The F-box protein ZEITLUPE controls stability and nucleocytoplasmic partitioning of GIGANTEA

The regulation of protein turnover and nucleocytoplasmic distribution is part of the molecular mechanism at the core of the circadian clock (Harmer, 2009; Herrero and Davis, 2012; Más, 2008; Tataroğlu and Schafmeier, 2010). F-box proteins are key components of SCF-type E3 ligases that target proteins for degradation via the 26S proteasome pathway and play a substantial role in all circadian systems (Godinho et al., 2007; He et al., 2003; Ko et al., 2002; Koh et al., 2006; Siepka et al., 2007; Somers et al., 2000). The cytosol-localized F-box protein ZEITLUPE (ZTL) is the defining component of the SCF^ZTL E3 ligase, which regulates degradation of TIMING OF CAB2 EXPRESSION (TOC1) and PSEUDO RESPONSE REGULATOR5 (PRR5), two evening-phased circadian clock components (Fujiwara et al., 2008; Kiba et al., 2007; Kim et al., 2007; Más et al., 2003). Additionally, PRR5 enhances TOC1 protein by facilitating nuclear import of TOC1, sequestering both proteins from proteolysis by SCF^ZTL (Wang et al., 2010). Hence, the cytosolic localization of ZTL highlights the importance of protein compartmentalization in the regulation of clock activity.

ZTL belongs to a small, unique family of F-box proteins [ZTL, FLAVIN BINDING, KELCH REPEAT, F-BOX1 (FKF1) and LOV KELCH PROTEIN2 (LKP2)] containing a LOV (light, oxygen or voltage sensing) domain at the N-terminus, followed by an F-box domain and six kelch repeats (Demarsy and Fankhauser, 2009; Ito et al., 2012). The ZTL LOV domain shares high homology with the LOV domains of plant phototropins and WHITE COLLAR-1 (WC-1) in Neurospora (Ballario et al., 1996; Crosson et al., 2003; Ito et al., 2012). The LOV domains in these proteins bind to flavins and function as a light-driven molecular switch in light-mediated protein activation (He et al., 2002; Imaizumi et al., 2003; Matsuoka and Tokutomi, 2005).

LOV domains also mediate protein-protein interactions, including homodimerization of WC-1 and the phototropins (Ballario et al., 1998; Nakasako et al., 2008). ZTL binding to GIGANTEA (GI) in blue light requires a functional LOV domain, and LOV mutations that abolish flavin binding diminish the interaction and eliminate blue light-enhanced interactions (Kim et al., 2007). The ZTL-GI interaction stabilizes ZTL protein in the light and ensures robust oscillation of its target proteins (Kim et al., 2007). Similar LOV domain-mediated interactions take place between FKF1 and GI to influence CONSTANS (CO) expression and protein abundance in the regulation of photoperiodic flowering (Imaizumi et al., 2005; Sawa et al., 2007; Song et al., 2012). In long days, an FKF1-GI complex forms in the nucleus, degrading transcriptional repressors of CO and activating CO expression, which, in turn, promotes FT expression and flowering (Imaizumi et al., 2005; Sawa et al., 2007).

In this study, we identify a new post-transcriptional step in GI regulation. We show that ectopic expression of the ZTL N-terminus (LOV domain) competitively interferes with endogenous GI-ZTL interactions, simultaneously promoting ZTL degradation and inhibiting GI functions. Furthermore, under endogenous expression levels, GI stabilization and cytoplasmic retention occurs through a LOV domain-mediated GI-ZTL interaction. These reciprocal co-enhancements in the cytosol help maintain a strong oscillation of the GI-ZTL complex, contributing to the robustness of the circadian system.

The 35S:LOV and 35S:LOV-F constructs to generate Arabidopsis transgenic line expressing the LOV (1-192 aa) and LOV-F-box (1-280 aa) polypeptides under the control of the 35S promoter were generated as described previously (Han et al., 2004). In short, a C-terminal c-myc-tagged 35S:LOV construct was created by digesting the full length ZTL cDNA in pRTL2 with BsoBI and XbaI to remove the C-terminal region and fusing the first 192 residues to a single c-myc tag (LOV). Similarly, XbaI and XmnI were used to fuse the N-terminal 280 residues (LOV-F) to a c-myc tag. These constructs were subcloned into pZP200 and transformed into Arabidopsis (Columbia) containing CAB2:LUC using standard techniques (Clough and Bent, 1998).

