Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Previous studies in Arabidopsis thaliana have identified several histone methylation enzymes, including ARABIDOPSIS TRITHORAX1 (ATX1)/SET DOMAIN GROUP 27 (SDG27), ATX2/SDG30, LSD1-LIKE1 (LDL1), LDL2, SDG8, SDG25, and CURLY LEAF (CLF)/SDG1, as regulators of the key flowering repressor FLOWERING LOCUS C (FLC) and the florigen FLOWERING LOCUS T (FT). However, the combinatorial functions of these enzymes remain largely uninvestigated.
Here, we investigated functional interplays of different histone methylation enzymes by studying higher order combinations of their corresponding gene mutants.
We showed that H3K4me2/me3 and H3K36me3 depositions occur largely independently and that SDG8-mediated H3K36me3 overrides ATX1/ATX2-mediated H3K4me2/me3 or LDL1/LDL2-mediated H3K4 demethylation in regulating FLC expression and flowering time. By contrast, a reciprocal inhibition was observed between deposition of the active mark H3K4me2/me3 and/or H3K36me3 and deposition of the repressive mark H3K27me3 at both FLC and FT chromatin; and the double mutants sdg8 clf and sdg25 clf displayed enhanced early-flowering phenotypes of the respective single mutants.
Collectively, our results provide important insights into the interactions of different types of histone methylation and enzymes in the regulation of FLC and FT expression in flowering time control.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Flowering in a correct time of year is critical to ensure plant reproductive success. This major developmental transition, from vegetative to reproductive growth, is fine-tuned by both intrinsic and environmental factors. In Arabidopsis thaliana, several distinct pathways, including autonomous, photoperiod, vernalization, and gibberellin (GA)-dependent pathways, form a complex regulatory network that integrates physiological information with environmental cues to control flowering time (Amasino, 2010; Crevillen & Dean, 2011). FLOWERING LOCUS C (FLC) encodes a MADS-box transcription factor and plays a central role in A. thaliana flowering time regulation via direct repression of flowering time integrators FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) (Amasino, 2010; Crevillen & Dean, 2011). Both autonomous and vernalization pathways repress FLC expression, which subsequently induces FT and SOC1 expression to promote flowering. Distinctively, long-day photoperiods activate FT expression via CONSTANS (CO) and FT further activates SOC1 expression to promote flowering (Lee & Lee, 2010). Genome-wide analysis of binding sites of the FLC protein has revealed that, among others, some genes involved in the GA pathway and in circadian rhythm are targets of FLC (Deng et al., 2011), implying an FLC-mediated connection with additional flowering pathways.
FLC regulates flowering time in a dosage-dependent manner, and FLC transcription is fine-tuned by a range of chromatin modification factors (Berr et al., 2011; Buzas et al., 2012; He, 2012). Among others, Polycomb Group (PcG) and Trithorax Group (TrxG) proteins play key roles in epigenetic repression and activation of FLC, respectively. Initially identified in Drosophila for their roles in regulating homeotic gene expression, PcG and TrxG are widely conserved in animals and plants and are involved in various developmental and cellular processes (Pien & Grossniklaus, 2007; Schuettengruber et al., 2007). The SET-domain protein CURLY LEAF (CLF)/SET DOMAIN GROUP 1 (SDG1) constitutes the catalytic core of Polycomb Repressive Complex 2 (PRC2), which acts as a histone methyltransferase (HMTase) catalysing trimethylation on histone H3 lysine 27 (H3K27me3) at FLC chromatin (Wood et al., 2006; Finnegan & Dennis, 2007; De Lucia et al., 2008; Jiang et al., 2008; Doyle & Amasino, 2009; Angel et al., 2011; Buzas et al., 2011; Coustham et al., 2012). Studies have shown that the SET-domain protein H3K27me3 abundance across the FLC body is inversely correlated with transcription in a range of A. thaliana ecotypes and mutants exhibiting different flowering times, and that PcG-mediated repression is crucial for the maintenance of FLC at low expression levels during vernalization and after return to warm temperatures. The CLF homologue SWINGER (SWN)/SDG10 may also contribute to FLC repression, in particular upon vernalization (Wood et al., 2006; De Lucia et al., 2008). Long noncoding RNAs (lncRNAs) have been identified from the FLC locus and proposed to act together with PRC2 to establish the FLC repressive chromatin state (Swiezewski et al., 2009; Helliwell et al., 2011; Heo & Sung, 2011). FLC reverts to an active state during gametogenesis and/or embryogenesis (Sheldon et al., 2008; Choi et al., 2009; Yun et al., 2011), and the process of reversion predictably involve both the removal of the repressive H3K27me3 and the establishment of an active chromatin state. The precise identity of chromatin modification factors involved in this reversion process remains to be established.
