FKF1 F-Box Protein Mediates Cyclic Degradation of a Repressor of CONSTANS in Arabidopsis

See allHide authors and affiliations

Science  08 Jul 2005:
Vol. 309, Issue 5732, pp. 293-297
DOI: 10.1126/science.1110586


The temporal control of CONSTANS (CO) expression and activity is a key mechanism in photoperiodic flowering in Arabidopsis. FLAVIN-BINDING, KELCH REPEAT, F-BOX 1 (FKF1) protein regulates CO transcription, although the molecular mechanism is unknown. We demonstrate here that FKF1 controls the stability of a Dof transcription factor, CYCLING DOF FACTOR 1 (CDF1). FKF1 physically interacts with CDF1, and CDF1 protein is more stable in fkf1 mutants. Plants with elevated levels of CDF1 flower late and have reduced expression of CO. CDF1 and CO are expressed in the same tissues, and CDF1 binds to the CO promoter. Thus, FKF1 controls daily CO expression in part by degrading CDF1, a repressor of CO transcription.

Photoperiodism manages a plant's flowering in response to changes in day length (13) through control of CONSTANS (CO) protein expression and activity (47). The protein FLAVIN-BINDING, KELCH REPEAT, F-BOX 1 (FKF1) governs the daytime expression pattern of CO (6) and is thought to be a component of a Skp1-Cullin-F-box (SCF) E3 ubiquitin ligase complex (810). This raises the possibility that FKF1 controls CO expression by degrading a regulator of CO. For most Arabidopsis F-box proteins, the C-terminal domains contain specific binding sites for substrates targeted for ubiquitination (11). With the use of a yeast two-hybrid screen to isolate proteins interacting with the FKF1 C-terminal kelch repeats, we identified three Dof transcription factors (12) (Fig. 1A). These Dof proteins also interacted with the LOV KELCH REPEAT PROTEIN 2 (LKP2) kelch repeats but not with the ZEITLUPE (ZTL) kelch repeats (Fig. 1A). Because the expression levels of the three Dof transcripts oscillated in constant light conditions (Fig. 1B), we named these genes CYCLING DOF FACTOR 1 (CDF1), CDF2, and CDF3 (Arabidopsis Genome Initiative gene codes At5g62430, At5g39660, and At3g47500, respectively) (fig. S1). To verify the interactions detected in yeast, we carried out binding studies using recombinant glutathione S-transferase (GST)–FKF1 and –LKP2 kelch fusions and the three His/T7-CDF proteins. The results of the in vitro pull-down assays were similar to those of the two-hybrid assays (Fig. 1C), validating the interactions observed in yeast.

Fig. 1.

FKF1 binds to the CDF1 protein, which affects flowering time. (A) Interaction between the FKF1, LKP2, and ZTL kelch domains and the CDF1, CDF2, and CDF3 proteins in yeast. Bait, a hybrid protein containing a DNA binding domain; prey, a hybrid protein containing an activation domain (18). (B) Expression patterns of CDF1, CDF2, and CDF3 transcripts under constant light conditions (LL). Wild-type plants were entrained under 12 hours light and 12 hours dark conditions for 10 days and then released to the LL condition. The data show the mean ± SEM of quantified expressions of three experiments. (C) In vitro pull-down assay between FKF1, LKP2, and ZTL kelch domains and CDF proteins. (D) Flowering time phenotype of the wild type (WT), CDF1-overexpressing lines (35S::CDF1), and fkf1 and lkp2 mutants under LD and SD conditions. Data are mean ± SEM of 16 plants.

If FKF1, LKP2, or both are involved in the degradation of CDF proteins in vivo, overexpression of these CDF proteins might produce flowering phenotypes similar to those of fkf1 and/or lkp2 loss-of-function mutant plants. Plants overexpressing CDF1 (35S::CDF1 plants) (fig. S3) flowered later than wild-type plants in long-day (LD) conditions (Fig. 1D). In contrast, overexpression of either CDF2 or CDF3 did not change flowering time under long- or short-day (SD) conditions (13). The flowering phenotype of the lkp2 mutant (fig. S2) resembled that of wild-type plants (Fig. 1D). Because the fkf1 mutant, but not the lkp2 mutant, exhibited a late-flowering phenotype similar to that of the 35S::CDF1 lines, this raised the possibility that FKF1 might be responsible for regulating CDF1 protein turnover in vivo.

