Arabidopsis homologs of components of the SWR1 complex regulate flowering and plant development

The development of eukaryotic organisms is controlled by dynamic and coordinated action of gene-specific transcription factors and chromatin modifiers (histone modification enzymes and ATP-dependent chromatin-remodeling enzymes) that establish and maintain spatial and temporal gene expression patterns (reviewed by Pole and Almouzni,2006). In Arabidopsis, molecular genetic analyses have revealed many chromatin modifiers regulating the transition to flowering. Flowering time is regulated in Arabidopsis by a complex genetic network that consists of at least four interdependent genetic pathways, the long day, vernalization, gibberellin (GA) and autonomous pathways (reviewed by Putterill et al., 2004; Henderson and Dean, 2004). Among the genes involved in this complex genetic network, FLOWERING LOCUS C (FLC) is a central flowering regulator integrating the autonomous and vernalization pathways. FLC acts as a floral repressor, downregulating the expression of common downstream target genes -the so-called flowering pathway integrators FT, SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1, AGL20) and LEAFY(Simpson and Dean, 2002). Ultimately, the expression level of these integrators determines the exact flowering time (Hepworth et al.,2002). Recently, the regulation of FLC through chromatin modifications has been intensively studied (reviewed by Lee, 2005).

Vernalization, a process accelerating flowering by sensing a long period of winter cold, suppresses the expression of FLC through a series of histone modifications, including deacetylation and the subsequent methylation of histone H3 at lysine 9 (K9) and lysine 27 (K27)(Sung and Amasino, 2004; Bastow et al., 2004). Deacetylation, and the subsequent K9 and K27 methylations, require VERNALIZATION INSENSITIVE 3 (VIN3), which encodes a PHD-domain protein, and the methylations (particularly K9) require VERNALIZATION 1 (VRN1) and VRN2, which encode a Myb-related DNA-binding protein and a Polycomb group protein homologous to Drosophila Su[z]12, respectively(Sung and Amasino, 2004; Bastow et al., 2004). After returning to warm temperature, the epigenetically repressed state of FLC is maintained by LIKE HETEROCHROMATIN PROTEIN 1(LHP1; also known as TFL2 - TAIR), through the maintenance of the increased levels of H3K9 methylation at FLC chromatin(Sung et al., 2006).

In summer annuals lacking the strong FLC activator FRIGIDA (FRI), FLC expression is repressed by autonomous pathway genes. If any of them is mutated, FLC expression is increased resulting in late flowering (reviewed by Henderson and Dean, 2004; Lee, 2005). Among the autonomous pathway genes, FVE (AtMSI4), FLOWERING LOCUS D (FLD) and RELATIVE OF EARLY FLOWERING 6(REF6) are involved in histone modification at FLC. Mutation in any one of them causes hyperacetylation of H3 and H4 tails, resulting in upregulation of FLC (He et al.,2003; Ausin et al.,2004; Noh et al.,2004). FVE encodes a homolog of yeast MSI (multicopy suppressor of IRA1) and mammalian retinoblastoma-associated proteins RbAp46 and RbAp48 (also known as Rbbp7 and Rbbp4, respectively), which have a function in the histone deacetylase (HDAC) complex(Ausin et al., 2004; Kim et al., 2004). FLD encodes an amine oxidase, homologous to the human LSD1(lysine (K)-specific demethylase 1; also known as AOF2 - Human Gene Nomenclature Database), whereas REF6 encodes one of the plant jumonji-family proteins, the mammalian and yeast homologs of which function as histone demethylases (He et al.,2003; Shi et al.,2004; Noh et al.,2004; Tsukada et al.,2006). Thus, it is likely that FLD and REF6 have roles in the deacetylation of FLC chromatin through histone demethylation.

It has also been reported that Arabidopsis homologs of components of the yeast PAF1 (RNA Polymerase Associated Factor 1) complex are required for trimethylation of H3 K4 at the promoter region of FLC, a hallmark of the active chromatin state (Krogan et al., 2003; Ng et al.,2003). Thus, mutation in any of the homologs of the PAF1 complex causes complete suppression of FLC, and consequently causes early flowering in winter annuals and autonomous pathway mutants(Oh et al., 2004; He et al., 2004). A similar suppression is observed in a mutation in EARLY FLOWERING IN SHORT DAYS (EFS), a homolog of the SET domain methyltransferase(Soppe et al., 1999; Kim et al., 2005). The efs mutation causes reduced trimethylation of H3 K4 or dimethylation of H3 K36 in FLC chromatin (Kim et al., 2005; Zhao et al.,2005).

