Functional domain studies uncover novel roles for the ZTL Kelch repeat domain in clock function

The small LOV/F-box/Kelch family of E3 ubiquitin ligases plays an essential role in the regulation of plant circadian clocks and flowering time by sensing dusk. The family consists of three members, ZEITLUPE (ZTL), LOV KELCH PROTEIN 2 (LKP2), and FLAVIN-BINDING KELCH REPEAT F-BOX PROTEIN 1 (FKF1), which share a unique protein domain architecture allowing them to act as photoreceptors that transduce light signals via altering stability of target proteins. Despite intensive study of this protein family we still lack important knowledge about the biochemical and functional roles of the protein domains that comprise these unique photoreceptors. Here, we perform comparative analyses of transgenic lines constitutively expressing the photoreceptor LOV domain or the Kelch repeat protein-protein interaction domains of ZTL, FKF1, and LKP2. Expression of each domain alone is sufficient to disrupt circadian rhythms and flowering time, but each domain differs in the magnitude of effect. Immunoprecipitation followed by mass spectrometry with the ZTL Kelch repeat domain identified a suite of potential interacting partners. Furthermore, the ZTL Kelch repeat domain mediates interaction with the LOV domain of ZTL and the ZTL homologs LKP2 and FKF1. This suggests that the Kelch repeat domain of ZTL may mediate homo- and hetero-dimerization of the three LOV/F-box/Kelch proteins and provide added insight into the composition of the protein complexes and an additional role for the Kelch repeat domain.


Introduction
consequences but exaggerate ZTL and FKF1 mutant phenotypes [11,12,14]. However, when LKP2 is expressed at high levels it can cause the clock to be arrhythmic, indicating its role in clock function [6].
In order to fully understand the overlapping and distinct functions of this important gene family, intense research has begun to investigate the structures and biochemical functions of the LOV and Kelch repeat domains [9,11,[22][23][24][14][15][16][17][18][19][20][21]. The N-terminal LOV domain is a blue light photoreceptor that is critical for the regulation of ZTL, LKP2, and FKF1 function. Regulation occurs through light-dependent interaction with the regulatory protein GIGANTEA (GI) which can promote or inhibit E3 ligase activity depending on the target protein [21,25,26]. GI binding is required for the E3 ubiquitin ligase function of FKF1, and it restricts this activity the light period of the day. GI inhibits the E3 ubiquitin ligase activity of ZTL, restricting it to the dark period [21,25,26]. Interestingly, GI is also required for the stability of the ZTL protein during the day and performs this action by recruiting the deubiquitinating enzymes UBP12 and UBP13 and acting as a co-chaperone with HSP90 [21,27,28].
In addition to its role in promoting protein interactions with regulatory proteins such as GI, the LOV domain is also involved in directly binding substrate proteins that are critical component of the SKP1/CULLIN/F-BOX (SCF) multi-subunit E3 ubiquitin ligase that is required for interaction with the other components of this complex [32][33][34]. We have previously demonstrated that the F-box domain is required for proper function of the LOV/ F-box/Kelch family of proteins, as the expression of "decoy" versions of ZTL and FKF1 that lack the F-box domain mimics published loss-of-function mutant phenotypes [13]. One role of the Kelch repeat domain is to promote interactions with substrates that are ubiquitylated. This information comes from studies of FKF1, where the Kelch repeat domain binds to and promotes the degradation of the floral repressors called the CYCLING DOF FACTORs (CDFs) [24]. Interestingly, ubiquitylation of these substrates relies on the interaction between FKF1 and GI [24,26].
In contrast, the ZTL Kelch repeat domain has not been demonstrated to interact with known ZTL substrates [13,[29][30][31]. However, the Kelch repeat domain is presumably important for ZTL function, as mutations in the Kelch repeat domain ablate ZTL function [10,20,30,35]. Additionally, expression of a truncated form of ZTL which contains only the LOV and F-box domains lengthens period similarly to a ztl loss-of-function mutant rather than shortening period as is observed in plants overexpressing full-length ZTL [7,10,22].
Together these data show that the Kelch repeat domain is important for ZTL's role in clock function, but its exact biochemical role remains unknown.
In this study, we investigate the genetic and biochemical functions of the ZTL Kelch role of the ZTL Kelch repeat domain using immunoprecipitation followed by mass spectrometry to identify a list of putative protein interacting partners. We find that FKF1 and LKP2 as well as the native ZTL protein are part of a complex with the ZTL Kelch repeat domain. Using yeast-two-hybrid analyses, we determine that the Kelch repeat domain interacts directly with the LOV domain of ZTL suggesting two possibilities: an antiparallel conformation for intermolecular homodimers or intramolecular interaction between the two domains. Our results suggest that a biochemical role of the ZTL Kelch repeat domain may be promoting homo-or hetero-dimerization with other LOV/F-box/Kelch proteins supporting previous data providing a possible mechanism for the role of ZTL in promoting auto-ubiquitylation and also mediating stability of FKF1. Furthermore, our genetic analysis suggests that the Kelch repeat domain may modulate the formation of higher order protein complexes that are essential for the function of LOV/F-box/Kelch proteins.

