EARLY IN SHORT DAYS 1 (ESD1) encodes ACTIN-RELATED PROTEIN 6 (AtARP6), a putative component of chromatin remodelling complexes that positively regulates FLC accumulation in Arabidopsis

The floral transition is highly regulated in many plant species to modulate flowering time in response to environmental and endogenous factors, and to ensure reproductive success. Arabidopsis thaliana is a facultative long-day (LD) species in which winter and summer annual accessions can be distinguished. In winter annual accessions, flowering time is regulated by the vernalization, photoperiod and gibberellin (GA) pathways(Boss et al., 2004; Komeda, 2004; Puterill et al.,2004; Amasino, 2005). Winter annuals require exposure to an extended period of cold (vernalization) to become flowering competent, thus preventing premature flowering in the fall(Michaels and Amasino, 2000; Henderson and Dean, 2004). This requirement is mainly conferred by dominant alleles at the FRIGIDA (FRI) (Johanson et al., 2000) and FLOWERING LOCUS C (FLC) loci(Michaels and Amasino, 1999; Sheldon et al., 1999), as well as by other FLC-related genes within the MAF clade(Scortecci et al., 2001; Ratcliffe et al., 2003; Werner et al., 2005). Active alleles of FRI increase FLC expression to levels that delay flowering (Michaels and Amasino,1999; Sheldon et al.,1999). FLC is a MADS box transcription factor that acts to delay flowering, in part by suppressing the expression of the floral promoters FT and SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1),which function as integrators of flowering signals(Kobayashi et al., 1999; Samach et al., 2000). Vernalization promotes flowering by overcoming the effect of FRI and repressing FLC expression; this repression is stably maintained after plants are returned to warm growth conditions, allowing plants to flower(Michaels and Amasino, 1999; Sheldon et al., 1999). The photoperiod pathway promotes flowering in response to LD through the activation of the floral integrators FT and SOC1. Mutations in photoperiod-pathway genes [e.g. constans (co), fd,fe, fha/cryptochrome2 (cry2), ft, fwa and gigantea (gi)] delay flowering in LD but have little effect on flowering time under short days (SD)(Searle and Coupland, 2004). The GA pathway is required for flowering in non-inductive photoperiods, and mutants with reduced GA levels are extremely delayed in flowering time under SD (Wilson et al., 1992).

Many summer annual accessions of Arabidopsis lack an active FRI allele (Johanson et al.,2000; Gazani et al., 2003; Shindo et al., 2005). Under these circumstances, FLC expression is low and flowering occurs rapidly without vernalization. In these accessions, the reduction of FLC expression depends on the function of the autonomous pathway(Michaels and Amasino, 2001). In fact, mutations in autonomous pathway genes [fca, flowering locus d (fld), fpa, fve, fy, flowering locus k (flk)and luminidependens (ld)] cause a flowering delay under any photoperiod (Boss et al., 2004)that is associated with higher FLC expression, and can be rescued by vernalization (Michaels and Amasino,1999; Sheldon et al.,1999; Michaels and Amasino,2001).

Thus, transcriptional regulation of the FLC repressor is a central checkpoint in both winter and summer annual accessions of Arabidopsis. Recently, the involvement of chromatin modification in FLC regulation has been described (for a review, see He and Amasino, 2005). In non-vernalized winter annual plants, FLC chromatin is in an active conformation and is enriched in modifications, such as the acetylation of histones 3 (H3) and 4 (H4), and the trimethylation of lysine 4 of H3 (H3-K4),which are hallmarks of active genes (He et al., 2003; Ausin et al.,2004; He et al.,2004). Late-flowering autonomous pathway mutants also have increased levels of H3-K4 trimethylation and histone acetylation compared with the rapid-flowering parental line (He et al., 2003; Ausin et al.,2004; He et al.,2004; Kim et al.,2005). Many early flowering mutations suppressing the late flowering phenotype of FRI-containing lines have identified components that are required to maintain high levels of FLCexpression. This is the case of mutants such as early flowering in short days (efs), photoperiod independent early flowering 1(pie1), early flowering 5 (elf5), vernalization independence3 (vip3) and frigida-like1(frl-1), and mutants in genes encoding components of the PAF1 complex(ELF7, VIP4, VIP5 and VIP6/ELF8)(Zhang and Van Nocker, 2002; Noh and Amasino, 2003; Zhang et al., 2003; Noh et al., 2004; He et al., 2004; Michaels et al., 2004; Oh et al., 2004; Kim et al., 2005). Most of these mutations also appear to affect flowering in an FLC-independent manner.

After exposure to an extended winter and the completion of vernalization,the level of modifications associated with `active' chromatin is reduced, and the histone tails of FLC chromatin are deacetylated and become enriched in methylation of lysine 9 (K9) and 27 (K27) of H3(Bastow et al., 2004; Sung and Amasino, 2004), which are hallmarks of repressed genes (Orlando,2003). Mutants that are unable to reduce FLC transcript levels by vernalization or to maintain the vernalised state have permitted the identification of some of the proteins participating in this process, such as the chromatin remodelling factors VERNALIZATION INSENSITIVE 3 (VIN3) and VERNALIZATION 2 (VRN2) (Gendall et al.,2001; Sung and Amasino,2004), and a plant-specific DNA-binding protein, VRN1(Levy et al., 2002).

In summer annual accessions, reduced expression of FLC depends on the autonomous pathway, and is associated with lower histone acetylation of FLC chromatin as a result of FVE and FLD function(He et al., 2003; Ausin et al., 2004). Mutations in both genes cause FLC chromatin to become more acetylated at H3 and H4 concomitantly with an increase in FLC expression(He et al., 2003; Ausin et al., 2004).

