, SWN and CLF are required for the vernalization-mediated FLC repression [50,51]. We sought to investigate PRC2-mediated FLC repression in Arabidopsis plants grown in normal conditions (i.e., without vernalization treatment). In addition, the expression of FLC relatives such as FLM, MAF4 and MAF5, like FLC expression, is also regulated by chromatin modification [14,15]; hence, it was also of interest to investigate whether PRC2-like complexes repress the expression of FLC relatives. First, we addressed the role of CLF in the regulation of FLC and FLC relatives. Transcript levels of these genes were examined in seedlings of the clf-81 mutant carrying a lesion in the SET domain of CLF [48]. We found that FLC, MAF4 and MAF5 were de-repressed in clf, whereas transcripts of FLM, MAF2 and MAF3 in clf remained at levels similar to wild-type Col ( Figure 1A); hence, CLF plays an essential role in repressing the expression of FLC, MAF4 and MAF5 during vegetative development. Secondly, we investigated the role of FIE in the regulation of FLC and FLC relatives using FIE-suppressed seedlings [55] (note that fie alleles can not be transmitted through the female gamete [56]). Consistent with a recent report [51], in FIE-suppressed seedlings FLC expression was de-repressed ( Figure 1B); furthermore, we found that MAF4 and MAF5 were also de-repressed, whereas FLM, MAF2 and MAF3 in these seedlings were expressed at levels similar to those in the wild type ( Figure 1B). Hence, like CLF, FIE also selectively represses the expression of FLC, MAF4 and MAF5.
CLF has been shown to directly interact with EMF2 and these two proteins may be part of a PRC2-like complex involved in the regulation of vegetative development in Arabidopsis [57]. We therefore examined transcript levels of FLC and FLC relatives in emf2 seedlings. Indeed, FLC, MAF4 and MAF5, but not FLM, MAF2 or MAF3, were de-repressed in emf2 ( Figure 1C). Hence, like CLF and FIE, EMF2 also selectively represses FLC, MAF4 and MAF5 expression during vegetative development. Together, these data suggest that there is a CLF-containing PRC2-like complex composed of at least EMF2 and FIE, which acts to repress FLC, MAF4 and MAF5 expression during vegetative development.

CLF and FIE also Repress FT Expression in Vegetative Development
The de-repression of FLC and MAFs in clf, emf2 and FIEsuppressed plants was expected to lead to late flowering because the elevated expression of these genes alone causes late flowering [3,4,13]; however, these mutant plants all are early-flowering [47,52,55]. These early-flowering phenotypes are likely due to increased or ectopic expression of genes that promote flowering. CLF and EMF2 have been shown to repress the expression of the flowering promoter AGL19 [49]; furthermore, ectopic expression of AG in clf and emf2 may also partly contribute to the earlyflowering phenotypes [47,54]. In addition, a very recent report shows that FT expression is upregulated in 21-day-old clf mutant plants grown under continuous light [58], indicating that FT derepression may partly account for the early-flowering phenotype of clf. We examined FT mRNA levels in young Col and clf seedlings to address whether FT is also de-repressed in clf mutants before the floral transition. Indeed, FT expression was greatly de-repressed in clf seedlings ( Figure 2A). These data together with recent findings [58] suggest that CLF represses FT expression throughout vegetative development.
Recently, it has been shown that FT mRNA levels are higher in emf2 relative to Col [54,58], but the role of EMF2 in FT repression is unclear [54]. We also examined FT mRNA levels in emf2 seedlings. Consistent with the recent reports [54,58], FT expression was de-repressed in emf2 ( Figure 2B). Because FIE may be part of the PRC2-like complexes containing EMF2 and CLF [41], we examined FT transcript levels in FIE-suppressed seedlings to determine whether FIE is also involved in FT repression, and found that FT is strongly de-repressed in these seedlings compared to the control Col-gl1 seedlings ( Figure 2C). Taken together, these data suggest that a PRC2-like complex containing CLF, EMF2 and FIE, represses FT expression in vegetative development to repress the floral transition.
Interestingly, although these PRC2 subunits repress both FLC and FT expression and FLC directly represses FT expression, loss or suppression of the functions of these subunits leads to a greater FT derepression compared to FLC derepression ( Figure 1 and Figure 2; also refer to Figure 3), suggesting that PRC2-like complexes have a repressive effect on FT expression much stronger than that on FLC expression.

