Regulation of Flowering Time by Histone Acetylation in Arabidopsis

See allHide authors and affiliations

Science  05 Dec 2003:
Vol. 302, Issue 5651, pp. 1751-1754
DOI: 10.1126/science.1091109


The Arabidopsis autonomous floral-promotion pathway promotes flowering independently of the photoperiod and vernalization pathways by repressing FLOWERING LOCUS C (FLC), a MADS-boxtranscription factor that blocks the transition from vegetative to reproductive development. Here, we report that FLOWERING LOCUS D (FLD), one of sixgenes in the autonomous pathway, encodes a plant homolog of a protein found in histone deacetylase complexes in mammals. Lesions in FLD result in hyperacetylation of histones in FLC chromatin, up-regulation of FLC expression, and extremely delayed flowering. Thus, the autonomous pathway regulates flowering in part by histone deacetylation. However, not all autonomous-pathway mutants exhibit FLC hyperacetylation, indicating that multiple means exist by which this pathway represses FLC expression.

The developmental transition from a vegetative to a reproductive phase in Arabidopsis is genetically controlled by several pathways, which integrate the endogenous developmental state of the plant and environmental cues (1). The autonomous and the vernalization pathways independently regulate the floral transition by repressing FLC expression (24). There are six genes in the autonomous pathway: FLD, FCA, FPA, FY, FVE, and LUMINIDEPENDENS (LD) (1, 2). FCA and FPA encode RNA binding proteins (5, 6), FY encodes a protein involved in mRNA 3′ processing (7), and LD encodes a homeodomain-containing protein (8).

We have previously identified FLD as an autonomous-pathway gene by an analysis of fld-1 (9). The fld-1 mutant, which is in the Columbia (Col) accession, was crossed with the Landsberg erecta accession to create a segregating population for the map-based cloning of FLD (fig. S1). FLD consists of two exons (Fig. 1A) that encode a protein of 789 amino acid residues (FLD is locus At3g10390, see The fld-1 lesion is a point mutation (C to T) that converts Arg10 (CGA) into a premature stop codon (TGA). After the identification of FLD, two additional fld mutants (fld-3 and fld-4) were obtained from the Salk Institute Genome Analysis Laboratory collection (10). Both alleles are the result of transferred DNA (T-DNA) insertions in the coding region (Fig. 1A), and neither expresses the full-length FLD mRNA (Fig. 1B).

Fig. 1.

Structure and expression of FLD. (A) FLD gene structure. Triangles indicate T-DNA insertions; the T-DNA was located 1.6 kb downstream of the translational initiation codon in fld-3 and 0.8 kb downstream of the initiation codon in fld-4. (B) FLD mRNA accumulation in seedlings determined by RNA blot analysis. About 20 μg of total RNA extracted from 10-day-old seedlings was loaded in each lane. R10, Arg10; stop, stop codon.

All of the fld mutants flowered much later than wild-type Col in inductive long-day photoperiods (Table 1 and Fig. 2A) and flowered rapidly after exposure to a prolonged period of cold (vernalization) (Table 1), which is characteristic of autonomous-pathway mutants (2). The two T-DNA mutants flowered later than fld-1 (Table 1); thus, although the fld-1 lesion is a premature stop codon, this allele may have residual activity. fld-3 and fld-4 mutants flower later than any other autonomous mutant in the Col genetic background (Table 1) (11), and extensive aerial rosettes, which are often associated with extremely delayed flowering, are produced at the early nodes of the main inflorescence stem (Fig. 2, A and B).

Fig. 2.

Phenotype of fld mutants. (A) fld mutants flower later than wild-type plants in long days. (B) A single node on the main inflorescence stem of fld-3 showing an aerialrosette. (C) FLC dependence of fld phenotype. The fld-1 flc-3 double mutant flowers early like flc-3. (D) Abundance of FLC mRNA in fld, fve, fca, and FRISf2-Col [FRISf2-Col is a line in which a functional FRI allele from the San Feliu-2 (Sf-2) accession was introgressed into the Col background (3)]. About 20 μg of total RNA extracted from 10-day-old seedlings was loaded in each lane, and the blot was hybridized first with a FLC probe and subsequently with an 18S rDNA probe. (E) FLC::GUS transgene expression pattern in Col, fld-3, and FRI backgrounds. Arrowheads indicate stained tissues. Scale bars, 1 mm. (F) FLD::GUS (translation fusion) expression pattern in fld-1. Scale bar, 1 mm.

Table 1.

Flowering times of plants in the long-day condition. Values shown are mean numbers ± SD. Numbers in parentheses are the total numbers of plants evaluated. vern, 44-day vernalization.