Here, we report the identification of EARLY IN SHORT DAYS1(ESD1), a gene that is required for the maintenance of FLCexpression. The esd1 mutation causes early flowering through the reduction of FLC expression, although the mutation also appears to affect flowering through other FLC-like repressors. Using a map-based approach, we have determined that ESD1 encodes ARP6, a member of the actin-related protein family that share moderate sequence homology and basal structure with conventional actins. Recently, ARPs and actins have been discovered in the nucleus as integral components of several chromatin remodelling and histone acetyltransferase (HAT) complexes(Schafer and Schroer, 1999; Galarneau et al., 2000; Rando et al., 2000; Shen et al., 2000; Olave et al., 2002; Blessing et al., 2004). We present evidence that ESD1 is needed to achieve the levels of both H3 acetylation and H3-K4 methylation required for high FLCexpression.

Mutant seed stocks used were in the Landsberg erecta(Ler) and Columbia (Col) genetic backgrounds, and were obtained from the Arabidopsis Biological Resource Centre (ABRC) of Ohio State University(Columbus, USA), the Nottingham Arabidopsis Centre (NASC) in UK and personal donations. The monogenic fve-1, fca-1, co-2 and gi-3 mutants were described by Koornneef et al.(Koornneef et al., 1991); flc-3 was described by Michaels and Amasino(Michaels and Amasino, 2001)and the Col FRI-Sf2 lines were described by Lee et al. (Lee et al.,1995). GA-deficient ga1-3 and ga2-1 mutants were described by Koornneef and van der Veen (Koornneef and van der Veen, 1980) and spy-5 by Jacobsen and Olszewski (Jacobsen and Olszewski,1993). The origin of the esd1-1 to esd1-9alleles is summarized in Table 1. The esd1-10 allele in Col corresponds to the T-DNA line Wisc Ds-Lox 289_29L8, and was kindly provided by the ABRC. We confirmed that all esd1 mutations were allelic by their failure to complement the early flowering phenotype in F1 plants derived from crosses between them. Plants were grown in plastic pots containing a mixture of substrate and vermiculite (3:1). Controlled environmental conditions were provided in growth chambers at 21°C and 80% relative humidity. Plants were illuminated with cool-white fluorescent lights (approximately 120 μE m^-2second^-1). LD conditions consisted of 16 hours of light followed by 8 hours of darkness; SD conditions consisted of 8 hours of light followed by 16 hours of darkness.

Total leaf number was scored as the number of main leaves in the rosette(excluding cotyledons) plus the number of leaves in the inflorescence at the time of opening of the first flower(Koornneef et al., 1991). Cauline, adult and juvenile leaves were scored independently. Rosette leaves lacking abaxial trichomes were considered as juvenile leaves(Telfer et al., 1997). Floral organs were analyzed by scanning electron microscopy as described(Ruiz-Garcia et al.,1997).

Double mutants were constructed by crossing the monogenic esd1-2mutant with lines carrying the mutations co-2, gi-3, fve-1, fca-1, ga1-3,ga2-1 or spy-5. esd1-3 was crossed with a line carrying the flc-3 mutation in Col and with Col FRI Sf-2(Lee and Amasino, 1995). Double mutants were isolated from selfed F2 progeny that showed the esd1 phenotype and that segregated for the second mutation.

The esd1-2 mutation was initially mapped to chromosome 3 between markers GAPab and nga6, using the cleaved-amplified polymorphic sequence(CAPS) and the simple sequence length polymorphism (SSLP) molecular markers indicated in Table S1 in the supplementary material. Additional analysis of 925 esd1-like F2 plants allowed us to locate ESD1 to a pericentromeric region of 1.4 cM, between the T8N9 and ATA1 markers (see Table S2 in the supplementary material). To fine map the esd1 mutation within the interval deleted in the esd1-1 and esd1-6 mutant plants, which is located between the 5F21A14 and 1T14A11 markers, we designed specific PCR molecular markers (see Table S3 in the supplementary material)that were used to amplify the genomic DNA of each esd1 mutant allele,in order to score the presence or absence of the amplified product. Southern blot hybridizations with genomic DNA were performed to confirm the PCR results(data not shown).

Four overlapping binary TAC clones (JAtY74I04, JAtY64M05, JAtY54G02,JAtY49O18) spanning the minimum deleted region in the esd1 alleles were obtained from the Genomic Arabidopsis Resource Network (GARNET) and introduced into the esd1-3 allele by Agrobacterium tumefaciens-mediated transformation using the floral-dip method(Clough and Bent, 1998). The Agrobacterium strain used was C58C1. Transformant plants were selected on soil by spraying seedlings with BASTA.

Only esd1-3 mutant plants transformed with the JAtY T74I04 TAC clone that spans from position 28823 bp of T4P3 BAC clone to 78776 bp of T14A11 BAC clone, showed complementation of the early flowering phenotype. To check whether the integration of the TAC clone was complete in transformed esd1-3 plants, we used a set of specific molecular markers (see Table S3 in the supplementary material) contained in the deleted region. We chose markers that amplify PCR products over genomic DNA extracted from wild-type plants, but not from esd1-3 mutant plants. In this way, we demonstrated that the genomic region of the JAtYT74I04 TAC clone integrated in the complementing transgenic plants contained only two ORFs predicted to encode proteins, At3g33520 and At3g33530. The rest of the ORFs present in this region correspond to pseudogenes and retrotransposon elements.

