FIE, a nuclear PRC2 protein, forms cytoplasmic complexes in Arabidopsis thaliana

Highlight FIE, a WD-40 subunit of the Arabidopsis PRC2 complexes known to take part in H3K27 methylation of nuclear chromatin, is also localized in cytoplasmic complexes.


Introduction
Polycomb group proteins (PcGs) are highly conserved chromatin modifiers that regulate key developmental pathways in plants (Kohler and Villar, 2008;Derkacheva and Hennig, 2014) and animals Pirrotta, 2007, 2008), selectively controlling temporal and spatial expression of numerous genes. PcG proteins mediate repression of gene expression through an epigenetic mechanism involving methylation of histone H3 at lysine 27, which subsequently leads to chromatin remodeling and condensation (Bouyer et al., 2011;Kim et al., 2012;Ikeuchi et al., 2015;Horst et al., 2016). In Arabidopsis thaliana, PcGs were found to play a major role in regulating various developmental processes, including the transition from the gametophytic to sporophytic phase, embryogenesis, and organogenesis, as well as the transition from the vegetative to reproductive phase and flower development. As much as 20-35% of the genes in Arabidopsis are potentially regulated by PcG proteins (Zhang et al., 2007;Turck et al., 2007;Bouyer et al., 2011;Kim et al., 2012).
Since FIE is encoded by a single-copy gene in Arabidopsis, it is probably an essential component of all PRC2 complexes that regulate different aspects of the plant life cycle . The FIE protein was shown to interact with each of the HMTase subunits: MEA (Luo et al., 2000;Yadegari et al., 2000;Kohler et al., 2003;Bracha-Drori et al., 2004), CLF Mosquna et al., 2009), and SWN (Luo et al., 2000;Chanvivattana et al., 2004), as well as with the MSI1 subunit (Kohler et al., 2003). FIE was shown to regulate the female gametophyte and seed development (Ohad et al., 1996;Chaudhury et al., 1997), to establish the anterior-posterior polar axis in the endosperm (Sorensen et al., 2001) and to inhibit endosperm overproliferation (Grossniklaus et al., 1998;Luo et al., 1999;Ohad et al., 1999;Kohler et al., 2003;Wang et al., 2006). Accordingly, mutants impaired in FIE function were shown to develop seed-like structures containing juvenile endosperm in the absence of fertilization, and arrested embryos surrounded by overproliferated endosperm when fertilization took place, leading to seed abortion in both cases (Ohad et al., 1996;Chaudhury et al., 1997;Grossniklaus et al., 1998;Luo et al., 1999;Kiyosue et al., 1999;Guitton et al., 2004). FIE also partakes in regulating the transition from embryonic to seedling phase (Bouyer et al., 2011), while during the vegetative stage FIE is required for regulating leaf development and flowering time (Kinoshita et al., 2001;Katz et al., 2004;Chanvivattana et al., 2004). fie-cosupressed mutants, which lose FIE expression during postembryonic stages, display multiple morphological and developmental aberrations, such as loss of apical dominance, curled leaves and precocious flowering . Additionally, leaves and flower organs in these mutants undergo homeotic transformation to carpel-like structures, demonstrating the vital role of FIE in regulating organogenesis (Yoshida et al., 2001;Kinoshita et al., 2001;Katz et al., 2004). Furthermore, fie null mutants display a progressive loss of cell differentiation states after germination, resulting in the formation of a disorganized callus-like plant body (Bouyer et al., 2011).
The fundamental function of FIE in regulating developmental programs along the plant life cycle, including phase transition from the gametophytic to the sporophytic phase, is highly conserved through plant evolution (Butenko and Ohad, 2011). In the moss Physcomitrella patens, the single-copy PpFIE gene controls cell proliferation and differentiation in both the recessive sporophytic and the dominant gametophytic stages of the plant life cycle (Mosquna et al., 2009;Okano et al., 2009;Pereman et al., 2016;Horst et al., 2016). While PpFIE protein was mostly evident in cells undergoing cell-fate transition (Mosquna et al., 2009), in flowering plants transcription of the FIE gene is ubiquitous (Springer et al., 2002;Danilevskaya et al., 2003;Katz et al., 2004;Zimmermann et al., 2004;Hermon et al., 2007;Rodrigues et al., 2008;Luo et al., 2009;Kapazoglou et al., 2010). While in dicots FIE is a single copy gene, two FIE-like genes were identified in the genomes of several monocot species (Springer et al., 2002;Danilevskaya et al., 2003;Luo et al., 2009). One FIE-like gene is expressed ubiquitously, while the other is expressed only in the endosperm (Springer et al., 2002;Danilevskaya et al., 2003;Hermon et al., 2007;Luo et al., 2009;Butenko and Ohad, 2011).
Though FIE was shown to be essential throughout the Arabidopsis life cycle, little is known regarding its accumulation pattern. To visually monitor FIE accumulation during the Arabidopsis life cycle, lines bearing a transgene expressing an FIE-green fluorescent protein (GFP) fusion protein from the endogenous FIE promoter were generated. FIE-GFP was able to fully complement fie mutant phenotype. As expected, FIE-GFP and endogenous FIE were found to be nuclear localized in all tissues tested. However, a significant fraction of FIE-GFP, as well as of the endogenous FIE, was localized in the cytoplasm. Moreover, MEA HMTase protein was found to interact with FIE in the cytoplasm and to form high-molecular-mass protein complexes. Taken together, these results suggest that, in addition to their nuclear function of maintaining chromatin transcriptional regulation, plant PRC2 complexes may have novel distinct functions in the cytoplasm.
Plants were grown in Percival incubators at 22 ± 2°C under white fluorescent illumination (100 μE m −2 s −1 ) in 16 h light-8 h dark cycles (long day). All seeds were sown on half-strength Murashige and Skoog (MS) medium (Murashige and Skoog, 1962) supplemented with 1% sucrose and B5 vitamins, under selection conditions when needed.
Generation of antibodies against plant PRC2 proteins FIE, MEA, CLF and SWN Antibodies were raised in New Zealand White rabbits. αMEA antibodies were generated against a polypeptide corresponding to amino acids (aa) 285-300 of MEA protein. The αMEA antibody recognizes the MEA protein of the expected 97 kDa size, and in addition polypeptides with smaller molecular mass (Fig. 5C,upper panel,arrows). αCLF antibodies were generated against a polypeptide corresponding to aa 1-130 of the CLF protein. αSWN antibodies were generated against polypeptides corresponding to aa 1-122 and to aa 552-568 of the SWN protein. All three antibodies were shown to detect the endogenous proteins at their expected sizes from extracts from wildtype plants, but not in respective mutants (see Supplementary Fig. S1 at JXB online). Generation and characterization of αFIE no. 61 antibodies was previously described .

