Introduction
Regulation of protein stability by the ubiquitin/proteasome system participates in a broad variety of physiologically and developmentally controlled processes in all eukaryotes (
Ciechanover et al, 2000;
Smalle and Vierstra, 2004). In this pathway, a critical step involves ubiquitin ligases (E3s), which facilitate the transfer of ubiquitin moieties to a substrate protein, the preparative step for degradation via the 26S proteasome. Among the different E3 enzymes, the composition of CUL4‐based E3 ligases (CRL4s) was only recently identified (
Higa and Zhang, 2007). CUL4 binds RBX1 to recruit a specific E2 ubiquitin‐conjugating enzyme, and also binds DDB1, an adaptor protein, which itself associates with a substrate receptor. Affinity purification of CLR4s from mammalian cells identified various WD40 proteins as possible substrate receptors (
Angers et al, 2006;
He et al, 2006;
Higa et al, 2006;
Jin et al, 2006). Many of these proteins, also called DDB1 and CUL4‐associated factors (DCAFs), contain WDxR motifs that are required for efficient DDB1 binding. However, for most of them, their roles and substrates remain unknown. In humans, about 90 different DCAFs have been predicted (
He et al, 2006), suggesting the existence of a large number of CRL4s. A similar number of WD40 repeat proteins harbouring at least one WDxR motif have been identified in the model plant
Arabidopsis thaliana (
Lee et al, 2008). One of the predicted Arabidopsis DCAFs is MULTICOPY SUPPRESSOR OF IRA1 (MSI1), which belongs to an evolutionary conserved protein family (reviewed in
Hennig et al, 2005), whose founding member is MSI1 from yeast (
Ruggieri et al, 1989). In both metazoans and plants, MSI1‐like proteins are part of several protein complexes involved in diverse chromatin functions (reviewed in
Hennig et al, 2005). In particular, MSI1 has been proposed to maintain epigenetic memory during development by targeting silencing complexes to chromatin.
In Arabidopsis, MSI1 is essential for plant reproductive development (
Köhler et al, 2003;
Guitton et al, 2004). In
msi1 mutants, seeds abort when the mutant allele is inherited from the mother regardless of the paternal contribution. In such seeds, the endosperm (an embryo nourishing tissue) does not cellularize, whereas the embryo exhibits cell‐cycle and developmental defects.
msi1 mutants have a strong penetrance of autonomous endosperm development in the absence of fertilization and form rare parthenogenetic embryos (
Köhler et al, 2003;
Guitton and Berger, 2005). MSI1 is part of the FIS–PRC2 complex together with at least three other proteins, MEDEA (MEA), FERTILIZATION‐INDEPENDENT SEED2 (FIS2) and FERTILIZATION‐INDEPENDENT ENDOSPERM (FIE), which is required for normal seed development (
Köhler et al, 2003).
MEA encodes a SET‐domain‐containing histone methyltransferase homologous to
Drosophila Enhancer of Zeste (
Grossniklaus et al, 1998) and regulates the imprinted expression of itself, as well as of its target gene
PHERES1 (
PHE1), encoding a MADS‐domain transcription factor (
Köhler et al, 2005). Imprinting regulation by FIS–PRC2 involves the silencing of the paternal allele of
MEA and the maternal allele of
PHE1, respectively (
Köhler et al, 2005;
Baroux et al, 2006;
Gehring et al, 2006;
Jullien et al, 2006). In contrast, auto‐repression of the maternal
MEA allele is FIS–PRC2 independent (
Baroux et al, 2006).
Here, we report that all WD40 repeat MSI1‐like proteins from various organisms carry at least one conserved WDxR motif, a signature of DCAFs. Arabidopsis MSI1 physically interacts with DDB1A and is part of a CUL4–DDB1A–MSI1 protein complex. Functional analysis revealed that CUL4, as well as the Arabidopsis DDB1 homologs, are essential for seed production. Importantly, the cul4 mutation leads to autonomous endosperm development and loss of parental MEA imprinting, that is reactivation of the paternal MEA allele, supporting a functional link of this E3 ligase and the FIS–PRC2 complex.
Discussion
MSI1 is a well‐characterized WD40 protein with several functions in the control of chromatin dynamics and gene expression (
Hennig et al, 2005). In particular, MSI1 was identified as a subunit of the FIS–PRC2 complex (
Köhler et al, 2003), which regulates parental imprinting during seed development. Here, we provide evidence that CUL4–DDB1 physically associates with MSI1 and is involved in the regulation of the FIS–PRC2 in Arabidopsis. However, because the MSI1 protein associates with additional protein complexes such as chromatin assembling factor 1 (CAF‐1) and a complex with the retinoblastoma‐related protein (
Exner et al, 2006;
Jullien et al, 2008), we do not exclude the possibility that CUL4–DDB1 acts at more than one level.