Total RNA was isolated using TRIzol (Invitrogen-Gibco), electrophoresed and transferred onto Hybond N⁺ membranes (Amersham), following described protocols. For the FLC probe, we used a 700 bp EcoRI/SphI fragment from pFLC lacking the MADS-box domain(Michaels and Amasino, 1999). As loading controls, we used a 305-bp EcoRI fragment of the cauliflower 18S ribosomal DNA gene. ARP6 transcript levels were assayed by RT-PCR. cDNA was prepared by reverse transcription of total RNA from Arabidopsis roots, stems, cauline leaves, floral buds and flowers, according to described procedures(Piñeiro et al., 2003). ARP6 gene-specific primers,5′-GAGCTTCGACCACTTGTCCCAGAT-3′ and 5′-GCATTACAATATACGACAAATAATGTG-3′, were designed to amplify the C-terminal end of the coding region, including the last intron and a portion of the 3′ untranslated region. For low abundance mRNAs, such as the MAF, FT and SOC1 genes, we also performed reverse transcriptase-mediated PCR, according to described procedures(Scortecci et al., 2001; Piñeiro et al., 2003; Ratcliffe et al., 2003). UBIQUITIN 10 (UBQ10) was used as control in these experiments.

esd1-2 fca-1 FLC:GUS plants were obtained by crossing esd1-2 with fca-1 plants carrying a 6 kb FLC:GUStranslational fusion construct (Sheldon et al., 2002). GUS activity in fca-1 and esd1-2 fca-1 FLC:GUS plants was revealed by incubation in 100 mM NaPO₄ (pH 7.2), 2.5 mM 5-bromo-4-chloro-3-indolyl-β-D-glucuronide, 0.5 mM K₃Fe(CN)₆, 0.5 mM K₄Fe(CN)₆ and 0.25% Triton X-100. Plant tissue was incubated at 37°C for 20 hours. After staining, chlorophyll was cleared from the samples by dehydration through an ethanol series.

ChIP assays were carried out as described(Ausin et al., 2004). Chromatin proteins and DNA were cross linked in 10-day-old Col, esd1-3, FRI, esd1-3FRI, Ler, esd1-2, fca-1, esd1-2 fca-1,fve-1 and esd1-2 fve-1 seedlings by formaldehyde fixation. After chromatin isolation, the H3 acetylated and methylated fractions were immunoprecipitated using specific antibodies to acetylated K9 and K14, and trimethylated K4, residues (06-599 and 07-473 from Upstate Biotechnology,respectively). Cross-links were reversed by incubations at 65°C for 2 hours, and DNA was purified with QIAquick spin columns (QIAGEN) and eluted in 40 μl of TE (pH 8.0). Semiquantitative PCR was used to amplify six different fragments of the FLC gene(Michaels and Amasino, 1999)(details and primer sequences are available on request). All PCR reactions and quantification of the amplified DNA were done as described previously(Ausin et al., 2004). We carried out three independent experiments and data provided in Fig. 7 are from one representative. UBQ10 served as an internal control for the ChIP analysis. To calculate the fold enrichment in H3 acetylation or methylation, FLC was first normalized to UBQ10 in each sample, and,subsequently, these values were normalized against their respective wild-type controls.

Mutants at the ESD1 locus were independently identified in screens for Ler mutations conferring early flowering under SD(esd1-1 and esd1-2) or for Col mutations that accelerate developmental phase transitions (esd1-3 to esd1-9)(Table 1). All of the selected alleles produced a similar array of phenotypes, independently of their genetic background. Plants homozygous for esd1 mutations were early flowering under both LD and SD photoperiods (Table 1), showing a more extreme phenotype under SD(Table 1, Fig. 1A,B). The fact that esd1 mutants flower earlier under inductive photoperiods indicates that the mutations do not abolish the photoperiod responses. Earliness was associated with a reduction in the length of all developmental phases of the plant (Fig. 1C), based on leaf shape and leaf trichome distribution(Telfer et al., 1997). This reduction was more dramatic in the case of adult rosette leaves, which were almost absent from esd1 mutants grown under LD and highly reduced in esd1 mutant plants under SD (Fig. 1C). This behaviour is similar to that exhibited by other early flowering mutants such as esd4 and ebs, which also show a major reduction of the adult vegetative phase(Gomez-Mena et al., 2001; Reeves et al., 2002).

A part from their flowering time phenotype, esd1 mutants also displayed a complex pleiotropic vegetative and reproductive phenotype. esd1 mutants show a moderate increase in hypocotyl length but have shortened inflorescence internodes. Furthermore, esd1 leaves are smaller and more curled than wild-type leaves, and usually have serrated margins (Fig. 1D). esd1 flowers are smaller than wild-type ones(Fig. 2C) and frequently bear extra perianth organs. This phenotype was more extreme under SD, where esd1 flowers contained an average of two extra sepals and two extra petals per flower (Fig. 2B,E,G). Mutant flowers showed a reduced fertility that was associated with a reduction in the amount of pollen and approximately a 50%reduction in seed set when compared with wild-type plants. Mutant carpels were much smaller and the filaments of the stamens of mutant flowers were shorter than those of wild-type plants (Fig. 2E). Concomitantly, siliques were approximately 50% shorter in esd1 mutants than in wild-type plants(Fig. 2F). Thus, mutations at the ESD1 locus directly or indirectly alter multiple aspects of plant development.