CLF Acts in Partial Redundancy with Part of the Autonomous Pathway to Repress FLC Expression in the Absence of Vernalization
The autonomous pathway constitutively represses FLC expression to promote flowering, and part of this pathway is involved in the generation of repressive histone modifications in FLC  chromatin [59]. The autonomous-pathway repressor FCA directly binds to the FLC locus and is involved in the H3K4 demethylation of FLC chromatin [18]. Recent studies in mouse embryonic stem cells have suggested the coordinated regulation of H3K4 demethylation and PRC2-mediated repressive histone modifications in maintaining transcriptional gene repression [60]. Hence, it was of interest to examine the genetic interaction of clf with fca. We introduced clf into the fca mutant, and quantified FLC transcripts in clf, fca and clf;fca seedlings by real-time quantitative PCR. Consistent with previous findings [7], FLC was highly expressed in fca mutants ( Figure 3A); however, FLC was further de-repressed in clf;fca and FLC mRNA levels in the double mutants were much higher than those in fca or clf ( Figure 3A). Hence, CLF acts in partial redundancy with FCA to repress FLC expression in the absence of vernalization. We further measured flowering times of fca and clf;fca mutants grown in long days. Although FLC was so highly expressed in clf;fca, the double mutants flowered much earlier than fca ( Figure 3B). As noted above, FT is de-repressed in clf; hence, it is likely that the early-flowering phenotype of clf;fca is partly due to FT derepression. We quantified FT transcript levels in clf, fca and clf;fca seedlings.  relative to Col, whereas FT expression was suppressed in fca because of elevated FLC expression ( Figure 3C). Furthermore, FT expression was partially suppressed in clf;fca, but FT transcript levels in the double mutant were still higher than those in fca ( Figure 3C), suggesting that the early-flowering phenotype of clf;fca is at least partly due to the elevated FT expression.
CLF Directly Interacts with the FLC, MAF4, MAF5 and FT Chromatin As noted above, CLF, EMF2 and FIE repress FLC, MAF4, MAF5 and FT expression, however, it was not known whether these PRC2 subunits acted directly on these genes or indirectly. Using chromatin immunoprecipitation (ChIP), we first examined whether CLF directly interacts with the FLC, MAF4, and MAF5 loci. Specifically, genomic DNA was immunoprecipitated using an antibody recognizing GFP from seedlings of a 35S:GFP:CLF clf transgenic line in which GFP:CLF fully functions and its distribution mimics that of the endogenous CLF [48], and subsequently, the genomic DNA was quantified by real-time PCR or examined by PCR if the amounts of DNA in a ChIP sample were too low to be quantified. We found that both the region (FLC-P2) around the transcription start site (TSS) and 59 part of Intron I of FLC (FLC-I) were greatly enriched, whereas a 59 promoter region 1.8 kb upstream from the TSS in FLC was not enriched ( Figure 4B and 4C). Moreover, we found that regions in the first introns of MAF4 and MAF5 were also enriched ( Figure 4B), whereas MAF3, a close relative of MAF4 and MAF5 located immediately upstream MAF4 ( Figure 4A), and At5g65090, the gene immediately downstream MAF5 (At5g65080), were not enriched ( Figure 4C). Together, these data suggest that CLF selectively binds to FLC, MAF4 and MAF5 in vivo to repress the expression of these genes.
To examine whether CLF directly interacts with the FT locus, using ChIP-PCR we checked the middle region of FT (FT-I; see Figure 4A), a region where FLC has been shown to bind [28]. As shown in Figure 4C, FT fragments were strongly enriched in the ChIP samples from the 35S:GFP:CLF clf transgenic line. Hence, CLF directly interacts with FT chromatin to represses FT expression during vegetative development.