Protein extraction and immunoblot analyses
Native extraction of cytosolic soluble proteins Extraction of soluble cytosolic proteins was performed as previously described . Plant tissues were ground to a fine powder with liquid nitrogen and then homogenized in protein extraction buffer (100 mM Tris pH 7.2, 10% sucrose, 5 mM MgCl 2 , 5 mM EDTA, 40 mM β-mercaptoethanol, and Complete Protease Inhibitor Cocktail (Roche)) at a ratio of 1 ml to 1 g tissue (initial weight). The homogenate was centrifuged twice at 14 000 g for 10 min and the supernatant was collected. The resulting native cytosolic protein extract was further used for sedimentation and size-exclusion chromatography assays, or mixed with sample buffer and analysed by protein immunoblotting as previously described .

Differential ultracentrifugation
Native cytosolic protein extract, prepared from 1 g of fresh tissue, was pelleted using a mini-ultracentrifuge (Beckman) at 150 000 g and 4 ºC for 1 h. In these conditions, all cell organelles, large protein complexes, non-soluble proteins, and cell membranes precipitate to the pellet, whereas the soluble cytosolic proteins remain in the supernatant. The soluble fractions were collected and analysed by protein immunoblotting using αFIE or αMEA antibodies.

Chromatin protein extraction
Extraction of chromatin proteins was performed on a nuclei-enriched fraction. For immunoblotting analyses, the nuclei-enriched pellet was incubated for 20 min at room temperature (RT) in equal volume (v/v) of ice-cold denaturative nuclei lysis buffer (50 mM Tris pH 7.5, 10 mM EDTA, 1% sodium dodecyl sulfate (SDS) and 0.2 mM PMSF and Complete Protease Inhibitor Cocktail (Roche)) with 1-2 μl DNase1 (Sigma-Aldrich) per sample. Extract was cleared by centrifugation for 10 min at 14 000 g and RT and the supernatant collected.
For size-exclusion chromatography analyses, the nuclei-enriched pellet was resuspended in equal volume (v/v) of ice-cold native chromatin extraction buffer (250 mM NaCl, 50 mM Hepes, 5 mM EDTA, 0.1 mM PMSF, 0.5 mM DTE, 5 mM MgCl 2 and Complete Protease Inhibitor Cocktail (Roche)) with 1-2 μl DNase1 (Sigma-Aldrich) per sample. The extract was flash-frozen in liquid nitrogen to disrupt the nuclear envelope and then incubated on an orbital mixer for 30 min at 4 ºC. The resulting nuclei lysate was centrifuged for 10 min at 14 000 g at RT and the supernatant was collected. The pellet was extracted again as described above, and the supernatant was combined with the previous one.