Active FLC transcription during vegetative growth is associated with high levels of H3K4me2/me3 and H3K36me2/me3, which are deposited by the TrxG family HMTases, including SDG8/EARLY FLOWERING IN SHORT DAYS (EFS)/ASH1-HOMOLOG2 (ASHH2) (Kim et al., 2005; Zhao et al., 2005; Xu et al., 2008), ARABIDOPSIS TRITHORAX1 (ATX1)/SDG27 and ATX2/SDG30 (Pien et al., 2008; Saleh et al., 2008), and SDG25/ATXR7 (Berr et al., 2009; Tamada et al., 2009). These different HMTases are recruited to transcriptional protein complexes during the activation, initiation and/or elongation processes of FLC transcription (Jiang et al., 2009; Ko et al., 2010; Berr et al., 2011; Choi et al., 2011). The LYSINE-SPECIFIC DEMETHYLASE1 (LSD1) family genes LDS1-LIKE1 (LDL1), LDL2 and FLOWERING LOCUS D (FLD) are involved in H3K4 demethylation at FLC chromatin (Jiang et al., 2007; Liu et al., 2007). Recently, it was reported that the JmjC domain protein JMJ18 binds FLC chromatin and acts in H3K4 demethylation (Yang et al., 2012). JMJ12/RELATIVE OF EARLY FLOWERING 6 (REF6) was also reported to be involved in H3K4 as well as H3K36 demethylation at FLC chromatin (Ko et al., 2010). In contrast, however, Lu et al. (2011) showed that JMJ12/REF6 is specifically involved in H3K27 demethylation.
FT encodes a small mobile protein considered to be a florigen that moves through the phloem from leaves to the shoot apex to promote floral meristem identity (Wigge et al., 2005; Kobayashi & Weigel, 2007). FT expression is also directly regulated by several chromatin-remodelling factors (He, 2012). CLF-PRC2 deposits H3K27me3 on FT chromatin to repress FT expression (Jiang et al., 2008; Farrona et al., 2011; Lopez-Vernaza et al., 2012). FT misexpression in clf mutants masks the effect of FLC misexpression and causes an early-flowering phenotype (Lopez-Vernaza et al., 2012). JMJ12/REF6 antagonizes the CLF-PRC2 repression through H3K27 demethylation at FT chromatin (Lu et al., 2011). While JMJ14 binds FT chromatin and mediates H3K4 demethylation to repress FT expression (Jeong et al., 2009; Lu et al., 2010; Yang et al., 2010), the enzymes catalysing H3K4 and H3K36 methylations at FT chromatin remain to be identified.
The multiple enzymes involved in the regulation of histone methylations/demethylations at FLC and/or FT chromatin provide an interesting opportunity to investigate cross-talk among these different regulators of histone methylation. Previous studies have investigated only a few combined mutants and in most cases have been limited to one or two types of histone methylation at FLC chromatin. In this study, we generated previously unavailable higher order combinations of A. thaliana gene mutants and gained a comprehensive insight into the interaction of these genes in the regulation of flowering time, flowering gene expression and histone methylation.
Materials and Methods
The single mutants sdg8-1, atx1-2, atx2-2, sdg25-1, ldl1-2, ldl2 and clf-29, and the double mutant ldl1-2 ldl2 have been previously described (references in Table 1). They are loss-of-function gene mutants, caused by a T-DNA insertion within the corresponding gene, and all are derived from the Arabidopsis thaliana (L.) Heynh Colombia (Col) ecotype. Seeds were produced under glasshouse conditions.
