One example of signal convergence of multiple floral induction pathways is the control of FLC expression. FLC is a MADS-box transcription factor that acts as a negative regulator of floral initiation and an integrator of the autonomous and vernalization pathways (16, 17). Another level of signal convergence for floral induction pathways is the regulation of SOC1 and FT expression. SOC1, a MADS-box transcription factor, and FT, a RAF-kinase inhibitor-like protein, are positive regulators of the expression of floral meristem identity genes and floral initiation (18-22). The expression of SOC1 and FT is negatively regulated by FLC but positively regulated by CO, which encodes a B-box zinc-finger transcription factor (6, 20). The autonomous and vernalization pathways both suppress the expression of FLC, resulting in a decreased expression of FLC and consequently increased expression of SOC1 and FT in a later developmental stage or after a prolonged exposure of plants to low temperature (14, 22-25). The photoperiodic pathway, acting through many components, including photoreceptors, the circadian clock, and transcription regulators, positively regulates expression of SOC1 and FT to stimulate flowering in Arabidopsis when the day length increases (4, 6, 20, 21, 26).

Transcription regulation is one mechanism underlying the control of flowering time and the interaction between different floral induction pathways. It has been shown that FLC binds to the CArG element in the promoter of SOC1 to antagonize the positive effect of CO at a separate cis-element of the SOC1 promoter (6). Another example of different signaling pathways converging at transcription regulation is in the control of LFY expression by the photoperiodic pathway and the gibberellin pathway. These two pathways have been shown to control LFY mRNA expression through different cis-elements of the LFY promoter (12). In addition to transcription controls, RNA processing has also been shown to play important roles in the regulation of flowering-time gene expression. Three (FCA, FY, and FPA) of the six (FCA, FY, FPA, FLC, LD, and FVE) classic autonomous pathway genes encode proteins that either contain RNA recognition motifs or interact with a RNA-binding protein (27-29). FCA expression is autoregulated by means of alternative utilization of polyadenylation sites in its premRNA. The FCA/FY complex interacts with the FCA premRNA to promote the selection of the promoter-proximal polyadenylation site, resulting in accumulation of the alternatively spliced FCA β-mRNA that encodes an inactive protein and inhibition of the production of FCA γ-mRNA that encodes the active FCA protein (29). The major function of FCA, FY, and FPA in floral initiation is to suppress FLC expression, but how these RNA-binding proteins affect FLC mRNA abundance remains unclear (3, 30).

The gene corresponding to the mutations was identified based the analysis of DNA sequences flanking the T-DNA inserts. This gene encodes a 577-residue protein that contains three KH motifs (Fig. 2A and Fig. 6A, which is published as supporting information on the PNAS web site). Because none of the four mutant alleles showed readily discernable morphological changes other than delayed flowering, we reasoned that the major function of the corresponding gene is to regulate flowering time and referred to it as FLK. The T-DNAs of the two flk alleles (flk-1 and flk-2) were inserted in the first intron, the T-DNAs of other two alleles (flk-3 and flk-4) were inserted in the second intron (Fig. 2A). It appears that all of the flk alleles identified are loss-of-function or null mutations, because none of the flk alleles tested expressed detectable amounts of FLK mRNA (Fig. 2B and data not shown). In the wild-type plant, FLK is expressed in various tissues, including flowers, leaves, roots, and siliques, and it is relatively abundant in the young inflorescence (http://mpss.udel.edu/at/java.html). The FLK gene and flk mutant have also been reported recently by others (34).

The KH domain is an evolutionarily conserved RNA-binding motif found in proteins of diverse organisms, including eubacteria, archaea, and eukaryotes (31, 38). The core consensus sequence V/IIGXXGXXI/V in the middle of the KH domain (≈50-70 residues) is perfectly conserved in all three KH domains of FLK (Fig. 6A). This core consensus sequence has been found to be important for RNA-binding activity of KH domain proteins. For example, a single amino acid substitution in the KH domain core sequence of the human FMR1 protein abolishes its RNA-binding activity and causes fragile X mental retardation syndrome (39-41). The first two KH domains of FLK are grouped near the N terminus, and the third KH domain is located at the C terminus (Figs. (Figs.2A2A and 6A). Such an overall architecture of KH domain arrangement is conserved in the PCBP (polyC-binding protein) type of KH domain RNA-binding proteins, including the founding member of the KH domain protein family, heterogeneous nuclear ribonucleoprotein K (38). The N terminus of FLK preceding the first KH domain is rich in glutamine (33/150 residues or 22%). This region also contains three perfect 8-residue repeats (LEPQQYEV), although these repeats are not strictly conserved in the putative rice ortholog of FLK (Fig. 6A).