It has also been reported that putative homologs of components of the ATP-dependent chromatin remodeling complexes are involved in the regulation of flowering time (reviewed by Reyes,2006). In particular, several homologs of components of the yeast SWR1 complex (SWR1C), a member of the SWI2/SNF2 superfamily, have been reported to regulate FLC expression in Arabidopsis. The biochemical function of yeast SWR1C is to catalyze the replacement of H2A with its variant H2AZ (Mizuguchi et al.,2004; Kobor et al.,2004). H2AZ deposition by SWR1C is required for transcriptional regulation, heterochromatic barriers and genome stability in yeast (reviewed by Kamakaka and Biggins,2005). For transcriptional regulation, the SWR1C deposits H2AZ into the transcriptional start site of a myriad of genes(Raisner et al., 2005; Zhang et al., 2005). Yeast SWR1C consists of 13 subunits and homologs of most of them exist in Arabidopsis, although SWC3 and SWC7 homologs have not been detected(see Table S1 in the supplementary material)(Meagher et al., 2005). The substrate of SWR1C, H2AZ, is also conserved in Arabidopsis: three homologs of H2AZ - HTA8 (At2g38810), HTA9 (At1g52740) and HTA11 (At3g54560) -are present (Yi et al., 2006). PHOTOPERIOD-INDEPENDENT EARLY FLOWERING 1 (PIE1) is the Arabidopsis gene most homologous to SWR1, although it contains a SANT domain, which is usually found in ISWI family members(Noh and Amasino, 2003). Two other reported homologs of components of SWR1C are ACTIN-RELATED PROTEIN 4 (AtARP4) and SUPPRESSOR OF FRIGIDA 3 (SUF3;also known as AtARP6 - TAIR)(Kandasamy et al., 2005; Choi et al., 2005; Deal et al., 2005; Martin-Trillo et al., 2006). The mutation in any of these homologs causes an early flowering phenotype through the reduction of FLC expression. Furthermore, the pie1 and suf3 mutants show similar developmental defects,including early flowering, leaf serration, production of extra petals and weak apical dominance.

To extend the study of SWR1C in Arabidopsis, we isolated a T-DNA insertion mutant in the homolog of SWC6 (AtSWC6), another component of yeast SWR1C, and produced RNAi-mediated H2AZ gene-knockdown transgenic plants. The mutation in AtSWC6 resulted in the same phenotypes as suf3. Chromatin immunoprecipitation assays showed that both AtSWC6 and SUF3 bind to the FLC promoter region. Consistently,when SUF3:GFP and AtSWC6:RFP constructs were co-transfected into protoplasts, complete colocalization of the two proteins was observed. In yeast two-hybrid and coimmunoprecipitation analyses, PIE1 interacted with SUF3, AtSWC6 and H2AZ, and AtSWC6 interacted with SUF3 and AtSWC2, another SWR1C component homolog. Finally, H2AZ gene-knockdown transgenic plants exhibited a similar phenotype to atswc6. Our results strongly suggest the presence of a homolog of SWR1C that regulates diverse aspects of plant development, including flowering.

The atswc6 mutant in a Col background was obtained from the Arabidopsis Stock Center (SAIL_1142/CS841940). For genetic analyses and construction of RNAi or amiRNA targeting H2AZs, we used the Arabidopsis Col:FRI^SF2 strain (FRI)(Michaels and Amasino, 1999; Lee et al., 2000). All flowering-time mutant alleles used in this study have been described previously (Lee et al., 2000; Moon et al., 2005; Choi et al., 2005). Plants were grown as previously described (Choi et al.,2005). Flowering time was measured by counting the number of rosette leaves from at least 20 plants.