The early flowering phenotype of esd1 mutants suggested that ESD1 could negatively interact with a flowering promoting pathway in Arabidopsis. To test this hypothesis, we analyzed the phenotype of double mutants carrying esd1 and mutations causing a delay in flowering time. We chose representative mutations for each of the photoperiod,GA and autonomous pathways. Within the photoperiod pathway, mutations at the CO and GI loci delay flowering mainly under LD(Koornneef et al., 1998)(Table 2). esd1-2 co-2and esd1-2 gi-3 double mutants flowered later than esd1-2mutants, and earlier than co-2 and gi-3 plants, and thus displayed an additive phenotype (Table 2). Similar to co-2 and gi-3 single mutants, esd1-2 co-2 and esd1-2 gi-3 double mutants lack the capacity to respond to inductive photoperiods, and flowered with a similar number of leaves under both LD and SD photoperiods(Table 2).

Mutations affecting GA synthesis, such as ga1-3(Sun and Kamiya, 1994) and ga2-1 (Yamaguchi et al.,1998), delay flowering in both LD and SD(Wilson et al., 1992). By contrast, mutations in SPINDLY (SPY) cause constitutive GA signalling and accelerated flowering time(Jacobsen and Olszewski,1993). To determine whether the GA synthesis and response pathways are required for the early flowering phenotype of esd1, we analyzed the phenotype of esd1-2 ga1-3, esd1-2 ga2-1 and esd1-2 spy-5double mutants. Under LD, the esd1-2 ga1-3 and esd1-2 ga2-1double mutants showed an additive flowering time phenotype, in that they flowered earlier than their late parent and later than their early parent(Table 2). Under SD conditions, esd1-2 ga2-1 also showed an intermediate flowering time phenotype;however, the esd1-2 ga1-3 double mutant was unable to flower under SD. This is similar to the phenotype of the ga1-3 mutant, and indicates that the early flowering of esd1 mutants requires GA biosynthesis under SD. In agreement with these results, esd1-2 spy-5double mutants also display an additive early flowering phenotype that is more readily observed under SD (Table 2).

To test the interaction between ESD1 and autonomous pathway genes,we analyzed the flowering phenotype of esd1-2 fve-1 and esd1-2 fca-1 double mutants (Table 2; Fig. 3A). Under LD, some of the esd1-2 fve-1 and esd1-2 fca-1 double mutants were indistinguishable from esd1, although, on average, esd1-2 fve-1 and esd1-2 fca-1 produced one and two leaves more than esd1, respectively (Table 2; Fig. 3A). Under SD, esd1-2 fve-1 and esd1-2 fca-1 mutants were also very similar to esd1-2 single mutants(Table 2), producing a few more leaves (16 and 21 leaves, respectively) than the early flowering parental plants (15 leaves); this result indicates that the late flowering phenotype of autonomous pathway mutations requires ESD1. In summary, these results suggest that ESD1 does not interact with the photoperiod and GA floral induction pathways, but shows an almost epistatic interaction with genes in the autonomous pathway.

The epistatic interaction with mutations in the autonomous pathway suggest that esd1 might cause early flowering either by increasing the activity of the autonomous pathway downstream of FCA and FVE, or by bypassing the requirement for the autonomous pathway by reducing FLC expression. Because other flowering pathways, besides the autonomous pathway, converge on the regulation of FLC expression,it was of interest to evaluate the interaction of esd1 with FRI, a dominant-positive regulator of FLC(Lee and Amasino, 1995; Michaels and Amasino, 1999; Johanson et al., 2000), and to study FLC expression in the different mutant backgrounds. When the FRI allele introgressed from the San Feliu-2 ecotype(FRI-Col) (Lee and Amasino,1995) was combined with the esd1-3 mutation in a Col genetic background, plants showed an additive phenotype in which the FRI late-flowering phenotype was only partially suppressed by esd1-3 (Table 3, Fig. 3B).

In order to check if esd1 suppresses the effect of the autonomous pathway mutations and FRI by reducing FLC mRNA levels, we compared the abundance of the FLC mRNA in wild-type, esd1-2,fca-1, esd1-2 fca-1, fve-1, esd1-2 fve-1, FRI and esd1-3 FRIseedlings (Fig. 3C). FLC mRNA was present at higher levels in both fca-1 and fve-1 mutants and in FRI-containing lines than in wild-type plants, as has been previously shown(Michaels and Amasino, 1999; Sheldon et al., 1999). FLC transcript levels in FRI and in fca-1 and fve-1 mutants were decreased by the esd1 lesion(Fig. 3C). FLC mRNA levels were also compared between wild-type plants and esd1 mutants. Because FLC is expressed at a low level in Ler wild-type plants, we could not observe a clear reduction in its expression in the esd1 alleles isolated in Ler background. However, we were able to detect a reduction in the FLC expression in the esd1alleles isolated in Col background, which bears an FLC allele expressed at higher levels (Fig. 3C). In summary, ESD1 is required to maintain high FLC expression levels, either as promoted by FRI or by mutations that impair the autonomous pathway, and, consistent with the genetic analysis, esd1 mutations suppress the increase in FLCexpression caused by autonomous pathway mutations more effectively than that caused by active FRI alleles. In agreement with this scenario, the expression of the floral integrator genes FT and SOC1,normally repressed by FLC (Moon et al., 2003), was upregulated in the esd1 mutants(Fig. 3D).

