An E3 ligase complex regulates SET-domain polycomb group protein activity in Arabidopsis thaliana

Abstract

Transcriptional repression via methylation of histone H3 lysine 27 (H3K27) by the polycomb repressive complex 2 (PRC2) is conserved in higher eukaryotes. The Arabidopsis PRC2 controls homeotic gene expression, flowering time, and gene imprinting. Although downstream target genes and the regulatory mechanism of PRC2 are well understood, much less is known about the significance of posttranslational regulation of PRC2 protein activity. Here, we show the posttranslational regulation of CURLY LEAF (CLF) SET-domain polycomb group (PcG) protein by the F-box protein, UPWARD CURLY LEAF1 (UCL1). Overexpression of UCL1 generates mutant phenotypes similar to those observed in plants with a loss-of-function mutation in the CLF gene. Leaf curling and early flowering phenotypes of UCL1 overexpression mutants, like clf mutants, are rescued by mutations in the AGAMOUS and FLOWERING LOCUS T genes, which is consistent with UCL1 and CLF functioning in the same genetic pathway. Overexpression of UCL1 reduces the level of CLF protein and alters expression and H3K27 methylation of CLF-target genes in transgenic plants, suggesting that UCL1 negatively regulates CLF. Interaction of UCL1 with CLF was detected in plant nuclei and in the yeast two-hybrid system. The UCL1 F-box binds in vivo to components of the E3 ligase complex, which ubiquitylate proteins that are subsequently degraded via the ubiquitin-26S proteasome pathway. Taken together, these results demonstrate the posttranslational regulation of the CLF SET-domain PcG activity by the UCL1 F-box protein in the E3 ligase complex.

Polycomb group (PcG) proteins epigenetically silence gene expression and play an important role in controlling eukaryote cell proliferation, stem cell identity, cancer, genomic imprinting, and X chromosome inactivation (1, 2). The term “polycomb” initially referred to mutations in Drosophila that caused improper specification of body segment identity. Three different PcG complexes, polycomb repressive complex 1 (PRC1), PRC2, and PcG-like PRC2, are present in animals and work together in a stepwise manner to repress expression of their target genes (3).

The PRC2 complex is highly conserved in Drosophila, mammals, and flowering plants (1, 4). The Drosophila PRC2 complex is composed of three PcG proteins, Extra Sex Combs (Esc), Suppressor of Zeste 12 [Su(z)12], and Enhancer of Zeste [E(z)], as well as additional core proteins, such as p55 (2). Esc and p55 are WD-40 proteins, Su(z)12 is a C2H2 zinc finger, and E(z) is a SET-domain protein that methylates histone H3 lysine 27 (H3K27), which silences target gene expression. In Arabidopsis, there is an Esc homolog, FERTILIZATION INDEPENDENT ENDOSPERM (FIE); p55 homologs, ARABIDOPSIS MULTICOPY SUPRESSPR OF IRA1 to 5 (MSI1 to MSI5); zinc finger-containing Su(Z) homologs, EMBRYONIC FLOWER 2 (EMF2), VERNALIZATION2 (VRN2) and FERTILIZATION INDEPENDENT SEED2 (FIS2); and SET-domain E(z) homologs, CURLY LEAF (CLF), SWINGER (SWN), and MEDEA (MEA) (4).

There are three distinct Arabidopsis PRC2 complexes, which are named after their respective zinc finger-containing Su(Z) homolog constituents: EMF2, VRN2, and FIS2. The EMF2 PcG complex, composed of FIE, MSI1, CLF/SWN, and EMF2, prevents early flowering and regulates vegetative growth (4). During the vegetative state of development, the EMF2 PcG complex interacts with CLF to repress the expression of a key floral promoting gene, FLOWERING LOCUS T (FT), as well as the floral organ identity genes AGAMOUS (AG) and AGAMOUS-LIKE 19 (AGL19) (5–7). Mutations in the CLF SET-domain PcG gene cause pleiotropic phenotypes, including hyponastic (curly) leaves, homeotic transformation of floral organs, and early flowering (8). The FIS2 complex is composed of FIE, MSI1, MEA, and FIS2, which are required for proper seed development. Mutations in the MEA SET-domain PcG gene cause seed abortion, the formation of nonviable embryos, and disruption of gene imprinting in the endosperm (4, 9).