Primer pairs for amplification of cDNAs or genomic DNAs for plasmid construction are detailed in Table 1. To construct AtSWC6 and SUF3 overexpressors,full-length AtSWC6 and SUF3 cDNAs amplified by PCR were cloned into a binary vector myc-pBA (Zhou et al., 2005). For 35S-AtSWC6:GFP construction,full-length AtSWC6 cDNA without the stop codon was cloned into the modified pPZP211 (KH24) vector with the 35S promoter, GFP coding sequence, and NOS terminator derived from p326-GFP (Choi et al., 2005). For generation of AtSWC6 p-AtSWC6:GUStransgenic plants, 329 bp of upstream sequence from the stop codon of At5g37060 (the neighboring gene) to the translational start site of AtSWC6, and the full-length genomic DNA of AtSWC6 except for the stop codon, were amplified and cloned into the T-DNA binary vector pDW137(Blázquez et al., 1997)as a translational fusion with the β-glucuronidase (GUS) gene. To prepare the construct expressing AtSWC6:GFP or AtSWC6:RFP fusion proteins for protoplast transient expression assays, the open reading frame of AtSWC6 was cloned into p326-GFP and p326-RFP(Choi et al., 2005). For yeast two-hybrid assays, AtSWC6, SUF3, AtSWC2 (At2g36740), HTA8, HTA9, HTA11, PIE1-N (1-1560 bp), PIE1-M (1560-3660 bp)and PIE1-C (3660-6168 bp) cDNAs were amplified and cloned into the BamHI site or BglII site (for AtSWC2) of pGBKT7 and pGADT7 vectors, respectively. To produce the H2AZ RNAi construct, an inverted-repeat construct including the HTA8, HTA9 and HTA11cDNAs was made using the pKANNIBAL vector in which a spliceable intron separates the two repeats (Helliwell and Waterhouse, 2003). Then, the region including the 35S promoter,two repeats, intron and OCS terminator was subcloned into the NotI site of the binary vector pART27 for transformation of FRI. Artificial miRNAs to specifically knockdown HTA8, HTA9, HTA11, and both HTA9 and HTA11, were prepared by overlapping PCR using the pGEM-T Easy vector containing MIR319a as template(Schwab et al., 2006). PCR products including the amiRNA sequences were inserted into the binary vector pCGN18 (Choi et al., 2005). For Agrobacterium-mediated transient expression, the full-length AtSWC2, HTA11 and PIE1 cDNAs were cloned into myc-pBA or the modified pPZP211 (KH24).

Ten-day-old seedlings grown under long-day conditions were used for the ChIP assay as described previously (Kim et al., 2006). Polyclonal anti-myc (Santa Cruz Biotechnology, sc-789)was used for immunoprecipitation. PCR detection of FLC regions and TUBULIN β-2 (TUB2) was performed with the primers shown in Table 1.

Total RNA was extracted from Arabidopsis using TRIZOL (Sigma). The digoxigenin (DIG) labeled-mRNA probes for FLC, SOC1, SUF3 and AtSWC6 were prepared as previously described(Choi et al., 2005). RNA blotting, hybridization and washes were performed as described in the DIG application manual (Roche). RT-PCR and primers for SOC1, FLC, FT and TUB2 are as previously described(Lee et al., 2000; Moon et al., 2005). Primers for AtSWC6, HTA8, HTA9 and HTA11 are: AtSWC6,5′-ATGGAGGAAGAGATGTCGAAC-3′ and 5′-CTATGCAACAAATTTCTGACA-3′; HTA8,5′-ATGGCTGGTAAAGGTGGGAAAG-3′ and 5′-TCAATCCTTGGTGACTTTGTTG-3′; HTA9,5′-ATGTCGGGGAAAGGGCTAAAG-3′ and 5′-CTATTCCTTGGCGGATTTGTTG-3′; HTA11,5′-ATGGCAGGCAAAGGTGGAAAAG-3′ and 5′-TCACTCCTTGGTGGTTTTGTTG-3′.

Plant tissues were harvested and ground in liquid nitrogen and solubilized in protein extraction buffer [10 mM HEPES pH 7.5, 100 mM NaCl, 1 mM EDTA, 0.2%(v/v) Triton X-100, 1 mM PMSF, and 1×protease inhibitor cocktail tablets(Roche)]. The debris was cleared by centrifugation at 15,000 gfor 10 minutes at 4°C. Western blot analyses were performed to determine the myc:AtSWC6 or AtSWC6:GFP protein levels in 35S-myc:AtSWC6 atswc6or 35S-AtSWC6:GFP atswc6 using anti-myc or anti-GFP antibodies.

Detection of GUS expression in AtSWC6p-AtSWC6:GUS transgenic plants was performed according to Jefferson et al.(Jefferson et al., 1987). Plant samples were incubated in 90% acetone at room temperature for 10 minutes. The samples were rinsed once with staining solution without X-Gluc(0.2% Triton X-100, 50 mM sodium phosphate buffer pH 7.2, 2 mM potassium ferrocyanide, 2 mM potassium ferricyanide). The staining solution was replaced by staining solution containing 2 mM X-Gluc, and the plant samples were incubated overnight at 37°C. Samples were then observed after incubating in 50% ethanol at room temperature for 30 minutes.

Rosette leaves of Col plants grown for 4 weeks in long-day conditions were sampled for the isolation and transformation of protoplasts as described[Sheen, J. (2002). A transient expression assay using Arabidopsismesophyll protoplasts. http://genetics.mgh.harvard.edu/sheenweb/]. Protoplasts were co-transformed with both GFP and RFP fusion constructs,totalling about 10 μg of plasmid DNA (prepared using the Qiagen Plasmid Maxi Kit) and cultured at 22°C in the dark. 12 hours after transformation,protoplasts were observed as previously described(Choi et al., 2005).