Size-exclusion chromatography
Size-exclusion chromatography (SEC) was performed on native cytoplasmic or chromatin-associated protein extracts from either rosette leaves of 17-day-old plants or inflorescences (flowers after anthesis). Total protein concentration was determined using the Bradford assay for cytoplasmic extracts, but not for low-protein-content nuclear extracts. Chromatography was carried out using an ÄKTA fastperformance liquid chromatograph (FPLC) (Amersham Pharmacia Biotech) at 4 ºC. Extracts (up to 0.5 ml and/or 500 mg of protein) were applied to an HR 10/30 Superose 6 column (Amersham Pharmacia Biotech) equilibrated in extraction buffer. Proteins were eluted at 0.3 ml min -1 under maximum pressure of 1 MPa. Fractions of 1 ml for rosette leaves and of 0.5 ml for inflorescences were collected. Size of eluted proteins was calibrated using protein size marker standards (Amersham Pharmacia Biotech). Proteins from each fraction were concentrated using 10 μl of Strata Clean Resin Beads (Stratagene) by incubating for 1 h at 4 ºC. Proteins were released from the beads by boiling for 5 min in sample buffer and analysed by immunoblotting.
Transformation of Arabidopsis Col-0 was performed by the floral dip method (Clough and Bent, 1998). Twenty independent kanamycin-resistant T1 plants were obtained and confirmed to contain the gFIE-GFP insert by PCR.
Testing the functionality of gFIE-GFP by complementing the fie+/-phenotype The complementation assay, adapted from (Kinoshita et al., 2001), measures the ability of the ProFIE:FIE gDNA -GFP transgene to complement the seed abortion phenotype associated with inheritance of a maternal mutant fie-1 allele. Independent female kanamycinresistant T1 ProFIE:FIE gDNA -GFP plants, each representing a distinct transgenic line, were crossed with FIE/fie heterozygous males. The resulting F1 seeds were germinated on kanamycin (50 mg l -1 ) to select for the ProFIE:FIE gDNA -GFP transgene, and resistant plants were genotyped via PCR to identify FIE/fie heterozygous plants. Self-pollinated siliques were examined for percentage of aborted seeds. In plants heterozygous for fie-1/FIE and hemizygous for the ProFIE:FIE gDNA -GFP transgene, full complementation by gFIE-GFP is expected to result in 25% seed abortion, instead of 50% observed in control fie-1/FIE heterozygous plants.
Molecular analysis for genotyping the endogenous FIE locus was performed by a two-step PCR reaction to amplify a sequence spanning the point mutation in the fie-1 allele. The PCR product was then subjected to Tsp 504I, a restriction enzyme that specifically recognizes the fie-1 mutated sequence, but not the wild-type sequence.

Bimolecular fluorescence complementation analysis
Protein-protein interaction in plants was examined by Bimolecular Fluorescence Complementation (BiFC) assay (Bracha-Drori et al., 2004;Ohad et al., 2007). FIE, MEA, CLF and SWN full-length cDNAs were cloned into the SpeI site of pSY 751 and pSY 752 binary vectors, which contain the N-terminal (YN) and the C-terminal (YC) fragments of yellow fluorescent protein (YFP), respectively (Bracha-Drori et al., 2004;Mosquna et al., 2009). Equal concentrations of Agrobacterium tumerfaciens strain GV3101/pMp90 containing the plasmids of interest were transiently coexpressed in Arabidopsis cotyledons or in Nicotiana benthamiana leaves. The procedure of transient transformation in Arabidopsis seedlings was based on protocols described by Marion et al. (2008) and Li et al. (2009) with the following modifications: Arabidopsis seedlings were grown on MS-agar in standard six-well plates, 20-30 seeds per well, for 5-6 days until the cotyledons emerged. Agrobacterium cells harboring the appropriate plasmids were grown at 28 ºC in Induction medium (Bracha-Drori et al., 2004) supplemented with antibiotics and 0.2 mM acetosyringone, to give a final OD 600 of 2-4. For transformation, the bacterial cultures were diluted in Induction medium to OD 600 of 1-1.5, mixed in 1:1 ratio and poured on the seedlings until covering them completely. Open plates were placed in a vacuum desiccator jar, and infiltration was performed by applying a vacuum (80 psi) for 2 min ×3. Then, the infiltration medium was removed and the plates were returned to a growth chamber. Protein expression was examined 48-72 h following infiltration by confocal laser scanning microscopy (CLSM).
N. benthamiana leaves were co-transformed with YC-HA-FIE and YN-GG-MEA constructs by leaf infiltration as described in Bracha-Drori et al. (2004). Protein expression was examined 24-48 h following injection by CLSM.