The ubiquitin-26S proteasome system is conserved in eukaryotes (10). Three ubiquitin-ligase activities (E1, E2, and E3) sequentially transfer ubiquitin to target proteins that are degraded in the 26S proteasome complex. The system degrades abnormal proteins that are produced by biosynthetic errors or have folded into nonnative conformations. It also controls the level of specific regulatory proteins that control cell cycle progression, signal transduction, gene transcription, programmed cell death, and development in animals as well as hormone synthesis and signaling, plant-pathogen interactions, self-incompatibility, circadian rhythms, morphogenesis, and histone modifications in plants (10–12). A well-characterized E3 ligase, SCF, is composed of multiple subunits: a CULLIN1 (CUL1) scaffold protein; a catalytic RING domain protein, RING-BOX1 (RBX1); an adaptor protein called S-PHASE KINASE-ASSOCIATED PROTEIN1 (SKP1) in animals and ARABIDOPSIS SKP1 HOMOLOG1/2 (ASK1/2) in Arabidopsis; and an F-box protein, which confers substrate specificity in the transfer of ubiquitin from the E2 ubiquitin-conjugating enzyme to the target protein.

Histone methylation by PcG protein complexes maintains gene silencing (2). Alterations in PcG complex composition and association to chromatin may be beneficial as organisms undergo developmental transitions and respond to their environment, however (4, 13). There is increasing evidence for regulation of PRC2 activity by posttranslational modifications of PcG proteins by E3 ubiquitin ligases (14, 15). Here, we show that CLF SET-domain activity is regulated at the posttranslational level by UPWARD CURLY LEAF1 (UCL1), the F-box component of an E3 ligase that binds CLF and targets it for degradation via the ubiquitin-26S proteasome pathway.

Results

Plants That Overexpress UCL1, an F-Box Protein Gene, Have Curly Leaves.

We mutagenized Arabidopsis plants using a T-DNA activation vector that activates flanking gene expression (16) and identified a mutation, upward curly leaf1-dominant (ucl1-D), whose semidominant curly leaf phenotype (Fig. 1A and Fig. S1A) was also observed in plants with a recessive loss-of-function mutation in the CLF gene (8). Plasmid rescue methods (16) revealed that the T-DNA had inserted in the intergenic region between At1g65740 and At1g65750 (Fig. 1B). Real-time quantitative RT-PCR (qRT-PCR) showed that expression of At1g65740, At1g65760, and At1g65770 was elevated in ucl1-D homozygous plants (Fig. S1B). Among these three genes, we observed the curly leaf phenotype only when overexpressing At1g65740 in transgenic plants using the Cauliflower Mosaic Virus (CaMV) promoter (Figs. S1 C and D and S2 B, D, F, and G). We designated At1g65740, which encodes an F-box protein, as UCL1.

Fig. 1.

Insertion of an activation T-DNA next to an F-box gene. (A) WT, heterozygous, and homozygous ucl1-D mutant plants. (B) Genomic region flanking the T-DNA insertion site in ucl1-D plants. bar, Basta-resistance gene; 4 Enhancers, CaMV enhancer tetrad; LB, T-DNA left border; RB, T-DNA right border.

UCL1 Overexpression and clf Loss of Function Produce Similar Phenotypes.

