Different Polycomb group complexes regulate common target genes in Arabidopsis

Grigory Makarevich, Olivier Leroy, Umut Akinci, Daniel Schubert, Oliver Clarenz, Justin Goodrich, Ueli Grossniklaus, Claudia Köhler

Author Affiliations

  1. Grigory Makarevich1,2,,
  2. Olivier Leroy1,2,,
  3. Umut Akinci2,,
  4. Daniel Schubert3,
  5. Oliver Clarenz3,
  6. Justin Goodrich3,
  7. Ueli Grossniklaus2 and
  8. Claudia Köhler*,1,2
  1. 1 ETH Zürich, Institute of Plant Sciences and Zürich‐Basel Plant Science Center, Plant Developmental Biology, LFW E 31, Universitätsstrasse 2, Zürich, 8092, Switzerland
  2. 2 Institute of Plant Biology, University of Zürich and Zürich‐Basel Plant Science Center, Zollikerstrasse 107, Zürich, 8052, Switzerland
  3. 3 Institute for Molecular Plant Sciences, University of Edinburgh, Daniel Rutherford Building, Mayfield Road, Edinburgh, EH9 3JR, UK
  1. *Corresponding author. Tel: +41 44 632 5959; Fax: +41 44 632 1044; E‐mail: koehlerc{at}
  1. These authors contributed equally to this work

View Abstract


Polycomb group (PcG) proteins convey epigenetic inheritance of repressed transcriptional states. Although the mechanism of the action of PcG is not completely understood, methylation of histone H3 lysine 27 (H3K27) is important in establishing PcG‐mediated transcriptional repression. We show that the plant PcG target gene PHERES1 is regulated by histone trimethylation on H3K27 residues mediated by at least two different PcG complexes in plants, containing the SET domain proteins MEDEA or CURLY LEAF/SWINGER. Furthermore, we identify FUSCA3 as a potential PcG target gene and show that FUSCA3 is regulated by MEDEA and CURLY LEAF/SWINGER. We propose that different PcG complexes regulate a common set of target genes during the different stages of plant development.


The regular development of multicellular organisms depends on the stable propagation of established gene expression states during mitotic cell divisions, and in both animals and plants Trithorax (Trx) and Polycomb group (PcG) proteins carry out this task. Some Trx and PcG proteins have intrinsic histone methyltransferase activity, which is mediated by the evolutionarily conserved 130‐residue SET domain (for SU[VAR]3–9, Enhancer of Zeste (E(Z)), TRX), suggesting that the maintenance of cellular memory involves methylation of histones (Cao & Zhang, 2004). However, the enzymatic activity of the PcG SET domain protein E(Z) was detectable only if the protein was assembled with ESC, SUZ12 and p55 in the Polycomb repressive complex 2 (PRC2; Cao & Zhang, 2004). The PRC2 complex suppresses Homeobox (Hox) genes and other target genes by setting trimethylation marks on histone H3 lysine 27 (H3K27me3). The histone methyltransferase activity of E(Z) is essential for target gene silencing, as an E(Z)H703K mutant protein, which has reduced enzymatic activity, fails to mediate Hox gene silencing (Müller et al, 2002).

In contrast to animals, plant PcG proteins are often encoded by small gene families. In Arabidopsis, there are three E(Z) homologues (MEDEA (MEA), CURLY LEAF (CLF) and SWINGER (SWN); Guitton & Berger, 2005), three SUZ12 homologues (FERTILIZATION INDEPENDENT SEED2 (FIS2), VERNALIZATION2 and EMBRYONIC FLOWER2; Guitton & Berger, 2005) and five p55 homologues (MULTICOPY SUPPRESSOR OF IRA1 (MSI1–5); Hennig et al, 2005). Genetic evidence suggests that there are different PcG complexes in plants, regulating different developmental pathways. One PRC2‐like complex, which contains MEA, MSI1 and the ESC homologue FERTILIZATION INDEPENDENT ENDOSPERM (FIE), has been detected in flowers and seeds (Köhler et al, 2003a). Interaction studies in yeast suggest that FIS2 is also part of this complex (Chanvivattana et al, 2004). Mutations in the corresponding genes result in the development of seed‐like structures without fertilization. If fertilization occurs, seeds that inherit a mutated maternal allele have proliferation defects in the embryo and endosperm and abort (Guitton & Berger, 2005). The only known direct target gene of MEA is the type I MADS‐box gene PHERES1 (PHE1); MEA regulates allele‐specific expression of PHE1 by repressing the maternal PHE1 allele (Köhler et al, 2003b, 2005). By contrast, CLF regulates leaf and floral organ development, and AGAMOUS (AG) and APETALA3 (AP3) are probably direct target genes (Goodrich et al, 1997; D.S. and J.G., unpublished data).