Plasmolysis assay
To induce plasmolysis, leaves or whole Arabidopsis seedlings were submerged in 0.8 M NaCl solution for several minutes. The leaf samples were mounted on microscope slides in the same solution and analysed by CLSM.
Generation of transgenic MEA-3HA line MEA cDNA (At1g02580) sequence was amplified without the stop codon and ligated in-frame to the 3′ end 3×hemagglutinin (3×HA) epitope tag, followed by a TGA stop codon, into a modified binary Ti plasmid pZP111. The resulting pZP111-35S:MEA-3×HA-T(Nos) construct was verified by sequencing and introduced into Agrobacterium tumefaciens strain GV3101 pMP90 (Koncz and Schell, 1986). Transformation of Arabidopsis Col-0 was performed by the floral dip method (Clough and Bent, 1998). The MEA-3×HA protein expressed efficiently in inflorescences and was used for coimmunoprecipitation (Co-IP) assay.

Co-immunoprecipitation of cytoplasmic proteins
The cytosolic fraction was isolated from inflorescences of MEA-3HA plants or N. benthamiana leaves using the method for native extraction of cytosolic soluble proteins. To remove insoluble material, extracts were centrifuged for 30 min at 20 000 g and 4°C. The supernatants were incubated with αHA (Covance, Cat No. MMS-101R; 1:100) or αGlu-Glu (Covance, Cat No. MMS-115R; 1:1000) monoclonal antibodies for 12 h at 4 ºC and then agitated with 25 μl protein G Sepharose beads (4-fast flow, Amersham Pharmacia Biotech) for 1 h to precipitate the antibody-antigen complexes. The beads were precipitated by centrifugation at 10 000 g and 4 ºC for 1 min, washed four times with 1 ml of phosphate-buffered saline then eluted from the beads by boiling them in sample buffer for 5 min. Equal volumes of immunoprecipitated and unbound proteins were used for immunoblotting. The membranes were then incubated with αGFP (Covance, Cat No. MMS-118P; 1:1000), αHA (1:3000) or αFIE antibodies.

Results
FIE mRNA was previously shown to accumulate in a wide range of tissues in Arabidopsis, encompassing both vegetative and reproductive organs (Kiyosue et al., 1999;Yadegari et al., 2000;Katz et al., 2004;Zimmermann et al., 2004). FIE transcript was detected during ovule development and after fertilization in both the embryo and endosperm (Kiyosue et al., 1999).

FIE protein accumulates in all Arabidopsis tissues and organs
As mRNA accumulation does not necessary correlate with presence of the protein product, FIE protein accumulation during plant development was characterized, using specific αFIE antibodies . Immunoblot analysis of nuclear protein extracts from different plant tissues demonstrated that FIE was present in all tissues tested, including juvenile rosette leaves, cauline leaves, flowering stems, inflorescences and siliques (Fig. 1A).
To resolve the patterns of FIE accumulation at the suborgan level, a construct constituting a genomic fusion between FIE and GFP under regulation of the native FIE promoter was used to generate transgenic plants. To ensure that FIE-GFP spatiotemporal expression mimics the endogenous FIE pattern, the entire endogenous FIE locus (At3g20740) was incorporated into the transgene, including all the introns, promoter, and genomic sequence downstream of the STOP codon (Fig. 1B). This construct, designated ProFIE:FIE gDNA -GFP, was introduced into wild-type Col-0 plants, giving rise to several transgenic lines.
To reveal whether the chimeric protein, designated gFIE-GFP, is fully functional, a complementation assay was performed, in which the ability of the transgene to rescue the fie seed abortion phenotype was tested. Independent ProFIE:FIE gDNA -GFP transgenic lines, each carrying the transgene at a single hemizygous locus in a wild-type background, were crossed to heterozygous fie-1 mutant plants. T1 plants bearing the fie-1 allele were identified based on seed abortion phenotype, and the percentage of aborted seeds in the siliques was determined. fie is a female gametophytic lethal mutation and therefore heterozygous FIE/fie plants abort ~50% of their seeds (Ohad et al., 1996;Katz et al., 2004). Accordingly, a single ProFIE:FIE gDNA -GFP transgenic allele introduced into the FIE/fie background should reduce the seed abortion rate to 25% if full complementation occurs. Such reduction was detected in plants derived from lines 141 and 142 (χ 2 test, P<0.05), indicating full complementation of the seed abortion phenotype (Table 1). Furthermore, mature fie/fie gFIE-GFP/gFIE-GFP plants exhibited wild-type phenotype (see Supplementary Fig. S2), and none of the developmental aberrations observed in FIE-deficient mutants Bouyer et al., 2011), thus confirming that the gFIE-GFP protein is biologically functional. All further analyses were performed on progeny of line 142, termed line TH142, which is homozygous for the ProFIE:FIE gDNA -GFP transgene in wild-type Col-0 background.