Loss of CLF Function Leads to Reduction in Global H3K27
Trimethylation, but not in H3K27 Dimethylation during Vegetative Development CLF is a plant homolog of the Drosophila E(z), an H3K27 methyltransferase in the Esc-E(z) PRC2 complex [61,62]. Previous studies have shown that E(z) and E(z)H2, the mammalian homolog of E(z), display PRC2-complex-dependent H3K27 methyltransferase activities on chromatin substrate (reviewed in [63]). It has been shown that CLF is partly required for H3K27me3 in CLF-target genes such as AG and STM [48]. We compared global histone methylation levels in clf and wild-type Col seedlings, including H3K27 dimethylation, H3K27 trimethylation and H3K4 trimethylation. Levels of trimethyl H3K27 were strongly reduced in clf relative to Col ( Figure 5A), whereas levels of dimethyl H3K27 and trimethyl H3K4 in clf were similar to those in Col ( Figure 5B and 5C), indicating that CLF is likely to be a histone methyltransferase catalyzing H3K27 trimethylation. Interestingly, lower levels of trimethyl H3K27 were still detected in clf mutant seedlings, which may be deposited by PRC2-like complexes containing CLF relatives including SWN and MEA.
CLF Mediates the Deposition of H3K27me3 in FLC, MAF4, MAF5 and FT As noted above, CLF mediates global H3K27 trimethylation during vegetative development; in addition, recent whole-genome analysis of H3K27 trimethylation in Arabidopsis has revealed that this modification is associated with FLC chromatin in the absence of vernalization treatment [39], which is likely deposited by a CLF-containing PRC2-like complex. It was of interest to examine the H3K27 trimethylation state in FLC, MAF4 and MAF5 in clf seedlings. As shown in Figure 6A, H3K27me3 was enriched in the promoter region FLC-P2 and 59 part of Intron I of FLC (FLC-I) in Col and loss of CLF activities significantly reduced the levels of trimethyl H3K27, consistent with the derepression of FLC in clf ( Figure 3A). Furthermore, H3K27me3 was also enriched in MAF4 and MAF5 in the wild type and strongly reduced in clf ( Figure 6B). In contrast, very little trimethyl H3K27 was detected in the neighboring genes including MAF3 and At5g65090 ( Figure 6B). In addition, we did not detect trimethyl H3K27 in FLM ( Figure 6B), another close relative of FLC and MAFs. Together, these data show that CLF mediates the deposition of trimethyl H3K27 selectively in FLC, MAF4 and MAF5, consistent with the selective de-repression of these three genes, but not FLM or MAF3 in clf.
We also found that H3K27me3 was enriched in FT chromatin in Col as reported previously [39], and that H3K27me3 in FT was nearly eliminated in clf ( Figure 6B), consistent with the drastic derepression of FT in clf ( Figure 3C). As described above, CLF, EMF2 and FIE may be part of a PRC2-like complex that represses FT expression. Together, these data suggest that a CLF-containing PRC2-like complex may be responsible for depositing repressive H3K27me3 in FT chromatin.