Although the effects of esd1 mutations on flowering time are more readily observed in the late-flowering FRI and autonomous pathway mutant backgrounds, as discussed above, the fact that esd1 mutants also flower earlier than the rapid-flowering wild-type strains Lerand Col (Fig. 1 and Table 1) suggests that, in addition to regulating FLC expression, ESD1 plays other roles in the control of flowering time. To determine the fraction of the esd1 early-flowering phenotype that is independent of the effect of the esd1 mutation on FLC expression, we analyzed the phenotypic effect of the esd1-3 mutation in an flc null(flc-3) genetic background(Michaels and Amasino, 1999)under both LD and SD. When combined with flc-3, the esd1mutation significantly reduces the number of leaves produced by flc-3under both photoperiods (Fig. 3E and Table 3),confirming that esd1 mutations have an FLC-independent effect on flowering time. Indeed, loss of function of ESD1 also resulted in downregulation of some other members of the FLC/MAF gene family, particularly MAF1, MAF4 and MAF5 (Fig. 3F). RT-PCR analysis indicated a modest but reproducible decrease in MAF1 gene expression and a marked silencing of the MAF4 and MAF5genes, suggesting that these MAF gene family members represent additional regulatory targets of ESD1.

To understand the molecular function of ESD1, we decided to identify the gene responsible for the observed phenotypes in the mutant by a map-based cloning approach. For this, 925 esd1 F2 plants derived from the cross between esd1-2 and Col were screened with the markers shown in Table S1 in the supplementary material. This allowed us to locate ESD1 south of the T8N9 marker and north of the ATA1 marker, close to the pericentromeric region of chromosome 3 (see markers used in Table S2 in the supplementary material). Owing to the severe suppression of recombination in the vicinity of the centromere and because this chromosomal region is almost completely sequenced, we designed an alternative strategy to complete the identification of the ESD1 gene, based on the observation that all of the isolated alleles harbour a deletion in the pericentromeric region of chromosome 3. We identified the shorter overlapping genomic region that was deleted in all of the esd1 alleles by using PCR molecular markers to amplify specific genomic DNA fragments from all of the esd1 alleles,and looked for the presence or absence of an amplified product (see Table S3 in the supplementary material). In this way, we delimited the ESD1locus to a deleted genomic region between the 5F21A14 and 1T14A11 markers. This region spans three overlapping BAC clones, F21A14, T4P3 and T14A11(Fig. 4A), and is enriched in retrotransposon and transposase elements, pseudogenes and highly repeated sequences.

Subsequently, different overlapping binary TAC clones spanning the deleted region were identified and introduced into the esd1-3 allele by Agrobacterium-mediated transformation, in order to identify those that complemented the mutant phenotype. Only certain transgenic esd1-3 mutant plants transformed with JAtYT74I04 TAC flowered at a similar time as wild-type plants, and had lost the pleiotropic phenotype of esd1 mutant plants (Fig. 4B). The analysis of these transgenic plants showed that TAC clone integration had been incomplete (see Materials and methods), and that only in those lines that contained the genomic region harbouring open reading frames At3g33520 y At3g33530 had the mutant phenotype been complemented, suggesting that one of those ORFs represent ESD1. To determine which one of them corresponded to ESD1, we searched for T-DNA insertions within the At3g33520 and At3g33530 ORFs and identified the Wisc Ds-Lox 289_29L8 line for At3g33520 and the SALK_003098 line for At3g33530(Alonso et al., 2003). We obtained seeds of these lines and identified plants homozygous for the T-DNA insertions. Only the plants that harbour a T-DNA insertion in At3g33520 flowered early under both LD and SD (producing around nine and 29 leaves, respectively; wild-type plants produce 14 leaves in LD and 66 in SD), and showed a pleiotropic phenotype similar to that of esd1mutants regarding leaf shape, extra perianth organs and small siliques(Fig. 4C). Reverse-transcription (RT-PCR) analyses showed no expression of the At3g33520 mRNA in these insertional mutant plants(Fig. 4C), indicating that the T-DNA insertion causes a loss-of-function allele. Complementation tests confirmed that this T-DNA mutation was allelic to esd1. Thus, we refer to the Wisc Ds-Lox 289_29L8 line as the esd1-10 allele, and conclude that the ESD1 locus corresponds to the At3g33520gene.

To determine the genomic structure of ESD1, a cDNA was identified and sequenced. The ESD1 gene possesses six exons and five introns,and encodes a protein of 421 amino acids(Fig. 5A). This protein corresponds to ARP6, a member of the actin-related protein family that shares moderate sequence homology and basal structure with conventional actins, but it has two peptide insertions that seemingly provide divergent surface features from actins (Fig. 5B). ARPs are normally grouped into several classes or subfamilies that are highly conserved in a wide range of eukaryotes, from yeast to plants and humans(Goodson and Hawse, 2002). Database searches with the AtARP6 protein sequence identified eight potential ARP proteins in Arabidopsis (ARP2-ARP9)(McKinney et al., 2002). In particular, AtARP6 is a likely ortholog of a group of less-characterized ARPs,including ARP6s from yeast, C. elegans, fruit fly and humans(Fig. 5B). RT-PCR analysis revealed that ARP6 mRNA is detected in most plant organs, with the highest levels found in roots and floral buds(Fig. 5C). Lower levels were detected in cauline leaves, stems and flowers. These results indicate that ARP6 is expressed ubiquitously.

FLC gene expression integrates signals coming from different pathways involved in the regulation of the floral transition(Michaels and Amasino, 1999; Sheldon et al., 2000; Amasino, 2005). Recent work has demonstrated the role of histone modification in the regulation of FLC expression through FRI, the autonomous and the vernalization pathways (He et al.,2003; Ausin et al.,2004; Bastow et al.,2004; He et al.,2004; Sung et al., 2004; Kim et al., 2005). These results have also identified the first intron of FLC as a relevant region for histone modification(He et al., 2003; Ausin et al., 2004; Bastow et al., 2004; He et al., 2004; Sung et al.,2004) and transcriptional regulation(Gendall et al., 2001; Sheldon et al., 2002).