Expression of the LOV domain of LOV/F-box/Kelch proteins disrupt the circadian clock and flowering time
We have previously shown that expressing ZTL, LKP2, and FKF1 without the F-box domain results in disruption of circadian clock function and flowering time [13]. We next wanted to explore the roles of individual LOV and Kelch repeat domains separately and determine their effects on circadian clock pacing and flowering time. To do this, we overexpressed affinity-tagged LOV and Kelch repeat domains of FKF1, LKP2, and ZTL in the CCA1p::Luciferase (CCA1 promoter driving expression of firefly Luciferase) background, and monitored circadian clock period and flowering time. We included CCA1p::Luciferase plants that express FKF1, LKP2, and ZTL decoys (LOV-Kelch fusion proteins which lack the F-box domain), which we have analyzed previously [13], and wild type CCA1p::Luciferase parental plants, as controls. In order to compare results from experiments performed separately, we use the difference between the period or flowering time of the individual T1 transgenic and the average period or flowering time of the concurrent wild type control plants for our statistical analyses [36,37]. The data generated in these experiments is displayed in Figure   1 and Tables 1 and 2. We track period and flowering time for a large number of individual T1 transgenic insertion plants allowing us to avoid potential pitfalls of following single insertion lines that may be affected by genomic insertion location.  The function of the LOV domain is well defined [13,19,[29][30][31], allowing us to predict that overexpressing the LOV domains of ZTL, FKF1, and LKP2 would be sufficient to disrupt the functions of the endogenous proteins [22]. We overexpressed affinity tagged LOV domains from ZTL, LKP2, and FKF1 and monitored circadian clock period and flowering time ( Fig 1A- and triangles) [13]. For clarity we define the majority subpopulation as the one with the larger number of individuals, and the minority subpopulation, as the subpopulation with the smaller number of individuals.
ZTL, LKP2, and FKF1 have all been shown to regulate clock function [7,10,11,13,30]. LKP2 decoy in our previous study [13], here we identify subpopulations for the plants that   pink points). This is consistent with the idea that the LOV domain can sequester GI from the nucleus as was shown previously, but also that the Kelch repeat does not perform this function [22]. This information also demonstrates that expressing the Kelch repeat domain of the LOV/F-box/Kelch proteins can have dramatic effects on the circadian clock and flowering time confirming that this domain has an important function.