The constructs of full-length genes, deletions of ZTL and FKF1 and their mutants for transient expression in Nicotiana benthamiana or in Arabidopsis mesophyll protoplasts were prepared using the gateway system (Invitrogen). Entry clones for ZTL, ZTL-KELCH, ZTL-LOV, ZTL(G46E)-LOV, FKF1-LOV and FKF1 were generated by PCR with Pyrococcus furiosus (Pfu) DNA polymerase (Stratagene), using plasmids containing full-length protein-coding sequences of ZTL, FKF1 and ZTL(G46E) (Kim et al., 2007) as templates and cloned into pCR-CCD-F. The primer pairs used for plasmid construction are listed (supplementary material Table S1). The final constructs were established by the LR recombinase reaction using each entry clone and the Gateway version of pCsVMV:GFP-C-999 or pCsVMV:HA-C-1300 as destination vectors for Arabidopsis protoplast expression and for tobacco infiltration, respectively. 35S:GI-GFP has been described (Kim et al., 2007).

phyB-101 (Col) and cry1-304 (Col) were obtained from the Arabidopsis Biological Resource Center and Dr C. Lin (UCLA, USA), respectively. ztl-103, GI:GI-HA and 35S:GI-GFP have been described previously (David et al., 2006; Kim et al., 2007). GI:GI-HA LOV-OX and 35S:GI-GFP ztl-103 were generated by genetic crossing between GI:GI-HA and LOV-OX (87-16) and between 35S:GI-GFP and ztl-103, respectively. All lines were confirmed by either PCR-based genotyping or their resistance to antibiotics.

Protein extraction, immunoblot and immunoprecipitation analyses were performed as described previously with minor modifications (Kim et al., 2007). For co-immunoprecipitation of GI and LOV, formaldehyde treatment was performed for tissue-crosslinking. Arabidopsis seedlings were grown for 10 days in a 12-hour light/12-hour dark photoperiod and then harvested at zeitgeber time (ZT)13. The seedlings were soaked in 20 ml of ice-cold 1× PBS (140 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄ and 1.8 mM KH₂PO₄) containing 1% formaldehyde and vacuum-infiltrated for 30 minutes in an ice bath. Ice-cold glycine was added to bring the final concentration to 300 mM and then the vacuum infiltration was continued for 30 minutes. Following the infiltration, the seedlings were washed three times in 1× PBS and then frozen in liquid nitrogen for further protein analysis. For immunoprecipitation, protein extracts were prepared using extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml leupeptin, 1 μg/ml aprotinin, 1 μg/ml pepstatin, 5 μg/ml antipain, 5 μg/ml chymostatin, 50 μM MG132, 50 μM MG115 and 50 μM ALLN) and immunoprecipitations were performed with anti-HA as described previously. Immunodetection using anti-HSP90 (1:50,000), anti-ADK (1:40,000), anti-Histone 3 (1:40,000), anti-ZTL (1:500) and anti-HA (1:1000) was performed as described previously except for different dilutions.

Isolation of nuclear and cytosolic proteins for tobacco tissue and nuclear proteins for Arabidopsis was performed with a CelLytic PN Isolation/Extraction Kit (Sigma-Aldrich) as previously described (Wang et al., 2010) with minor modifications. Isolation of cytosolic protein from Arabidopsis was carried out as previously described (Kim et al., 2007) but using 1× nuclei isolation buffer of the CelLytic PN Isolation Extraction Kit (Sigma-Aldrich). Purification of each compartment was validated by immunodetection using cytosolic (ADK) and nuclear (Histone H3) marker proteins.

For the hypocotyl elongation assay, seeds plated on medium with 1× MS salts, 3% sucrose and 0.8% agar were incubated at 4°C for 3 days and germinated under white light at 23°C for 8 hours before being transferred into a growth chamber with constant blue or red light. Different light intensities were achieved by covering the plates with neutral density filters [Roscolux 397 (Rosco Laboratories, Stamford, CT)]. Hypocotyl length was measured after 7 days incubation using a ruler.

The free-running period of transgenic plants was determined using CAB2:luciferase (CAB2:LUC) as a bioluminescent reporter as previously described (Somers et al., 2004).