We compared clf and CaMV::UCL1 phenotypes throughout the Arabidopsis life cycle. Homozygous clf mutant plants display curly leaves, floral homeotic changes attributable to the ectopic expression of AG, reduced internode and inflorescence height (8), and early flowering under reduced day-length conditions attributable to the derepression of FT (7). Likewise, homozygous CaMV::UCL1 plants display curly rosette leaves (compare Fig. S2 C and D); reduced internode and inflorescence height (Fig. S3A); defects in flower morphology (compare Figs. S3 B and C and D–I); and homeotic changes of floral organs, such as petals with stamen-like features (Fig. S3 F and I) and ovule-bearing sepals with stigmatic papilla (Fig. S3 G–I). CaMV::UCL1 plants flowered early, especially when grown under short-day conditions (Fig. S3J). Thus, clf and CaMV::UCL1 plants display similar pleiotropic phenotypes.

We generated CaMV::UCL1; ft-1/ft-1 plants and CaMV::UCL1; ag-1/ag-1 plants and found that the early flowering and curly leaf phenotypes in CaMV::UCL1 plants were suppressed by the ft-1 and ag-1 mutations, respectively (Fig. S3 K and L). This suggests that CaMV::UCL1, like clf mutants, causes early flowering and leaf curling by derepression of FT and AG, respectively. These results suggest that CaMV::UCL1 and clf cause pleiotropic phenotypes by the same genetic pathways.

UCL1 Overexpression Activates Expression of CLF-Target Genes and Alters Their H3K27 Methylation Profiles.

We analyzed CLF-target gene expression by isolating RNA from CaMV::UCL1 and WT seedlings, which were independently hybridized to Affymetrix microarrays (Table S1). A list of genes showing significant changes in expression includes AG and AGL genes as well as flowering time regulators FT and FLOWERING LOCUS C (FLC). We verified by real-time qRT-PCR analysis that well-known CLF-target genes AG, APETALATA3 (AP3), AGL17, FT, FLC, KNOTTED-LIKE FROM ARABIDOPSIS THALIANA 2 (KNAT2), MEA, and SEPALLATA3 (SEP3) were overexpressed in rosette leaves of ucl1-D and CaMV::UCL1 plants (Fig. 2A). By contrast, expression of non–CLF-target genes SEUSS (SEU), LEUNIG (LEU), APETALATA1 (AP1), and CLF was not significantly changed by UCL1 overexpression (Fig. 2A). Likewise, overexpression of the At1g65770 gene, which is adjacent to UCL1, did not activate CLF-target genes (Fig. S4A). These results suggest that CLF-target genes are specifically up-regulated in plants that overexpress UCL1.

Fig. 2.

Overexpression of UCL1 activates CLF-target genes and alters histone methylation patterns. (A) Real-time qRT-PCR analysis in WT, ucl1-D, and CaMV::UCL1 plants. Values are plotted relative to the TUBULIN-β-CHAIN2 (TUB2) reference gene and represent the average of duplicate measurements ± SD. AG-I::GUS activity in CaMV::UCL1 (B, arrow), in ucl1-D (C, arrow), and in WT (D) rosette leaf. (E) Levels of H3K27me3 at AG, FLC, and FT chromatin in ucl1-D and CaMV::UCL1 seedlings relative to WT. DNA concentration after precipitation with H3K27me3 antibody was quantified by real-time qRT-PCR and normalized to an internal reference (TUB2). Values in E are plotted relative to WT for each DNA, which was set at 1.0, and represent the average of triplicate measurements ± SD. TUB, tubulin β-chain 2.

We introduced the pAG-I::GUS gene, consisting of the AG promoter with the AG first and second introns fused to the β-glucuronidase (GUS) reporter gene (17), into the CaMV::UCL1 or ucl1-D genetic background. In pAG-I::GUS; CaMV::UCL1 plants, we detected ectopic GUS expression in young rosette leaves (Fig. 2B and Fig. S5A), floral stems, and petals (Fig. S5D) compared with the control pAG-I::GUS plants (Fig. 2D and Fig. S5 C and F). In pAG-I::GUS; ucl1-D plants, we detected a slight increase in ectopic GUS expression in young leaves (Fig. 2C and Fig. S5B) and in the floral stem (Fig. S5E, arrow). These results are consistent with the real-time qRT-PCR data shown in Fig. 2A. By contrast, we could not detect altered GUS expression in control pAG-I::GUS; CaMV::At1g65770 plants (Fig. S4 B–D). These results show that UCL1 overexpression, similar to loss-of-function clf mutations (18), up-regulates AG transcription.