Line Rosette leaf number Node numberView inline
Col 12.7 ± 1.2 (9) 2.9 ± 0.8
fld-1 60.4 ± 14.2 (8) 8.5 ± 0.6
fld-1 + vern 14.8 ± 1.7 (12) 3.9 ± 0.5
fld-3 93.5 ± 21.8 (19) 12.5 ± 4.6
fld-3 + vern 17.1 ± 1.1 (13) 5.5 ± 0.7
fld-4 86.0 ± 18.0 (20) 12.2 ± 0.9
FRISf2-Col 78.2 ± 8.8 (7) 10.8 ± 0.8
FRISf2-Col + vern 15.9 ± 1.3 (7) 4.4 ± 1.3
fve-4 40.8 ± 3.7 (6) 6.5 ± 1.0
fld-1 flc-3 12.1 ± 0.7 (7) 2.7 ± 0.9
flc-3 10.2 ± 1.1 (9) 2.1 ± 0.6
  • View inline* Total node number of the main inflorescence stem.

  • Steady-state FLC mRNA levels are much higher in the fld mutants than in the parental Col line (Fig. 2D). Increased FLC expression is characteristic of other autonomous mutants such as fca and fve, and lines containing FRIGIDA (FRI), which is an activator of FLC that confers the vernalization-responsive late-flowering habit in many winter-annual accessions of Arabidopsis (12) (Fig. 2D). In addition, the late-flowering phenotype of fld is completely suppressed by the loss of FLC; the fld-1 flc-3 double mutant flowers like flc-3 in long days (Table 1 and Fig. 2C). Thus, the role of FLD in flowering-time control is to repress FLC expression.

    To examine the FLC spatial expression pattern in fld mutants, a translational FLC::GUS fusion (13) was introduced into fld-3. FLC was preferentially expressed in shoot and root apical regions, which are enriched in dividing cells (Fig. 2E). This FLC expression pattern is identical to that of FLC::GUS in a FRI-containing line (Fig. 2E). Thus, an fld lesion causes FLC expression to increase in the same spatial pattern as that caused by FRI in winter annuals. As might be expected for a regulator of FLC, the pattern of FLD expression is identical to that of FLC, as determined by a translational FLD::GUS fusion (Fig. 2F) (13) and by reverse transcription polymerase chain reaction analyses of the native mRNAs (11).

    FLD is a plant homolog of the human protein KIAA0601 (14); FLD is 42% identical to KIAA0601 over a region of 568 amino acids, and the similarity between these two proteins over the same region is about 60% (fig. S2). Much of FLD and KIAA0601 (14) consist of a region that is similar to human polyamine oxidase 1 (HsPAO1) (15), maize polyamine oxidase (ZmPAO) (16), and a predicted Arabidopsis polyamine oxidase (AtPAO, At5g13700; see (Fig. 3A and fig. S2). Crystal structure–derived residues of ZmPAO that are involved in binding the cofactor FAD (16) are conserved in all five proteins (fig. S2). However, FLD and KIAA0601 contain an additional region—recently referred to as a SWI3p, Rsc8p, and Moira (SWIRM) domain (17)—in the N-terminal region (Fig. 3A and fig. S2). SWIRM domains are found in a range of proteins involved in chromatin remodeling (17). The joining of a SWIRM domain to a polyamine oxidase is found in FLD, two related genes in Arabidopsis (At1g62830 and At3g13682, see, and three related genes in rice (fig. S3) (11). In human, mouse, and Drosophila, only one such gene in each has been identified to date (KIAA0601, LOC230843, and CG17149, respectively) (fig. S3), and two have been identified in Caenorhabditis elegans (18) (fig. S3).

    Fig. 3.

    Domain architecture of FLD and acetylation of FLC chromatin. (A) Domain architectures of FLD and related proteins. FLD and Hs-KIAA0601 consist of a SWIRM domain joined to a region similar to maize and human polyamine oxidases (ZmPAO and HsPAO1). GenBank accession numbers: NM_111874, AtFLD; BC048134, Hs-KIAA0601; CAA05249, ZmPAO; AAK55763, HsPAO1. (B) ChIP analyses of the acetylation state of histone H4 in FLC chromatin. The input is Colchromatin before immunoprecipitation. No AB, control sample lacking antibody; co, a late-flowering mutant in the photoperiod pathway, in which FLC expression level is not changed (26), that served as a control. ACTIN served as an internalcontrol of the ChIP analysis. The fold enrichment of a mutant compared to Colwas calculated as follows: FLC and ACTIN in the mutant sample were first normalized to FLC and ACTIN in the Col sample; the normalized FLC was then divided by the normalized ACTIN to obtain the fold enrichment in the mutant. The fold enrichment of mutants over Col is shown.

