Author Correction: The F-box protein FKF1 inhibits dimerization of COP1 in the control of photoperiodic flowering

The previously published version of this Article contained errors in Figure 5. In panel c, the second and fourth blot images were incorrectly labeled 'α-Myc' and should have been labelled 'α-HA'. These errors have been corrected in both the PDF and HTML versions of the Article.

M ost flowering plants bloom in response to seasonal changes in environmental factors such as day length and temperature. In the model dicot plant Arabidopsis thaliana, flowering time is mainly regulated by the photoperiodic, autonomous, gibberellin, and vernalization pathways 1 . These signaling pathways converge to induce expression of the florigen gene FLOWERING LOCUS T (FT), which encodes a mobile protein that can induce the shoot apical meristem to make the transition from vegetative to reproductive development 2,3 . In the photoperiodic pathway, CONSTANS (CO) has a major role in inducing FT transcription, although other regulators also independently affect FT expression [4][5][6] . CO encodes a zinc finger-type transcription factor containing two B-boxes and a CO, CO-LIKE, and TOC1 (CCT) domain 7 . CO directly binds to the FT promoter and activates its transcription 8 . Levels of CO mRNA are regulated in a circadian manner: the CO mRNA is abundant during the daytime under long-day (LD) conditions and during the nighttime under short-day (SD) conditions 1,9 . However, FT transcription, controlled by CO, differs remarkably between LD and SD conditions because light signals tightly regulate CO at the posttranslational level. Elucidation of these regulatory mechanisms showed that FLAVIN-BINDING, KELCH REPEAT, F-BOX1 (FKF1) and CONSTITUTIVE PHOTOMORPHOGENIC1 (COP1) control CO stability [10][11][12] .
FKF1 is a key component of the SKP1/CUL1/F-box (SCF)-type E3 ligase complex and has three domains (LOV, F-box, and KELCH-repeat). FKF1 functions as a blue-light receptor in which the LOV domain binds to a flavin mononucleotide chromophore 13 . In photoperiodic flowering, FKF1 positively regulates CO at both the transcriptional and posttranslational levels 11,12 . In the morning, CYCLING DOF FACTOR (CDF) transcription factors (CDF1, CDF2, CDF3, and CDF5) repress the expression of CO and FT. In the afternoon of LD, FKF1 is expressed and lightactivated FKF1 interacts with GIGANTEA (GI) to degrade CDFs; this induces transcription of CO and FT, leading to early flowering. In SD, by contrast, FKF1 is mainly expressed after dusk, and unilluminated FKF1 has a decreased affinity for GI, resulting in the persistence of CDFs and thus a failure to induce CO transcription 11 . As recently reported, light-activated FKF1 also interacts with and stabilizes CO to activate FT transcription 12 .
In darkness, CO is ubiquitinated by COP1 in the nucleus and is degraded by 26S proteasome-dependent proteolysis 10,14,15 . Weak mutants of COP1 (a strong cop1 allele is lethal) exhibit very early flowering in SD and accumulate high levels of CO in darkness 15 . COP1 is a RING-type E3 ubiquitin ligase containing three domains (RING, coiled-coil (CC), and WD40-repeat) 16 . COP1 partitions between the nucleus and cytoplasm in a lightdependent manner 17 and forms homodimers through the CC domain; the homo-dimerization is required for COP1 function and subcellular localization 18,19 . COP1 associates with SUP-PRESSOR OF PHYTOCHROME A 1 (SPA1) to form a (COP1) 2 (SPA1) 2 tetramer. Homo-and hetero-dimerization of COP1 have important roles in COP1 function 19,20 . The (COP1) 2 (SPA1) 2 tetramer is involved in poly-ubiquitination and destabilization of CO in the dark 21 . In addition to CO, COP1 promotes destabilization of several nuclear proteins involved in flowering time and photomorphogenesis [21][22][23] .
Thus, the E3 ubiquitin ligases FKF1 and COP1 play critical roles in controlling photoperiodic flowering by directly regulating CO stability 10,12,15 , as FKF1 stabilizes CO in the light and COP1 destabilizes CO in the dark. Other important regulators also affect the stability of CO: (i) in the morning, HOS1 and phytochrome B (phyB) decrease CO stability 14,24,25 , (ii) in the afternoon of LD, phyA, cryptochrome 2 (cry2), and FKF1 increase CO stability 10,12,26,27 , and (iii) in darkness, COP1 mediates degradation of CO 15 .