To characterize the flowering time phenotype of the cdf1 mutants, we used double-stranded RNA interference (RNAi) to reduce CDF1 expression in wild-type and fkf1 plants. CDF1 expression in CDF1 RNAi (cdf1-R) lines (wild-type and fkf1) was reduced to less than 15% of that observed in nontransgenic lines (fig. S3). The cdf1-R lines flowered earlier than wild-type plants in LD conditions (fig. S4), indicating that CDF1 may contribute to the regulation of flowering time. In fkf1 cdf1-R lines, reduction of CDF1 expression also resulted in early flowering but did not completely rescue the fkf1 late-flowering phenotype (fig. S4). This suggests that CDF1 is not the sole effector of FKF1 control over flowering time, athough the CDF1 gain-of-function experiments indicate a genetic interaction between FKF1 and CDF1.

To determine whether FKF1 regulates CDF1 protein stability in vivo, we examined the expression of CDF1 protein in the wild type and in the fkf1 mutant. We used transgenic plants constitutively expressing hemagglutinin (HA)–tagged CDF1 (35S::HA-CDF1) to distinguish between transcriptional and post-transcriptional control of CDF1 protein levels. Plants harboring the 35S::HA-CDF1 construct exhibited a similar late-flowering phenotype to those with 35S::CDF1 (13). Although the expression of the CDF1 transcripts was constitutively high in both LD and SD conditions, HA-CDF1 protein levels were diminished from zeitgeber time (ZT) 10 to ZT22 in LD and ZT7 to ZT19 in SD conditions (Fig. 2, A and B), indicating that CDF1 protein abundance is post-transcriptionally regulated. Consistent with this notion, 35S::HA-CDF1 plants treated with the proteasome inhibitor MG-132 from ZT8 to ZT16 exhibited elevated levels of HA-CDF1 proteins at ZT16 (fig. S5), demonstrating that the stability of CDF1 is controlled by the proteasome-dependent pathway. In the fkf1 mutant background, HA-CDF1 protein was more stable between ZT10 to ZT22 in LD and ZT7 to ZT19 in SD conditions, compared to the wild type (Fig. 2, C and D, and fig. S6). These results indicate that FKF1 functions as an E3 ubiquitin ligase involved in CDF1 protein turnover.

Fig. 2.

FKF1 regulates the stability of the CDF1 protein. (A to J) Representative images of a Western blot of HA-CDF1 protein profile and quantification of CDF1 transcript expression and HA-CDF1 protein expression, under LD and SD conditions, in [(A) and (B)] 35S::HA-CDF1, [(C) and (D)] fkf1/35S::HA-CDF1, [(E) and (F)] CDF1::HA-CDF1, [(G) and (H)] fkf1-2/CDF1::HA-CDF1, and [(I) and (J)] CDF1::HA-CDF1 K253A plants. The expressions of CDF1 transcripts are the sum of both native CDF1 gene expression and HA-CDF1 gene expression. An anti-hemagglutinin antibody cross-reacting band (∼30 kD) was used as a loading control. The mean ± SEM of quantified values of three independent experiments are shown.

Because the 35S::HA-CDF1 construct does not reflect the endogenous expression pattern of CDF1, we used an HA-CDF1 cDNA driven by the CDF1 promoter (CDF1::HA-CDF1) to analyze the native expression of CDF1. CDF1::HA-CDF1 (24) and fkf1-2/CDF1::HA-CDF1 (29) lines with similar CDF1 transcript levels (fig. S3) were used in our analysis. In both lines, HA-CDF1 protein levels remained high in the morning in both LD and SD conditions. In the CDF1::HA-CDF1 line, CDF1 protein decreased rapidly to undetectable levels between ZT7 and ZT10 in LD and between ZT4 and ZT10 in SD conditions (Fig. 2, E and F). By contrast, we observed a more gradual decrease in CDF1 protein abundance in the fkf1-2/CDF1::HA-CDF1 line during these time periods (Fig. 2, G and H). Because FKF1 protein begins to accumulate during this part of the day, the data suggest that FKF1 mediates the acute reduction of CDF1 protein observed in the CDF1::HA-CDF1 line.

To further verify that FKF1 is directly involved in the degradation of CDF1, we analyzed the in vivo protein stability of a CDF1 variant harboring a point mutation (K253A) that attenuates the binding to FKF1 in yeast (fig. S6). When the CDF1 K253A transcript was driven by the constitutive 35S promoter, the CDF1 K253A protein was more stable in transgenic lines than nonmutated CDF1 protein at ZT13 (the time of the peak expression of FKF1) (fig. S6). In addition, when the CDF1 K253A cDNA was introduced in plants with a CDF1 minigene context (CDF1::HA-CDF1 K253A plants) (fig. S3), the expression profiles of CDF1 K253A protein resembled those of CDF1 protein expressed in fkf1 mutants (Fig. 2, I and J). These results indicate that the physical interaction with FKF1 renders the CDF1 protein unstable in vivo. We also examined whether FKF1 controls CDF1 transcription and determined that CDF1 expression in fkf1 was similar to that in wild-type plants (fig. S3), indicating that FKF1 is not likely a regulator of CDF1 transcription. Together with the data demonstrating that FKF1 regulates HA-CDF1 protein stability, these results suggest that endogenous CDF1 abundance is higher in the fkf1 mutant compared to the wild type, especially in late afternoon.