In flowering plants, imprinting has been mostly studied in the endosperm, which is a terminal tissue developing after fertilization of the central cell (reviewed in
Grossniklaus, 2005;
Feil and Berger, 2007;
Köhler and Weinhofer‐Molisch, 2010). Thus far, several genes have been found to be maternally expressed but paternally silenced including
MEA,
FIS2 and
FWA (
Vielle‐Calzada et al, 1999;
Kinoshita et al, 1999,
2004;
Jullien et al, 2006). In particular, the paternal silencing of
MEA requires the activity of the FIS–PRC2, which mediates trimethylation of histone H3 at the lysine residue 27 (H3K27me3) (
Gehring et al, 2006;
Jullien et al, 2006;
Makarevich et al, 2006). In contrast, the MADS‐box gene
PHE1 is predominantly expressed from the paternal allele while the maternal allele is downregulated (
Köhler et al, 2005). The maternal
PHE1 allele is repressed through the combined action of the FIS–PRC2 containing MEA and the unmethylated DNA state of a DMR in the 3′ region of
PHE1 (
Köhler et al, 2005;
Makarevich et al, 2006). In sperm cells, the DMR is most likely methylated by the maintenance DNA methyltransferase MET1, preventing silencing and leading to an active paternal
PHE1 allele (
Makarevich et al, 2008).
In line with a role of CUL4–DDB1 in the regulation of the FIS–PRC2 complex, we could show that reduced
CUL4 activity leads to autonomous endosperm division, although at a lower penetrance than in some other
fis class mutants (
Ohad et al, 1996;
Chaudhury et al, 1997;
Köhler et al, 2003;
Guitton et al, 2004). However, the percentage of seeds with a
fis phenotype observed in
cul4 (3.6–16%) is rather similar to that of
mea mutants ranging from 3 to 20% (
Grossniklaus and Vielle‐Calzada, 1998;
Kiyosue et al, 1999). For
mea this low penetrance can be explained due to functional redundancy with its paralogue
SWINGER (
Wang et al, 2006a). Moreover, we showed that paternal
MEA silencing was lost when
CUL4 function was compromised. Consistently, H3K27me3 repressive marks were significantly reduced in the
cul4‐1 knockdown mutant. A similar reduction in H3K27me3 marks was also observed at the
PHE1 locus, albeit this was not sufficient to abolish downregulation of the maternal
PHE1 allele, most likely because of the presence of additional regulatory mechanisms depending on DNA methylation (
Makarevich et al, 2008).
Several observations suggest that CUL4–DDB1 is not required for the establishment of MEA paternal silencing, but rather for its maintenance. First, we never observed paternal MEA reactivation in the pollen of cul4 knockdown or null mutants (data not shown). Second, we only visualized a strong paternal MEA expression 2–3 DAP, but not at 1 DAP. Third, ChIP experiments failed to reveal a loss of H3K27me3 marks at the MEA locus in the young floral buds, but only became evident after fertilization in young siliques. Finally, we also found that CUL4 participates in the maintenance of MEA repression at a later developmental stage (e.g. in 17‐day‐old plants), which depends on another form of PRC2 containing CLF as the histone methyltransferase.
Whether CUL4 and DDB1 only associate transiently or are more stable components of the FIS–PRC2 complex will need further investigations. However, it is noteworthy that MSI1 together with FIE, MEA and FIS2 were found in a very large protein complex of about 650 kDa (
Köhler et al, 2003), leading the authors to speculate that other proteins associate with FIS–PRC2. Another intriguing observation is that the PRC2 core component FIE, which is also part of various other forms of PRC2 in Arabidopsis (
Pien and Grossniklaus, 2007), was predicted to interact with DDB1 based on its structure (
Lee et al, 2008). Thus, not only one but even two PRC2 components may recruit CUL4 to the FIS–PRC2. Moreover, because metazoan homologs of Arabidopsis MSI1‐like proteins, such as the retinoblastoma‐binding proteins P55 in Drosophila and RbAp48 in mammals, have structural features of typical CUL4 substrate receptors, it is probable that our findings will extend to other organisms beyond plants.