CLF-Dependent H3K27 Trimethylation Suppresses H3K4 Trimethylation in its Target-Gene Chromatin
As noted above, PRC2 subunits repress but do not fully silence FLC and FT expression because both genes are still expressed at low levels in wild-type seedlings. It has been shown that active H3K4me3 is associated with FLC chromatin in Arabidopsis accessions which lack of FRI such as Col and Wassileskija (Ws) in which FLC expression is repressed [14,64], and repressive H3K27me3 is also associated with FLC chromatin in these accessions in the absence of vernalization treatment [39,64] (also see Figure 6A). However, it remains unknown whether FLC chromatin can simultaneously carry these two modifications as it is formally possible that these modifications could occur in two subpopulations of FLC chromatin and not in the same physical region of FLC. To examine whether FLC chromatin concomitantly carries both H3K4me3 and H3K27me3, we performed a sequential ChIP in which FLC chromatin from seedlings was immunoprecipitated first with anti-trimethyl H3K4 and second with anti-trimethyl H3K27. Both the region around TSS (FLC-P2) and 59 part of Intron I of FLC (FLC-I) in part of the FLC chromatin concomitantly harbor H3K4me3 and H3K27me3 ( Figure 7A). Similarly, using sequential ChIP we also found that the 59 transcribed region (FT-E) and the middle of FT (FT-I) in part of the FT chromatin simultaneously harbor H3K4me3 and H3K27me3 ( Figure 7A). In addition, we did not detect any DNA fragments from a heterochromatic locus Ta3 [65] that lacks of H3K4me3 or from a constitutive expressed house-keeping gene ACTIN 2 (ACT2) carrying abundant H3K4me3 (data not shown) but lacking of H3K27me3 ( Figure 7A). Together, these data show that part of the FLC and FT chromatin simultaneously possesses the bivalent chromatin marks of active H3K4me3 and repressive H3K27me3.
We further investigated the interaction of H3K27 trimethylation with H3K4 trimethylation in FLC and FT chromatin. The H3K4 trimethylation state in these two loci was examined in clf seedlings by ChIP. Levels of trimethyl H3K4 in 59 genomic FLC including FLC-P1 and FLC-P2 regions and in the 59 transcribed  region of FT (FT-E) and the middle of genomic FT (FT-I) were increased upon loss of CLF activities ( Figure 7B and 7C), consistent with FLC and FT derepression in clf. Furthermore, the levels of trimethyl H3K4 in MAF4 and MAF5 were also increased in clf relative to Col ( Figure 7B), in line with MAF4 and MAF5 derepression in clf. Together, these data suggest that the CLFdependent H3K27 trimethylation suppresses H3K4 trimethylation in FLC, MAF4, MAF5 and FT. Interestingly, the global levels of trimethyl H3K4 in clf were similar to those in Col ( Figure 5C), indicating that CLF-containing PRC2-like complexes only suppresses the H3K4 trimethylation in their target-gene chromatin.

Discussion
Our studies reveal that the Arabidopsis PRC2-like complex subunits CLF, EMF2 and FIE repress the expression of FLC and FLC relatives including MAF4 and MAF5, and that CLF directly binds to and mediates the deposition of repressive H3K27me3 in these three loci. Furthermore, we show that during vegetative development CLF and FIE strongly repress FT expression, and that CLF directly interacts with and mediates the deposition of H3K27me3 in FT chromatin. Our results suggest that CLFcontaining PRC2-like (CLF-PRC2) complexes containing EMF2 and FIE directly interact with and deposit into the FLC, MAF4, MAF5 and FT chromatin repressive trimethyl H3K27 leading to the suppression of active H3K4me3 in these loci, and thus repress the expression of these flowering genes. Given the central roles of FLC and FT in flowering-time regulation in Arabidopsis, these findings suggest that CLF-PRC2 complexes play a significant role in control of the Arabidopsis flowering.