The vectors and yeast strains (Matchmaker GAL4 Two-Hybrid System 3) were obtained from Clontech. The yeast two-hybrid assay was performed as previously described (Kim et al., 2006)and according to the manufacturer's instructions. Binary constructs presented in Table 1 were transiently expressed in Nicotiana benthamiana leaves and the coimmunoprecipitation assay performed as previously described(Kim et al., 2006).

The yeast ARP6 protein, a homolog of SUF3 in Arabidopsis, is a component of SWR1C, which consists of 13 subunits including SWR1 and SWC6/VSP71 (Kobor et al.,2004; Mizuguchi et al.,2004). Because there is a single homolog of SWC6, At5g37055, in Arabidopsis, we analyzed a T-DNA insertion mutant(SAIL_1142/CS841940) (Fig. 1). This mutant has a T-DNA insertion in the second intron of At5g37055,which results in the complete loss of expression(Fig. 1A-C). We named this gene Arabidopsis thaliana SWC6 (AtSWC6). The AtSWC6 gene encodes a HIT-type zinc-finger protein whose homologs, SWC6 and ZNHIT1, are subunits of the yeast SWR1 and mammalian SRCAP (SWI2/SNF2-related CBP activator protein) complexes, respectively(Cai et al., 2005; Mizuguchi et al., 2004). All three proteins have seven cysteines and one histidine highly conserved in a C-terminal region (see Fig. S1 in the supplementary material). The atswc6 mutant showed an early flowering phenotype as well as leaf serration, flowers with extra petals, small siliques and weak apical dominance(Fig. 1D-H; Table 2; Table 3). When the atswc6 mutation was introduced into FRI-containing Col(Col:FRI^SF2, referred to as FRI below)(Michaels and Amasino, 1999; Lee et al., 2000), which has very late flowering phenotype, the atswc6 suppressed FRI-mediated late flowering (Fig. 1D,F; Table 3). The atswc6 FRI showed much earlier flowering than FRI, but similar flowering with Col. When FLC expression was checked, the atswc6 FRI line showed an approximately 50% reduction as compared with FRI, but fivefold higher expression than Col(Fig. 1C). Such a partial suppression of FLC was also observed in suf3 FRI(Fig. 1C)(Choi et al., 2005). Thus, the phenotypes caused by atswc6 were almost identical to that of suf3 reported previously (Choi et al., 2005). Consistently, the double mutant of atswc6 suf3 was indistinguishable from any of the single mutants, irrespective of the presence of FRI (Fig. 1F). However, the expression analyses showed that the level of SUF3 was not affected by atswc6 and that of AtSWC6was not affected by suf3, indicating that the two genes do not regulate each other at the transcriptional level(Fig. 1B,C).

To confirm whether the mutation of At5g37055 is responsible for the atswc6 phenotype, a full-length cDNA of At5g37055 fused with 6 mycs or a GFP coding sequence driven by the 35S promoter was introduced into atswc6 mutants. We generated four transgenic lines of 35S-myc:AtSWC6 atswc6 and 21 lines of 35S-AtSWC6:GFP atswc6(Fig. 2). Western blot analysis showed the accumulation of myc:AtSWC6 and AtSWC6:GFP proteins in 35S-myc:AtSWC6 atswc6 and 35S-AtSWC6:GFP atswc6 (see Fig. S2 in the supplementary material), suggesting that the fusion proteins were functional. All 35S-myc:AtSWC6 atswc6 showed complete rescue, but 35S-AtSWC6:GFP atswc6 showed partial rescue of flowering phenotype(Fig. 2A-C; Table 3). The other phenotypes,such as serrated leaves, weak apical dominance and extra petals, were completely rescued in all of the transgenic lines (data not shown). It is noteworthy that overexpression of AtSWC6 in atswc6 did not cause an additional delay in flowering and showed the same expression levels of FLC, FT and SOC1 as those of wild-type Col(Fig. 2C,D). Consistently, the AtSWC6 overexpression lines in Col did not show any additional phenotype (data not shown).

Because the target of SUF3 is FLC(Choi et al., 2005; Deal et al., 2005), we used chromatin immunoprecipitation (ChIP) to check whether AtSWC6 and SUF3 bind to the chromatin of the FLC promoter region. For this, we used 35S-myc:AtSWC6 atswc6 and 35S-myc:SUF3 suf3 transgenic lines showing complementation of the mutant phenotype(Fig. 2E). Both transgenic lines showed enrichment of AtSWC6 and SUF3 proteins at the proximal (FLC-B and FLC-C) region of the FLC promoter, but not at the distal (FLC-A)region of the promoter or first intron (FLC-V), suggesting that these proteins bind to the promoter region of FLC chromatin.