FIE protein localizes to the cytoplasm
Interestingly, the fluorescence signal from the gFIE-GFP protein was not confined only to the nucleus, but was also clearly evident in the cytoplasm of all cells tested (Fig. 1C). In mature tissues, the intracellular GFP fluorescence was visualized as a thick band in the cell periphery, typical of cytoplasmic appearance in mature cells with large vacuoles (Fig. 1C, 1, 2, 4, 6, 7). In young cells with small vacuoles, such as cells located in budding lateral root, in which the cytoplasm is not pressed to the cell wall, gFIE-GFP protein was observed in the entire volume of the cell (excluding the vacuole) (Fig. 1C,  15-16).
A plasmolysis analysis performed on leaf epidermal cells of the transgenic plants established that the extranuclear gFIE-GFP protein localized to the cytosol, but not to the cell wall or vacuole (Fig. 2). As compared with non-treated cells ( Fig. 2A), plasmolysed cells displayed a characteristic retraction of the plasma membrane from the cell wall, and gFIE-GFP fluorescence was detected within the wall-detached cytoplasm, but not in the cell wall or in the vacuole (Fig. 2B). The chloroplasts were clearly located on the inside of the GFP-labeled band (Fig. 2B), indicating that gFIE-GFP was not located in the tonoplast or in the vacuole. This suggested that the novel extranuclear compartment in which gFIE-GFP protein resides is the cytoplasm.
As cytoplasmic localization of PRC2 proteins was not reported previously in plants, further characterization of this phenomenon was pursued. Cytosolic and nuclear protein fractions, extracted from various tissues of wild-type and gFIE-GFP plants, were analysed by immunoblotting for the presence of native and transgenic FIE proteins. Both the gFIE-GFP and the endogenous FIE proteins were detected in nuclear and in cytoplasmic fractions (Fig. 3A), which is in line with the gFIE-GFP fluorescence distribution (Fig. 1C). The purity of the cytosolic fraction was confirmed by the detection of actin only in the cytosolic fraction and H3K27me 2 only in the nuclear fraction (see Supplementary Fig. S4A). Endogenous FIE, as well as gFIE-GFP, was detected in the cytoplasmic fractions in all examined tissues ( Fig. 3B and Supplementary Fig. S4B), including seedlings, mature rosette leaves, stems, cauline leaves and inflorescences. Endogenous FIE was detected in two closely migrating forms with an estimated size of ~45 kDa (Fig. 3A, B), except in rosettes, suggesting that alternative forms of the FIE protein exist in the cytosol in vivo. This polymorphic appearance may be due to posttranslational modification.
Further characterization of cytosolic FIE protein extracted from wild-type rosette leaves by ultracentrifugation revealed that FIE was mostly present in the soluble supernatant fraction (Fig. 3C). As this method sediments large protein complexes and insoluble organelles, it is likely that a large portion of FIE is present as a soluble protein in the cytosol.

FIE forms a high-molecular-mass complex in the cytoplasm
To examine whether cytosolic FIE forms complexes in vivo, cytoplasmic protein extracts from WT and gFIE-GFP transgenic plants were analysed by size-exclusion chromatography (SEC). Analysis of cytoplasmic protein extracts from rosette leaves showed FIE and gFIE-GFP proteins to be present in fractions 1-2 and 5-9, corresponding to high-molecular-mass proteins (Fig. 4), with molecular masses of ~1.2 MDa and 150-500 kDa, respectively. Monomeric gFIE-GFP and FIE were expected to elute between fractions 11 and 13, as their calculated masses are 66 and 45 kDa, respectively. Therefore, detection of FIE proteins in fractions representing larger molecular mass indicates that FIE takes part in a cytosolic protein complex. However, the majority of the cytoplasmic FIE eluted in the fractions corresponding to low molecular mass proteins (fractions 10-14), which is in agreement with FIE enrichment in the soluble fraction after sedimentation from rosette extracts (Fig. 3C).
A similar allocation of gFIE-GFP protein to fractions corresponding to high-molecular-mass complexes was observed in a parallel SEC analysis of nuclear extracts, purified from rosette leaves of a similar age (see Supplementary Fig. S5). Two HMTases, CLF and SWN, were detected in the same fractions with FIE ( Supplementary Fig. S5), using specific antibodies generated in this study (Supplementary Fig. S1). This is in line with a previous report demonstrating that CLF, SWN and FIE take part in the same nuclear complex (Wood et al., 2006).

