Post-translational Regulation of the Arabidopsis Circadian Clock through Selective Proteolysis and Phosphorylation of Pseudo-response Regulator Proteins*

The circadian clock controls the period, phasing, and amplitude of processes that oscillate with a near 24-h rhythm. One core group of clock components in Arabidopsis that controls the pace of the central oscillator is comprised of five PRR (pseudo-response regulator) proteins whose biochemical function in the clock remains unclear. Peak expression of TOC1 (timing of cab expression 1)/PRR1, PRR3, PRR5, PRR7, and PRR9 are each phased differently over the course of the day and loss of any PRR protein alters period. Here we show that, together with TOC1, PRR5 is the only other likely proteolytic substrate of the E3 ubiquitin ligase SCFZTL within this PRR family. We further demonstrate a functional significance for the phosphorylated forms of PRR5, TOC1, and PRR3. Each PRR protein examined is nuclear-localized and is differentially phosphorylated over the circadian cycle. The more highly phosphorylated forms of PRR5 and TOC1 interact best with the F-box protein ZTL (ZEITLUPE), suggesting a mechanism to modulate their proteolysis. In vivo degradation of both PRR5 and ZTL is inhibited by blue light, likely the result of blue light photoperception by ZTL. TOC1 and PRR3 interact in vivo and phosphorylation of both is necessary for their optimal binding in vitro. Additionally, because PRR3 and ZTL both interact with TOC1 in vivo via the TOC1 N terminus, taken together these data suggest that the TOC1/PRR3 phosphorylation-dependent interaction may protect TOC1 from ZTL-mediated degradation, resulting in an enhanced amplitude of TOC1 cycling.

The rotation of the earth on its axis yields dramatic diurnal changes in the environment. This periodic environmental change has driven the evolution of an internal timekeeping mechanism, the circadian clock, to allow the prediction of impending environmental change and synchronization with this 24-h environmental periodicity. The possession of an internal circadian clock that resonates with the environmental period enhances fitness and impaired clock function confers a striking loss of fitness (1). In plants this includes greatly reduced net photosynthesis and increased mortality (2,3).
At the core of the eukaryotic clock are one or more autoregulatory negative feedback loops: clock genes are transcribed and transcripts are translated into proteins that inhibit further transcription. Degradation of both mRNA and protein relieves this inhibition and the cycle renews. One of the feedback loops of the Arabidopsis thaliana clock has two partially redundant single Myb domain transcription factors, circadian clock associated 1 (CCA1) 6 and late elongated hypocotyl (LHY), negatively regulating the transcription of TOC1 (4). Current models of the Arabidopsis clock propose two to three additional negative feedback loops interlocked with the initial CCA1/LHY/ TOC1 loop (5,6).
TOC1 encodes a pseudo-response regulator (PRR), and is the founding member of a family of PRR genes that contribute to the Arabidopsis circadian clock. Transcript abundance of each of five PRR genes oscillates in circadian fashion, peaking sequentially from shortly after dawn to approximately dusk in the order PRR9-PRR7-PRR5-PRR3-TOC1/PRR1 (7). Loss of function for each of the PRR genes alters circadian period phenotypes, with toc1, prr3, and prr5 mutations shortening periods and prr7 and prr9 mutations lengthening periods (8 -16). CCA1 and LHY serve as positive regulators of PRR7 and PRR9, which in turn negatively regulate CCA1 and LHY (12,14,17). PRR7 and PRR9 are partially redundant with each other and with PRR5, because the double prr7 prr9 mutant shows extreme period lengthening, more than can be accounted for by additive interaction of the two single mutants (11,12,14), and the triple prr5 prr7 prr9 mutant is arrhythmic and exhibits constitutive CCA1 expression.