The H3K27me3 silencing mark is reduced at CLF-target genes in clf mutant plants (7, 19). We found that H3K27me3 levels at the AG intron D and E regions (Fig. S5G) were significantly reduced in UCL1 overexpression plants compared with WT (Fig. 2E and Fig. S5H). Reduced H3K27me3 levels were also detected at specific intron sites (Fig. S5G) in the FLC and FT loci (Fig. 2E). Thus, clf mutations (7, 19) and UCL1 overexpression reduce H3K27me3 at CLF-target genes.

UCL1 Binds CLF in the Yeast Two-Hybrid System and in Plant Nuclei.

We used the yeast two-hybrid system to determine if the UCL1 F-box protein binds to CLF. We failed to detect the interaction when we used full-length UCL1 or CLF (Fig. 3 A and B and Discussion). CLF amino acids 1–300 interacted with UCL1 amino acids 201–371 and activated reporter β-galactosidase (lacZ) gene expression (Fig. 3 A and B). We also checked the binding of UCL1 to MEA, a SET-domain PcG protein that is closely related to CLF. MEA amino acids 1–280, corresponding to CLF amino acids 1–300, did not interact with UCL1 amino acids 201–371 (Fig. 3 A and C). Thus, domains within the CLF and UCL1 proteins specifically interact in yeast.

Fig. 3.

UCL1 specifically binds to CLF but not to MEA in the yeast two-hybrid assay. (A) Schematic diagram of UCL1, CLF, and MEA. The amino acid numbers of proteins used in the yeast two-hybrid analysis are shown. (B) Interaction between UCL1 and CLF in the yeast two-hybrid assay. (C) Interaction between UCL1 and MEA in the yeast two-hybrid assay. (D) Interaction test between UCL1Δ and CLF, as well as between UCL1Δ and MEA, in the yeast two-hybrid assay. F, clones encoding full-length proteins; lacZ, β-galactosidase enzyme activity; SD-LW, synthetic dextrose minimal medium without leucine and tryptophan; SD-LWAH, synthetic dextrose minimal medium without leucine, tryptophan, adenine, and histidine.

To determine UCL1 localization, we generated CaMV::UCL1:GFP transgenic plants. GFP activity was detected in root cell nuclei (Fig. 4A), indicating that UCL1 is a nuclear protein like CLF (20). We investigated the in vivo interaction between full-length UCL1 and CLF using bimolecular fluorescence complementation (BiFC) (21). We detected a strong reconstituted YFP signal in the nucleus when CaMV::nEYFP:CLF and CaMV::cEYFP:UCL1 were cotransformed into Arabidopsis protoplasts (Fig. 4B, Lower). These results demonstrate that UCL1 interacts with CLF in plant nuclei.

Fig. 4.