Here, we provide evidence that FKF1 acts as an upstream negative regulator of COP1. FKF1 and COP1 regulate CO stability and photoperiodic flowering. FKF1 can interact with COP1 and reduce COP1 activity in a day-length-dependent manner. We suggest that posttranslational control of CO stability, mediated by negative regulation of COP1 by FKF1, promotes early flowering in LD.
FKF1 interacts with COP1 in vivo. The genetic analysis indicated that FKF1 and COP1 act in the same pathway of photoperiodic flowering. Since both COP1 and FKF1 function as E3ubiquitin ligases in proteasome-mediated proteolysis of their target proteins 13,21 , we examined whether FKF1 directly regulates COP1. We first tested whether FKF1 and COP1 can physically interact and found that FKF1 interacts with COP1 in yeast twohybrid assays, primarily via the RING domain of COP1 (Supplementary Fig. 1). This interaction between FKF1 and COP1 was also observed in planta by co-immunoprecipitation (Co-IP) following transient expression in Nicotiana benthamiana 30 and by bimolecular fluorescence complementation (BiFC) assays in onion epidermal cells (Fig. 1c, d). Next, to map the interacting domains of FKF1 and COP1, we separated COP1 into three domains: (1) RING, (2) CC, and (3) WD40. Also, we separated FKF1 into four domains: (1) LOV, (2) LOV + F-box, (3) F-box + KELCH, and (4) KELCH domains. In yeast two-hybrid assays, we found that both the F-box and KELCH domains of FKF1 interacted with the RING domain of COP1 (Fig. 1e). The domain interaction between FKF1 and COP1 was further confirmed by BiFC ( Supplementary Fig. 2) and Co-IP assays ( Supplementary  Fig. 3), consistent with the interaction results from yeast twohybrid assays. Together with the genetic data, these findings suggest that FKF1 interacts with and negatively regulates COP1 function in flowering.
CO is stabilized in the cop1 mutant independently of FKF1. Previous work reported that both FKF1 and COP1 E3-ubiquitin ligases interact with CO to control its function in LD-dependent early flowering antagonistically 12,15 . FKF1 and COP1 increase and decrease the stability of CO, respectively, because CO levels decrease in fkf1-2 mutants and increase in cop1-4 mutants. Thus, we tested whether the presence or absence of FKF1 activity affects CO stability in the cop1-4 background (Fig. 2a, b). We first analyzed CO levels at ZT15 (1 h before darkness) when FKF1 expression and activity are the highest in LD. In cop1-4 fkf1-t mutants, CO accumulated to levels similar to those in cop1-4, but CO was not detected in either WT or the fkf1-t mutant (Fig. 2a). Next, we analyzed CO levels every 4 h during the course of a day in LD, and found that CO levels were not altered in the cop1-4 background regardless of FKF1 genotype. CO levels were nearly constant during day and night in cop1-4, cop1-4 fkf1-t, and 35S:: Myc-FKF1/cop1-4 plants, but CO was hardly detectable in WT, fkf1-t, or 35S::Myc-FKF1 #3 plants (Fig. 2b).
Finally, we analyzed the relative abundance of CO mRNA in different mutant backgrounds (Fig. 2c, d). CO mRNA levels decreased in fkf1-t, and increased in 35S::Myc-FKF1, consistent with the observation that FKF1 functions to degrade CDF1, a negative regulator of CO transcription 11 . Interestingly, despite high accumulation of CO, the CO mRNA levels were lower in the cop1 mutant background, including in cop1-4, cop1-4 fkf1-t, and 35S::Myc-FKF1 #3/cop1-4 (Fig. 2c). However, it did not appear that high accumulation of CO negatively affected CO transcription, because native CO mRNA levels in 35S::CO-GFP were similar to those of WT (Fig. 2d). Taking these observations and those of a previous study 12 together, we suggest that FKF1 increases CO stability by reducing COP1 function in the late afternoon of LD to induce flowering. FKF1 does not affect COP1 stability. To examine whether FKF1 negatively regulates the stability of COP1, because FKF1 has E3-ubiquitin ligase activity 31 , we generated an anti-COP1 polyclonal antibody, and analyzed COP1 levels in fkf1-t and 35S::FKF1 #18 plants over the course of a day. Unexpectedly, we found that steady-state levels of COP1 persisted in both fkf1-t and 35S::FKF1 #18 compared with WT ( Fig. 3; Supplementary Fig. 4), indicating that FKF1 does not destabilize COP1. Similarly, we found that FKF1 levels were not significantly altered in cop1-4 or 35S::TAP-COP1 plants ( Supplementary Fig. 5). These results indicate that the FKF1-COP1 interaction does not affect the stability of either protein.