MEA interacts with FIE in the cytoplasm
To examine the possibility that the cytoplasmic FIE-containing complex may comprise additional known PRC2 subunits, nuclear and cytoplasmic protein extracts from wild-type Arabidopsis tissues were analysed by immunoblotting using specific anti-MEA, CLF and SWN antibodies. Surprisingly, endogenous MEA was identified in the cytoplasmic fractions from inflorescences (Fig. 5A), revealing a novel accumulation  of MEA outside its previously reported nuclear site. In contrast, CLF and SWN were detected only in the nuclear protein enriched fraction (Fig. 5B). Further characterization of inflorescence cytoplasmic extracts by a sedimentation assay showed that both MEA and the large portion of FIE were sedimented from the soluble fraction (see Supplementary Fig. S6A), suggesting that they may take part in a large protein complex. As FIE and MEA are known to form a high-molecularmass complex in the nucleus (Kohler et al., 2003), the presence of a similar complex in the cytoplasmic fraction was examined. In SEC analysis of cytoplasmic extract from inflorescences of gFIE-GFP transgenic plants, MEA was found to co-elute with gFIE-GFP in fractions 11-20, corresponding to complexes with an expected size of 150-400 kDa (Fig. 5C). gFIE-GFP was also detected in fractions 2-9, corresponding to large protein complexes of 600-1200 kDa (Fig. 5C), similarly to the observed complex in rosettes (Fig. 4).
As in rosettes, a parallel nuclear PRC2 complex was detected in inflorescences of gFIE-GFP plants, where FIE, gFIE-GFP and MEA were found to co-elute in fractions  The blot was probed with αFIE antibody. FIE was detected as a double band in all tissues, aside from rosette leaves. (C) Non-denaturated cytosolic protein extracts from rosette leaf tissue were sedimented by ultracentrifugation. Equal volumes from supernatant soluble fraction before and after (−/+) sedimentation were analysed by SDS-PAGE and probed with αFIE antibody. Absence of RuBisCO large chain protein band at ~53 kDa (marked with asterisk) following ultracentrifugation demonstrated successful sedimentation of large protein complexes. Ponceau staining in panels (A-C) was used to assess equal loading of samples.
To test whether FIE and MEA directly interact in the cytoplasm, a bimolecular fluorescence complementation assay (BiFC) (Bracha-Drori et al., 2004;Walter et al., 2004;Ohad et al., 2007) was performed in Arabidopsis, which is the endogenous environment for the examined proteins. Wild-type Arabidopsis seedlings transiently coexpressing YC-FIE and either MEA, CLF or SWN methyltransferases, fused to the YN portion, were analysed. In agreement with previous reports, FIE indeed interacted with each of the tested HMTases. FIE and MEA interaction was detected both in the nucleus and in the cytosol (Fig. 5D). On the other hand, the interaction of FIE with CLF or SWN was detected only in the nucleus (Fig. 5D), which coincides with the presence of CLF and SWN with FIE only in nuclear high molecular complexes (see Supplementary Fig.  S5; Wood et al., 2006). Co-immunoprecipitation experiments further demonstrated the ability of FIE and MEA to interact in the cytoplasm. Transiently expressed YN-MEA and YC-FIE proteins were precipitated together from a cytoplasmic extract ( Supplementary Fig. S7A), as reported previously by Bracha-Drori et al. (2004). Furthermore, transgenic plants expressing MEA protein fused with 3×HA-tag ( Supplementary Fig. S7B) were used in an independent co-IP assay. The MEA-3×HA protein was immunoprecipitated from inflorescence cytoplasmic extract using αHA antibody. Co-immunoprecipitation of the endogenous FIE was then detected by immunoblot analysis with the αFIE (Supplementary Fig. S7C).

Discussion
The epigenetic role of plant Polycomb proteins in regulating gene expression has been explored intensively over the recent years using genetics, molecular biology and biochemical approaches (Butenko and Ohad, 2011;Bemer and Grossniklaus, 2012;Derkacheva and Hennig, 2014;de la Paz Sanchez et al., 2015). In Arabidopsis, members of the three PRC2 complexes were shown to play a vital role in nearly every stage of the plant life cycle, as evident from the severe phenotypes of pcg mutants (Mozgova et al., 2015). As FIE is a single copy gene encoding the WD-40 subunit common to all nuclear PRC2 complexes, it is expected to be indispensable for all PRC2 functions.

FIE protein is present in the nuclei of all cells and tissues
Previous reports demonstrated that FIE has a central role in major developmental processes in Arabidopsis (Ohad et al., 1996;Kiyosue et al., 1999;Yadegari et al., 2000;Katz et al., 2004;Wood et al., 2006;De Lucia et al., 2008;Bouyer et al., 2011;Kim et al., 2012;Ikeuchi et al., 2015). Accordingly, FIE transcripts were detected in numerous vegetative and reproductive plant tissues (Genevestigator database; Zimmermann et al., 2004). In the current study, FIE protein was found in the nuclei of all examined cell types, organs, and tissues, demonstrating a wide temporal and spatial expression pattern (Fig. 1), which overlaps with the known distribution of FIE transcripts. Thus, this study provides additional molecular and biochemical evidence in support of the extensive role of FIE throughout the entire plant life cycle.
The functionality of the gFIE-GFP transgenic protein used in this study was demonstrated by its ability to fully complement the fie phenotype (Table 1 and Supplementary Fig. S2). Interestingly, the double-homozygous gFIE-GFP transgenic lines, bearing two WT endogenous and two transgenic FIE alleles, displayed a WT phenotype (not shown). In contrast, full or partial loss of FIE leads to a variety of developmental impairments (Ohad et al., 1996;Kiyosue et al., 1999;Luo et al., 2000;Katz et al., 2004). This is consistent with FIE playing a role as part of a multi-protein complex, where excess of an individual subunit would not interfere with the complex function, while loss of any subunit may abolish complex functionality.