UCL1 forms an SCF complex, binds CLF, and reduces CLF protein level. (A) Subcellular localization of UCL1:GFP fusion protein in root cells. Subcellular localization of UCL1:GFP fusion protein in CaMV::UCL1:GFP transgenic plants in the root tip of light-grown seedlings is shown by confocal laser scanning microscopy using a GFP filter set. A merged image using DIC, PI, and GFP filter sets is shown. (Inset) Close-up view using GFP and PI filter sets. (B) BiFC assay for protein-protein interaction of UCL1and CLF in Arabidopsis cells. (Upper) No YFP signal in the nucleus when CaMV::nEYFP:CLF and CaMV::cEYFP (without UCL1) were cotransfected as a negative control. (Lower) Reconstituted YFP signal in the nucleus when CaMV::nEYFP:CLF and CaMV::cEYFP:UCL1 were cotransfected. The known transcription factor FES, fused with CFP, was used as a nucleus marker after cotransformation. DIC, differential interference contrast microscopy. (C) Yeast two-hybrid assay for interaction of UCL1 with ASK1 and ASK2. UCL1Δ has a deletion (amino acids 1–43) and does not contain most of the N-terminal F-box motif. −LW, synthetic dextrose minimal (SD) medium without leucine and tryptophan; −LWU, synthetic dextrose minimal (SD) medium without leucine, tryptophan, and uracil; lacZ, β-galactosidase enzyme activity. (D) Coimmunoprecipitation of UCL1, CUL1, and ASK1. Proteins from protoplasts transformed with constructs expressing GFP, UCL1-HA, CUL1-HA, and ASK-GFP were precipitated (Output) or not precipitated (Input) with anti-GFP antibody, run on SDS/PAGE gels, blotted, and reacted with Anti-HA or Anti-GFP antibodies. *Neddylated CUL1. (E) Protein blot with anti-GFP antibody in CaMV::GFP:CLF 15-d-old seedlings or from F1 seedlings generated by crossing CaMV::GFP:CLF with the indicated lines. For a loading control, ribulose-bisphosphate carboxylase (RbcS) protein was stained with Ponceau. GFP fluorescence in the root elongation (F and H) and differentiation zone (G and I) from F1 plants generated by crossing CaMV::GFP:CLF with WT (F and G) or with CaMV::UCL1 (H and I). (Scale bars: A, 20 μm; B, 10 μm; G–I, 20 μm.)

UCL1 Binds Proteins in the E3 Ligase Complex.

We checked the interaction of F-box UCL1 with other polypeptides in an SCF E3 ubiquitin ligase complex with the yeast two-hybrid system. UCL1 interacted with ASK1, an Arabidopsis SKP1 homolog, and also with ASK2, albeit with weaker binding activity (Fig. 4C). UCL1Δ with most of its F-box deleted (amino acids 44–371) (Fig. 3A) failed to interact with ASK1 (Fig. 4C), indicating that the F-box motif in UCL1 is essential for ASK1 binding activity. As expected, UCL1 did not directly interact with CUL1 (Fig. 4C), which needs ASK1 to mediate UCL1 F-box protein binding to the CUL1 scaffold protein to form an SCF complex (22, 23).

We performed coimmunoprecipitation experiments to determine if UCL1 is part of the SCF complex in vivo. Using Arabidopsis protoplasts, UCL1 and CUL1 with hemagglutinin (UCL1-HA and CUL1-HA) were coimmunoprecipitated when GFP-tagged ASK1 (ASK1-GFP) was coexpressed and reacted with anti-GFP antibodies (Fig. 4D). These results show that UCL1 is a component of the SCF complex in plant nuclei.

UCL1 Overexpression Reduces CLF Protein Level.

F-box proteins bind target proteins that are ubiquitylated and subsequently degraded in the proteasome (10, 24). In protein blot experiments with GFP antibody, we detected GFP:CLF fusion protein in control homozygous CaMV::GFP:CLF seedlings as well as in F1 seedlings generated by crossing CaMV::GFP:CLF with WT plants (Fig. 4E). By contrast, the level of GFP:CLF fusion protein was significantly reduced in F1 seedlings generated by crossing CaMV::GFP:CLF with CaMV::UCL1 plants (Fig. 4E). To confirm that GFP:CLF RNA levels are not altered in F1 seedlings, we measured RNA expression in these plants by qRT-PCR. CLF and GFP RNA is expressed at a similar level in F1 seedlings generated by crossing CaMV::GFP:CLF with WT plants and with CaMV::UCL1 plants (Fig. S6 A and B).

We checked UCL1 and AG RNA expression as a positive control, as well as SWN and MSI1 expression as a negative control (Fig. S6 C–F).