PRC2 Subunits-Mediated Repression of FLC and FLC Relatives
Previous studies indicate that a PRC2-like complex containing VRN2, FIE and SWN or CLF might be involved in FLC repression in Arabidopsis plants grown in normal conditions [51]. In this study, we show that CLF is an essential component for FLC repression because CLF directly binds to FLC chromatin and loss of CLF function leads to a reduction in H3K27me3 and FLC derepression. SWN, a CLF relative, may also play a role in FLC repression because low levels of trimethyl H3K27 in FLC chromatin have still been detected in clf seedlings ( Figure 6A) and simultaneous co-suppression of SWN and CLF leads to FLC derepression [51], though swn mutants do not display a phenotype [57]. In addition, we have found that EMF2, a CLF-interacting partner [57], represses FLC expression. Previously it has been shown that VRN2, an EMF2 relative, also interacts with CLF and represses FLC expression in the absence of vernalization [51,66]. EMF2 and VRN2 can act in partial redundancy in PRC2-like complexes [57]; hence, these two proteins may act in partial redundancy to repress FLC expression. Furthermore, we have found that CLF, EMF2 and FIE also repress the expression of MAF4 and MAF5. Together, these findings suggest that these PRC2 subunits may form a CLF-PRC2 complex that directly represses FLC, MAF4 and MAF5 expression.
The Drosophila PRC2 complex contains four core components including E(z), Esc, Su(z)12 and p55, and these components are evolutionarily conserved in animals and plants (reviewed in [40,41]). CLF and SWN, EMF2 and VRN, and FIE are homologs of E(z), Su(z)12, and Esc respectively. Arabidopsis has five homologs of p55 including MSI1 and FVE. MSI1 is part of a PRC2-like complex that regulates seed development [43], but is not involved in FLC repression [67]. FVE, a component in the autonomous pathway, represses FLC expression to promote flowering [5]. fve mutants grown under normal conditions, are phenotypically wild type except for late flowering [5], whereas clf mutants, emf2 mutants and FIE-suppressed plants display pleiotropic phenotypes [47,53,55], suggesting that these three genes play a role in plant development that is much broader than that played by FVE. Interestingly, like CLF, EMF2 and FIE, FVE also represses MAF4 and MAF5 expression ( Figure S1). Together, these findings are consistent with a model in which a CLF-PRC2 complex composed of CLF, EMF2, VRN2, FIE and FVE selectively represses the expression of FLC, MAF4 and MAF5 to promote the floral transition in the absence of vernalization. In addition, SWN might also be part of this complex and may partially substitute for CLF. It is noteworthy that FVE can directly interact with a plant retinoblastoma protein (see the discussion below) [5], and future biochemical experiments are required to assess whether FVE is part of a CLF-PRC2 complex.
A CLF-PRC2 Complex May Act in Concert with the Autonomous-Pathway Repressors to Repress FLC Expression in the Absence of Vernalizaition The autonomous pathway includes six classic loci such as FCA, FLD and FVE, and these genes do not form a linear pathway [68]. This pathway is so named because mutations in these genes lead to late flowering in all photoperiods due to the elevated FLC expression (reviewed in [2]). FLD, a plant homolog of the human Lysine-Specific Demethylase 1 that has been found in histone deacetylase co-repressor complexes, is involved in the H3K4 demethylation (a mechanism associated with gene repression) and deacetylation of FLC chromatin [8,17]. In addition, recent studies have shown that FCA functions closely with FLD, and that like FLD, it is involved in H3K4 demethylation of FLC chromatin [18]. In this study, we have found that removing CLF and FCA function leads to the synergistic FLC derepression, indicating that the CLF-PRC2 complex-mediated H3K27me3 acts in partial redundancy with the FCAand FLD-mediated chromatin repression in FLC suppression in the absence of vernalization. In addition, our studies also suggest that the CLF-dependent H3K27 trimethylation may antagonize H3K4 trimethylation in FLC chromatin, indicating that H3K27 trimethylation may facilitate H3K4 demethylation in FLC chromatin. Furthermore, the Drosophila PRC2 complex has been shown to be associated with histone deacetylases, suggesting that histone deacetylation is also linked to the PRC2-mediated gene repression [69]. Interestingly, recent studies have shown that FVE can directly interact with a plant retinoblastoma protein of which the human homolog has been found to be associated with a histone deacetylase complex [70], and that FVE is indeed involved in the deacetylation of FLC chromatin [5]. Taken together, it is likely that a CLF-PRC2 complex may act in concert with the autonomous-pathway repressors such as FCA and FLD, and histone deacetylases to generate a repressive chromatin environment through histone deacetylation, H3K4 demethyaltion and H3K27 trimethylation, and thus represses FLC expression.