For flowering time measurement, seeds were sterilized and plated on MS medium (0.8% agar) supplemented with 3% sucrose and cold-treated for 4 days before being moved to long days (16-hour light/8-hour dark, 70-100 μmol m^-2) or short days (8-hour light/16-hour dark, 100 μmol m^-2). After 7 days of growth, seedlings were transplanted to soil and were grown in the same light condition until flowering. Total numbers of rosette and cauline leaves were counted when the bolts were about 1 cm in length.

Analysis of transcript level was performed using RNA gel blot analysis for ZTL, semi-quantitative RT-PCR for CO and FT as previously described (Somers et al., 2004) and quantitative real-time PCR for GI-HA as previously described (Kim et al., 2008). Arabidopsis seedlings were grown for 10 days in a 12-hour light/12-hour dark photocycle and then harvested as indicated. Total RNA was extracted using Trizol Reagent (Invitrogen) or TRI reagent (MRC) according to the manufacturer’s instructions, and further treated with DNase (Ambion) before RT-PCR and q-PCR analysis. An equal amount (∼400 ng) of total RNA was used for generating cDNA using the SuperScript III Kit (Invitrogen). Real-time PCR was performed using IQ SYBR Green Supermix (Biorad) and an iQ5 real-time PCR detection system (Bio-Rad). The reaction protocol for the q-PCR reaction proceeded as follows: 94°C for 2 minutes, followed by 40 cycles of 94°C for 15 seconds and 60°C for 34 seconds. The relative expression of target genes between samples was normalized to ACT2 expression, and further normalized to the maximum level among their relative expression for each trial. The primer pairs used in the PCR analyses are listed (supplementary material Table S1).

Protoplast preparation and DNA transfection was performed as previously described with a minor modification (Kim et al., 2008; Kim and Somers, 2010). GI:GI-HA plants were grown for 20-30 days in a 12-hour light/12-hour dark photoperiod. Preparation of Arabidopsis protoplasts started at ZT7 and DNA transfection ended at ZT11-12. About 400 μl protoplast solution with 10⁵ protoplast cells and ∼90-140 μg DNA for effectors depending on plasmid size were used per reaction. After a 24-hour incubation in the same chamber where plants were grown, protoplasts were harvested.

For microscopy imaging, transient expression in N. benthamiana was performed using p19-enhanced Agrobacteria infiltration (Voinnet et al., 2003). Agrobacteria containing the plasmids 35S:GI-GFP alone (plus 35S:H2B-RFP and 35S:P19) or 35S:GI-GFP and CsVMV:HA-ZTL-LOV (plus 35S:H2B-RFP and 35S:P19) were co-infiltrated into N. benthamiana plants. Images were captured 2 days after infiltration using an E600 fluorescence microscope (Nikon, Japan) with proper filter sets and Advanced SPOT Software (Diagnostics Instruments). The intensity of fluorescent signals was plotted using ImageJ software.

The ZTL LOV domain is essential for stabilizing ZTL levels through interaction with GIGANTEA (GI) (Kim et al., 2007). To investigate further the functional role of the ZTL LOV domain, we generated transgenic plants ectopically expressing the LOV (ZTL-LOV; 1-192 aa) and LOV-F (ZTL-LOV-F; 1-280 aa) domains. Because the F-box portion of all F-box proteins interacts with the SKP1 protein within the context of the SCF ubiquitin ligase complex, we tested whether this domain might affect the phenotype, possibly by sequestering LOV-F into SCF complexes. This could have a great effect on SCF ubiquitin ligase function in the plant, by ‘poisoning’ SCF complexes in all contexts of its ligase function. We characterized four transgenic lines expressing the 20-kDa LOV domain (LOV-OX) and three lines expressing the 26-kDa LOV-F (LOV-F-OX) domain (supplementary material Fig. S1) for effects on circadian cycling, flowering time and photomorphogenesis.

Disruption of clock function often alters circadian period and/or amplitude, accompanied by changes in hypocotyl elongation and flowering time (Yakir et al., 2007). ZTL loss-of-function mutations lengthen the free-running period by several hours, whereas constitutive expression of full-length ZTL shortens this period or causes arrhythmicity (Somers et al., 2004). We tested several LOV and LOV-F transgenic lines for period alteration under constant red or blue light after entrainment in 12-hour light/12-hour dark cycles. Under constant red light, the LOV-OX and LOV-F-OX lines have a free-running period ∼0.5-2 hours longer than that of wild type (WT) (Fig. 1A). Under constant blue light, the period was also lengthened by 1-2.5 hours (Fig. 1B).