We also measured in vivo GFP fluorescence in root tissues. Control F1 plants generated by crossing CaMV::GFP:CLF with WT displayed strong GFP fluorescence in the root elongation zone (Fig. 4F) and root tip (Fig. 4G and Fig. S4H). By contrast, we detected a significant reduction of GFP fluorescence in the root elongation zone (Fig. 4H) and root tip (Fig. 4I and Fig. S4I) in F1 plants generated by crossing CaMV::GFP:CLF with CaMV::UCL1. In control experiments, ectopic expression of CaMV::At1g65770 did not diminish GFP fluorescence (Fig. S4J). Taken together, these results indicate that overexpression of UCL1 specifically decreases the level of GFP:CLF protein in planta.

UCL1 Function During Endosperm and Stamen Development.

qRT-PCR experiments revealed significant UCL1 expression in flowers, particularly in siliques with developing seeds and in male reproductive structures, stamens (Fig. S7A). We measured UCL1 promoter activity by analyzing transgenic plants with UCL1::GUS (transcriptional fusion) or UCL1::UCL:GUS (translational fusion) genes. Cytoplasmic GUS activity in UCL1::GUS plants was detected in young stamens (Fig. S7 B and C) and in early endosperm development (Fig. S7 D–F). In UCL1::UCL1:GUS plants, GUS activity was detected soon after fertilization in the primary endosperm nucleus (Fig. 5A) and in proliferating endosperm nuclei (Fig. 5 B–D). We did not detect GUS activity in the zygote or embryo (Fig. 5 A–D).

Fig. 5.

Ectopic expression of CLF in the endosperm causes mea-like phenotypes. GUS activity in a mature UCL1::UCL1:GUS ovule right after fertilization (A) and in seeds 4 (B), 18 (C), and 36 (D) h after self-pollination. Em, embryo; En, endosperm nucleus; Pen, primary endosperm nucleus; Z, zygote. (E) Real-time qRT-PCR analysis of CLF RNA level in WT and MEA::CLF transgenic plants at the indicated days after self-pollination (DAP). Values are plotted relative to expression of CLF in WT at 2 DAP, which was set at 1.0, and represent the average of duplicate measurements ± SD. (F) MEA::CLF transgenic plants containing abnormal aborted seeds (arrowhead) and small white unfertilized ovules. AS, aborted seed; Ov, ovule. (G) WT open silique. WT seed (H) or MEA::CLF seed (I). Unfertilized WT ovule (J) or MEA::CLF ovule (K) at 4 d after removal of anthers is shown. Cc, central cell nucleus; Ec, egg cell nucleus. Arrowheads point to nuclei of proliferating endosperm cells. (Scale bars: A–D and H–K, 50 μm;F and G, 500 μm.)

To understand the function of UCL1 in WT plant development, we examined the phenotype of ucl1 loss-of-function mutations caused by insertion of a T-DNA in the UCL1 gene. No developmental abnormalities were detected in homozygous ucl1-1 seed (Fig. S2E). It is possible that the extensive genetic redundancy within the F-box gene family (12) might have masked the effects of the ucl1-1 mutation (Discussion).

To gain clues about UCL1 function, we examined the effect of overexpressing its target, CLF, in a region of the plant where UCL1 is expressed, the endosperm. We used the endosperm-specific MEA promoter to express CLF and found a high level of CLF expression in seeds of transgenic MEA::CLF plants compared with WT (Fig. 5E). Interestingly, we detected phenotypes associated with loss-of-function mea mutations (25, 26), including aborted seeds (Fig. 5F), excessive endosperm cell proliferation (Fig. 5I), proliferation of unfertilized central cells, and seed-like structure formation in the absence of fertilization (Fig. 5K), which were not detected in WT seeds (Fig. 5 G, H, and J). We also detected ectopic expression of the AGL62 transcription factor that controls endosperm cellularization, which is also observed in mea mutant seed (27) (Fig. S7G). Control experiments showed that the MEA::CLF transgene did not affect MEA gene expression in the seed (Fig. S7H). These data show that ectopic expression of CLF in the endosperm is detrimental to seed development. One possibility is that overexpression of CLF in the endosperm suppresses the activity of the closely related MEA PcG protein.