The expression of AtSWC6 was examined by RT-PCR, northern blot analysis and histochemical GUS reporter assay in AtSWC6p-AtSWC6:GUStransgenic lines. AtSWC6 transcripts were detected in all tissues tested at variable levels (Fig. 3A). AtSWC6 expression was strongly detected at early stage, and increased slightly with developmental age(Fig. 3B,E). The level of AtSWC6 transcripts did not show any daily rhythm, although that of SUF3 increased at night (Fig. 3C). Furthermore, AtSWC6 transcript levels were not affected by photoperiod or vernalization(Fig. 3D). In the transgenic plants expressing the translational fusion of AtSWC6 genomic DNA with GUS (AtSWC6p-AtSWC6:GUS), GUS expression was detected strongly in actively dividing cells such as root and shoot apex, lateral root primordia, inflorescence, flowers, axillary bud, developing siliques and premature seeds (Fig. 3E). However, GUS expression was rarely detected in mature seeds and siliques, old leaves and stems (Fig. 3E).

To investigate whether AtSWC6 expression is regulated by other flowering time genes, we examined the transcript level of AtSWC6 in flowering-time mutants. The presence of FRI did not change the transcript level nor did an flc mutation(Fig. 1C; see Fig. S3 in the supplementary material). When we checked all the mutants involved in positive regulation of FLC (suf4, vip4, pie1, suf3 and abh1), the autonomous pathway (ld, fld, fve, fca, fy and fpa), the photoperiod pathway (gi and co) and flowering pathway integrators (ft and soc1), the AtSWC6 transcript level remained the same (see Fig. S3 in the supplementary material).

Double mutant analysis showed that the atswc6 mutation reduced the fca- or ld-mediated high levels of FLC expression,thus resulting in earlier flowering and higher expression of SOC1(Fig. 4A,B,E). The flc atswc6 double mutant flowered earlier than the flc single mutant, indicating that AtSWC6 regulates other floral repressor(s) in addition to FLC (Fig. 4C). Additionally, atswc6 partially suppressed late flowering of ft and caused upregulation of SOC1, which could be due to a decrease in FLC or in an additional repressor(s) or both in the atswc6 mutant (Fig. 4D,F).

To investigate the cellular localization of AtSWC6, we examined the florescence of GFP in the root of 35S-AtSWC6:GFP atswc6 transgenic plants. The AtSWC6:GFP proteins were localized in the nuclei of root cells(see Fig. S4 in the supplementary material). We also performed protoplast transient assays using constructs expressing AtSWC6:GFP, AtSWC6:RFP and SUF3:GFP fusion proteins under the control of the 35S promoter so as to compare in detail the localization of AtSWC6 and SUF3. Consistent with the localization of AtSWC6:GFP in roots, both AtSWC6:GFP and AtSWC6:RFP were localized in the nuclei of leaf protoplasts. When they were transfected simultaneously, the fluorescence overlapped entirely and the AtSWC6 proteins formed nuclear speckles (Fig. 5, upper panels). By contrast, when AtSWC6:RFP proteins were expressed with SUF3:GFP, most of the AtSWC6:RFP proteins moved to where SUF3:GFP proteins were localized, such that the fluorescence overlapped(Fig. 5, lower panels). Previously, it was reported that SUF3 is localized at the nuclear periphery(Choi et al., 2005). Our data thus indicate that AtSWC6:RFP relocates to the nuclear periphery where SUF3:GFP is localized, and strongly suggests that the two proteins interact.