Table 1. Arabidopsis thaliana mutant lines used in this study
To obtain higher order combinations of mutants, genetic crosses were performed. Seeds (F1) were collected from individual siliques produced by hand pollination. They were grown and self-pollinated to obtain the next segregating populations. Genomic DNA was isolated from leaf sections of individual plants using a previously described method (Berendzen et al., 2005). To determine the genotypes of individual plants, DNA was analysed by PCR amplification using gene-specific primers (sequences available in Supporting Information Table S1). PCR was carried out using the following conditions: pre-heating at 95°C for 10 min, followed by 30 cycles of 15 s at 95°C, 30 s at 60°C and 15 s at 72°C.
Flowering time analysis was conducted on plants grown in soil in a glasshouse at 22°C under different photoperiods: long day (LD; 16 h light : 8 h dark), medium day (MD; 12 h light : 12 h dark) and short day (SD; 8 h light : 16 h dark). For vernalization tests, 7-d-old seedlings were maintained at 4°C for 40 d and then grown under different photoperiod conditions. Flowering time was measured from a developmental perspective as the total number of rosette leaves at bolting.
RNA extraction and RT-PCR
Total RNA was extracted from the aerial parts of 16-d-old seedlings grown under MD conditions with the Nucleospin RNA Plant kit (Macherey-Nagel; http://www.mn-net.com). First-stand cDNA synthesis was performed using 2 μg of RNA with Impro-II Reverse Transcriptase (Promega; http://www.promega.com). Quantitative real-time PCR was performed on a LightCycler 480II (Roche; http://www.roche-applied-science.com), using the SYBER Green Master mix. Each sample was analysed in triplicate, and ACTIN2 (ACT2) and GLYCERALDEHYDE-3-PHOSPHATE DEHYDROGENASE C2 (GAPC2) were used as internal reference genes (Pien et al., 2008; Saleh et al., 2008; Berr et al., 2010; Ko et al., 2010). Gene-specific primers are listed in Table S1.
Sixteen-day-old seedlings grown under MD conditions were used in chromatin immunoprecipitation (ChIP) analysis. ChIP was performed according to the previously described protocol (Berr et al., 2010). Antibodies used in this study were anti-trimethyl-histone H3K4 (07-473; Millipore; http://www.millipore.com), anti-dimethyl-histone H3K4 (pAb-003-050; Diagenode; http://www.diagenode.com), anti-trimethyl-histone H3K27 (07-449; Millipore) and anti-trimethyl-histone H3K36 (Ab9050; Abcam; http://www.abcam.com). ChIP DNA was analysed by quantitative real-time PCR using gene/region-specific primers listed in Table S1. ACT2 and FUSCA3 (FUS3) were used as internal standards for normalization of active (H3K4me2/me3 and H3K36me3) and repressive (H3K27me3) marks, respectively (Jiang et al., 2007; Pien et al., 2008; Jeong et al., 2009; Kwon et al., 2009; Berr et al., 2010; Ko et al., 2010).
Production of higher order combination mutants
We produced higher order mutant combinations through genetic crosses of previously available mutants (Table 1). Seeds (F1) from crosses were collected from individual siliques on the parent plant, grown and self-pollinated to obtain the segregated F2 population. More than 90 individual F2 plants from each cross were genotyped by genomic PCR to identify double mutants. For triple mutants, segregation and PCR genotyping were continued until the F4 generation to obtain all three gene mutant alleles in a homozygous state. In this study, we obtained the double mutants sdg8-1 atx1-2, sdg8-1 atx2-2, sdg25-1 clf-29 and sdg8-1 clf-29, and the triple mutants sdg8-1 ldl1-2 ldl2 and sdg25-1 ldl1-2 ldl2. Hereafter, we omit the mutant allele suffix to shorten the mutant name; for example, sdg8-1 is shortened to sdg8 and sdg8-1 atx1-2 to sdg8 atx1. More details concerning the alleles can be found in Table 1
sdg8 overrides atx1, atx2 and ldl1 ldl2 in flowering time regulation
Under our glasshouse growth conditions with LD photoperiods, sdg8 but not atx1, atx2 or ldl1 ldl2 plants showed a drastically reduced size compared with the wild-type Col (Fig. 1a). Remarkably, sdg8 atx1, sdg8 atx2 and sdg8 ldl1 ldl2 plants had a small size comparable to that of sdg8, whereas sdg25 and sdg25 ldl1 ldl2 plants had a normal size comparable to that of Col (Fig. 1a). We next measured the flowering times of these mutants by the leaf number at bolting. Consistent with their descriptions in previous studies (see references in Table 1), the sdg8, atx1, atx2 and sdg25 mutants were early flowering, whereas ldl1 ldl2 was late flowering, compared with Col (Fig. 1b). Interestingly, the sdg8 atx1, sdg8 atx2 and sdg8 ldl1 ldl2 mutants showed an early-flowering phenotype similar to that of sdg8, whereas sdg25 ldl1 ldl2 showed a flowering time later than that of sdg25 but earlier than that of ldl1 ldl2 (Fig. 1b). Similar mutant effects on flowering time were also found in plants grown under MD or SD photoperiods and with (+V) or without (−V) vernalization treatment (Fig. 1b). Taken together, these data indicate that, under all plant growth conditions studied, sdg8 overrides atx1, atx2 and ldl1 ldl2 in flowering phenotype determinacy, whereas sdg25 and ldl1 ldl2 act antagonistically in a rather synthetic manner in flowering time determinacy.