Although CUL4–DDB1 is involved in FIS–PRC2 functions,
CUL4 loss‐of‐function mutations do not phenocopy all aspects of
fis class mutants. In particular, we did not observe in
cul4 mutants as strong penetrance of autonomous endosperm development as in
msi1 mutants (
Köhler et al, 2003;
Guitton et al, 2004), nor the presence of parthenogenetic embryos (
Guitton and Berger, 2005). This could be explained by at least two different features that distinguish
cul4 from the
fis class mutants. First, recent work suggests that Arabidopsis
CUL4 is also involved in cell‐cycle regulation (
Marrocco et al, 2010;
Roodbarkelari et al, 2010), as it has been shown in metazoans (
Jin et al, 2006;
Abbas et al, 2008;
Havens and Walter, 2009). Thus, instead to promote cell proliferation in the endosperm as observed in the
fis class mutants, the loss of
CUL4 restricts cell division in this tissue counteracting its
fis phenotype. Second, the
cul4‐1 knockdown does not affect parental imprinting of all FIS–PRC2 targets. In particular, we did not observe the de‐repression of the maternal
PHE1 allele, although we found a decrease in H3K27 methylation at this locus.
PHE1 encodes a MADS‐domain transcription factor (
Köhler et al, 2003), while misexpression in
fis class mutants, but not in
cul4‐1, could explain some phenotypic differences during seed development.
Ubiquitylation has already been linked to
Polycomb‐mediated repression (reviewed in
Niessen et al, 2009). Indeed, the human PRC1 complex exhibits an E3 ligase activity for histone H2A (
Wang et al, 2004), which is triggered by two of its subunits, RING1 and RNF2 (also referred to RING1B or RING2) (
de Napoles et al, 2004;
Buchwald et al, 2006). In the prevailing model, PRC1 binds to histone H3K27me3 to catalyse monoubiquitination of histone H2A, which in turn could interfere with the transcriptional machinery or chromatin remodelling proteins to repress transcription of target genes (
Stock et al, 2007;
Zhou et al, 2008). Moreover, it was recently shown that a tight balance between histone H2A ubiquitylation and deubiquitylation is important for
Polycomb‐mediated repression in Drosophila (
Scheuermann et al, 2010). Arabidopsis also contains RING‐domain proteins that, together with LHP1, may fulfil PRC1‐like functions (
Xu and Shen, 2008), most likely via histone H2A ubiquitylation (
Bratzel et al, 2010).
Our finding that the CUL4–DDB1MSI1 E3 ligase is required in the maintenance of FIS–PRC2‐dependent parental imprinting in Arabidopsis raises the question which substrate(s) are targeted for ubiquitylation in this process. One possibility is that CUL4–DDB1MSI1 ubiquitylates directly one of the FIS–PRC2 subunits.
We initially speculated that MSI1 could be either a substrate or a substrate receptor of this E3 ligase, or even both. Thus, we introgressed the
pMSI::MSI1–RFP reporter gene (
Chen et al, 2008) into the
cul4‐1 mutant background and checked for MSI1 protein accumulation. However, when
cul4‐1 pMSI::MSI1–RFP plants were self‐pollinated, the MSI1–RFP protein level was only slightly higher than in a wild‐type background (
Supplementary Figure S8), suggesting that MSI1 protein turnover is not controlled by CUL4.
In contrast, we noticed that MEA protein does not decay after pollination in cul4 mutants. Thus, it is possible that ubiquitylation controls its stability, although we cannot exclude that this accumulation mainly results from the persistence of the MEA transcript. Nevertheless, unscheduled MEA protein accumulation may alter FIS–PRC2 activity by, for example, titering some of its components or associated proteins. This may also explain the paradox why in the presence of more MEA protein, repression of the paternal MEA allele is lost.
Finally, it is also possible that CUL4–DDB1
MSI1 acts at the FIS–PRC2 level by a mechanism that does not imply protein degradation. In this respect, it is well established that CRL4 E3 ligases trigger different kinds of non‐proteolytic ubiquitylation reactions, including the assembly of K63‐linked polyubiquitin chains and monoubiquitylation. Thus, in the process of nucleotide excision repair after UV damage, a CUL4–DDB1
DDB2 E3 ligase (DDB2 being a WD40 substrate receptor) triggers non‐proteolytic ubiquitylation of XPC (xeroderma pigmentosum complementation group C) to permit its binding to damaged DNA (
Sugasawa et al, 2005) where it induces histone ubiquitylation, presumably to modify the chromatin structure at the sites of DNA lesions (
Kapetanaki et al, 2006;
Wang et al, 2006b). Interestingly, the fission yeast Cul4 associates with Clr4 histone methyltransferase and is required for RNAi‐mediated heterochromatin formation (
Hong et al, 2005;
Jia et al, 2005), although the CUL4‐mediated modifications, which are involved in this process, remain unknown. Future investigations will identify CUL4–DDB1
MSI1 E3 substrates and clarify its function(s) in PRC2‐dependent epigenetic regulation.