We performed yeast two-hybrid assays to determine whether the Arabidopsis homologs of the yeast SWR1C components directly interact with each other. For this assay, we included three homologs of H2AZ - HTA8,HTA9 and HTA11 (Yi et al.,2006) - and a homolog of SWC2 (At2g36740), another component of SWR1C, which directly binds to H2AZ in yeast(Wu et al., 2005). The results showed that AtSWC6 interacts with SUF3 and AtSWC2, and revealed homodimerization of AtSWC6 (Fig. 6A). By contrast, SUF3 showed neither homodimerization nor interaction with AtSWC2, suggesting that AtSWC6 links SUF3 and AtSWC2. For PIE1 interaction analysis, we divided the PIE1 protein into three regions: an N-terminal region (PIE1-N, 1-520 aa) containing the HSA domain; a middle region (PIE1-M, 520-1220 aa) containing two ATPase domains; and a C-terminal region (PIE1-C, 1220-2055 aa) containing the putative SANT domain(Noh and Amasino, 2003). It has been suggested that the SANT domain is required for interaction with both HATs or HDACs and chromatin (Sterner et al., 2002; Noh and Amasino,2003). PIE1-C interacted with both SUF3 and AtSWC6 but not with AtSWC2, and PIE1-N interacted with the three H2AZs (HTA8, HTA9, HTA11) but not with the other proteins tested (Fig. 6D,E). PIE1-M did not show any interaction, despite the fact that it contains ATPase domains. AtSWC2 also interacted with all three H2AZs(Fig. 6B,C). This suggests that AtSWC2 and PIE1-N together are involved in binding to H2AZ. To confirm the yeast two-hybrid result, we performed coimmunoprecipitation analysis using tobacco leaves transiently expressing myc:AtSWC6, AtSWC6:GFP, SUF3:Flag,myc:AtSWC2, HTA11:GFP and PIE1:GFP (Fig. 6F,G,H; see Fig. S6 in the supplementary material). The interactions of AtSWC6-SUF3, AtSWC6-AtSWC2, AtSWC2-HTA11, PIE1-SUF3, and PIE1-AtSWC6 were confirmed by the in planta coimmunoprecipitation assay. As expected, the negative control SUF4 failed to interact with AtSWC6(Fig. 6G). Together, our results show interactions among PIE1, SUF3, AtSWC6, AtSWC2 and H2AZ.

To ascertain whether the knockdown of H2AZ genes would cause a similar phenotype to that observed in pie1, atswc6, and suf3. We generated RNAi-mediated knockdown transgenic plants targeting all three Arabidopsis H2AZ homologs (Fig. 7). Inverted repeats of the three tandemly fused H2AZ DNA fragments separated by a spliceable intron were inserted between the 35S promoter and the OCS terminator, and the construct was used for transformation of FRI (Fig. 7A). Among 62 T1 transformants, eight plants flowered with fewer than 30 rosette leaves, and 11 plants flowered with 30-40 rosette leaves. A total of 30 T1 plants (about 50%) flowered with fewer than 50 leaves, which is earlier flowering than FRI (Fig. 7B,C). We further analyzed three lines showing the early flowering phenotype (flowered with less than 35 leaves). RT-PCR analysis showed that the transcript levels of HTA8 and HTA11 were reduced by approximately 50%, and that of HTA9 was slightly less reduced(Fig. 7D,E). Furthermore, RNA blot analysis showed that the early flowering lines had reduced FLCexpression compared with FRI, suggesting that H2AZ is essential for full activation of FLC (Fig. 7D).

To confirm the requirement of individual H2AZ genes for high-level FLC expression, we generated artificial microRNA (amiR)-mediated knockdown transgenic plants specifically targeting mRNA of HTA8(amiR-HTA8), HTA9 (amiR-HTA9) and HTA11(amiR-HTA11) in FRI, as previously described(Schwab et al., 2006). Although the targeted H2AZ was specifically knocked down, none of the amiR-HTA8 and HTA11 lines (48 and 64, respectively), and none (48 lines) of the amiR-HTA9 lines except two, showed any obvious flowering phenotype, and the FLC expression level was the same as that of the FRI (Fig. 7F). However, two lines of amiR-HTA9 showed earlier flowering (flowered with 38 and 40 leaves), and all amiR-HTA9 lines showed some degree of leaf serration similar to that seen in suf3 and atswc6.

When we checked the tissue specificity of each H2AZ, they were expressed in almost every tissue and there was no difference in the expression pattern between them, although HTA9 showed higher expression than the other two (see Fig. S5 in the supplementary material). Thus, the absence of an obvious flowering phenotype in the amiR-HTA lines is likely to be due to functional redundancy among the three H2AZs. To confirm this, we made amiR-HTA9&11 transgenic lines that knocked down both HTA9 and HTA11 (Fig. 7G). These lines showed early flowering, reduced expression of FLC, leaf serration and flowers with extra petals(Fig. 7F). Among 32 transformants of the amiR-HTA9&11, 16 T1 plants flowered with 15-20 rosette leaves, which is similar to that of suf3 FRI or atswc6 FRI, and the rest also flowered before producing 30 rosettes. Taken together, our results strongly suggest that HTA8, HTA9 and HTA11 are functionally redundant and that FLC expression is regulated by the overall level of the three H2AZs.