We next examined flowering time in these lines when grown under long (16-hour light/8-hour dark) or short (8-hour light/16-hour dark) days. Similar to ZTL overexpressors (ZTL-OXs), both LOV-OX and LOV-F-OX lines flowered late in long days compared with WT, with little effect in short days (Fig. 1C). Photoperiodic control of flowering time is mainly regulated by CONSTANS (CO) and FLOWERING LOCUS T (FT) (Mouradov et al., 2002). In both transgenic types, CO and FT expression was strongly reduced during long photoperiods (Fig. 1D,E). These reductions are similar to that found in ZTL-OXs (Somers et al., 2004), or other late-flowering mutants with altered clock activity (Mouradov et al., 2002).

Overexpression of ZTL lengthens hypocotyls in either red or blue light (Somers et al., 2004). We tested the effect of LOV and LOV-F overexpression on hypocotyl growth under these conditions. In red light, both types of plants have slightly longer hypocotyls at all but the very highest fluence rates (Fig. 1F,G). These effects are phytochrome B (phyB)-dependent, as phyB is epistatic to both OX types (Fig. 1F,G). Similar lengthening effects were obtained under blue light (Fig. 1H,I) except that the cryptochrome1 (cry1) mutation additively lengthens the effect of LOV/LOV-F backgrounds. This indicates that blue light signaling via a LOV/LOV-F pathway is parallel to CRY1 signaling, consistent with LOV/LOV-F disruption of phytochrome signaling (Neff and Chory, 1998).

Taken together, the three phenotypes examined in the LOV-OX and LOV-F-OX lines cannot be explained by simple dominant-negative or gain-of-function effects of LOV-OX on ZTL. The lengthened period of LOV- and LOV-F-OXs is similar to that of ztl loss-of-function mutants, but the late flowering and hypocotyl lengthening phenocopies ZTL overexpression (Somers et al., 2004). Instead, these results might arise from a combination of effects of the LOV and LOV-F domains on ZTL and/or an interacting protein, such as GI.

We measured the level of endogenous ZTL protein in the LOV-OXs and LOV-F-OXs to determine whether the longer period resulted from lower endogenous ZTL. In all lines tested, ZTL levels were strongly reduced or undetectable compared with WT plants grown under a 12-hour light/12-hour dark (LD) photoperiod (Fig. 2A) or under constant blue light (Fig. 2B). In LD cycles, the relative decrease was stronger at ZT13, compared with ZT1, especially in LOV-OX lines (Fig. 2A). However, in constant light, ZTL levels were similarly very low in all lines (Fig. 2B). ZTL mRNA in these transgenic lines showed no consistent differences from WT (Fig. 2C), indicating a post-transcriptional mechanism.

Because ZTL is stabilized in blue light by binding to GI via the LOV domain (Kim et al., 2007), we next tested whether lower ZTL resulted from the competitive binding of LOV and LOV-F proteins to GI. Co-immunoprecipitations (co-IPs) were performed using plants expressing both ZTL-LOV and GI-HA, ZTL-LOV alone, or GI-HA alone. ZTL-LOV polypeptides co-immunoprecipitated with GI-HA only in plants expressing both ZTL-LOV and GI-HA (Fig. 2D), confirming that the LOV domain of ZTL alone is sufficient to bind GI in vivo (Kim et al., 2007). Additionally, the amount of endogenous ZTL co-immunoprecipitated with GI was significantly reduced in the GI:GI-HAxLOV-OX line compared with the GI:GI-HA line (Fig. 2E), even though precipitated GI is much higher in GI:GI-HAxLOV-OX. This suggests that a strong ZTL-LOV interaction with GI competes with ZTL for GI binding, reducing access of endogenous ZTL to GI and resulting in ZTL degradation and a longer circadian period.

GI protein, but not GI message, is reduced in ztl mutants, suggesting that GI levels might be regulated by ZTL post-transcriptionally (Kim et al., 2007). We performed immunoblot analyses to determine the effect of LOV overexpression on GI protein accumulation. GI-HA levels are significantly increased in lines expressing LOV polypeptide relative to the untransformed GI:GI-HA line over the entire LD time course (Fig. 3A). GI-HA mRNA remains unchanged in the LOV-OX background (Fig. 3B; supplementary material Fig. S2), indicating that ZTL-LOV stabilizes GI post-transcriptionally.