FKF1 is involved in the regulation of proper daytime CO expression (6), and CO transcripts are expressed in specific tissues (14, 15). If CDF1 is a transcription factor involved in the regulation of CO expression, the spatial pattern of CDF1 expression is likely to overlap with that of CO. To examine the subcellular localization of CDF1 protein, we generated plants expressing a CDF1–yellow fluorescence protein (YFP) fusion and determined that the protein was localized in the nucleus (Fig. 3). We then analyzed the spatial expression pattern of CDF1 using a CDF1 promoter–driven β-glucuronidase (GUS) gene construct (CDF1::GUS). CDF1::GUS activity was observed mainly in the vascular tissues of cotyledons, leaves, and hypocotyls, as well as in stomata, and the staining patterns were similar between LD- and SD-grown plants (Fig. 3). No GUS activity was detected in the roots under either growth condition (13). FKF1::GUS is expressed in whole leaves and strongly in stomata (16), whereas CO::GUS is expressed in vascular tissues (14, 15). The overlapping expression patterns of FKF1, CDF1, and CO support the notion that these genes are involved in the same regulatory pathway.

Fig. 3.

CDF1 is a nuclear protein and expresses mainly in vascular tissues. (A) Subcellular localization of CDF1-YFP protein in root. (B) Nuclear staining of the same cell shown in (A). Scale bars in (A) and (B), 10 μm. (C to F) Staining of GUS activity in CDF1::GUS plants. CDF1::GUS plants were grown under LD or SD conditions. Scale bars in (C) to (E), 2 mm. Scale bar in (F), 50 μm.

Considering that CDF1 may be involved in the regulation of CO expression, we examined CO and FLOWERING LOCUS T (FT) expression in both CDF1 gain- and loss-of-function mutants. In 35S::CDF1 plants, CO expression was suppressed at all ZT times under both LD and SD conditions (Fig. 4A and fig. S7), indicating that CDF1 functions as a transcriptional repressor of CO. FT transcript was not induced under either LD or SD conditions in 35S::CDF1 plants (fig. S8), which is likely due to the severely depressed CO levels. In cdf1-R lines, both the CO and FT expression patterns were unaltered (figs. S7 and S8), suggesting the existence of an additional suppressor(s) of CO. In fkf1 cdf1-R lines, we observed an increase in CO expression under LD conditions from ZT10 to ZT19 (Fig. 4B and fig. S7), the time period in which CDF1 protein is more abundant in the fkf1 mutant versus wild-type plants. In addition, the slight increase in CO transcription in fkf1 cdf1-R lines under LD conditions appeared to be sufficient to elevate the peak level of FT expression, a sensitive readout for the level of CO induction (fig. S8). These results support the notion that CDF1 is a suppressor of CO. However, the daytime peak of CO, which is present in the wild type, was not observed in the fkf1 cdf1-R plants; thus, FKF1-mediated degradation of CDF1 is not the only mechanism by which FKF1 controls the daytime CO expression.

Fig. 4.

CDF1 functions as a CO repressor and binds to specific sequences in the CO promoter. (A to D) CO expression under LD conditions in (A) wild-type and 35S::CDF1 plants, (B) fkf1 and fkf1 cdf-R plants, (C) wild-type and CDF1::HA-CDF1 plants, and (D) wild-type and CDF1::HA-CDF1 K253A plants. Representative blotting images of individual transgenic lines and the mean ± SEM of quantified values of blotting images derived from three independent experiments are shown in all panels. UBQ, ubiquitin (46). (E) The sequences of the probe and the competitors for DNA binding assay. The Dof binding motifs are shaded in black, and the nucleotide changes in the mutated fragments are shaded in gray. The sequences of the complementary strands are shown for the –173/–135 fragments. (F and G) Sequence specificity of CDF binding. We used (F) two sequences present in the CO promoter or (G) two mutated sequences as cold competitors. The molar ratios of probe to competitors are 1:1, 1:10, and 1:100. (H) Schematic drawing of effectors and reporters used in the transient assay. (I) The effects of CDF1-VP64 protein on the two CO promoter activities. The activities of the firefly luciferase (luc) were normalized by the activities of Renilla luciferase. The mean ± SEM of normalized values of four independent samples are shown.