Recruitment of PRC2 Subunits to the Target Loci
FLC, FLM and MAF2-5 are close relatives and have similar genomic structures [3,13]. Particularly, MAF2, MAF3, MAF4 and MAF5 are arrayed in a gene cluster (a tandem array) located at the bottom of Chromosome 5 [13]; however, CLF represses only MAF4 and MAF5, but not MAF2 or MAF3 in this gene cluster. The CLF-dependent H3K27me3 occurs in MAF4 and MAF5, but is absent from MAF3 and At5g65090 (the gene immediately downstream MAF5), suggesting that the H3K27 trimethylation in MAF4 and MAF5 is not the result of spreading from the neighboring genes. Furthermore, CLF specifically binds to MAF4 and MAF5, but not to MAF3 or At5g65090. This suggests that CLF is specifically recruited to the MAF4 and MAF5 loci, indicating that there are cis-regulatory DNA elements in these two genes that may function similarly to Polycomb-group response elements in Drosophila [40] to recruit a PRC2-like complex.

PRC2 Subunits-Mediated FT Repression
PRC2 subunits CLF, EMF2 and FIE all strongly repress FT expression during vegetative development, suggesting that a PRC2-like complex containing CLF, EMF2 and FIE represses FT expression. To date, all known PRC2 complexes in animals and plants contain four core components including p55 or a p55 homolog (reviewed in [40,41]); however, the p55 homolog directly involved in FT repression still remains elusive. FVE, a p55 homolog and an FLC repressor, is not directly involved in FT repression because FT is strongly repressed in fve due to the elevated FLC expression [67], indicating that the PRC2-like complex repressing FT expression might be different from the one involved in FLC repression. Consistent with this notion, we have found that H3K27 trimethylation in FT chromatin is nearly eliminated in clf, whereas low levels of trimethyl H3K27 in FLC chromatin have been detected in clf, indicating that CLF relatives such as SWN may partially substitute for CLF in the deposition of H3K27me3 in the FLC locus, but not in the FT locus.
Our studies suggest that the putative CLF-PRC2 complex directly deposits repressive H3K27me3 in FT chromatin to repress FT expression. FT chromatin can be simultaneously marked with active H3K4me3 and repressive H3K27me3; the CLF-dependent H3K27 trimethylation suppresses, but does not eliminate H3K4 trimethylation in FT chromatin ( Figure 7A and 7C), consistent with that FT is repressed but not fully silenced by PRC2 subunits in vegetative development. Recent studies suggest that LHP1 specifically recognizes and binds to H3K27me3 deposited by PRC2-like complexes to maintain stable transcriptional gene repression [37,38]. LHP1 has been shown to directly bind to the FT locus and loss of LHP1 activities leads to FT derepression and early flowering [36,37]. Hence, the CLF-dependent H3K27me3 in FT chromatin may be 'read' by LHP1 resulting in stable FT repression during vegetative development.