We then tested whether a direct interaction between ZTL-LOV and GI is necessary to stabilize GI protein and whether ZTL has a specific role in this process. We analyzed GI-HA protein levels at ZT12 in GI:GI-HA protoplasts transiently transfected with ZTL- and FKF1-related constructs as effectors including GFP, GFP-ZTL-LOV, GFP-ZTL(G46E)-LOV, GFP-FKF1-LOV, GFP-ZTL and GFP-ZTL(G46E). FKF1 also binds GI through its LOV domain, but is nuclear-localized and controls flowering time without affecting circadian period (Sawa et al., 2007). A predicted, nuclear localization signal exists at the N-terminus of FKF1 (supplementary material Fig. S3A); this was confirmed by observing nucleus-enriched fluorescent signals when GFP-tagged FKF1-LOV (GFP-FKF1-LOV) or FKF1 (GFP-FKF1) were transiently expressed in N. benthamiana. These localization patterns are very different from the cytosol-enriched fluorescent signals seen with GFP-ZTL-LOV and GFP-ZTL (supplementary material Fig. S3B). We also tested the effect of the G46E mutation, which disrupts the ZTL-GI interaction (Kim et al., 2007).

The abundance of GI-HA protein varied depending on the effector, although all effector proteins were expressed well, with higher levels of LOV polypeptides and lower levels of full-length proteins (Fig. 3C). Compared with the GFP-transfected controls, GI-HA levels were markedly increased in GI:GI-HA protoplasts transfected with GFP-ZTL-LOV and GFP-ZTL, with fourfold and twofold enrichments, respectively (Fig. 3D,E). By contrast, the non-interacting GFP-ZTL(G46E) and GFP-ZTL(G46E)-LOV mutants showed little or no GI enhancement relative to the effects of the GFP control (Fig. 3D,E). In addition, FKF1-LOV had little effect on GI stabilization, similar to that of ZTL(G46E)-LOV and much less than that of ZTL-LOV, although all three polypeptides were similarly strongly expressed (Fig. 3C,D). These results, together with our previous report of GI-mediated ZTL stabilization, suggest that ZTL and GI co-stabilize through LOV domain-mediated direct interaction.

GI is stabilized more effectively by cytosolic ZTL-LOV than by nuclear FKF1-LOV, so we hypothesized that ZTL regulates both the stability of cytosolic GI and its nucleocytoplasmic distribution. We measured the ratio of cytosolic to nuclear GI by fractionating cytosolic and nuclear proteins from N. benthamiana leaves transiently expressing GI-GFP in the presence of ZTL- or FKF1-related derivatives as effectors (Fig. 4A,B). Co-expression of GI-GFP with HA-ZTL-LOV increases cytosolic GI levels relative to nuclear protein levels, whereas expression of GI-GFP with HA-ZTL-KELCH, HA-ZTL(G46E)-LOV or HA-FKF1-LOV have little or no effect on nuclear and cytosolic GI levels (Fig. 4A,C). Similar changes in GI nucleocytoplasmic partitioning were observed when full-length HA-ZTL was expressed, but not when HA-ZTL(G46E) or HA-FKF1 were expressed as effectors (Fig. 4B,C). Both full-length ZTL and the ZTL-LOV domain increase the relative cytosolic distribution of GI-GFP significantly, with the effect of the ZTL-LOV domain being markedly stronger than full length ZTL. By contrast, full-length FKF1 and the LOV domain of FKF1 have a slight effect on GI localization. However, neither the ZTL C-terminus (KELCH) nor ZTL mutant (G46E) altered GI nucleocytoplasmic partitioning. These results tie the effectiveness of cytosolic retention of GI to its ability to interact with ZTL. We confirmed these fractionation experiments in planta by observing more GI-GFP fluorescence in the cytosol and less fluorescence in the nucleus when ZTL-LOV was expressed as an effector (Fig. 4D). We also found that light quality had no effect on ZTL-LOV-dependent partitioning of GI (supplementary material Fig. S4).