We first analyzed the effect of esd1 mutations on a FLC:GUS translational fusion containing all the FLC promoter and intron elements required for proper regulation(Sheldon et al., 2002). For this purpose, we introduced the FLC:GUS construct into the fca-1 esd1-2 background and analyzed five independent lines for GUSexpression. In contrast to the pattern of GUS expression in the fca-1 background, all of the fca-1 esd1-2 FLC:GUS lines we examined showed undetectable FLC:GUS expression in the shoot apical meristem (SAM) and in the root apical meristem (RAM)(Fig. 6). These results indicate that ARP6 is required for the high level of FLCexpression in the SAM and the RAM.

Because esd1 suppresses the late-flowering phenotype of autonomous pathway mutants such as fve and fca, and FVErepresses FLC transcription through a histone deacetylation mechanism, we initially speculated that ARP6 could be required for the acetylation of histones necessary to activate FLC expression. In fact, as mentioned above, there is considerable evidence implicating nuclear ARPs in chromatin remodelling complexes(Cairns et al., 1998; Boyer and Peterson, 2000; Olave et al., 2002).

To determine whether ARP6 promotes histone acetylation of the FLCchromatin, we used chromatin immunoprecipitation (ChIP) assays(Fig. 7B). High levels of expression of FLC in the fve mutant were correlated with H3 and H4 hyperacetylation at the FLC locus(He et al., 2003; Ausin et al., 2004). Chromatin of esd1-2, fve-1 and fve-1 esd1-2 plants was immunoprecipitated by using antibodies against acetylated H3, and PCR was used to amplify six DNA fragments spanning the promoter, the first exon and the first intron of FLC from the precipitated chromatin. For five out of the six probes assayed, FLC amplified sequences were consistently less abundant in DNA from precipitated chromatin of fve-1esd1-2double mutants than from chromatin of the fve-1 mutant plants(Fig. 7B). Thus, in fve-1 esd1-2 plants, FLC chromatin shows a reduction in acetylated H3 in comparison to the fve-1 mutant, indicating that ARP6 affects the levels of H3 acetylation of FLC. We concluded that ARP6 is required to activate FLC expression through a mechanism involving the histone acetylation of FLC chromatin. We extended this assay to other genetic backgrounds with high levels of FLC expression, such as fca-1 and FRI. For this, we focused our analysis on the FLC V and FLC IX probes, because they were among those that consistently showed the biggest effect of esd1 on the histone acetylation of FLC chromatin in a fve background. In agreement with previous data, we only detected very small changes in acetylated H3 in fca-1 and FRI backgrounds, when compared with those observed in the fve-1 mutant(He et al., 2003; Ausin et al., 2004). These differences were suppressed to a certain degree when fca-1 or FRI was combined with esd1(Fig. 7C).

Because esd1 mutations reduced FLC expression in the fca and FRI background as shown, we hypothesized that ARP6 might be required for other chromatin modifications, in addition to histone acetylation, that are involved in the regulation of FLC expression. To further explore this hypothesis, we examined if ARP6 has an effect on histone methylation at the FLC locus. It has been shown recently that H3-K4 hypertrimethylation is associated with actively transcribed FLCchromatin (He et al., 2004),being elevated in FRI-containing winter annuals and autonomous pathway mutants. Given the fact that esd1 mutations reduce FLC expression in these backgrounds, we wondered whether ESD1 was required for the elevated trimethylation of H3-K4 in FLC chromatin. Compared with wild-type plants, the trimethylated H3-K4 levels were elevated in a FRI-containing line and in autonomous pathway mutants, as reported previously(He et al., 2004; Kim et al., 2005)(Fig. 7D). Introduction of esd1 into FRI, fca and fve consistently eliminated the H3-K4 trimethylation increase in FLC chromatin associated with FRI and the autonomous pathway mutations(Fig. 7D). These data indicate that ESD1 is also required for the hypertrimethylation of H3-K4 in FLC chromatin.

esd1 mutants were selected from multiple screens for early flowering and accelerated phase change plants in Arabidopsis. Phenotypical analyses of these mutants revealed a complex pleiotropic phenotype affecting vegetative and reproductive development, together with a reduction in flowering time and phase length. The results of our genetic analyses revealed that the early flowering phenotype of esd1 mutants is almost completely epistatic over the flowering time delay caused by mutations in the autonomous pathway, and that esd1 partially suppresses the late flowering phenotype conferred by active FRIalleles (Fig. 3, Tables 2, 3). These epistatic effects correlate at the molecular level with a decrease in the steady state levels of FLC mRNA in lines carrying esd1 mutant alleles. Together,these results indicate that ESD1 is required for the expression of FLC. Thus, mutations of ESD1 behave like mutations at the EFS, ESD4, PIE1, ELF5, VIP3, ELF7, VIP4, VIP5 and VIP6/ELF8 loci (Soppe et al., 1999; Reeves et al.,2002; Zhang and Van Nocker,2002; Noh and Amasino,2003; Zhang et al.,2003; Noh et al.,2004; He et al.,2004; Oh et al.,2004; Kim et al.,2005), all of which are also required for high FLCexpression and flowering repression.