Recently, the function of the ATP-dependent chromatin remodeling complex SWR1C has been intensively studied in yeast, a single-cell organism. The subunits of SWR1C, and of the mammalian homolog SRCAP, have been biochemically identified and analyzed (Kobor et al.,2004; Mizuguchi et al.,2004; Cai et al.,2005), but the presence of a homologous complex in plants has not been established. In addition, how SWR1C homologs affect development in higher eukaryotes remains largely unknown. Homologs of most SWR1C components are present in Arabidopsis, and thus the function of the relevant complex can be genetically dissected. In this work, we characterized AtSWC6 and analyzed its interaction with PIE1, SUF3 and AtSWC2, the homologs of components of SWR1C. We also analyzed the function of ArabidopsisH2AZ, a substrate of SWR1C. Our results strongly support the presence of an SWR1C-like complex in Arabidopsis that ensures proper development,including floral repression through full activation of FLC.

Mutations in three homologs of SWR1C components - pie1, suf3 and atswc6 - cause similar developmental defects(Noh and Amasino, 2003; Choi et al., 2005) (see also this study). All three mutants show leaf serration, weak apical dominance,flowers with extra petals, short siliques and early flowering. Genetically,all three mutants cause suppression of late flowering in autonomous pathway mutants as well as in an FRI-containing line. They also show earlier flowering than an flc-null and an approximately 50% reduction in FLC expression in an FRI-containing line. The FLClevel in suf3 FRI or atswc6 FRI is fivefold higher than that in Col although the flowering time is similar, suggesting that SUF3and AtSWC6 regulate additional floral repressor(s). In addition, the expression patterns of SUF3, AtSWC6 and PIE1 are similar(Fig. 3)(Noh and Amasino, 2003; Choi et al., 2005). Although all three genes are involved in the regulation of flowering time, their expression is affected neither by photoperiod nor vernalization, nor by any of the flowering-time mutants. It is notable, however, that pie1 has a somewhat stronger effect than suf3 or atswc6(Noh and Amasino, 2003). pie1 also has the additional phenotypes of reduced fertility and short flowering stem in the Col background, phenotypes that were not obvious in suf3 or atswc6 in the same genetic background(Noh and Amasino, 2003; Choi et al., 2005; Deal et al., 2005). Thus, PIE1 might have a function that is at least partially independent from that of SUF3 and AtSWC6.

In yeast, ARP6 and SWC6 are the most-tightly coupled components in SWR1C(Wu et al., 2005). They are necessary for the association with SWC2 and for nucleosome binding. Consistent with this, SUF3 and AtSWC6 have the most similar developmental function in Arabidopsis. The phenotypes of suf3 and atswc6 mutants are indistinguishable, and the suf3 atswc6 double mutant has the same phenotype as each single mutant. The only difference we found between SUF3 and AtSWC6was in the daily rhythm of expression; SUF3, but not AtSWC6,showed a daily rhythm, the significance of which is not clear because the suf3 mutant does not show any conspicuous rhythmic defect. Together with the fact that overexpression of SUF3 or AtSWC6 does not cause any phenotype, our results strongly suggest that SUF3, AtSWC6 and PIE1 act together as a protein complex. The protein interaction analyses confirmed that the three proteins physically interact(Fig. 6).

Although we have not purified the SWR1C homolog, three lines of experimental evidence strongly support its presence in Arabidopsis. Firstly, protoplast transfection assays clearly showed that AtSWC6 and SUF3 are colocalized. AtSWC6 alone disperses throughout the nucleoplasm as speckles(Fig. 5). But when AtSWC6 and SUF3 are expressed together, AtSWC6 moves to the nuclear periphery where SUF3 is localized, as previously reported (Choi et al., 2005). Such a relocation of AtSWC6 strongly supports their interaction. In addition, the localization analyses may indicate that the nuclear periphery is the place where chromatin remodeling occurs, as previous reports suggested that gene activation or gene silencing occurs at the nuclear periphery (Misteli, 2004; Casolari et al., 2004). Consistent with this, both AtSWC6 and SUF3 bind to the proximal region of the FLC promoter in a ChIP assay. More direct evidence was provided by protein interaction analyses (Fig. 6), which showed that all of the putative components of the Arabidopsis SWR1C homolog are interconnected. Finally, when the expression of H2AZ was reduced to below a certain threshold level by RNAi or amiRNA, the plants showed similar developmental defects to those shown by suf3, atswc6 and pie1, as expected(Fig. 7). Thus, although we have not directly measured the biochemical H2AZ replacement activity, our results strongly suggest that the complex including PIE1, SUF3 and AtSWC6 interacts with H2AZs biochemically and genetically.