    The human protein KIAA0601 is a component of human histone deacetylase 1,2 (HDAC1/2) co-repressor complexes (14, 19). These complexes are involved in initiating repression of gene expression by deacetylation of histone residues (19). However, whether KIAA0601 is required for the deacetylation activity of this complex is not known. To explore whether FLD has a role in deacetylation, we examined the acetylation state of histone H4 at the FLD target, FLC, by chromatin immunoprecipitation (ChIP) (13). Compared with acetylation of FLC chromatin in the wild-type plant, H4 in fld is hyperacetylated (Fig. 3B). The hyperacetylation occurs in the region close to the transcription initiation point, the first intron, and in a region of the second intron (fig. S4). The acetylation difference is not detectable in a region about 2 kb 5′ to the translation initiation point or in a region of the last exon (fig. S4C). Thus, FLC hyperacetylation in fld occurs in a defined region similar to that reported for another plant gene (20).

    In contrast to the situation in fld, there is no change in FLC acetylation in several other mutants of autonomous-pathway genes (fca, fpa, and ld) or in a FRI-containing line (Fig. 3B and fig. S4). However, like fld, the autonomous-pathway mutant fve exhibits an increase in FLC acetylation, although this increase is not as large as that in fld (Fig. 3B). Consistent with the weaker hyperacetylation phenotype of fve-4 compared with fld, fve-4 is also not as delayed in flowering as are the fld mutants (Table 1). The weaker flowering and acetylation phenotypes of fve might result from redundancy of FVE function in the Arabidopsis genome or partial activity of fve-4 allele. These data indicate that both FLD and FVE are involved in FLC repression by histone deacetylation, perhaps as components of a HDAC co-repressor complex.

    The FLC hyperacetylation phenotype of fld indicates that FLD is required for the ability of certain HDACs to regulate FLC. The Arabidopsis genome contains four homologs of human HDAC1/2 (AtHDA1, AtHDA6, AtHDA7, and AtHDA9; see However, AtHDA mutants that have a phenotype similar to fld have not been found. Thus, either the HDAC that participates in FLC regulation with FLD is also involved in other processes and a mutation in it is highly pleiotropic or lethal, or there is redundancy among the Arabidopsis HDACs such that no single mutant will have a large effect on flowering time. Suppression of AtHDAs by an antisense construct that is likely to have partially suppressed the expression of several HDACs does cause delayed flowering (21).

    HDAC complexes are typically recruited to target genes through an interaction with cis-regulatory DNA elements. For example, the KIAA0601-CoREST HDAC1/2 co-repressor complex represses neuron-specific genes in nonneuronal cells by means of the cis-regulatory element known as repressor element 1 (22). To determine whether a specific region of the FLC locus is required for histone deacetylation, we created a series of internal deletions in the region 5′ of the translation start codon and the first intron of FLC because FLC expression is regulated by these regions (23) (Fig. 4A), and we introduced these constructs into the flc-3 null mutant in the Col background. Deletion of a region required for FLC deacetylation might produce a phenocopy of the fld mutant. Indeed, deletion of a 295–base pair region of Intron I caused extremely late flowering in transgenic plants (Fig. 4, B and C), and in these transgenic lines, this FLC transgene was both highly expressed (Fig. 4D) and hyperacetylated (Fig. 4E). Deletions in other regions of FLC did not cause hyperacetylation of FLC transgenes (Fig. 4E) or an extreme delay of flowering (11). Thus, a specific region in the first intron of FLC is required for deacetylation.

    Fig. 4.

    Deletion of a region (Δ295) in the first intron of FLC phenocopies the fld phenotype. (A) Internal deletions in the FLC genomic region. The translation initiation point is +1. The filled boxes represent exons, and the open boxes represent introns. (B) Flowering times of first transformed generation (T1) transgenic plants derived from construct –2362 (a fully functional FLC transgene). (C) Flowering times of T1 transgenic plants derived from intron deletion Δ295. (D) Abundance of FLC mRNA in T1 transgenic plants derived from the construct –2362 and the Δ295 deletion. (E) ChIP analyses of the acetylation state of histone H4 in FLC transgene chromatin. The input is the chromatin of flc-3 transformed with construct –2362 before immunoprecipitation. No AB, control sample lacking antibody. The transformant selection marker gene NPTII served as an internal control for the ChIP analyses. The fold enrichment (calculated as described in Fig. 3B legend) of the deletions compared with the –2362 construct is shown.

    The modification of the level of acetylation in FLC chromatin suggests a mechanism by which certain autonomous-pathway genes regulate FLC expression. The autonomous pathway was originally defined as a set of genes with a specific mutant phenotype: vernalization and photoperiod-responsive late flowering (24). Genetic analyses indicate that the autonomous pathway may in fact be composed of genes that control flowering by more than one mechanism (25). Our observation that hyperacetylation of FLC chromatin is only observed in fld and fve mutants supports a model in which FLD and FVE regulate FLC expression by a mechanism distinct from other autonomous-pathway genes. Given the centrality of FLC in flowering-time control, it is not surprising that FLC is subject to multiple independent regulators.

    Supporting Online Material

    Materials and Methods

    Figs. S1 to S4


    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article