Additionally, the residual early flowering phenotype observed in esd1-3 flc-3 double mutants, especially under SD, indicates an additional role of ESD1 in the repression of flowering time that is independent of FLC. The most conservative hypothesis is that ESD1 is also required for the expression of FLC-related repressors, such as some of the MAF genes, which is consistent with our results showing a decreased expression of MAF1, MAF4 and MAF5, previously shown to play a role in flowering repression in Arabidopsis under certain environmental conditions(Scortecci et al., 2001;Ratclife et al., 2003). Finally, until triple and quadruple mutants carrying lesions at FLC, ESD1 and these MAF genes are analyzed, we cannot discard possible additional effects of ESD1 on flowering time through additional genes. The pleiotropic phenotype of esd1 mutants together with the broad expression pattern detected for this gene suggest that its function could be required in other developmental processes apart from flowering time.

Positional identification of the genomic region deleted in esd1alleles and the complementation of the esd1 phenotype by a genomic clone containing both At3g33520 and At3g33530 ORFs, together with the lack of genetic complementation between a T-DNA insertion line in At3g33520 and esd1-3, identified ESD1 as encoding ARP6. Recently, two publications have also described the characterization of early flowering mutants affected in the ARP6 gene, proposing its role in the maintenance of FLC expression and repression of flowering in Arabidopsis (Choi et al.,2005; Deal et al.,2005). ARP6 belongs to the actin-related protein family that shares moderate sequence homology and basal structure with actins. In Arabidopsis and rice, four divergent ARP classes (ARP4, ARP5, ARP6 and ARP9) are sequence homologs of ARPs, which are nuclear located in animals and fungi (McKinney et al.,2002; Kandasamy et al.,2004). Most of the nuclear ARPs are essential components of large multiprotein chromatin-modifying complexes(Blessing et al., 2004). The fruit fly ARP6-related protein ARP13E is associated with heterochromatin and may also play a role in chromatin structure(Frankel and Mooseker, 1996; Kato et al., 2001). ARP13E colocalizes with heterochromatin protein 1 (HP1)(Frankel et al., 1997), which is also linked to heterochromatin-mediated gene silencing and chromatin structure (Eissenberg and Elgin,2000). Moreover, in nuclei expressing mutant forms of HP1, the localization patterns of HP1 and dARP6 are altered in a parallel fashion(Frankel et al., 1997),implying that dARP6 interacts with HP1 directly or indirectly, and that they play a role in the organization of heterochromatin together. Mutants with a defect in an Arabidopsis HP1 ortholog, LIKE-HETEROCHROMATIN PROTEIN 1 (LHP1), also show an early flowering phenotype(Gaudin et al., 2001), raising the possibility that both proteins might be involved in the same chromatin-remodelling complexes in Arabidopsis.

Covalent modification of chromatin histones constitutes a code for maintaining states of gene activation and repression, and is a major component in the transcriptional regulation of FLC(Gendall et al., 2001; He et al., 2003; Ausin et al., 2004; Bastow et al., 2004; He et al., 2004; Sung et al.,2004). High levels of expression of FLC in autonomous pathway mutants are correlated with H3 and H4 hyperacetylation and trimethylation of H3-K4 at the FLC locus (He et al.,2003; Ausin et al.,2004; He et al.,2004; Kim et al.,2005). Furthermore, ARP6-like proteins have been found in other organisms as part of large protein complexes involved in chromatin remodelling(Krogan et al., 2003; Kobor et al., 2004; Mizuguchi et al., 2004). Because esd1 mutations suppress the late-flowering phenotype of fve mutants, and FVE represses FLC transcription through a histone deacetylation mechanism, we initially hypothesized that ESD1 could be required to activate FLC expression to levels that inhibit flowering, participating in chromatin remodelling complexes involved in histone acetylation of FLC chromatin. The lack of expression of GUS in esd1 fca plants expressing the FLC:GUS translational fusion, already suggested that if ESD1was required for active expression of FLC, this had to take place at the FLC sequences present in the construct used (promoter, first exon and first intron) (Sheldon et al.,2002). In fact, the results of ChIP experiments directed to that chromosomal region of FLC demonstrated that it is hypoacetylated in the esd1 fve mutant compared with the fve mutant(Fig. 7B). Thus, we conclude that ESD1 is required for histone acetylation at FLC, probably through its participation in HAT complexes. However, esd1 mutations also reduce both the late-flowering phenotype and FLC expression in FRI-containing lines and fca mutants, despite the fact that in these backgrounds the levels of acetylated H3 of FLC chromatin did not show significant changes in comparison to fve(Fig. 7C). This raised the possibility that ARP6 would participate in other mechanisms besides histone acetylation; our results indicate that the hypermethylation of H3-K4 in FLC chromatin is one of these mechanisms(Fig. 7D). It remains to be determined whether the effect of esd1 on the expression of other MAF genes takes place through similar mechanisms.