An enhanced ratio of cytosolic to nuclear GI by ZTL-LOV or full-length ZTL protein could result from the protection of cytosolic GI from degradation and/or from the sequestration of GI to the cytosol. We next tested the effect of different ZTL-LOV levels on the accumulation of GI-GFP protein in the nucleus and cytosol. As the relative amount of ZTL-LOV was increased, there was a corresponding increase of cytosolic GI and a concomitant reduction in nuclear GI protein (Fig. 4E, compare levels 0 with 4). These different compartmental responses of GI are likely to be due to ZTL effects on both GI cytosolic retention and protein stabilization.

The long period phenotype in the LOV and LOV-F lines results from reduced ZTL levels (Fig. 2). However, the late flowering and long hypocotyls in these lines are not related to ZTL, because ztl null mutants have short hypocotyls and normal flowering in long days (Somers et al., 2004). Instead, these phenotypes are similar to plants lacking GI (Fowler et al., 1999; Huq et al., 2000; Martin-Tryon et al., 2007; Park et al., 1999). Because GI protein increases in LOV-OX, a change in availability and/or subcellular partitioning of GI might be responsible for the flowering time and hypocotyl phenotypes.

To address this, we tested the nucleocytoplasmic distribution of GI-HA in a stably transformed LOV-OX background. Relative to GI:GI-HA alone, cytosolic GI was increased, whereas nuclear GI was diminished (Fig. 5A, compare lanes 1 and 4 with 2 and 5). This confirmed the results of the transient assays (Fig. 4). The lower level of nuclear GI probably reduces the level of GI-FKF1 nuclear complexes in this background, leading to later flowering. Additionally, some ZTL-LOV is present in the nucleus (Fig. 4A), which might interfere with nuclear FKF1-GI interactions, further delaying flowering.

Finally, we tested how GI cytosolic and nuclear distribution is affected in the absence of ZTL by measuring GI-GFP partitioning in a 35S:GI-GFP ztl-103 background. Consistent with enhanced cytosolic GI distribution in LOV-OX plants, we observed more nuclear accumulation and cytosolic depletion of GI-GFP protein in the ztl mutant, relative to WT (Fig. 5B, compare lanes 1 and 3 with 2 and 4). Taken together, our data show that ZTL facilitates the retention and stabilization of GI in the cytosol (Fig. 6).

In this study, we uncovered two unexpected new roles for the F-box protein ZTL. We first observed that constitutive ectopic expression of the LOV and LOV-F domains of ZTL causes lengthening of the circadian period, delayed flowering in long days and elongated hypocotyls in red and blue light. Although endogenous ZTL levels are reduced in these lines, this combination of developmental and circadian defects is not associated with ztl mutants. ZTL deficiency is marked by long period, short hypocotyls in red light and normal to slightly early flowering in long days (Somers et al., 2000). Surprisingly, ectopic LOV domain expression also markedly increased levels of GI. However, neither can these three phenotypes be explained by high GI expression. GI-OX has little effect on flowering in long days and tends to slightly shorten both period and hypocotyl length (Mizoguchi et al., 2005). Instead, we propose that these phenotypes result from the sequestration of GI into a LOV-GI complex. This prevents GI from carrying out its normal function and skews GI residence to the cytosol, depleting the nucleus of active GI. The developmental and circadian phenotypes of these lines phenocopy gi loss-of-function mutants, supporting this model (Fowler et al., 1999; Huq et al., 2000; Martin-Tryon et al., 2007; Park et al., 1999).

The first unexpected finding is the role of ZTL in the stabilization of GI. We have previously established that GI stabilizes ZTL protein and provided evidence that GI levels were diminished in a ztl background (Kim et al., 2007). Here, we confirm and extend this finding by showing that GI protein levels are significantly higher when a LOV-containing polypeptide is ectopically expressed at high levels in Arabidopsis plants, although GI mRNA transcript levels are unchanged (Fig. 3A,B). This supports the notion that LOV-domain interactions with GI are needed for GI accumulation. Additionally, transient expression of full-length ZTL and the ZTL LOV domain enhanced endogenous GI protein level by two to fourfold, whereas a non-GI interacting ZTL mutant (G46E) and the FKF1-LOV domain had little to no effect on GI accumulation (Fig. 3B-D). Taken together, these results show a specific and direct effect of ZTL on GI stabilization and indicate that ZTL is likely to be a limiting factor in controlling GI levels post-transcriptionally.