FIE localizes to the cytoplasm
The surprising outcome of this study is the discovery that FIE protein accumulates in the cytoplasm (Figs 1C and 3), distant from of its acknowledged site of action on the chromatin. The systematic examination of the subcellular localization of FIE determined that cytoplasmic accumulation of FIE is a general phenomenon in Arabidopsis, taking place in every cell where FIE is produced, and is not limited to specific tissues (Figs 1C and 3B,and Supplementary Fig. S3). Overall, cell fractionation experiments (Figs 1A and 3) together with live cell imaging ( Fig. 1C and Supplementary Fig. S3) indicated that endogenous and transgenic FIE proteins localize in both the nuclear and the cytoplasmic compartments.
The presence of the typically nuclear Polycomb proteins in the cytoplasm has not been previously reported in plant cells. However, the unique cytosolic accumulation of the PRC2 components may indicate that plant PcGs have alternative non-nuclear functions. This suggestion is in agreement with accumulating evidences from studies in metazoa, demonstrating cytoplasmic localization and function of PcG proteins (Rietzler et al., 1998;Ogawa et al., 2003;Witte et al., 2004;Su et al., 2005;Bryant et al., 2008;Philipp et al., 2010;Follmer et al., 2012;Roy et al., 2012).

FIE and MEA take part in a cytoplasmic highmolecular-mass complex
In both vegetative and floral tissues, cytoplasmic FIE was found to take part in high-molecular-mass protein complexes (Figs 4 and 5C). These findings support the suggestion that FIE has a functional role in the cytoplasm, rather than simply being detected in this compartment prior to its translocation to the nucleus. The molecular masses of these cytoplasmic FIE-containing complexes were distributed across a wide range, from 150 to 1200 kDa (Figs 4 and 5C). In both rosettes and inflorescences, cytoplasmic FIE was detected in two main peaks: one corresponding to a molecular mass of 150-500 kDa and the second corresponding to a molecular mass of ~1-1.2 MDa (Figs 4 and 5C). Therefore, it is possible that in these two plant tissues, FIE forms more than one cytoplasmic complex, each with a different protein composition. In rosette leaves, however, the majority of cytosolic FIE appeared in a monomeric form, as FIE eluted as monomers in the size-exclusion chromatography (Fig. 4), and remained in the soluble fraction during sedimentation (Fig. 3C). The significance of this phenomenon is not yet clear.
In contrast to rosettes, in inflorescences a large portion of the cytosolic FIE was found in a multiprotein complex ( Supplementary Fig. S5A). This complex also included MEA, an HMTase subunit of the nuclear PRC2 complex in floral reproductive organs. The above is based on the fact that endogenous MEA protein was detected in the cytoplasmic fraction isolated from inflorescences (Fig. 5A), and co-eluted with FIE in the fractions corresponding to the 150-500 kDa complex (Fig. 5C). Moreover, direct FIE and MEA interaction in the novel cytoplasm compartment was demonstrated in planta via a BiFC assay in Arabidopsis leaves (Fig. 5D), and further substantiated by the ability of a tagged MEA protein to co-precipitate the endogenous FIE from cytosolic extract (see Supplementary  Fig. S7C). The cytoplasmic interaction between MEA and FIE was previously observed using a BiFC assay in N. benthamiana leaves (Bracha-Drori et al., 2004;Mosquna et al., 2009), which was attributed to overexpression in a non-endogenous system. However, based on the current results using three independent approaches to demonstrate FIE and MEA endogenous localization and formation of high-molecular-mass complexes in the cytoplasm in Arabidopsis cells, it is proposed that PRC2 complexes might have cytosolic functions.
Nuclear PRC2 complexes were detected, as expected, in both rosettes and inflorescences, ranging between 150 and 1200 kDa ( Supplementary Figs S5 and S6B). Nuclear FIE co-eluted with MEA in inflorescences ( Supplementary Fig.  S6B), and with SWN and CLF in rosettes (see Supplementary  Fig. S5). In both cases, these fractions corresponded to highmolecular-mass protein complexes, including fractions larger than 600 kDa ( Supplementary Fig. S5). These results are in line with previous independent reports (Kohler et al., 2003;Wood et al., 2006;De Lucia et al., 2008). Overall, our findings indicate that nuclear and cytoplasmic PcG complexes coexist in parallel in Arabidopsis cells.
Interestingly, in the inflorescences the sizes of FIE-MEA complexes varied between the cytoplasm and the nucleus. The nuclear complex corresponded to ~500-1200 kDa (see Supplementary Fig. S6B), in contrast to the putative cytoplasmic FIE-MEA complex that ranged between 150 and 500 kDa (Fig. 5D), suggesting a qualitative and/or quantitative difference in the composition of the cytoplasmic and nuclear complexes.