Our observation that ARP6 regulates the activation of FLCexpression by promoting both histone acetylation and methylation is consistent with a role for plant ARPs in chromatin-mediated transcriptional regulation. ARP4 is also likely to be involved in transcriptional regulation via chromatin remodelling, as it is a component of the human SWI/SNF and yeast INO80 complexes that are involved in chromatin remodelling, transcriptional regulation and DNA damage repair (Zhao et al., 1998; Shen et al.,2003). Other ARP4-containing complexes, such as yeast NuA4 and human TIP60, are suggested to have roles in chromatin-mediated epigenetic control of transcription through modifications of core histones(Galarneau et al., 2000; Ikura et al., 2000). Yeast Arp4 interacts with all four core histones(Harata et al., 1999), and recent findings have shown that Arp4 and Arp6 are also part of the Swr1 chromatin-remodelling complex, which catalyzes the exchange of conventional histone H2A for the histone H2A.Z variant in nucleosome arrays(Krogan et al., 2003; Kobor et al., 2004; Mizuguchi et al., 2004). These histone variants are involved in the regulation of gene expression and the establishment of a buffer to the spread of silent heterochromatin(Meneghini et al., 2003). Indeed, a human H2A.Z complex, equivalent to the yeast Swr1 complex has histone acetyl transferase activity, which might help to understand the role of ESD1 in histone acetylation(Owen-Hughes and Bruno, 2004). In the same way, the fact that components of the Swr1 complex were found to interact genetically with the PAF1 complex might explain the role of ESD1 in the trimethylation of H3-K4 in FLC chromatin(Krogan et al., 2002; Mueller and Jaehning, 2002; Squazzo et al., 2002; Krogan et al., 2003; Krogan et al., 2004). Like the yeast PAF1 complex, the PAF1-like complex in Arabidopsis may also recruit an H3-K4 methyl transferase to FLC to regulate its expression(Kim et al., 2005). Indeed,mutations in Arabidopsis homologs of the components of the PAF1 complex cause a decrease in the trimethylation of H3-K4 in FLCchromatin, and provoke early flowering and small leaves, similar to the esd1 mutation (He et al.,2004), raising the possibility that all of these genes are in the same pathway and regulate similar targets.

In agreement with the pleiotropic phenotype of esd1 mutants, the general pattern of expression of ESD1/ARP6 suggests that this gene is required in additional vegetative and reproductive developmental processes in which protein complexes harbouring ARP6 might play a relevant regulatory role. Given the molecular identity of ESD1, it seems reasonable to propose that loss-of-function alleles will cause a great effect on transcription,interfering with the expression of genes controlling various developmental pathways and thereby provoking changes in the morphology of different organs throughout the development of Arabidopsis. Among them, organ number in the perianth, which increases in esd1 mutants, is affected in a similar way in pie1 mutants. PIE1 encodes a protein similar to the ATP-dependent, chromatin remodelling proteins of the ISWI and SWI/SNF2 family, and it is a close homolog to the Swr1 ATPase, the core subunit of the yeast Swr1 complex that harbours Arp6(Mizuguchi et al., 2004). Loss of function of the PIE1 gene causes strikingly similar phenotypes to those of the esd1 mutant (Noh and Amasino, 2003), apart from the development of extra petals. In addition, pie1 mutations also cause early flowering and suppress FLC-mediated delay of flowering as a result of the presence of FRI or of mutations in autonomous pathway genes, suggesting that PIE1 and ARP6 may act in the same genetic pathways and might be part of the same protein complexes. However, in contrast to esd1 mutations, which suppress FLC expression in both SAM and RAM(Fig. 6), the effect of pie1 lesions is restricted to the shoot apex(Noh and Amasino, 2003),suggesting that the root tip expression of FLC requires ARP6and probably other root-expressed relatives of PIE1, and that the level of FLC expression in the shoot apex, but not in the root apex,influences flowering behaviour.

Recent analyses of knockdown AtARP4 expression in Arabidopsis have also revealed dramatic pleiotropic phenotypes, both similar to and entirely different from those of esd1/arp6(Kandasamy et al., 2005a). For example, silencing of the expression of ARP4 or loss of function of ARP6 caused early flowering; however, silencing of the expression of ARP4 but not ARP6 induced specific phenotypes, such as the altered organization of plant organs, delayed flower senescence and high levels of sterility (Kandasamy et al.,2005a), suggesting that both of these proteins may also be involved in the same and in different chromatin modifying complexes in Arabidopsis. Another ARP member, AtARP7 is required for normal embryogenesis, plant architecture, root growth and floral organ abscission(Kandasamy et al., 2005b), and may be also involved in chromatin-remodelling complexes.

In summary, our results demonstrate that ESD1/ARP6 is required for both FLC and FLC-like gene expression in the shoot and the root apex, and for the activity of a floral repressor pathway. The role of ESD1 in FLC regulation is to ensure competence for a high level of expression of this gene. We propose that ARP6 is required to activate FLC transcription through mechanisms involving both histone acetylation and methylation. We have determined that FLC, and maybe the FLC paralogs MAF1, MAF4 and MAF5 are targets of ARP6-containing chromatin-remodelling complexes, and that some components of the autonomous pathway might affect the activity of such complexes. Moreover, the pleiotropic phenotype observed for esd1mutants suggests a crucial role for the Arabidopsis ARP6 protein in the regulation of several leaf and flower development stages, probably through chromatin-mediated regulation of gene expression. Further functional studies,such as the identification of the proteins within ARP6-containing complexes,as well as the identification of additional genes regulated by these complexes, will help us to understand the crucial role of ARP6 in Arabidopsis development.

We are grateful to Dr George Coupland for providing us with esd1-1seeds and Dr Jean Finnegan for fca-1 seeds harbouring the FLC:GUS construct. We would like to thank to Dr Richard Amasino for kindly providing flc-3 and FRI-Col lines, to the ABRC and NASC for supplying the T-DNAs lines used in this study, to the GARNET for providing us with the binary TAC clones and to GENOSCOPE for the ARP6 cDNA clone. This work was supported by grants BIO2001-00391 to J.M.M.-Z. and BIO2002-00753 to J.A.J. from the Spanish Ministerio de Ciencia y Tecnología. A.L. is funded by a predoctoral fellowship from Direccion General de Investigación, Secretaria General de Política Científica y Tecnológica (Spain). We thank to all members of Jarillo's lab for helpful discussions.