sdg8 and sdg25 enhance clf phenotype
The clf mutant has upward-curled rosette leaves and small plant size (Fig. 1a; Goodrich et al., 1997; Xu et al., 2008). The sdg8 clf plants showed a greater reduction in rosette size and more severely upward-curled leaves, compared with the clf or sdg8 mutant (Fig. 1a). Similarly, the sdg25 clf plants, albeit slightly bigger than the sdg8 clf plants, also showed a greater reduction in rosette size and more severely upward-curled leaves, compared with the clf or sdg25 mutant (Fig. 1a). Flowering time investigation revealed that the sdg8 clf and sdg25 clf mutants displayed an enhanced early-flowering phenotype compared with clf, sdg8 or sdg25 under all growth conditions studied: with different photoperiods and with or without vernalization treatments (Fig. 1b). These data contrast with the previously observed reversion to wild-type phenotype of the atx1 clf double mutant (Saleh et al., 2007) and suggest that clf and sdg8 or sdg25 act in parallel on flowering regulatory pathways that together enhance the mutant early-flowering phenotype.
Flowering time genes FLC, FT and SOC1 are misregulated in the mutants
To investigate the molecular mechanism underlying the flowering phenotypes of the mutants, we analysed the expression levels of several flowering time genes by quantitative RT-PCR (Fig. 2). FLC expression was detected at reduced levels in sdg8, atx1, atx2 and sdg25 but at increased levels in ldl1 ldl2 and clf, compared with that in Col. Remarkably, sdg8 atx1, sdg8 atx2, sdg8 ldl1 ldl2 and sdg8 clf showed an FLC expression level roughly similar to that of sdg8, whereas sdg25 ldl1 ldl2 showed an FLC expression level between those of sdg25 and ldl1 ldl2, and sdg25 clf a level between those of sdg25 and clf. These data indicate that sdg8 but not sdg25 overrides atx1, atx2, ldl1 ldl2 or clf in the alteration of FLC expression. In contrast to FLC, MADS AFFECTING FLOWERING 2 (MAF2) expression was unchanged in all our studied mutants (Fig. 2).
FT is repressed by FLC and acts downstream to promote flowering. In sdg8, atx1, atx2, sdg8 atx1 and sdg8 atx2, FT was up-regulated to different levels that inversely correlated with FLC expression levels (Fig. 2). In sdg8 ldl1 ldl2, however, the up-regulation of FT was weaker than in sdg8, although similar FLC levels were found in sdg8 and in sdg8 ldl1 ldl2. This observation suggests that LDL1/LDL2 may directly suppress FT expression. In support of this assumption, in ldl1 ldl2 the elevated FLC expression failed to repress FT and instead a slight increase in FT expression was detected (Fig. 2). In sdg25, the reduced FLC level did not affect FT expression, indicating that SDG25 is necessary in promoting FT expression. LDL1/LDL2 repression seems to dominate over SDG25 activation because a similar increase in FT expression was detected in sdg25 ldl1 ldl2 as in ldl1 ldl2 (Fig. 2). Finally, although it had an increased FLC level, a drastic increase in FT expression was detected in clf, suggesting that FLC depends on CLF in FT repression. The increase in FT expression was reduced in sdg8 clf and sdg25 clf compared with that in clf, indicating that SDG8 and SDG25 may be directly involved in FT activation. Taken together, our data reveal that the FT expression patterns cannot be fully explained by altered FLC levels in our studied mutants and that at least SDG8, SDG25, LDL1/LDL2 and CLF seem to simultaneously regulate both FLC and FT.