FKF1 can inhibit COP1 homo-dimerization. Since FKF1 does not affect the stability of COP1, we assumed that the FKF1-COP1 interaction decreases COP1 activity. COP1 interacts with SPA1 to form a (COP1) 2 (SPA1) 2 tetramer, and homo-and heterodimerization of COP1 is important for its biological function 19,20 . Therefore, we speculated that the FKF1-COP1 interaction prevents COP1-COP1 dimerization, the COP1-SPA1 interaction, or both, thus decreasing COP1 activity and increasing CO stability in the late afternoon of LD. To test these possibilities, we performed Co-IP assays in N. benthamiana. We found that COP1 dimerization occurred under both light and dark conditions in the absence of FKF1. Surprisingly, FKF1 overexpression diminished COP1 dimerization in the light but not in the dark (Fig. 4a). In the light, COP1 dimerization was severely decreased and, instead of forming homodimers, COP1 interacted with FKF1. In the dark, COP1 dimerization occurred normally and COP1 did not interact with FKF1. Next, we tested whether the in vivo interaction of FKF1 with COP1 depends on light. For this, we used Co-IP assays with 35S::FKF1 #18 plants, which revealed that the in vivo interaction requires light (Fig. 4b).
We further used yeast three-hybrid assays to test whether FKF1 inhibits COP1-COP1 homo-dimerization and/or COP1-SPA1  Nuclear protein-enriched fractions were immunoblotted using an anti-CO antibody (α-CO) to measure CO levels and an anti-H3 antibody (α-H3) for a loading control. Data are means ± s.d. from at least three biological repeats. c Relative abundance of CO mRNA during a day in 10-day-old seedlings. d Native CO mRNA levels during a day in 35S::CO-GFP transgenic plants. To detect the native CO mRNA, a specific primer was designed from the 3′-UTR region of the CO mRNA sequence. c, d For RT-qPCR, the relative expression level of each gene was normalized to the mRNA level of ACTIN (AT3G18780) as a loading control. Data are means ± s.d. from three biological replicates hetero-dimerization. In these assays, FKF1 transcription was controlled by the Met-repressible pMET25 promoter. We analyzed the inhibition of COP1 homo-dimerization by FKF1 under blue light or in darkness, and found that FKF1 inhibits COP1 homo-dimerization more under blue light than in darkness ( Fig. 4c; Supplementary Fig. 6). COP1 homo-dimerization was completely inhibited by FKF1 under blue light and was reduced in methionine-deficient conditions in the dark, based on both yeast colony survival and β-galactosidase activity. Moreover, we found that FKF1 does not inhibit the COP1-SPA1 interaction, regardless of light conditions. These results strongly suggest that in yeast, FKF1-mediated inhibition of COP1 dimerization is promoted by blue light, although some of the activity remains in darkness. Taken together, these results suggest that in Arabidopsis, blue-light-activated FKF1 can interact with and attenuate COP1 homo-dimerization.  Table 2). The hypocotyls of fkf1-t seedlings were as long as those of WT, regardless of daylength conditions (LD/SD/constant darkness; DD). Interestingly, hypocotyls of both 35S::FKF1 #18 and 35S::Myc-FKF1 #3 were significantly shorter than those of WT and fkf1-t in SD and slightly shorter in LD, but this was not statistically significant, and they were as long as WT in DD. These results suggest that FKF1 overexpression negatively regulates COP1 in hypocotyl elongation only in SD.