Possible Role of the CLF-PRC2 Complex-Mediated FT Repression in the Regulation of FT by Photoperiod
The PRC2-mediated transcriptional gene repressing mechanisms are conserved in animals and plants (reviewed in [40,41]). Our studies suggest that during vegetative development, Arabidopsis exploits these evolutionarily conserved ancient gene-repressing mechanisms to control FT expression; specifically, a CLF-PRC2 complex is utilized to repress, but not to fully silence FT expression in vegetative development. In the absence of PRC2 subunits, FT is highly activated; for instance, levels of FT transcripts in clf seedlings are about 200 fold of those in the wild type. It has been shown that in the wild type FT is expressed in vasculature such as veins of leaves where day length is perceived (reviewed in [1]). Previous studies show that loss of CLF activities leads to a strong derepression of AG throughout the leaf including veins and mesophyll cells [47]; hence, loss of CLF-PRC2-complex activities may well lead to FT derepression throughout the leaf including veins. Overexpressing FT via a strong constitutive viral promoter (35S) has been shown to give rise to extremely early flowering independent of the photoperiods [29,30]. Thus, it is critical for plants to keep FT to be expressed at low levels for preventing precocious flowering and for the regulation of FT by the photoperiods. PRC2 subunits, likely functioning in the context of a CLF-PRC2 complex, maintain FT expression at basal lower levels in vegetative development, which may serve to provide some room for the elevated FT expression in response to photoperiods and thus enable the photoperiodic control of flowering time in plants.

RNA isolation, reverse transcription and quantitative PCR assays
Total RNAs from aerial parts of 7 to 10 day-old seedlings grown in long days were extracted as described previously [17]. cDNAs were reverse-transcribed from total RNAs with Moloney murine leukemia virus reverse transcriptase (Promega).
Real-time quantitative PCR was performed on an ABI Prism 7900HT sequence detection system using SYBR Green PCR master mix (Applied Biosystems) as described previously [17]. Each sample was quantified at least in triplicate and normalized using TUB2 (At_5g62690) as the endogenous control. Primers used are specified in Table S1.

Histone extraction and immunoblotting
Histone protein extraction and Western analysis were performed as described previously [18,71]. Briefly, total histones were extracted from about 10-day-old seedlings, separated in an SDS-PAGE gel, and subsequently were transferred to a 0.2-mm nitrocellulose membrane (Bio-Rad). The protein blots were first probed with anti-trimethyl H3K27, anti-dimethyl H3K27 (Millipore) and anti-trimethyl H3K4 (Abcam), and followed by anti-H3 (Millipore). The chemiluminescent SuperSignal West Pico system (Pierce) was used to develop the protein blots according the manufacturer's instructions.

Chromatin immunoprecipitation (ChIP)
The ChIP experiments were performed as described previously [65] using seedlings. Rabbit polyclonal anti-trimethyl-histone H3 (Lys 4) (Abcam), anti-trimethyl-histone H3 (Lys 27) (Upstate) and anti-GFP (Invitrogen) were used in immunoprecipitation experiments. Amounts of the immunoprecipitated genomic DNA were examined by PCR or quantified by real-time quantitative PCR. The PCR amplification of a genomic region was usually tried at several cycle numbers to identify a cycle number at which the amplification of DNA fragments in the immunoprecipitated DNA samples did not reach the plateau phase. Quantitative measurements of various regions of FLC, MAF4, MAF5 and FT were performed using SYBR Green PCR master mix (Applied Biosystems). Primers used to amplify FLC-P1, FLC-P2, ACTIN 2, TUB2 and TUB8 were described previously [17,72], and other primers used are specified in Table S1. Each of the immunoprecipitations was repeated independently once, and each sample was quantified in triplicate.

Sequential ChIP analysis
The sequential ChIP experiments were performed as previously described [73] with modifications. Briefly, chromatin from Ws seedlings was immunoprecipitated with anti-trimethyl H3K4, subsequently eluted in a solution of 500 mM NaCl, 30 mM DTT and 0.1% SDS at 37uC, and was further diluted in a lysis buffer [65] supplemented with 16 Roche protease inhibitor cocktails (-EDTA). The eluted chromatin was subsequently immunoprecipitated with anti-trimethyl H3K27; DNA fragments were recovered and purified for PCR analysis. Figure S1 FVE represses MAF4 and MAF5 expression. Total RNAs were extracted from Col, fve and fca seedlings grown in long days. MAF4 and MAF5 were de-repressed in fve, but not in fca.