SOC1 encodes a MADS-box transcription factor and acts in parallel with FT in promoting flowering. In addition to being repressed by FLC, SOC1 expression is also activated by FT (Yoo et al., 2005). Consistently, in our studied mutants, changes in SOC1 expression were less pronounced compared with changes in FT expression (Fig. 2). The observed SOC1 expression patterns could be explained by combined alterations of FLC and FT levels in the different mutants (Fig. 2), making it less clear whether or not any of our studied enzymes are directly involved in SOC1 transcription regulation.
H3K4me2/me3 and H3K36me3 levels are altered in a largely independent manner at flowering time genes in the mutants
We next investigated histone methylation levels at flowering time genes to reveal mutant effects. As illustrated in Fig. 3, three regions of FLC (a, b and c), two of SOC1 (a and b), one of MAF2, and one of FT were analysed in ChIP assays using antibodies specifically recognizing a type of histone methylation. We found that in sdg8 the levels of H3K36me3 but not H3K4me2/me3 were reduced at FLC and FT, and very slightly at SOC1 and MAF2 (Fig. 4). In atx1, reduced H3K4me3 levels and increased H3K4me2 levels were detected at FLC, FT and SOC1 but not at MAF2, whereas in atx2 reduced H3K4me2 but unchanged H3K4me3 was detected at FLC (regions b and c) and FT (Fig. 4). The double mutants sdg8 atx1 and sdg8 atx2 showed alterations of H3K36me3 and H3K4me2/me3, which were largely additive of the changes observed in the respective single mutants (Fig. 4), suggesting independence of deposition of these histone methylation marks.
The relationship between H3K36me3 and H3K4me2/me3 depositions was further examined using ldl1 ldl2 and higher order mutants (Fig. 5). Increased levels of H3K4me2/me3 were detected at FLC and FT but not at SOC1 and MAF2 in ldl1 ldl2. We also detected an increase of H3K36me3 at FLC and FT in ldl1 ldl2 (Fig. 5). In sdg8 ldl1 ldl2, the H3K4me2/me3 increase was detected at levels largely similar to those in ldl1 ldl2, whereas H3K36me3 was reduced to similar levels as in sdg8. This indicates that SDG8 is responsible for the H3K36me3 increase observed in ldl1 ldl2. Distinctively, while sdg25 showed reduced levels of both H3K4me3 and H3K36me3 at FLC and FT, sdg25 ldl1 ldl2 showed H3K4me3 and H3K36me3 levels comparable to those of Col at FLC and to those of ldl1 ldl2 at FT (Fig. 5). Thus, the ldl1 ldl2-associated H3K4me3 and H3K36me3 increases at FT were independent of SDG25, and their levels at FLC in sdg25 ldl1 ldl2, which were similar to those in Col, may represent an average between the decreased and increased levels observed in sdg25 and ldl1 ldl2, respectively. In all our studied cases, no dependence/cross-talk between H3K36me3 and H3K4me2/me3 could be detected.
Reciprocal antagonism of H3K4me2/me3 and H3K36me3 with H3K27me3 deposition operates at flowering time genes
Increased levels of H3K27me3 accompanying H3K4me2/me3 and/or H3K36me3 reductions were detected at FLC and FT in sdg8, atx1, sdg8 atx1, sdg8 atx2 and sdg25 (Figs 4, 5), suggesting that H3K4me2/me3 and H3K36me3 antagonize H3K27me3 deposition. We further investigated the relationship of H3K27me3 with H3K4me3 and H3K36me3 in clf and higher order mutants (Fig. 6). We found that clf exhibits a reduction in H3K27me3 accompanied by increased levels of H3K4me3 and H3K36me3 at FLC and FT, suggesting that H3K27me3 antagonizes H3K4me3 and H3K36me3 deposition. As expected, the increase in H3K36me3 in clf was attenuated in sdg8 clf, which showed a reduction in H3K36me3 similar to that in sdg8 but a reduction in H3K27me3 and an increase in H3K4me3 similar to those in clf (Fig. 6). The sdg25 clf mutant showed an H3K27me3 reduction similar to that in clf and H3K4me3 and H3K36me3 reductions slightly less severe than those in sdg25 (Fig. 6). Taken together, our data suggest a reciprocal antagonistic cross-talk between deposition of the repressive mark H3K27me3 by CLF and deposition of the active marks H3K36me3 by SDG8 and H3K4me3 by SDG25 and other HMTase(s).