Discussion

Posttranslational Regulation of PcG Proteins.

In Arabidopsis, the ubiquitin-26S proteasome system's role as a regulatory system is comparable in breadth and depth to transcriptional gene regulation (12). The Arabidopsis and rice F-box gene families each have ∼700 members (28). By comparison, there are 68 and 74 F-box genes in the human and mouse genomes, respectively (29). Some Arabidopsis F-box proteins have clear animal homologs, but most, like UCL1, are plant-specific.

Cullin serves as the backbone of the SCF complex and interacts with RBX1 via its C terminus and with the SKP1 (ASK1) adapter protein via its N terminus (23). SKP1 (ASK1) protein also binds to the F-box motif of F-box proteins to form the complete complex (24). In F-box proteins, the 60-aa F-box motif is at the N terminus and the domain that binds to the target protein is at the C terminus (30, 31). Likewise, the well-conserved N-terminal F-box motif of UCL1 interacts with ASK1 protein (Fig. 4), and its C-terminal domain interacts with its target protein, CLF (Fig. 3).

Full-length UCL1 or CLF protein failed to interact in the yeast two-hybrid system (Fig. 3). Similar results have been reported for the binding of full-length F-box proteins and their targets in yeast, however. For example, the Arabidopsis full-length F-box protein, UNUSUAL FLORAL ORGANS (UFO), and its full-length target, the LEAFY (LFY) transcription factor, bind at a low level in the yeast two-hybrid system (32). Deletion of the F-box domain resulted in more than an eightfold enhancement of binding, however, and it was concluded that the protein-protein interaction domain in the UFO F-box became accessible when the F-box domain was deleted. It has also been suggested that the unoccupied F-box motif antagonizes the target-binding activity (33). Likewise, full-length UCL1 (amino acids 1–371) did not bind CLF amino acids 1–300, whereas UCL1 with most of its F-box deleted (amino acids 44–371) bound CLF amino acids 1–300 (Fig. 3D). Indeed, it is a common problem with the yeast two-hybrid system that full-length proteins have buried domains not available for interactions with other proteins (34). This likely explains why full-length UCL1 and its target, CLF, fail to interact in the yeast two-hybrid system. Nevertheless, we successfully detected full length-protein interaction using BiFC in planta (Fig. 4B).

Within the UCL1 C-terminal domain is a conserved DOMAIN OF UNKNOWN FUNCTION 295 (DUF295) that may be responsible for binding CLF. If so, there is considerable genetic redundancy because there are ∼40 genes in Arabidopsis that encode proteins with an F-box domain and a DUF295 domain. There are three domains in CLF: a C-terminal SET domain involved in H3K27 methylation (19); a central C5 domain that interacts with the VEFS domain of EMF2, FIS2, and VRN2 PcG proteins (18); and an N-terminal domain of unknown function. It is the CLF N-terminal domain, whose sequence is specific to CLF, that binds to UCL1 in the yeast two-hybrid system (Fig. 3). Thus, the specificity of the interaction in Arabidopsis likely resides in the interaction between the redundant UCL1 DUF295 domain and the unique CLF N-terminal domain.

Function of UCL1 During Arabidopsis Development.