Genome-wide analyses in yeast showed that H2AZ is preferentially enriched at repressed promoters (Raisner et al.,2005; Zhang et al.,2005). However, H2AZ is more susceptible to loss from yeast chromatin upon increasing ionic strength, and the acetylated form of H2AZ is preferentially detected in active promoters(Zhang et al., 2005; Millar et al., 2006). Thus, it was proposed that H2AZ is deposited at repressed promoters and facilitates rapid activation, probably through acetylation of H2AZ by the NuA4 complex(Zhang et al., 2005; Millar et al., 2006). Consistent with this, H2AZ promotes full activation of target genes but does not impact upon repression. For example, the yeast deletion mutant of H2AZ shows approximately twofold attenuation in YDC1 activation when induced by heat sock (Zhang et al.,2005). The deletion mutant also causes twofold reduction in PHO5 and GAL1 expression when induced by phosphate starvation and galactose treatment, respectively, but it always shows higher expression than the wild type before induction(Santisteban et al., 2000). We observed a similar result in the mutants of Arabidopsis homologs of SWR1C components. In the presence of FRI, FLC is strongly activated(Michaels and Amasino, 1999). However, if any of the SWR1C homologs are mutated, the expression of FLC is reduced twofold but is still higher than that in the plants without FRI (Fig. 1)(Noh and Amasino, 2003; Choi et al., 2005). Consistently, the knockdown lines of H2AZ showed a similar reduction(Fig. 7). Therefore, our results suggest that the SWR1C homolog in Arabidopsis promotes full activation of a target gene, FLC, as is proposed for the function of SWR1C in yeast.

Histone variant H2AZ is highly conserved in a wide variety of species including yeast, Tetrahymena, Drosophila, chicken, Xenopus,mouse and human (reviewed by Kamakaka and Biggins, 2005). Arabidopsis also has three highly homologous H2AZs (Yi et al.,2006). H2AZ is essential for survival in Tetrahymena,Drosophila and mouse, but is not essential for yeast(Santisteban et al., 2000; Faast et al., 2001; Kamakaka and Biggins, 2005). Whether H2AZ is essential for survival of Arabidopsis is currently unknown owing to genetic redundancy. We also failed to knockdown all three H2AZ genes by RNAi, and thus it remains an open question. However, suf3 and atswc6 mutants are not embryo lethal, although SUF3 and AtSWC6 are single-copy genes in Arabidopsis. Because the homologs of SUF3 and AtSWC6 are found only in SWR1C and are not shared with other chromatin modification complexes, such as yeast INO80 or NuA4 or mammalian TRRAP/TIP60, this suggests that the SWR1C homolog in Arabidopsis is not essential for survival. It also indicates that the SWR1C homolog does not regulate the whole genome, but rather a certain group of genes that includes floral repressors.

The Arabidopsis homolog seems to be divergent from yeast SWR1C. Firstly, PIE1 is divergent from SWR1 in that it contains a SANT domain at the C-terminus, which is usually found in ISWI-family proteins(Noh and Amasino, 2003). Interestingly, AtSWC6 and SUF3 interact with the C-terminal region of PIE1 that contains the SANT domain (Fig. 6). By contrast, SWR1 interaction with SWC6 and ARP6 is dependent on the region between two ATPase domains(Wu et al., 2005), which corresponds to the middle region of PIE1 in our analysis. Secondly, SWC2 could not pull down SWC6 and ARP6 in a yeast swr1 deletion mutant(Wu et al., 2005), but AtSWC2 directly interacts with AtSWC6, a SWC6 homolog, indicating that AtSWC2 is connected with AtSWC6 and SUF3 in Arabidopsis. Thirdly, the binding of H2AZ to the N-terminal region of SWR1 is not suggested in yeast, but in Arabidopsis H2AZ interacts with the N-terminal region of PIE1 and with AtSWC2. This therefore indicates the divergence of Arabidopsisand yeast SWR1C homologs. Further biochemical purification of the SWR1C homolog and functional studies using microarray and chromatin immunoprecipitation-coupled chip hybridization will help us to understand the crucial role of H2AZ and the SWR1C homolog in leaf and flower development and in flowering time control in Arabidopsis.

We thank ABRC for providing SAIL_1142/CS841940 seeds and R. Amasino for critical reading of the manuscript. This work was supported partially by the Korea Science and Engineering Foundation through the National Research Laboratory Program (No. M10600000164-06J0000-16410) and Global Research Laboratory Program (2006-03870), a grant from Seoul R&BD Program, and a grant from the Korea Science and Engineering Foundation through the Plant Metabolism Research Center, Kyung Hee University. K.C., C.P. and J.L. were supported by the Brain Korea 21 program.