In this study, we have investigated the interplay of several types of histone methylation regulator and have demonstrated that sdg8 overrides atx1, atx2 or ldl1 ldl2 in FLC expression and flowering time regulation. The decrease in H3K36me3 levels observed at FLC in sdg8 is consistent with previous reports showing that SDG8 encodes an H3K36-specific methyltransferase (Zhao et al., 2005; Dong et al., 2008; Xu et al., 2008; Grini et al., 2009; Berr et al.,2010). It has also been reported that SDG8 exhibits dual substrate specificity for both H3K36 and H3K4 methylation (Ko et al., 2010). However, in our study we did not detect any significant change of H3K4me2/me3 in sdg8. Previously, a reduction in H3K4me3 associated with loss of SDG8 function was detected at FLC primarily in a FRIGIDA (FRI)-containing winter-annual accession (Kim et al., 2005; Ko et al., 2010). FRI serves as a scaffold protein recruiting SDG8, ATX1 and other proteins (Jiang et al., 2009; Ko et al., 2010; Choi et al., 2011); thus, loss of SDG8 might indirectly have caused a reduction in H3K4me3 because of impaired recruitment of ATX1 or another H3K4-methyltransferase. Our study was performed in the summer-annual accession Col, which lacks a functional FRI (Johanson et al., 2000). In agreement with ATX1 catalysing H3K4me3 deposition and ATX2 catalysing H3K4me2 deposition (Pien et al., 2008; Saleh et al., 2008), we observed reductions of H3K4me3 and H3K4me2 at FLC in atx1 and atx2, respectively. In addition, atx1 showed increased levels of H3K4me2, which probably resulted from defects in converting H3K4me2 to H3K4me3.
The SDG8 protein was previously found to contain a CW domain binding H3K4me1/me2, leading to the hypothesis that H3K4me1/me2 may serve as docking sites to recruit SDG8 for downstream H3K36 methylation (Hoppmann et al., 2011). However, this hypothesis is challenged by several observations in our study. At FLC and FT, we observed that levels of H3K36me3 were largely unaffected by the increase in H3K4me2 and the decrease in H3K4me3 in atx1 or by the decrease in H3K4me2 in atx2. Furthermore, the double mutants sdg8 atx1 and sdg8 atx2 displayed H3K36me3 and H3K4me2/me3 alterations largely additive of those of the single mutants. The human LSD1 specifically demethylates H3K4me1 and H3K4me2 (Shi et al., 2004). Consistent with a previous report (Jiang et al., 2007), we detected increases in both H3K4me2 and H3K4me3 at FLC as well as at FT in ldl1 ldl2. The H3K4me3 increase may be caused by the conversion from H3K4me2 (Jiang et al., 2007). Our study additionally revealed increases of H3K36me3 at FLC and FT in ldl1 ldl2. Although we cannot exclude the possibility that LDL1/LDL2 may also demethylate H3K36me2, which could explain H3K36me3 increases in a similar way as for H3K4me3 increases, it is possible that the increase in H3K36me3 was an indirect effect, simply associated with high levels of FLC and FT expression in ldl1 ldl2. Because sdg8 ldl1 ldl2 shows levels of reduced H3K36me3 similar to those of sdg8, SDG8 is believed to be responsible for the H3K36me3 increases observed in ldl1 ldl2. Lastly, reductions of H3K4me1/me2/me3 and H3K36me2 have been previously reported in sdg25 (Berr et al., 2009; Tamada et al., 2009), and we found that both H3K4me3 and H3K36me3 levels were reduced at FLC as well as at FT in sdg25. Examination of sdg25 ldl1 ldl2 indicated that SDG25 and LDL1/LDL2 independently regulate H3K4me3 and H3K36me3 levels. Taken together, our data suggest a largely independent deposition of H3K36me3 and H3K4me2/me3 at the flowering genes FLC and FT.