Second, we showed that transient expression of cytosol-localized ZTL or ZTL-LOV, but not nuclear-localized FKF1 nor non-interacting ZTL deletions and mutants, increases the cytosolic accumulation of GI while diminishing nuclear levels (Fig. 4A-D) and that these effects are ZTL-LOV dosage dependent (Fig. 4E). These transient assay results were supported by the observation that subcellular partitioning of GI in the LOV-OX background is shifted to the cytosol (Fig. 5A,B), and by the finding that GI nuclear levels are significantly higher than in the cytosol in ztl-103 (Fig. 5C,D). These results indicate that a second novel and specific regulatory role of ZTL is the control of the nucleocytoplasmic partitioning of GI through LOV-mediated cytosolic retention (Fig. 6).

Interestingly, this dual effect of ZTL on GI is reminiscent of the relationship between FREQUENCY (FRQ) and an RNA helicase (FRH) in the Neurospora clock. The formation of a FRQ-FRH complex (FCC) helps stabilize and retain FRQ in the cytosol, with increased FRQ nuclear levels when the interaction is disrupted or FRH levels are low (Cha et al., 2011). Conversely, the mammalian clock component mCRY helps retain the nucleocytoplasmic protein mPER2 in the nucleus and away from ubiquitylation and degradation in the cytosol (Yagita et al., 2002). In both cases, one clock component simultaneously preferentially localizes and stabilizes a partner protein. These heterodimeric interactions between clock proteins appear to be a conserved feature among circadian systems, and might build a greater robustness into the network by increasing mutual dependencies and stabilities.

Previous work has shown that ZTL, LKP2 and FKF1 have different primary roles in the control of flowering time and circadian period. ztl mutants have a significantly lengthened circadian period, with only slightly early flowering in short days, whereas fkf1 mutants delay photoperiodic flowering time with marginal effects on circadian clock regulation (Baudry et al., 2010; Imaizumi et al., 2003; Kevei et al., 2006; Somers et al., 2000). lkp2 mutants have little effect on either flowering time and circadian period, but overexpression of LKP2 phenocopies ZTL overexpression (Imaizumi et al., 2003; Schultz et al., 2001; Takase et al., 2011).

Partial overlapping and redundant functions among these members in the regulation of the clock have been suggested by results showing ztl fkf1 and ztl fkf1 lkp2 genotypes have much longer periods and higher levels of TOC1 and PRR5 accumulation than do ztl mutants (Baudry et al., 2010). However, the functional relationships among these family members in flowering time control is less clear. Whereas fkf1 mutants flower very late, ztl lkp2 or ztl genotypes are early flowering in short days and these defects revert to very late flowering in the triple mutant (Fornara et al., 2009; Nelson et al., 2000; Takase et al., 2011).

Our findings now indicate that the early flowering phenotypes of the ztl lkp2 and ztl genotypes are likely to be indirect effects of these mutations. In the absence of ZTL, GI partitioning shifts to the nucleus, even though total GI levels in the ztl mutant are lower (Fig. 5C) (Kim et al., 2007). These higher nuclear GI levels would enhance the formation of GI-FKF1 complexes and promote flowering. By contrast, both plants with ectopic LOV expression (Fig. 1) and ZTL-OX plants (Somers et al., 2004) are late flowering, consistent with increased GI sequestration to the cytosol, which would reduce nuclear GI-FKF1 complex formation. This would result in higher CDF accumulation and lower transcript levels of CO and FT.

ZTL OX also reduces the level of TOC1 and PRR5 (Kiba et al., 2007), which might explain the similarity to the late flowering phenotype of the toc1 prr5 double mutant (Ito et al., 2008; Ito et al., 2012), For ZTL OX it is possible that both GI sequestration and reduced TOC1 and PRR5 act together to cause late flowering. However, ectopic expression of the ZTL-LOV domain, which also causes late flowering, cannot be attributed to reduction of TOC1 and PRR5 because it lacks the protein-interacting kelch domains.