So far, it is unclear how plant PcG complexes regulate their target genes and whether different PcG complexes act on distinct targets. Our results show that PcG‐mediated gene repression is established by an evolutionarily conserved mechanism and that different PcG complexes share a common subset of target genes during the different stages of plant development.


The PHERES1 locus contains H3K27 methylation marks

To test whether MEA binding at the PHE1 locus causes histone modifications, we analysed the MEA‐binding and histone methylation profiles at the PHE1 locus by performing chromatin immunoprecipitation (ChIP) experiments, using flowers harvested before pollination, open pollinated flowers, siliques at 2–3 days after pollination (DAP) and siliques at 4–6 DAP. We found preferential binding of MEA to PHE1 regions 1 and 2, with a peak of MEA binding detected in region 2 before fertilization and a second peak in regions 1 and 2 in siliques at 2–3 DAP (Fig 1A,B). We detected two peaks of MEA expression: one before fertilization and the other 2 days after fertilization (Fig 1C), which shows that MEA messenger RNA levels correlate with MEA binding to the PHE1 locus. As MEA binding occurred preferentially to region 2 of PHE1, we investigated the presence of different histone modifications in this region using closed flowers as the input material (peak 1 in the time course). We found strong enrichment of H3K27me3 as well as weak enrichment of H3K9me2. No enrichment was detected for H3K27me2 or the euchromatic mark H3K4me2 (Fig 1D).

Figure 1.

Dynamics of chromatin modifications at the PHE1 locus during development are correlated with the expression profile of MEA as well as with the binding of MEA to promoter regions of PHE1. (A) Schematic diagram of the PHE1 locus indicating the regions analysed by quantitative PCR after chromatin immunoprecipitation (ChIP). The numbers indicate the position relative to the translational start site of PHE1. (B) ChIP analysis of MEA binding to the PHE1 locus during reproductive development. Material used: closed flowers for 0 days after pollination (DAP), open flowers for 1 DAP and siliques for 2–6 DAP. For relative enrichment values, the ratios between PHE1 and UBQ11 enrichments were calculated. (C) Relative MEA messenger RNA levels in wild‐type flowers before fertilization (0 DAP) and siliques harvested at different DAP. (D) ChIP analysis of different histone modification marks at region 2 of PHE1 using closed flowers. (E) ChIP analysis of H3K27me3 marks at the PHE1 locus during reproductive development. Material used: closed flowers for 0 DAP, open flowers for 1 DAP and siliques for 2–6 DAP. (F) ChIP analysis of H3K27me3 at region 2 of PHE1 in wild type (WT) and mea/mea mutants using closed flowers. MEA, MEDEA; PHE1, PHERES1.

Despite anti‐H3K27me3 antibodies also recognizing H3K27me2, and anti‐H3K27me2 antibodies recognizing H3K27me3, both have a clear preference for their main substrates (Peters et al, 2002). Thus, the main methylation mark on the PHE1 locus is H3K27me3. We analysed whether MEA binding correlates with the presence of H3K27me3 at the PHE1 locus and found the strongest enrichment of H3K27me3 in closed flowers at region 2, overlapping with the peak of MEA binding. However, we also detected H3K27me3 at fragments 1, 3 and 4, which indicated that MEA also binds to these regions, but at a lower frequency. After fertilization, lower levels of H3K27me3 were detected, followed by a second peak at 2–3 DAP, also correlating with stronger MEA binding at this stage (Fig 1E). MEA expression, MEA binding to PHE1 and enrichment of H3K27me3 at PHE1 all had two peaks during reproductive development, prompting us to speculate that this reflects separate functions of the MEA PRC2 complex before and after fertilization.

The correlation of MEA binding and H3K27me3 at the PHE1 locus suggests that MEA is mediating PHE1 repression by setting H3K27me3. To test this hypothesis, we analysed H3K27me3 in closed flowers of homozygous mea mutants. In mea mutant flowers, H3K27me3 was reduced to half of the wild‐type levels (Fig 1F). This indicates that MEA is indeed mediating trimethylation of H3K27 at the PHE1 locus. However, the substantial levels of H3K27me3 remaining in mea mutants suggest that H3K27 in PHE1 chromatin is also methylated by different methyltransferases.