Next, we analyzed HY5 levels in 35S::Myc-FKF1 #3 and fkf1-t plants ( Supplementary Fig. 7b), since HY5 is one of major regulators of hypocotyl elongation, although other COP1 target proteins are also involved in this process 21 . HY5 stability in WT depends on the light period, as HY5 is more stable in LD than in SD, and not detected in DD. However, we could not find any evidence that HY5 becomes more stable in 35S::Myc-FKF1 #3 in SD. Thus, we concluded that FKF1 negatively affects COP1 function in hypocotyl growth in SD when FKF1 is constitutively overexpressed, and this is seemingly not related to the regulation of HY5 stability.
COP1 mutants that are unable to dimerize do not promote flowering. COP1 forms a homodimer and/or a heterodimer with SPA1 through the CC domain and finally forms a (COP1) 2 (-SPA1) 2 tetramer for its functional activity [18][19][20] . When COP1 dimerization is prevented, it is not functional in photomorphogenesis. To examine the effect of COP1 dimerization on flowering, we prepared mutated cDNAs using WT (Col-0) COP1 (COP1 WT ), and the mutant versions COP1 L105A and COP1 L170A which were previously reported to undergo normal or poor dimer formation, respectively 19 (Fig. 5a). First, we tested the binding between COP1 and mutated COP1, and FKF1 and mutated COP1 in N. benthamiana. The COP1 WT -COP1 L105A Co-IP signal was nearly the same as that of COP1 WT -COP1 WT , while that of COP1 WT -COP1 L170A was much weaker consistent with a prior publication 19 . This indicates that COP1 homo-dimerization    Fig. 5b; Supplementary  Fig. 8). Similarly, FKF1 also interacted with COP1 WT and COP1 L105A much more strongly than with COP1 L170A (Fig. 5c).
Finally, we examined whether the COP1 variants differed in their ability to destabilize HY5 for hypocotyl elongation in the dark (Fig. 5g, h; Supplementary Fig. 9, Supplementary Table 4). The COP1 WT -GFP and COP1 L105A -GFP fusion proteins, which form dimers normally, complemented the short-hypocotyl and cotyledon-expansion phenotypes of the cop1-4 mutant. However, COP1 L170A -GFP, which forms dimers poorly, did not rescue the defect. In cop1-4 mutants, HY5 was almost completely degraded in darkness by either COP1 WT -GFP or COP1 L105A -GFP, but not by COP1 L170A -GFP (Fig. 5i). Although we cannot rule out that these effects may be due to some other effect of the L170A mutation, these results are consistent with the level of COP1 dimerization correlating with the level of its functional activity in the timing of flowering and photomorphogenesis.

Discussion
For successful reproduction, most flowering plants bloom in a certain season, which they recognize mainly by sensing changes in temperature and day length. In Arabidopsis, CO is a key positive regulator of FT transcription in an LD-dependent manner, although FT expression is finely controlled by many regulators in other flowering pathways 2,4,6 . FKF1 and COP1 are direct positive and negative regulators, respectively, of the stability of CO 12,15 . Here, we demonstrate a direct link between FKF1 and COP1, in which FKF1 negatively regulates COP1 by the posttranslational regulation of CO. First, FKF1 genetically acts as an upstream negative regulator of COP1, as the late-flowering phenotype of the fkf1 mutation is not present in the cop1 background (Fig. 1a, b; Supplementary Table 1). Second, neither FKF1 overexpression nor fkf1 mutation alters CO abundance in the cop1 background (Fig. 3). Third, FKF1 strongly interacts with COP1 in the presence of light (Figs. 1c-e and 4a, b). Fourth, COP1 mutant variants that are unable to dimerize are unable to function in flowering as well as photomorphogenesis (Fig. 5). Finally, the interaction between FKF1 and COP1 can inhibit COP1 homo-dimerization in a light-dependent manner (Fig. 4a,  c). In summary, but, our findings show that the two important regulatory pathways for photoperiodic flowering of Arabidopsis, the FKF1-CO and COP1-CO pathways, that have previously been thought to act independently can act in the same pathway to regulate CO stability (Fig. 6).