The FKF1-dependent early flowering phenotypes of ztl or ztl lkp2 can be explained by higher accumulation of GI, and a GI-FKF1 complex, in the nucleus of these mutants caused by loss of ZTL and LKP2 as sequestration factors. As well, increased cytosolic retention of FKF1 by direct interaction with cytosolic ZTL and/or LKP2 has been suggested (Takase et al., 2011) and would additionally contribute to controlling the nuclear levels of a GI-FKF1 complex by affecting FKF1, rather than GI, partitioning. These interpretations suggest that, unlike its role in period control, ZTL has no endogenous proteolytic targets directly related to flowering time control, but acts indirectly through complex formation and cytosolic retention of GI and, possibly, FKF1. These findings also act as a cautionary note when interpreting results from ectopic expression.

ZTL is well established as a target-recruiting element of the E3 SCF complex, responsible for circadian-phase specific degradation of TOC1 and PRR5 (Somers et al., 2000; Han et al., 2004; Más et al., 2003; Kiba et al., 2007; Fujiwara et al., 2008). By contrast, GI is a biochemically unknown protein without well-characterized domains (Park et al., 1999; Fowler et al., 1999). gi mutations have no additional effect on period length in ztl backgrounds and some ZTL mutations (e.g. G119D and G46E) reduce interaction with GI and compromise clock function (Kevei et al., 2006; Kim et al., 2007). These findings indicate that a direct interaction between GI and ZTL is necessary for clock control. However, it is not clear if GI and ZTL interact with targets together as a complex or whether they function separately as protein regulators. Overexpression of the LOV domain enhances cytosolic GI protein, but this stabilized GI is non-functional through sequestration by the LOV polypeptide from endogenous ZTL (Fig. 5A; Fig. 2D,E). This suggests that GI and ZTL might form, via the LOV domain, a strong and stable complex. We have previously shown that mild ZTL overexpression can still considerably shorten period by 2-4 hours (Somers et al., 2004). This might result from an effect of GI, through a GI-ZTL complex, on ZTL activity beyond simple stabilization and increase of ZTL levels.

Alternatively, a GI-ZTL complex might function as a reservoir of ZTL in the cytoplasm. Light conditions combined with clock-controlled expression of GI might indirectly regulate interactions between ZTL and its other protein partners. In darkness, the ZTL-GI interaction is weaker and the complex may dissociate more easily and favor ZTL substrate (TOC1, PRR5) or ZTL-ASK interactions predominantly, or newly dark-synthesized ZTL might preferentially complex with these factors over GI.

Whereas cytoplasmic ZTL affects stability and nucleocytoplasmic distribution of GI, nuclear-localized FKF1 has a minor effect on GI (Fig. 3D; Fig. 4C). This is consistent with little or no clock-related phenotypes in fkf1 mutants or FKF1 overexpressors (Imaizumi et al., 2003). In addition, FKF1 overexpressors have nearly normal flowering time, whereas overexpression of both FKF1 and GI induces dramatically early flowering (Sawa et al., 2007), indicating that FKF1 alone is unlikely to affect GI levels or activity strongly. As with ZTL, GI stabilizes FKF1 (Fornara et al., 2009), but the levels of GI in the fkf1 mutant or FKF1 overexpressor have not been tested.

Although ZTL targets TOC1 and PRR5 for degradation, our results show that ZTL can also stabilize proteins (GI). In addition to interacting with the ZTL protein family, GI-ELF4 complexes regulate GI subnuclear positioning and GI is de-stabilized by an ELF3-COP1 complex in the nucleus (Kim et al., 2013; Yu et al., 2008). Considering that ELF3 and COP1 exist predominantly in the nucleus, ZTL might enhance overall GI protein levels by facilitating cytosolic retention of GI protein and sequestering it from proteolysis by SCF^COP1. ZTL might additionally stabilize GI by further sequestration from degradation by an unknown cytosolic mechanism (Fig. 6).

A stabilizing effect on CO by FKF1 has been reported recently (Song et al., 2012). A blue light-dependent FKF1-CO interaction drives accumulation of CO protein in the late afternoon, which mediates control of day length-dependent flowering time. At the same time, FKF1 destabilizes CDFs through kelch-mediated interactions, mostly at the end of day (Imaizumi et al., 2005). It appears, then, that the ZTL family of F-box proteins can have contrasting effects on interaction partners, either to recruit them towards or sequester them from proteolysis.

We thank Dr Hong Gil Nam for the kind donation of the Gateway version of pCsVMV:GFP-C-999 and pCsVMV:HA-C-1300 and Dr Chentao Lin for cry1-304.