Potential role of the cytoplasmic PcG complex
The identification of plant PcG proteins in the cytoplasm raises the question of their potential non-nuclear function. Although only two PRC2 proteins were identified in the cytosol in the current study, the putative cytoplasmic PcG complex(es) may include additional core members of the nuclear PRC2, namely zinc-finger and MSI1 proteins, or any of the nearly 20 noncore PRC2 proteins identified in Arabidopsis (Pikaard and Mittelsten Scheid, 2014). This suggestion is based on several independent studies in metazoa that demonstrated the presence of multiple PRC2 components in the cytoplasm of different cell types. In Drosophila, nine PcG proteins, including the core PRC2 subunits E(z), Su(12)z, and PSC, were identified in the cytoplasm of interphase-stage S2 embryonic cells (Follmer et al., 2012). Moreover, the PRC2 subunits EZH1 and EZH2 (HMTases), Su(z)12 (zinc finger) and EED (FIE homolog) were detected in the cytoplasm of various mammalian cells (Rietzler et al., 1998;Ogawa et al., 2003;Su et al., 2005;Bryant et al., 2008). It would thus be interesting to determine the composition of the plant cytoplasmic FIE-containing complexes in vegetative and reproductive tissues, and compare them with nuclear PRC2 complexes.
In contrast to the wide accumulation pattern of FIE in Arabidopsis, MEA protein was so far detected only in reproductive tissues (Kohler et al., 2003;Jullien et al., 2006). The putative cytoplasmic FIE-MEA-containing PRC2 complex is therefore expected to be limited to specific cells in developing ovules and seeds. It was reasonable to assume that analogous cytoplasmic complexes in vegetative tissues would contain either of the two remaining HMTases, namely CLF or SWN, as the catalytic subunit. However, in contrast to MEA, localization of CLF and SWN appeared to be strictly nuclear (Fig. 5A, B). It is thus unclear whether the cytosolic FIE-containing complexes detected in rosette leaves are genuine PRC2 complexes that contain a methyltransferase subunit. Alternatively, in the vegetative organs, cytoplasmic FIE may interact with non-PcG proteins to form a completely new type of complex.
Protein function can be elucidated from the observation of mutant phenotype; however, no specific cytosolic function is known to be associated with fie and mea KO mutant phenotypes. Given the functional conservation of the nuclear PcG function between plant and animal kingdoms, it is possible that such conservation exists for the cytoplasmic functions as well. In metazoan cells, various PRC2 subunits were shown to function in the cytosol downstream of activated growth-factor receptors, to promote proliferation, differentiation and dynamic cytoskeletal responses (Ogawa et al., 2003;Su et al., 2005;Bryant et al., 2008;Philipp et al., 2010;Roy et al., 2012). Mammalian EED and EZH2 proteins actively translocate from the nucleus to cytoplasm in response to stimulation from plasma membrane tumor necrosis factor (EED; Philipp et al., 2010), integrin (EED; Witte et al., 2004) and Notch receptors (EZH2; Roy et al., 2012). Notably, a catalytically active Polycomb complex containing the nuclear PRC2 subunits EZH2, EED and Su(z)12 was isolated from a cytoplasmic fraction of mammalian T cells and fibroblasts. The lysine-methylating activity of this complex was shown to be crucial for receptor-induced actin reorganization and proliferation (Su et al., 2005). Most of the direct interactors of mammalian cytosolic PRC2 that take part in regulating actin organization (Ogawa et al., 2003;Su et al., 2005;Roy et al., 2012), such as the GTP/GDP exchange factor Vav1, do not have orthologs in Arabidopsis. However the mammalian receptor for activated C kinase 1 (RACK1), which was shown to interact with EED in the cytoplasm (Philipp et al., 2010), has homologs in Arabidopsis, which act in a signal transduction pathway similar to their mammalian counterparts (Islas-Flores et al., 2015). Hence, it would be interesting to test whether the Arabidopsis cytoplasmic FIE or FIE-MEA-containing complexes can interact with cytoplasmic components, such as the actin-regulating cellular machinery, or play roles in signal transduction pathways in a similar manner to that reported in mammalian cells, though not necessarily through the conserved homologous components.

Supplementary data
Supplementary data are available at JXB online. Figure S1. Characterization of PRC2 specific antibodies. Figure S2. Morphology of gFIE-GFP complemented fie-1 -/plants is indistinguishable from WT plants. Figure S3. gFIE-GFP protein accumulates in all Arabidopsis tissues and organs. Figure S4. Endogenous FIE and gFIE-GFP proteins are present in the cytoplasm. Figure S5. Nuclear FIE complexes in vegetative tissues of Arabidopsis plant. Figure S6. FIE and MEA in reproductive tissues of Arabidopsis plant. Figure S7. FIE and MEA proteins co-immunoprecipitation from cytoplasmic fraction.