FT is rhythmically expressed, with a peak at the end of day (around ZT16) only in LD, a few hours after the first peak of CO expression 9 . COP1 is a light-dependent nucleocytoplasmic partitioning protein 17,32 ; however, its nuclear exclusion in darkness occurs very slowly, taking approximately 24 h 33,34 . Although it has been reported that COP1 degrades CO during nighttime, many studies, including ours (Fig. 2a, b), showed that COP1 does not function only in darkness, because CO is more stable in cop1-4 mutants than in WT during the daytime 15,35 . COP1 levels are not altered in FKF1 overexpressor or fkf1-t plants (Fig. 3). Since FKF1 accumulation is rhythmic and peaked at around ZT12-ZT16 in both LD and SD, the blue-light receptor FKF1 becomes active and interacts with COP1 in LD, but not in SD. In this scenario, it is highly possible that in the presence of COP1, CO is stabilized by FKF1 in LD enough to induce FT transcription; however, it is degraded rapidly and completely in SD.
Homo-dimerization of COP1 occurs through the CC domain 18,19 and SPA1 also binds to the CC domain of COP1 36,37 . The molecular weight of a COP1 tetramer (COP1) 2 (SPA1) 2 is approximately 440 kDa, but this tetramer is present in several multi-complexes much larger than 440 kDa in vivo 20,21 . In fact, FKF1 function (inhibiting COP1 homo-dimerization) produces different effects from that of the COP1 L170A mutation: FKF1 and COP1 L170A mutation both inhibit COP1 homo-dimerization, but FKF1 does not inhibit the COP1-SPA1 interaction (Fig. 4). Some photoreceptors, such as PHYs and CRYs, inhibit COP1 function although it is not clearly understood how they inhibit COP1 activity 21 . It has been reported that photo-excited CRY2 interacts  with SPA1 and enhances the CRY2-COP1 interaction, resulting in suppression of COP1 activity and CO degradation for early flowering 27 . It is possible that FKF1 and CRY2 work together to inhibit the formation of COP1 complexes, in which CRY2 inhibits COP1-SPA1 hetero-dimerization, and FKF1 inhibits COP1 homo-dimerization; the formation of (COP1) 2 (SPA1) 2 is totally inhibited in the late afternoon of LD. We suggest that a specific COP1 complex is destabilized by light-activated FKF1 and/or CRY2 thus stabilizing CO in a light-dependent manner. These two light-dependent regulatory mechanisms could have an important role in the regulation of COP1 complex formation for photoperiodic flowering, but this remains to be determined. There are many possible mechanisms to explain how FKF1 inhibits COP1 activity. We provide evidence that FKF1 may negatively regulate COP1 activity by inhibiting its dimerization, but other regulatory mechanisms may exist. First, FKF1 may compete with other E2-ubiquitin conjugating enzymes (such as AtUBC9) 22 because FKF1 binds to the RING domain of COP1 (Fig. 1d). Many E2 enzymes bind to the RING domain of RINGtype E3 ligases and this interaction plays an important role in E3 activity 38 . Studies of the relationship between FKF1 and other E3 enzyme(s) are needed to understand the regulatory mechanisms of E3-ubiquitin ligases. Second, FKF1 may be involved in the nuclear exclusion of COP1. We analyzed COP1 protein accumulation in nuclear and cytoplasmic fractions from Col-0, fkf1-t, and 35S::FKF1, and found that FKF1 alone is not involved in the light/dark-induced movement of COP1 ( Supplementary Fig. 10). Instead, we further found that all ZTL family members (ZTL, LKP2, and FKF1) interact with COP1 ( Supplementary Fig. 11), suggesting that the ZTL family may also be related to COP1 function throughout development. The functions of ZTL family members in other COP1-mediated regulatory mechanisms during growth and development remain to be determined.
Finally, experimental observations and mathematical modeling indicated that COP1 function is repressed in the light by a photoreceptor-related inhibitor termed "I" 39 . Here, we demonstrate that the blue-light receptor FKF1 is a strong candidate among the hypothesized inhibitors, because FKF1 interacts with and inhibits COP1 homo-dimerization in a light-dependent manner. It seems that FKF1-mediated regulation of both CDF1 stability and COP1 activity are required to regulate lightdependent and internal rhythm-dependent control of protein expression 40,41 . Based on these findings, we propose a new model involving an FKF1-COP1-CO cascade (Fig. 6); the inhibition of COP1 homo-dimerization by light-activated FKF1 stabilizes CO in the afternoon of LD, resulting in early flowering. In SD, however, FKF1 expression mainly occurs after dusk and an inactive form of FKF1 cannot interact with COP1, resulting in high levels of COP1 homodimers that can degrade CO completely, preventing FT transcription, which leads to late flowering. This FKF1-COP1-CO regulatory cascade could be another layer in previously suggested models of the FKF1-GI-CDF1-CO pathway 11,12 .