Determining the protein interaction profile of the ZTL Kelch repeat domain
Our genetic results indicate that the ZTL Kelch repeat domain plays an important role in the regulation of the circadian clock and flowering. However, to our knowledge no biochemical function has been attributed to this domain, and it is not believed to interact with known ZTL ubiquitylation substrates or regulatory partners. We hypothesize that the ZTL Kelch repeat domain may interact with unknown substrates or regulatory partners that may help elucidate its biochemical function. Thus, we performed an immunoprecipitation followed by mass spectrometry (IP-MS) experiment in our transgenic plants constitutively expressing a HIS-FLAG tagged ZTL Kelch repeat domain. We collected samples from plants grown in 12 hours light/12 hours dark growth conditions at three hours before dusk (ZT9) and three hours after dusk (ZT15) (S1 Table). As controls, we tag itself. Using this approach, we were able to identify 159 and 129 ZTL peptides from the Kelch repeat domain at ZT9 and ZT15, respectively suggesting that we were effectively immunoprecipitating the appropriate protein domain.
In our previous IP-MS studies using the ZTL decoy, we were able to identify peptides corresponding to the ZTL substrates TOC1, PRR5, and CHE and the regulatory proteins GI, UBP12, UBP13, and HSP90 [13]. While interaction studies in yeast have suggested that the Kelch repeat domain is not involved in these interactions [13,[28][29][30][31], it remains possible that the ZTL Kelch repeat domain could interact with known interacting partners in planta.
Thus, we searched our IP-MS results for peptides corresponding to the known ZTL interactors. We were unable to identify peptides corresponding to the majority of the known ZTL substrates and interacting partners (S2 Table). The only characterized interacting partners for which we were able to identify peptides were HSP90.1, HSP90.2, HSP90.4, and HSP90.5. However, we also identified peptides corresponding with these proteins in the controls. In order to determine whether the interactions between the ZTL Kelch repeat domain and HSP90 proteins was statistically significant, we performed SAINTexpress analysis [38,39] on our IP/MS results (S3 Table). We found that the interactions with HSP90.1 and 90.2 were statistically significant (SAINT score > 0. 5

and Log
Odds Score > 3), while the interactions with HSP90.4 and 90.5 were not statistically significant. The identification of a statistically significant interaction between the ZTL Kelch domain and HSP90 suggests that ZTL may be able to interact directly with HSP90 in the absence of GI, in addition to the ZTL-GI-HSP90 tri-partite complex that had been previously suggested [27,40]. However, lack of any peptides from other published ZTL interactors in our IP/MS results suggests that, consistent with previously published results [13,[29][30][31] and interacting partners of ZTL. It is possible that our assay was not sensitive enough to detect these interactions, but it is also possible that the ZTL Kelch domain plays an unknown role in the function of the protein through interaction with a unique group of protein partners.
We were unable to identify peptides corresponding to known ZTL targets and regulatory partners in our IP/MS results performed with the ZTL Kelch repeat domain. We next wanted to determine if other known clock or flowering time regulators interact with the ZTL Kelch repeat domain. We identified 640 statistically significant interacting proteins at ZT9, and 405 statically significant interacting proteins at ZT15. Of those proteins, 152 were identified at both ZT9 and ZT15 ( Fig 3A). We had previously performed IP-MS analysis at these same time points using plants expressing the ZTL decoy. As the ZTL decoy contains the Kelch repeat domain, we would expect that high-confidence Kelch repeat interactors would immunoprecipitate with the ZTL decoy and Kelch repeat [13]. For this reason, we compared the statistically significant interactors of the ZTL decoy with the interactors we identified in this study (Fig 3B, S4 Table). We identified 50 proteins that interacted with both the ZTL Kelch repeat domain and ZTL decoy at ZT9, and 40 proteins that interacted with both ZTL isoforms at ZT15. Of those proteins, 15 were identified as statistically significant interactors of both isoforms at both time points (Table 3). Six of those 15 proteins were subunits of the T-complex, molecular chaperones that assist in protein folding [41]. We also identify metabolic enzymes, a component of the 26S   We have identified a group of time and light independent ZTL Kelch repeat interacting proteins, but we also wondered if there are time dependent interactors [13,14,31]. There were 50 high-confidence interactors of both the ZTL Kelch repeat domain and the ZTL decoy at ZT9 and 40 at ZT15 ( Figure 3B, S4 Table). These proteins include numerous biosynthetic enzymes, additional components of the T-complex, and, at ZT9, HSP90.1 (S4 Table). FKF1 was also identified as a statistically significant interactor of the ZTL Kelch repeat domain at ZT9 (29 peptides), but not at ZT15 (0 peptides; Fig 4A-B). This aligns well with previous reports that show the ZTL Kelch repeat domain promotes interaction with FKF1 in heterologous systems [13,14,31]. We did not observe lightdependency for the interaction between the ZTL Kelch repeat and LKP2, as equal numbers of peptides at ZT9 and ZTL15 (20 and 21 peptides, respectively) were observed (Fig 4C-D).
These results confirm previous studies that suggest the ZTL Kelch repeat domain can promote heterodimerization in planta, but also expand on this idea and suggest that the interaction with FKF1 may be light dependent. The ZTL decoy is capable of interacting with the native ZTL protein, suggesting that ZTL is also capable of homodimerization [13]. However, it is unclear whether the ZTL Kelch repeat domain is sufficient to drive homodimerization. In order to determine whether we identify any peptides belonging to the native ZTL protein, we aligned each ZTL peptide identified by IP-MS to the ZTL protein sequence, and mapped it to the corresponding domain (Fig 4 E-F