By contrast, an antagonistic interplay of H3K36me3 and H3K4me2/me3 with H3K27me3 deposition was observed at FLC as well as at FT. In the sdg8, atx1, sdg8 atx1, sdg8 atx2 and sdg25 mutants exhibiting reductions of H3K4me3 and/or H3K36me3, H3K27me3 levels were elevated. This is in agreement with in vitro experiments showing that animal and A. thaliana PRC2 methylates H3K27 more efficiently on nucleosomal substrates containing unmethylated H3 than on those containing H3K4me3 or H3K36me2/me3 (Schmitges et al., 2011; Yuan et al., 2011). In A. thaliana, the Su(z)12 homologues EMBRYONIC FLOWER 2 (EMF2) and VERNALIZATION 2 (VRN2) form two distinct PRC2 complexes and the EMF2–PRC2 but not the VRN2–PRC2 complex is inhibited by H3K4me3 and H3K36me2/me3 (Schmitges et al., 2011). Our study further provides in vivo evidence for CLF acting as a key component of PRC2 in H3K4me3 and H3K36me2/me3 inhibition, because the sdg8 clf and sdg25 clf mutants show low H3K27me3 levels similar to those in clf. The sdg8 emf2 mutant also diplays low H3K27me3 levels at seed maturation loci similar to those in emf2 (Tang et al., 2012). A drastic elevation of H3K4me3 was detected at seed maturation loci in the double mutant sdg8 emf2 but not in the single mutants sdg8 and emf2 (Tang et al., 2012). This synergistic effect of sdg8 and emf2 remains mechanistically obscure. In agreement with a previous study (Jiang et al., 2008), our analysis revealed H3K4me3 elevations at FLC in clf, suggesting a reciprocal antagonistic effect of H3K27me3 on H3K4me3 deposition. H3K27me3 also inhibited H3K36me3 deposition at FLC, in an SDG8-dependent way; H3K36me3 was elevated in clf but reduced to similar low levels in sdg8 clf as in sdg8. It is as yet unknown whether H3K27me3 directly inhibits SDG8 and SDG25 in catalysing H3K36 and/or H3K4 methylation. Nonetheless, the antagonistic interplay between deposition of the active marks H3K4me3 and H3K36me2/me3 and deposition of the repressive mark H3K27me3 could explain in part why many genes in the genome contain either the active or the repressive mark but rarely both types (Roudier et al., 2011).
Consistent with H3K4me3 and H3K36me2/me3 function in transcription activation, the sdg8, atx1, atx2 and sdg25 mutants showed reductions in FLC expression to various degrees and the ldl1 ldl2 mutant showed an increase in FLC expression. Remarkably, the sdg8 atx1, sdg8 atx2 and sdg8 ldl1 ldl2 mutants displayed a reduction in FLC similar to that of sdg8. This differs drastically from findings for sdg25, which enhanced the reduction in FLC when combined with sdg8 or atx1 (Tamada et al., 2009) and attenuated the increase in FLC when combined with ldl1 ldl2 (this study). It is generally considered that H3K4me2/me3 is involved in transcription initiation and H3K36me2/me3 is involved further downstream in transcription elongation (Berr et al., 2011). Our data suggest that H3K36me2/me3, compared with H3K4me2/me3, play a more decisive role in the final FLC transcription outcome, and that an additive down-regulation of H3K36me3 levels caused by sdg8 and sdg25 may explain the enhanced FLC reduction in sdg8 sdg25, as reported in Tamada et al. (2009). The sdg8 clf mutant also showed a low FLC level close to that of sdg8, whereas sdg25 clf showed an increase in FLC, but this increase was smaller than that in clf. A balance between the active marks H3K36me3 and H3K4me2/me3 and the repressive mark H3K27me3 probably determines the outcome of FLC transcription levels. Our combinatorial mutant study revealing SDG8 as the key HMTase regulating FLC expression suggests that further investigation of the SDG8 protein complex(es) is warranted to gain insights into the epigenetic mechanisms of FLC transcription regulation.