To further analyze the role of CDF1 as a suppressor of CO, we used CDF1::HA-CDF1 plants with varying levels of CDF1 expression. Increasing CDF1 dosage resulted in an incremental delay in flowering time under LD conditions (fig. S10). Because FKF1 is expressed in these lines, it is expected that higher levels of CDF1 would be observed in the morning (Fig. 2 and fig. S9). The levels of CO expression in these lines were reduced in the early morning (from ZT22 to ZT1) (Fig. 4C and figs. S7 and S9). Induction of FT was also delayed, and the daily amount of FT expression was reduced (figs. S8 and S9). This alteration in FT expression is likely to contribute to the late-flowering phenotype observed in these lines. In the CDF1::HA-CDF1 K253A line, CO expression was reduced at all ZT time points under LD and SD conditions (Fig. 4D and fig. S7), and FT expression was diminished under both LD and SD conditions (fig. S8). In addition, two CDF1::HA-CDF1 K253A lines (expressing a stable CDF1 variant) (Fig. 2) with less than threefold overexpression of CDF1 displayed a late-flowering phenotype (fig. S10) similar to that observed in 35S::CDF1 lines with greater than 50-fold overexpression. Thus, the removal of CDF1 protein in the afternoon by FKF1 regulates CO expression and subsequent flowering events.

To determine if CDF1 directly regulates CO expression, we analyzed the CO promoter for CDF1 binding sites. Twenty putative Dof binding sites (AAAG) lie within 1 kilobase pair upstream of the CO transcription start site. Two regions in the CO promoter contained clusters of three or four Dof binding sites (–397/–358 and –173/–135 regions) (Fig. 4E). CDF1 protein binds to both fragments but exhibited a 10-fold greater preference for the –173/–135 fragment over the –397/–358 fragment (Fig. 4F). The sequences of these fragments differ not only in the number of the Dof binding motifs but also in the makeup of the sequences flanking the motifs (TGT versus nonconserved sequences). To assess whether the TGT flanking sequences confer specificity for CDF1 binding to the –173/–135 region, we mutated the –173/–135 fragment in both the consensus Dof binding site (Mut 1) and the flanking sequences (Mut 2) (Fig. 4E). The Mut 1 probe did not compete effectively with the nonmutated probe, and the Mut 2 probe was 10 times less effective as a competitor than the original probe (Fig. 4G). Thus, the core Dof binding sequence is necessary for CDF1 binding, whereas the flanking TGT sequence contributes to the specificity of binding.

To verify the direct binding of CDF1 to the Dof binding sites in the CO promoter in vivo, we performed a transient expression assay in which seedlings were bombarded with particles coated with constructs encoding CDF1 proteins and CO promoter–driven luciferase (luc) gene reporters (Fig. 4H and fig. S11). Both CDF1 and CDF1 K253A reduced the CO promoter activity (fig. S11). However, the reduction in luc expression may have been caused by physical damage due to bombardment. To avoid this complication, a translational fusion of CDF1 and the tetrameric VP16 activation domains (VP64) was constructed to convert this molecule to a transcriptional activator (17). The CDF1-VP64 fusion protein activated the CO (2.5 kb)::luc reporter, whereas neither CDF1 ΔDof-VP64 or VP64 alone stimulated expression of the reporter (fig. S11); thus, the Dof domain of CDF1-VP64 is required for activation. CDF1-VP64, but not CDF1 ΔDof-VP64, also induced luc expression driven by a smaller CO promoter (0.5 kb) (Fig. 4I). We then examined the ability of CDF-VP64 to activate a CO promoter with 11 Dof-binding site mutations (Fig. 4H). The basal level of luc expression of the mutated CO reporter was twofold higher than that of the nonmutated one (Fig. 4I), suggesting that endogenous Dof transcription factor(s) are involved in the repression of CO promoter activity. Importantly, CDF1-VP64 failed to induce the expression of the mutated CO promoter::luc reporter (Fig. 4I). Together with the findings from the in vitro DNA binding assay, these results indicate that CDF1 directly binds to the Dof binding sites in the CO promoter to regulate CO transcription.

Control of CO expression and activity is crucial in the regulation of photoperiodic flowering in Arabidopsis (2, 7). We have determined that, in the afternoon, FKF1 F-box protein degrades the transcription factor CDF1, a suppressor of CO. We propose that CO is repressed in the morning partially because of the accumulation of CDF1 binding to the CO promoter. Late in the afternoon, as FKF1 levels rise, CDF1 protein is quickly degraded in an FKF1-dependent manner to reduce the suppression of CO. Concomitantly, FKF1 also promotes CO transcription by a currently unknown mechanism(s) to produce the CO peak during the critical time period in which CO activity is required to promote flowering.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article