To study the function and cellular localization of the FLK protein, we prepared transgenic plants overexpressing the GFP-FLK fusion protein under the constitutive 35S promoter (see Supporting Materials and Methods). The 35S::GFP-FLK transgenic plants flowered earlier than the wild type (Fig. 2C). Because the loss-of-function flk mutant flowered later and 35S::GFP-FLK transgenic plants flowered earlier, we concluded that FLK is a positive regulator of floral initiation. The GFP-FLK fusion protein was enriched in the nucleus, although it was also found in the cytosol (Fig. 2D). The nuclear localization of GFP-FLK is consistent with a proposition that FLK may be involved in the regulation of flowering-time genes.

We investigated whether the flk mutation affects gene expression by analyzing genome-wide expression profiles of three different samples: 16-day-old seedlings grown in long days and harvested 16 h after light-on (Exp. 1), 16-day-old seedlings grown in continuous light (Exp. 2), and 7-day-old seedlings grown in long days (18 h light/6 h dark) but harvested 1 h after light-on (Exp. 3). Affymetrix (Santa Clara, CA) ATH1 genechip arrays representing ≈24,000 Arabidopsis genes were used in all three experiments. The complete DNA microarray results of the three experiments can be found online (www.ncbi.nlm.nih.gov/geo, accession no. GSE1512). We also used Q-PCR and RT-PCR to reexamine the gene expression changes detected by the DNA microarray studies. Results of the three microarray experiments are summarized in Table 1, for which three statistical criteria (see Table 1) were used to determine a putative misexpression event in the flk mutant. Although a number of genes showed altered expression in at least one microarray experiment (referred to as one misexpression event and marked by asterisks in Table 1), only eight genes showed altered mRNA expression in the flk mutant in all three microarray experiments (Table 1, type A, and www.ncbi.nlm.nih.gov/geo). The two genes for which the expression showed the most profound change in the flk mutant are FLC and FLK. The level of FLK expression is close to the background level in the flk mutant as demonstrated by RT-PCR (Fig. 2B) and Q-PCR (see Fig. 4A). The mRNA level of FLC was ≈6- to 10-fold higher in the flk mutant than in the wild type in all three microarray experiments. Increased FLC expression in the flk mutant was confirmed by both RT-PCR (Fig. 3A) and Q-PCR (see Fig. 4A). Because KH domain proteins are often associated with RNA metabolism, we examined whether the elevated FLC mRNA level in the flk mutant was due to a defect in FLC mRNA turnover (Fig. 3B). In this experiment, tissues were excised and incubated with transcription inhibitors (cordycepin, cycloheximide, and actinomycin D), and the FLC mRNA level was examined at different time points after the inhibitor treatment. Fig. 3B shows that FLC mRNA level decreased gradually in tissues treated with transcription inhibitors in both the wild type and flk mutant, but there was no obvious difference of the rate of FLC mRNA decay between the two genotypes. This result suggests that FLK may regulate FLC mRNA expression by means of a mechanism other than RNA turnover. The FLK protein expressed and purified from the in vitro translation system or from Escherichia coli also failed to bind FLC mRNA or polyribonucleotides in various conditions tested (X.Y. and C.L., unpublished data).

In addition to FLK, two other KH domain genes also showed altered expression profiles in the flk mutation in all three microarray experiments (Table 1). The mRNA level of one KH domain gene (At5g06770) was ≈100% higher in the flk mutant than in the wild type, whereas the other KH domain gene (At3g32940) showed ≈30-40% decreased expression in the flk mutant in all three microarray experiments (Table 1, type A). In contrast, the expression of other KH domain genes, including HEN4, which regulates floral organ development, showed normal expression in the flk mutant in all microarray experiments (Table 1, type A). The modest misexpression of At5g06770 was confirmed by RT-PCR, in which this KH domain gene showed modestly elevated expression in three different flk mutant alleles tested (Fig. 7, which is published as supporting information on the PNAS web site). A phylogenetic analysis indicates that at least one of the two KH domain genes affected by the flk mutation is closely related to FLK, although the sequence similarity between these KH domain proteins are not particularly strong (Fig. 6B and C). The function of neither gene is known. Because the expression of these two KH domain genes is affected by the flk mutation, it is speculated that their functions may be associated with FLK. The functions of the other four genes listed in Table 1 and their relationship with FLK are not clear.