Recently post-transcriptional and post-translational processes, such as phosphorylation and clock-regulated proteolysis of oscillator proteins, have been found to be critical for proper clock function in a number of eukaryotes (18 -24). In Arabidopsis, CCA1 phosphorylation by casein kinase 2 is essential for its role in the clock (25)(26)(27). Similarly, proteasome-dependent proteolysis post-translationally controls the levels of ancillary or core clock components in Arabidopsis, such as LHY, a subunit of casein kinase 2 (CKB4), GIGANTEA (GI), and ZTL (28 -31). Additionally, the E3 ubiquitin ligase SCF ZTL targets TOC1 and PRR5 proteins for degradation (32)(33)(34)(35). As the five PRR proteins share two highly conserved domains, the pseudo-receiver (PR) and the CONSTANS, CONSTANS-LIKE, and TOC1 (CCT) domains, we investigated whether ZTL may target a greater set of these proteins for proteasome-dependent degradation, and how their phosphorylation state affects their stability.
Here we establish that the five PRR family proteins cycle with a slightly lagging phase relative to respective transcripts. All of the PRR proteins examined are nuclear-localized and phosphorylated, and many of them become progressively phosphorylated over the circadian cycle, with maximum phosphorylation prior to degradation. We show that TOC1/PRR1 and PRR5 are likely the only PRR proteins of this family proteolytically controlled directly by SCF ZTL , but that ZTL indirectly controls PRR9 protein levels. A second post-translational process, phosphorylation, controls TOC1 and PRR3 interaction in vivo via the TOC1 N terminus. Our data suggests a complex post-translational relationship between TOC1, PRR3, and ZTL that helps establish a strong amplitude to TOC1 cycling.

EXPERIMENTAL PROCEDURES
Plant Materials and Growth Conditions-Construction of the TOC1-YFP minigene (TMG) in the Arabidopsis wild type (WT) (TMG WT) and ztl mutant (TMG ztl-3) backgrounds have been described previously (33). The genomic fragments for PRR3, PRR5, PRR7, and PRR9 (generically PRRn), containing the full promoter, 5Ј-untranslated region, and coding sequences up to the last codon before the STOP codon, were amplified by PCR with ExTaq (Takara Bio USA, Madison, WI) and subcloned into pGEM-T Easy (Promega, Madison, WI). The PRR:PRRn genomic fragments were then subcloned into the Gateway Entry vector pENTR-1A, and placed upstream of the green fluorescent protein (GFP) by LR recombination (Invitrogen) into the binary vector pMDC206 (36) according to the manufacturer's instructions. The genomic sequences cloned upstream of GFP were: Ϫ1431 to ϩ2285 bp (PRR3), Ϫ5116 to ϩ2340 bp (PRR5), Ϫ1194 to ϩ3176 bp (PRR7), and Ϫ1541 to ϩ2025 bp (PRR9). Plants were transformed by floral dip (37) via Agrobacterium tumefaciens strain GV3101. Primary transformants were selected for hygromycin resistance, and allowed to self-pollinate. PRR:PRRn-GFP constructs were transformed into WT Arabidopsis (Col) and ztl-1 (Col), backgrounds as described in the text, and multiple transformants were identified and characterized for each construct. One to three similarly expressing lines were chosen for the experiments presented.
Cycloheximide Treatments-Plants grown under LD cycles for 10 days were transferred to continuous white, red, blue light, or dark at ZT0 and transferred into liquid MS media with 0.01% Triton X-100 and 100 M cycloheximide at CT7 or CT19. Plants were immersed in the media and shaken slowly for the indicated durations before harvesting.
Protein Extraction and Immunoblot Analysis-Protein extraction for TOC1-YFP, PRR3-GFP, PRR5-GFP, PRR7-GFP, and PRR9-GFP detection was performed as described previously (42). Protein extraction for ZTL detection was described previously (31). Immunoprecipitations of PRR5-GFP, PRR3-GFP, or TMG and co-immunoprecipitation of ZTL were performed as described previously with minor modifications (42). 0.5 l of anti-GFP polyclonal antibody (ab290, Abcam) was incubated with 12 l of protein A-agarose (Invitrogen) made to a final volume of 25 l at 4°C for 2 h. Proteins were extracted from 500 l of ground tissue using 500 l of immunoprecipitation buffer, and the extract was added to the affinity matrix and incubated for 1.5 h with gentle rotation. Immunoblot analysis was performed as described previously (42).