The SET domain of MEA is necessary for PHE1 repression

To test whether MEA‐dependent H3K27me3 on PHE1 is required for MEA function, we mutated the catalytic centre of the SET domain of MEA by introducing a histidine 621 to lysine (H621K) point mutation. The equivalent mutation of the conserved histidine in the catalytic centre of the MEA homologue E(Z) markedly impairs the histone methyltransferase activity of the E(Z) protein (Müller et al, 2002; Fig 2A). The mutated complementary DNA, driven by the MEA promoter, was introduced into mea mutant plants. No transgenic line out of 27 primary transformants showed complementation of the mea seed abortion phenotype (seed abortion ratios between 45% (χ2=2.1<χ20.05[1]=3.84) and 55% (χ2=1.6<χ20.05[1]=3.84). By contrast, 5 out of 18 lines containing the wild‐type MEA sequence (MEAwt) showed complementation of the seed abortion phenotype (seed abortion ratios between 31% (χ2=2.5<χ20.05[1]=3.84) and 29% (χ2=2.3<χ20.05[1]=3.84; Fig 2B). The H621K mutation does not affect protein stability, as homozygous mea mutant plants expressing the MEAH621K version contain even higher MEA levels than wild‐type plants, confirming that MEA expression is regulated by a negative feedback loop (Fig 2C; Gehring et al, 2006; Jullien et al, 2006). We then addressed whether the MEA H621K mutation also impairs PHE1 repression. We crossed three transgenic lines expressing MEAH621K and three lines expressing MEAwt with a PHE1∷β‐GLUCURONIDASE (GUS) reporter line and assayed GUS activity in seeds. In mea mutant seeds of all three lines expressing MEAwt, PHEGUS expression was restricted to the chalazal region of the endosperm, as in the wild type. By contrast, in mea mutant seeds expressing MEAH621K, PHEGUS expression was detectable in the embryo and endosperm at the heart stage (Fig 2D; Supplementary Table SI online). These data indicate that the histone methyltransferase activity of MEA is necessary for normal MEA function, including PHE1 repression.

Figure 2.

The SET domain of MEA is necessary for PHE1 repression. (A) Alignment of conserved parts of the SET domains of MEA, CLF, SWN, E(Z) and KYP. Black boxes indicate identical amino acids and grey boxes indicate conservative changes. The conserved histidine residues are indicated by a star. Numbers indicate the positions of amino acids. (B) Open siliques of wild‐type (WT), mea/MEA and mea/MEA plants containing WT MEA (MEAwt) or the MEAH621K construct. (C) Western blot analysis of MEA expression in WT, mea/mea and mea/mea plants containing the MEAH621K construct. The Coomassie stain of a replicate gel is shown as a loading control. (D) PHEGUS expression in mea seeds containing the MEAwt or the MEAH621K construct. Scale bars, 100 μm. CLF, CURLY LEAF; E(Z), ENHANCER OF ZESTE; GUS, β‐GLUCURONIDASE; KYP, KRYPTONITE; MEA, MEDEA; PHE1, PHERES1; SWN, SWINGER.

CLF and SWN redundantly regulate PHE1 repression

Although MEA is in part responsible for the H3K27me3 at PHE1, the significant residual H3K27me3 enrichment in flowers of homozygous mea mutants shows that MEA is not exclusively responsible for PHE1 methylation. Therefore, we tested whether PHE1 is also methylated in sporophytic tissues (leaves), in which MEA is not expressed (Kiyosue et al, 1999). We found that H3K27me3 was also strongly enriched in PHE1 chromatin from leaves (Fig 3A). CLF and SWN are two close homologues of MEA; therefore, we tested whether sporophytic H3K27me3 on PHE1 requires one or both proteins. However, H3K27me3 at the PHE1 locus was unchanged in clf leaves (Fig 3A) and in mea/MEA;swn/swn leaves (data not shown). By contrast, clf swn double mutants had much less H3K27me3 on PHE1 than wild‐type leaves (Fig 3A). Thus, CLF and SWN are redundantly needed for trimethylation of H3K27 at the PHE1 locus. As PHE1 is not expressed in wild‐type leaves, we asked whether PHE1 repression involves H3K27me3. We analysed PHE1 expression in the leaves of clf and swn single mutants and in clf swn double mutants. PHE1 was not expressed in the leaves of wild type, clf and swn single mutants (Fig 3B). By contrast, PHE1 expression was upregulated in clf swn double mutants (Fig 3B). However, PHE1 expression levels were much lower in clf swn than in wild‐type siliques, which suggests that, in addition to the loss of H3K27me3, some activating signals are needed to completely activate PHE1 expression.

Figure 3.