The ZTL Kelch repeat domain interacts with the ZTL LOV domain
Our results suggest that the ZTL Kelch repeat domain is capable of interacting with native ZTL protein, and its identification as possessing the strongest circadian effect in our phenotypic assays suggests that its expression may disrupt higher order ZTL complexes.

Discussion Summary
It has long been known that ZTL is an essential E3 ligase for controlling proper periodicity in the circadian clock of Arabidopsis thaliana. However, the precise function of each protein domain has not been fully elucidated. We had previously investigated the role of the F-box domain in this protein by characterizing plants that express "decoy" forms of ZTL and its homologs that lack the F-box domain [13]. Here, we continued this process by The ability of the ZTL LOV domain and FKF1 Kelch repeat domain to inhibit native protein function is simplest to interpret: these characterized substrate interaction domains will preferentially interact with substrates and prevent their degradation. Similar interactions likely explain the ability of the LKP2 and FKF1 LOV domains to delay period when expressed, as both domains interact with TOC1 and PRR5 [30], although the different magnitudes of these phenotypes likely represent different affinities for these substrates.
Similarly, the ability of the LKP2 Kelch repeat domain to interact with the CDF proteins may cause the late flowering phenotype [24].  [19]. However, the ability of the native FKF1 protein to degrade the CDFs is dependent on the interaction with GI [26]. Overexpressing the FKF1 LOV domain may prevent the native FKF1 protein from interacting with GI, thus preventing degradation of CDFs and leading to delayed flowering. A similar effect may explain the extremely late flowering phenotypes of the minority population of ZTL LOV domain and LKP2 LOV domain expressing plants, as increased levels of the ZTL LOV domain drives GI localization towards the cytoplasm, preventing interaction between FKF1 and GI, which only occurs in the nucleus [22,26]. The absence of plants that exhibit an extremely late flowering phenotype when the ZTL Kelch repeat domain is expressed supports this hypothesis, as the ZTL Kelch repeat domain cannot directly interact with GI, and thus formation of the GI-ZTL or GI-FKF1 complex should be unaffected in a ZTL-Kelch repeat domain overexpression line [21].
To our knowledge, this study represents the first identification of a circadian defect dependent solely on the ZTL Kelch repeat domain. The identification of a large number of mutations in the Kelch repeat domain that ablate ZTL function suggests that the Kelch repeat domain is necessary [10,20,30,35]. However, previous studies have hypothesized that these mutations may affect ZTL protein function by destabilizing the structure of the entire ZTL protein [20]. Here, we have shown that the ZTL Kelch repeat domain is directly involved in circadian regulation, even in the absence of the LOV domain.