While MAF2 largely displayed unchanged expression levels as well as unchanged H3K36me3, H3K4me2/me3 and H3K27me3 at its chromatin, FT chromatin showed histone methylation changes very similarly to those for FLC in our studied mutants. This discovery of the same enzymes regulating both FLC and FT provides opportunity to further examine the interplay between a transcription factor and different histone methylations in transcription regulation. The FLC protein directly binds to the CArG motif within the first intron of FT and represses FT expression (Helliwell et al., 2006; Searle et al., 2006). The derepression associated with the reduction in FLC bypasses the decreased levels of H3K36me3 and/or H3K4me2/me3, resulting in FT up-regulation in the sdg8, atx1, atx2, sdg8 atx1 and sdg8 atx2 mutants. This provides an example contrasting with the general positive correlation between the H3K36me3 and/or H3K4me2/me3 enrichment and the gene transcription level. Transcriptome analyses have revealed that a few hundreds of genes are up-regulated, representing about half the total number of differentially expressed genes in atx1 (Alvarez-Venegas et al., 2006) or sdg8 (Xu et al., 2008; Grini et al., 2009). It is reasonable to speculate that some of these up-regulated genes may be like FT and also contain reduced H3K36me3 and/or H3K4me2/me3 levels. In spite of an elevated FLC level, the clf mutant showed a drastic increase in FT expression associated with H3K27me3 reduction, suggesting that FLC-mediated repression depends on CLF-mediated H3K27me3 deposition at FT. The results from the sdg8 clf and sdg25 clf mutants further indicate that, in the absence of CLF-mediated H3K27me3 deposition, H3K36me3 and H3K4me2/3 levels are critical for the dramatic increase in FT expression. SOC1 is also repressed by FLC (Helliwell et al., 2006; Searle et al., 2006) but activated by FT (Yoo et al., 2005). Consistently, compared with FT, SOC1 displayed compromised alterations of expression and histone methylation in our studied mutants. Notably, the atx2, sdg25 and ldl1 ldl2 mutants did not show a detectable change in H3K4me2/me3 and H3K36me3 at SOC1 chromatin, suggesting that ATX2, SDG25 and LDL1/LDL2 are not directly involved in SOC1 regulation. An increase of H3K27me3 associated with a reduction in SOC1 expression was observed in ldl1 ldl2; however, the underlying molecular mechanism requires further investigation.
The observation of several histone methylation enzymes simultaneously involved in the regulation of FLC and FT, and also SOC1 in some cases, underlines the complexity of flowering time control. A strict linear correlation between flowering time and expression level changes of FLC, FT or SOC1 could not be established in our analysed mutants. It is more likely that FLC, FT, SOC1 and other as yet unstudied target genes together ultimately determine the mutant flowering phenotype. The dominance of sdg8 over atx1, atx2 or ldl1 ldl2 in phenotype determinacy is consistent with the key role of SDG8 in the regulation of FLC, FT and SOC1 expression. More remarkably, enhanced effects were observed between sdg8 or sdg25 and clf in phenotype determinacy. Such types of interaction contrast with the generally known antagonistic relationship between the functions of TrG and PcG, as exemplified by the atx1 clf mutant, which displays a reversed wild-type phenotype (Saleh et al., 2007). The FT and SOC1 expression level changes cannot explain the enhanced early-flowering phenotype of sdg8 clf and sdg25 clf, implying involvement of additional/other target genes downstream from or independent of FLC. Consistent with this, mutations in FT as well as in one of the three other genes (AGAMOUS, FPA and SEPALLATA3) can largely suppress the clf phenotype (Lopez-Vernaza et al., 2012). Finally, various other chromatin factors, including enzymes involved in histone acetylation or monoubiquitination, the ATP-dependent chromatin remodelling factor SWR1 which is involved in deposition of the histone variant H2A.Z, and the histone chaperone FACT, are known to regulate FLC and/or FT expression (reviewed in He, 2012). It is reasonable to speculate that more functional interactions of histone methylation with other chromatin modifications will be uncovered in flowering time control.
The authors thank Dr Yuehui He (Temasek Life Sciences Laboratory, Singapore) for providing the ldl1 ldl2 seeds and the Arabidopsis Biological Resource Center for other mutant seeds. This work was supported by the French Centre National de la Recherche Scientifique (CNRS) and in part by the French Agence Nationale de la Recherche (ANR). S.S. received a PhD fellowship from the Higher Education Commission (HEC) of Pakistan.