Regulation of PHE1 repression in sporophytic tissues by CLF and SWN. (A) Chromatin immunoprecipitation (ChIP) analysis of H3K27me3 at region 2 of PHE1 in wild type (WT), clf and clf swn mutants. Light grey bars and dark grey bars indicate samples of closed flowers and leaf samples, respectively. (B) Relative messenger RNA levels of PHE1 in sporophytic tissues of WT, clf, swn and clf swn mutants. PHE1 mRNA levels in WT seeds at 3 days after pollination indicate maximum PHE1 expression of 1.0. Light grey bars and dark grey bars indicate samples of seeds and leaf samples, respectively. CLF, CURLY LEAF; PHE1, PHERES1; SWN, SWINGER.

MEA, CLF and SWN regulate common target genes

Our data establish that CLF and SWN repress the MEA target gene PHE1 by histone methylation in sporophytic tissues. Next, we asked whether other plant PcG target genes are repressed by different PcG complexes as well. AG and AP3 are probably direct target genes of CLF and SWN (Goodrich et al, 1997; D.S. and J.G., unpublished data); therefore, we analysed whether the expression of these genes is also upregulated in mea mutant seeds. Upregulation of AG and AP3 was observed in clf swn, but not in the seeds of mea/MEA mutant plants (data not shown). However, both genes were not repressed but highly expressed in wild‐type seeds at 3 DAP; thus, the mea mutation probably has no impact on AG or AP3 expression at this developmental stage. Therefore, we tested the expression of other potential CLF and SWN targets. clf swn double mutants form somatic embryos, suggesting a misexpression of embryonic regulators such as LEAFY COTYLEDON1 (LEC1), LEC2 and FUSCA3 (FUS3; Lotan et al, 1998; Stone et al, 2001; Chanvivattana et al, 2004). Indeed, FUS3 was highly upregulated, and LEC1 and LEC2 were mildly upregulated in clf swn mutants (Fig 4A). Similarly, we also observed strong increased expression of a FUS3GUS reporter gene in clf swn mutants (Fig 4B). Thus, the embryonic regulators LEC1, LEC2 and FUS3 are potential target genes of CLF and SWN. We then tested whether LEC1, LEC2 and FUS3 are also upregulated in mea mutant seeds. FUS3 expression was increased in siliques of mea/MEA mutant plants at 2–3 DAP (Fig 4C), whereas LEC1 and LEC2 expression was at very low levels in both mutant and wild‐type seeds (data not shown). In a parallel approach, we analysed the expression of FUS3GUS and LEC2GUS reporter constructs in mea mutant and wild‐type seeds. In wild‐type seeds, expression of FUS3GUS was restricted to the embryo until the heart stage, whereas in mea mutant seeds, GUS expression was detectable in the endosperm as well (mea/MEA; FUS3GUS/FUS3GUS; 114:156, strongly staining seeds:weakly staining seeds; 1:1; χ2=6.5<χ20.05[1]=6.635, Fig 4D). By contrast, LEC2GUS expression was similar in wild‐type and mea mutant seeds and remained restricted to the embryo (data not shown). To test whether FUS3 is a direct target gene of MEA, we performed ChIP experiments with anti‐MEA and anti‐H3K27me3 antibodies using flowers harvested before pollination. We detected a clear enrichment for FUS3 with both antibodies, which demonstrates that FUS3 is a direct target gene of MEA (Fig 4E,F). These data establish that a subset of plant PcG target genes is regulated by different PcG complexes during the different stages of development.

Figure 4.

FUS3 is regulated by CLF, SWN and MEA. (A) Relative messenger RNA levels of FUS3, LEC1 and LEC2 in clf swn double mutants in comparison with wild‐type (WT) leaves. (B) FUS3GUS expression in WT and clf swn mutant seedlings. Scale bars, 1 mm. (C) Relative mRNA levels of FUS3 in siliques of WT and mea/MEA mutants at 3 days after pollination (light grey bars). Expression of FUS3 in clf swn and WT leaves is shown as a comparison (dark grey bars). (D) FUS3GUS expression in WT and mea mutant seeds. Scale bars, 100 μm. (E) Chromatin immunoprecipitation (ChIP) analysis of MEA binding to the FUS3 locus in closed flowers. (F) ChIP analysis of H3K27me3 marks at the FUS3 locus in closed flowers. CLF, CURLY LEAF; FUS3, FUSCA3; GUS, β‐GLUCURONIDASE; MEA, MEDEA; SWN, SWINGER.