Mutations in the CLF and MEA genes primarily cause defects during vegetative and reproductive development, respectively, suggesting that they function primarily in distinct developmental stages (4). Consistent with this idea, it was recently reported that CLF might be the ancestral PcG gene and that MEA may have acquired a unique reproductive role in seed development during evolution (35). It has previously been shown that CLF functions in the vegetative stage of development and is likely to be regulated transcriptionally (18). Surprisingly, the expression patterns of CLF and MEA were shown not to be mutually exclusive. For example, CLF and MEA have overlapping expression patterns during plant reproduction (35) and both are expressed in the endosperm, where UCL1 promoter activity is detected (Fig. 5). Ectopic expression of CLF from the MEA promoter results in mea-like mutant phenotypes (Fig. 5), suggesting that CLF competes with MEA in the formation of PRC2 complexes. This is consistent with the fact that both CLF and MEA are able to interact in vitro with the same Su(z) homologs, including FIS2, that are expressed in endosperm (18, 36). So, the question arises, how is CLF and MEA function separated during endosperm development when their respective RNA expression patterns and ability to interact with PcG proteins overlap? One possibility is that UCL1, expressed on fertilization in the primary endosperm cell and in actively dividing endosperm cells, specifically binds and ubiquitinylates CLF, resulting in its degradation, leaving MEA free to form PRC2 complexes in the endosperm that are essential for seed viability (4).

Understanding the validity of the above model at the molecular level requires a more thorough knowledge of the levels of MEA and CLF RNA and protein in the Arabidopsis endosperm than is currently available as well as the relative affinity of FIS2 binding to MEA and CLF. Generating multiple KO mutants for endosperm-specific F-box proteins that target CLF could be used test this model at the genetic level, however. In silico and other molecular analyses (e.g., binding to CLF in the yeast two-hybrid system) could be used to find those F-box proteins. Plants with loss-of-function mutations in the appropriate F-box genes could be generated by genetic crosses to determine whether they show defects in seed development.

In summary, we have identified a previously undescribed posttranslational mechanism for the control of PRC2 PcG activity and have demonstrated an interaction between the ubiquitin-26S proteasome pathway and PcG-mediated gene silencing during Arabidopsis development.

Materials and Methods

Full details of methods used in this study are presented in SI Materials and Methods.

Plant Materials and Growing Conditions.

The ucl1-D mutant was isolated from an activation-tagging mutant library as described previously (37). Plants were grown under previously described conditions (38).

Histochemical GUS Staining and Microscopy.

Methods for GUS staining, fixing tissues, and microscopy are as described previously (38).

Recombinant Plasmid Construction.

Methods for generating CaMV::UCL1, 4× Enh::UCL1, CaMV::UCL1:GFP, UCL1::GUS, UCL1::UCL1:GUS, and MEA::CLF constructs are described in SI Materials and Methods.

Microarray Analysis.

Total RNA was isolated using an RNeasy Mini kit (Qiagen). Probe synthesis, detection, and scanning were performed according to protocols from Affymetrix, Inc.

Real-Time qRT-PCR.

RNA levels were quantified by real-time qRT-PCR (iQ5; Bio-Rad), and data were analyzed with iCycle iQ system software (Bio-Rad). Primer sequences are listed in Table S2.

ChIP.

ChIP was performed as described previously (39) with minor modifications. Sonication was performed at 30% for 5 × 10 s, 0.6 s on/0.4 s off in each 1 s, with a Fisher Scientific Sonic Dismembrator 500.

BiFC Assay.

The plant expression vectors pSAT4-nEYFP-C1 (E3801) and pSAT4-cEYFP-C1-B (E3802) were used to generate constructs, which were introduced in pairs into Arabidopsis protoplasts by PEG transfection as described previously (40).

Yeast Two-Hybrid Assay.

The pGBKT7 bait vector and pGADT7 or pGAD424 prey vector in the Matchmaker Two-Hybrid system (Clontech Laboratories) were used. Assay conditions were as described by the manufacturer.

Coimmunoprecipitation Assay.

Coimmunoprecipitation was carried out as described previously (41). Protein was immunoprecipitated with polyclonal rabbit anti-GFP antibody (Invitrogen) and detected with monoclonal mouse anti-GFP (Clontech Laboratories) and monoclonal rat anti-HA (Roche) monoclonal antibodies.