Yeast two-and three-hybrid assays. Yeast two-hybrid assays were performed using the Matchmaker GAL4 two-hybrid system (Clontech). The full and partial cDNAs of each gene were cloned into the pGADT7 and pGBKT7 vectors as prey and bait, respectively. The full and partial (RING, aa 1-104; CC, aa 121-213; WD40, aa 371-675) cDNAs of COP1 were cloned into the pGBK vector (as baits) 23 . FKF1 was cloned into the pGAD vector (as prey) with full and partial cDNAs (LOV, aa 1-174; LOV+F-box, aa 1-283; F-box+KELCH, aa 174-618; KELCH, aa 283-618) 11 . The clones were co-transformed into the yeast strain AH109. The pBridge vector (Clontech) was used for yeast three-hybrid assays. COP1 or SPA1 cDNAs were cloned into the multi-cloning site I of the pBridge vector, in which the binding domain BD-COP1 or BD-SPA1 fusion protein was expressed. Then, FKF1 was cloned into multi-cloning site II of the pBridge vector, in which FKF1 expression was controlled by the Met-repressible pMET25 promoter. These vectors were cotransformed into the yeast strain AH109. Yeast transformation was performed according to the Yeast Handbook (Clontech). The colonies were used for yeast cell growth assay, and a liquid assay using chlorophenol red-β-D-galactoside (CPRG) was used to measure β-galactosidase activity.
Generation of anti-COP1 antibody. To generate COP1 antibody, COP1 antigen was designed to target the N-terminal region of COP1 (COP1-N; 1-305 aminoacid including the RING and CC domains). The partial cDNA encoding COP1-N was amplified from the full-length COP1 cDNA and cloned into the pGEX-4T-3 vector to produce GST-tagged COP1 protein. The GST-COP1-N constructs were transformed into Escherichia coli BL21 (DE3) strain. Cells carrying the plasmids were grown at 37°C to an OD 600 of 0.8, and then the expression of GST-COP1-N protein was induced by adding 1 mM isopropyl-1-thio-β-D-galactopyranoside (IPTG) for 3 h. The proteins were predominantly in the pellet and protein was purified by elution from SDS-polyacrylamide gels using an electro-eluter (BIO-RAD). The purified GST-COP1-N protein (500 μg) was injected into rabbits every 2 weeks, and after the fourth injection, blood was gathered and the serum was separated. For affinity purification of COP1 antibody, the cDNA encoding COP1-N was cloned into the pET28a vector for 6xHis tagging, and 6xHis-COP1-N protein was induced by IPTG treatment in E. coli BL21 (DE3) strain, and purified. An anti-COP1 antibody was then purified by affinity binding with 6xHis-COP1-N recombinant protein. Quantitative real-time PCR. Total RNA was isolated from 10-day-old seedlings, using the Plant RNA Isolation Kit (Macrogene). For reverse transcription, the firststrand cDNA was prepared from 2 μg of total RNA using an M-MLV reverse-transcriptase (Promega). Relative gene expression levels were analyzed by qPCR using the Light Cycler 2.0 (Roche Diagnostics). Relative mRNA levels of each gene were normalized to the expression of ACTIN (AT3G18780) as a loading control. The gene-specific primers are CO (LP 5′-GCCTACTTGTGCATGAGCTG-3′, RP 5′-GTTTATGGCGGGAAGCAAC-3′), native CO (LP 5′-GGATATGGGATTGTT CCTTC-3′, RP 5′-CAAACCCATTTGCACAACAG-3′), and ACTIN (LP 5′-TG GGATGAACCAGAAGGATG-3′, RP 5′-AAGAATACCTCTCTTGGATTGT GC-3′).
Data availability. The authors declare that all data supporting the findings of this study are included in the manuscript and Supplementary Information files or are available from the corresponding author upon request.