PHE1 chromatin is strongly enriched in H3K27me3, and also contains some H3K9me2. Both marks are involved in transcriptional repression in many eukaryotes (Martin & Zhang, 2005), and gene silencing by PRC2 in animals involves H3K27me3 (Cao & Zhang, 2004). In Arabidopsis, H3K27me3 localizes primarily to euchromatin, and H3K27me2 localizes to euchromatin and heterochromatin, whereas H3K9me2 localizes almost exclusively to heterochromatin (Lindroth et al, 2004). The strong enrichment of the euchromatic mark H3K27me3 on the PHE1 locus suggests that silencing of PHE1 occurs in euchromatin and differs from the heterochromatic silencing of transposons. In plants, methylation of H3K27 has been also reported for FLOWERING LOCUS C (FLC; Bastow et al, 2004; Sung & Amasino, 2004), which is a possible target of the VRN2 PcG protein, as well as of MEA itself (Gehring et al, 2006; Jullien et al, 2006); however, it is not known whether H3K27me2 or H3K27me3 is the dominating repressive mark. Our results establish that, in the case of PHE1, PRC2 repression involves mainly H3K27me3 and not H3K27me2. In addition to H3K27me3, the PHE1 locus contains H3K9me2 in closed flowers but not at later stages (data not shown). Similarly, the FLC locus contains H3K9me2 (Bastow et al, 2004), and it has been proposed that methylation of both H3K9 and H3K27 is required for recruiting the CMT3 methyltransferase to heterochromatic loci (Lindroth et al, 2004).

PHE1 is a target of the plant PcG protein MEA, and H3K27me3 on PHE1 depends in part on MEA. H3K27me3 seems to be necessary for PHE1 repression, as an H621K mutation in the catalytic centre of the SET domain of MEA causes PHE1 de‐repression. The equivalent mutation in E(Z) impairs histone methyltransferase activity and Hox gene silencing (Müller et al, 2002), suggesting that MEA and E(Z) operate through similar mechanisms. MEA binding was not uniform along PHE1 chromatin but was strongest around the translation start. This is consistent with the idea that PcG silencing blocks transcription initiation in Drosophila (Dellino et al, 2004). Thus, repression of PcG target genes by H3K27 trimethylation seems to be conserved in animals and plants.

We show that PHE1 is redundantly methylated by CLF and SWN in sporophytic tissues, and low levels of H3K27me3 methylation in clf swn double mutants correlate with upregulated PHE1 expression. Thus, PHE1 repression depends on MEA‐containing, as well as CLF‐ or SWN‐containing, PRC2 complexes. In addition, we identified FUS3 as a common target gene of MEA as well as CLF and SWN, indicating that at least a subset of plant PcG target genes are under repressive control of MEA‐containing, as well as CLF‐ and SWN–containing, PRC2 complexes. FUS3 controls the transition from embryonic to vegetative development through terpenoid hormone signalling (Gazzarrini et al, 2004), and the functional connection between PRC2‐like complexes and FUS3 will be the subject of future investigations.

We conclude that the PcG genes MEA, CLF and SWN share at least a subset of common target genes during the different stages of plant development. MEA is required for PcG target repression during gametophyte and early seed development, whereas CLF and SWN take over this function during later sporophytic development.


Plant material and growth conditions. The mea mutant used in this study was the mea‐1 allele described by Grossniklaus et al (1998). The clf and swn mutants are T‐DNA insertion mutants SALK_021003 and SALK_050195, respectively. The PHE1GUS line has been described previously by Köhler et al (2003b). The LEC2GUS and FUS3GUS lines were kindly provided by François Parcy. Plants were grown in a greenhouse at 70% humidity and daily cycles of 16 h light at 21°C and 8 h darkness at 18°C.

Plasmid constructs. The MEA complementation construct was generated in pCAMBIA 3300 by fusing 3.8 kb upstream sequence of the MEA translational start to the MEA cDNA. The MEAH621K mutation was created using the QuikChange Site Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA) according to the manufacturer's instructions.

Quantitative reverse transcriptase–PCR, GUS expression and protein analysis. See the supplementary information online for details.

Chromatin immunoprecipitation. We followed the ChIP protocol of Johnson et al (2002); the details are given in the supplementary information online.

Supplementary information is available at EMBO reports online (

Supplementary Information

Supplementary Information [embor7400760-sup-0001.doc]


We thank F. Parcy for LEC2GUS and FUS3GUS lines, T. Jenuwein for anti‐H3K27me2 antibodies, V. Gagliardini for sequencing and L. Hennig for helpful comments on the manuscript. This research was supported by the University of Zürich, grants of the Swiss National Science Foundation to C.K. (3100A0‐104343 and PP00A‐106684/1) and U.G. (3100‐064061), and a grant from the Zürich‐Basel Plant Science Center to C.K. C.K. is supported by the EMBO Young Investigator Award.


View Abstract