Results of the expression comparison for genes that play roles in the autonomous pathway controlling flowering time or genes that regulate floral development, such as floral organ identity, are summarized in Table 1, type B. The expression profiles of genes known or likely to be involved in photoperiodic control of flowering time are listed in the Table 1, type C. Table 1, type B, shows that, with few exceptions, genes associated with autonomous pathway or floral development generally were expressed normally in the different samples analyzed in all three microarray experiments. For example, two autonomous pathway genes involved in RNA metabolism and regulation of FLC expression, FCA and FY, showed no change of expression in the flk mutant in all three microarray experiments (FPA is not represented on the ATH1 array). The expression of other genes that are known to regulate FLC expression, such as FRI, FLD, FVE, VRN1, and VRN2, were also not apparently affected by the flk mutation. Genes known for their roles in the regulation of floral organ development, such as AGAMOUS, AP1, AP2, CLV1, and SUP, also demonstrated normal expression in the flk mutant in all three microarray experiments (Table 1, type B).

In contrast to the genes associated with autonomous pathway or floral development (Table 1), the expression profiles of genes associated with photoperiodic regulation of flowering time showed more complicated patterns in the flk mutant (Table 1, type C). For example, among the 26 genes listed in Table 1, type B, only two putative misexpression events (marked by asterisks in Table 1) were detected according to the selection criteria (see Table 1). In contrast, a total of 15 putative misexpression events were detected among 13 genes listed in Table 1, type C. At least one putative misexpression event was detected for almost every gene listed in Table 1 (and 4 of 13 genes showed misexpression in two microarray experiments). Among the genes listed in Table 1, type C, FT and SOC1 are known to be positively regulated by CO and negatively regulated by FLC in response to different signals (3-5). One or two putative misexpression events were detected for the FT or SOC1 genes, respectively, both showing decreased expression in the flk mutant (Table 1, type C). The expression of FT and SOC1 were reexamined with RT-PCR or Q-PCR analyses. Fig. 3A shows that FT and SOC1 both were expressed at lower levels in the flk mutant in samples collected from 12 to 22 days after germination. A modestly decreased mRNA accumulation of FT and SOC1 was also demonstrated by using a Q-PCR assay (Fig. 4A). Because FLC expression was significantly increased in the flk mutant, our results are consistent with FLC being a negative regulator of the mRNA expression of FT and SOC1, resulting in delayed flowering. We concluded that FLK is a negative regulator of FLC expression that suppresses FLC expression to derepress floral initiation.

As described previously, most genes involved in photoperiodic regulation of flowering time registered putative misexpression events in at least one of the three microarray experiments, although the expression changes were modest (Table 1, type C). It is noted that most of the putative misexpression events were detected in 16-day-old samples grown in continuous light (Table 1, type C, Exp. 2), but not in 7-day-old samples grown in long days and collected 1 h after light-on (Table 1, type C, Exp. 3). Most of the gene expression changes detected in the 16-day-old flk mutant seedlings were confirmed by Q-PCR experiment (Fig. 4). For example, AGL24, APRR3, APRR5, APRR7, and CO genes each registered at least one decreased expression in the flk mutant (Table 1, type C) and similarly decreased expression of these genes were detected in the Q-PCR experiment (Fig. 4A). In contrast, CCA1 and LHY showed slightly increased expression in the 16-day-old flk mutant in both microarray and Q-PCR experiments (Table 1, type C, and Fig. 4A). This discrepancy does not seem to be due to the developmental difference between the wild type and the flk mutant that flower later, because few of the 26 genes, which are associated with autonomous/vernalization control of flowering time or floral development and are also known to increase expression at later developmental stages, showed similar bias (Table 1, type B, Exp. 2). Moreover, the genes that showed decreased expression in the flk mutant, such as CO, FT, SOC1, APRR5, and APRR7, are known positive regulators of floral initiation (20, 42, 43), whereas the genes that showed increased expression in the flk mutant, such as CCA1 and LHY, are known negative regulators of flowering (44, 45). In both cases, the function of those genes for which the expression was modestly altered by the flk mutation correlated with the delayed flowering phenotype of flk. This analysis again suggests that the misexpression of these genes in the flk mutant may not be due to random experimental variations. The function of RVE2 is not known, but it is a MYB protein related to CCA1 and LHY. Moreover, a recently reported protein called EPR1 that shares ≈57% amino acid similarity to RVE2 was found to suppress flowering (46).