ZTL Kelch repeat interaction profiles
In this study, we identified a large suite of proteins which may potentially interact with the ZTL Kelch repeat domain, of which 15 were identified as statistically significant interactors in both our samples here and our previous study with on the ZTL decoy [13]. Of the ZTL homolog LKP2, is likely to play a role in circadian function. While not in our highconfidence list due to potential light-dependence, we also identify FKF1 and the native ZTL protein as putative interactors with the ZTL Kelch repeat domain.
We have noted previously that complex formation between E3 ubiquitin ligases and their homologs may be a common feature of this class of proteins, and the ability of ZTL, FKF1, and LKP2 to heterodimerize has been previously reported [13,14,31,36,37]. We demonstrate that the homodimerization takes place between the LOV and Kelch repeat domains, suggesting that the interactions between ZTL and FKF1/LKP2 may also occur in this manner. This suggests that a function of the ZTL Kelch repeat domain is to interact with the LOV domain and promote higher-order complex formation, thus modulating LOV/ F-box/Kelch protein activity.
We hypothesize that the interaction between the LOV and Kelch repeat domains of ZTL are required for its function. In support of this hypothesis, plants overexpressing a truncated form of ZTL containing only the LOV and F-box (ZTL LOV-F) are phenotypically indistinguishable from plants overexpressing the ZTL LOV domain alone [22]. If the Kelch repeat domain was dispensable for substrate ubiquitylation, one would expect that the plants overexpressing ZTL LOV-F would shorten period like plants overexpressing the fulllength ZTL protein [10,22]. However, as the expression of the ZTL LOV-F protein lengthens the period, it suggests this truncated form is non-functional, suggesting that the presence ZTL Kelch repeat domain is required for proper substrate ubiquitylation. We have demonstrated that the ZTL Kelch repeat domain can promote hetero-and homo-dimerization. By incorporating these interactions into models of ZTL protein function, we may begin to explain a structural conundrum of ZTL function. As F-box proteins typically have their substrate recognition domains on the C-terminus of the protein, the LOV domain is not located in a typical location for substrate ubiquitylation [32,42] and thus may be too spatially distant from the E2 conjugating enzyme to ubiquitylate LOV substrates (Fig 6A). An interaction between the LOV and Kelch repeat domains would bring the LOV domain into proximity with the E2 conjugating enzyme, and thus substrate ubiquitylation would occur ( Figure 6B  We cannot currently distinguish whether the LOV-Kelch repeat interaction occurs inter-or intra-molecularly. In the intermolecular model, the LOV domain and Kelch repeat domain of the same ZTL molecule interact with one another, folding the protein into a "closed" conformation to bring the substrates into proximity with the E2 conjugating enzyme ( Figure 6B). This model is consistent with published data stating that ZTL occurs as a monomer in planta [43]. In the intramolecular model, two ZTL proteins align with one another in an anti-parallel fashion, and two substrate molecules are shared between the two ZTL proteins ( Figure 6C). This model is more consistent with recent IP-MS data which suggests that ZTL interacts with itself to form higher-order complexes [13]. Furthermore, heterodimers of ZTL and LKP2 or FKF1 could form in the same manner as the anti-parallel ZTL homodimers. Future work will be required to distinguish between the inter-and intramolecular models of the LOV-Kelch repeat interaction.

The ZTL LOV-Kelch repeat interaction model
It is interesting to note that we do not identify any peptides that correspond to GI in our ZTL Kelch repeat domain IP-MS experiments despite the observed interactions with the native FKF1, LKP2, and ZTL proteins. While this may be due to technical limitations, it may also be that the ZTL LOV-Kelch repeat interaction disrupts the LOV-GI interaction.
Sequential co-immunoprecipitation experiments may prove whether the LOV-GI and LOV-Kelch repeat interactions are mutually exclusive. However, as interaction with GI inhibits ZTL E3 ligase activity [21], this suggests that the LOV-Kelch repeat conformation is the active ZTL conformation.

Plant materials
The creation of the ZTL, LKP2, and FKF1 decoy was described previously [13]. PCR was used to amplify the LOV and Kelch repeat domains of ZTL, LKP2, and FKF1, including everything N-terminal of the F-box domain within the LOV constructs and everything Cterminal of the F-box domain within the Kelch repeat constructs, using the primers in S5 table.
The amino acid numbers of the F-box domain can be found in Figure 4. PCR products were cloned into pENTR/D-TOPO vectors (Invitrogen, catalog no. K240020). The domains were then fused to FLAG and His tags at the N terminus and under the control of a cauliflower mosaic virus 35S promoter by recombination into the plant binary pDEST vector pB7-HFN [44,45] using LR recombination. The decoy constructs were transformed into Arabidopsis (Arabidopsis thaliana) Col-0 expressing the circadian reporter CCA1p::Luciferase [46] by the floral dip method [47] using Agrobacterium tumefaciens GV3101.

Phenotypic analysis
Control pCCA1∷Luciferase and transgenic seeds were surface sterilized in 70% ethanol and 0.01% Triton X-100 for 20 minutes prior to being sown on ½ MS plates (2. light panels (Heliospectra L1). Hourly images were acquired for approximately six and a half days. Each hour, lights are turned off for a total of eight minutes in order to capture a 5 minute exposure on an Andor iKon-M CCD camera; lights are off two minutes prior to the exposure and remain off for one minute after the exposure is completed. After imaging is complete, the lights return to the normal lighting regime. The CCD camera was controlled using Micromanager, using the following settings: binning of 2, pre-amp gain of 2, and a 0.05 MHz readout mode [48]. Data collected between the first dawn of constant light and the dawn of the sixth day are used for analyses.
The mean intensity of each seedling at each time point was calculated using ImageJ [49]. The calculated values were imported into the Biological Rhythms Analysis Software System (BRASS) for analysis. The Fast Fourier Transform Non-linear Least Squares (FFT-NLLS) algorithm was used to calculate the period, phase, and relative amplitude from each individual seedling [50].
Following luciferase imaging, seedlings were transferred to soil (Fafard II) and

Data normalization and statistical analysis
To allow for comparison across independent imaging experiments, data was normalized to the individual wild type control performed concurrently. The average value of the wild type control was calculated for every experiment, then this average was subtracted from the value of each individual T1 insertion or control wild type plant done concurrently. This normalized value was used for statistical analyses.
Welch's t-test was used to compare each normalized T1 insertion population or subpopulation to the population of normalized control plants. In order to decrease the number of false positives caused by multiple testing, we utilized a Bonferroni corrected α as the p-value threshold. The α applied differs between experiments, and is noted throughout.

Immunoprecipitation and mass spectrometry of plants expressing the ZTL Kelch repeat domain
Individual T1 pB7-HFN-ZTL-Kelch transgenics in a Col-0 background and control Col-0 and pB7-HFC-GFP were grown as described for phenotype analysis. Seven-day old seedlings were transferred to soil and grown under 16 hours light/8 hours dark at 22 °C for 2-3 weeks. Prior to harvest, plants were entrained to 12 hours light/12 hours dark at 22°C for 1 week. Approximately 40 mature leaves from each background was collected and flash frozen in liquid nitrogen, such that each sample was a mixture of leaves from multiple individuals to reduce the effects of expression level fluctuations. Tissue samples were ground in liquid nitrogen using the Mixer Mill MM400 system (Retsch). Immunoprecipitation was performed as described previously [44,45,51]. Briefly, protein Trapping was performed at 5µl/min, 97% Buffer A for 3 min using a Waters Symmetry® Laboratory at Yale University using MASCOT [52]. Data was searched against the SwissProt_2015_11.fasta Arabidopsis thaliana database with oxidation set as a variable modification. The peptide mass tolerance was set to 10 ppm, the fragment mass tolerance to 0.5 Da, and the maximum number of allowable missed cleavages was set to 2.
To determine statistically significant interactors, we removed all proteins that only occurred in the controls, then performed SAINTexpress using interface available on the CRAPome website [38,39]. Proteins with a SAINT score of greater than 0.5 and a Log Odds Score of greater than 3 were considered statistically significant.