Phosphatase Assays-For binding assays, protein was extracted with ice-cold buffer A with 2.5 mM MnCl 2 added. 4,000 units of -protein phosphatase (New England Biolabs) were added to 1 ml of soluble extracts. The mixture was incubated at 30°C for 15 min. Samples were loaded into 8% SDS-PAGE (acrylamide:bisacrylamide, 149:1) and analyzed by immunoblot. For PRR phosphorylation tests, protein extracts were prepared in 1ϫ -phosphatase buffer supplemented with 2.5 mM MnCl 2 , 0.5% Triton X-100, 0.4% Nonidet P-40, 2.5 g/ml antipain, 2.5 g/ml chymostatin, 1 g/ml pepstatin, 5 g/ml leupeptin, 5 g/ml aprotinin, 25 mM phenylmethylsulfonyl fluoride and proteasome inhibitors as described above. Forty to fifty-microliter aliquots of the resulting protein extracts were incubated with 400 units of -protein phosphatase at 30°C for 15 min. Calf intestine alkaline phosphatase (CIP) assays were performed using protein extract prepared in New England Biolabs buffer 3 supplemented with protease inhibitors and proteasome inhibitors as described above. Fortymicroliter aliquots of the resulting protein extracts were incubated with 10 units of CIP (New England Biolabs) with or without phosphatase inhibitors (NaF/Na 3 VO 4 ) at 37°C for 15 min.
Transient Expressions and Immunoprecipitations-Assays using co-infiltrated Agrobacterium-mediated protein expression in Nicotiana benthamiana have been described previously (42). All constructs were 35S cauliflower mosaic virus promoter-driven, except ZTL, which was a native promoter-driven genomic clone (39). For TAP-TOC1/PRR3 assays, total N. benthamiana protein was extracted post-infiltration with ice-cold buffer A using 300 l of tissue, and 600 l of cleared supernatant was incubated with 25 l of human IgG-agarose at 4°C for 90 min. After rinses, the resin was further treated with or without protein phosphatase prior to subsequent immunoprecipitations. The resin was rinsed with protein phosphatase buffer, and incubated with -protein phosphatase (400 units) at 30°C for 15 min. For immunoprecipitations, total protein was extracted from PRR3-GFP Arabidopsis at the required time points. For dephosphorylation tests of PRR3-GFP, CIP was used as described above. Extracts were incubated with the previously prepared TAP-TOC1-loaded resin. After incubation at 4°C (90 min), the resin was rinsed and Precission protease (0.5 l) was added to release TOC1-TAP and associated proteins, and the supernatant was separated by SDS-PAGE. Immunoblot detection was as previously described.
Accession Numbers-Arabidopsis Genome Initiative locus identifiers for the genes mentioned in this study are as follows:

Cyclic Expression of the PRR Proteins Lags behind the Respective Message
Rhythms-TOC1 is the founding member of a five-member family of PRR and is a proteolytic target of the clock-related F-box protein ZTL. To determine whether other PRR/TOC1 family members are additional targets, we first characterized their endogenous protein expression pattern. C-terminal GFP fusions were made to the genomic sequences of PRR3, 5, 7, and 9 and stably transformed into Arabidopsis under the expression of each respective native promoter (PRR::PRRn-GFP). Multiple independent transformants were obtained for each construct and representative expression patterns for each are shown in Fig. 1. The period of each transgenic line chosen is slightly different from WT and generally in the opposite direction of the period effects of the known loss-offunction mutants of each gene (supplemental Table S1), consistent with the expected effects of slightly elevated expression. Free-running periods of PRR3::PRR3-GFP transformants are 1-1.5 h longer than WT in red and blue light, consistent with the slightly shorter period of the prr3 mutant (10). PRR5::PRR5-GFP transformants are 2-2.5 h longer than WT, consistent with the more severe shortening effect seen in prr5 mutants (8,10,16). The PRR7::PRR7-GFP transformants are slightly longer than WT in red and blue light, similar to long period of prr7 mutants. This unexpected finding has also been reported in plants strongly (43) and weakly (44) overexpressing PRR7, suggesting a complex relationship between PRR7 levels and period. PRR9::PRR9-GFP plants are similar to WT in red light, and slightly shorter in blue light, consistent with the long period of prr9 mutants in white light. Together these results indicate that each fusion protein is functional.
The abundance of each of the PRR proteins oscillates under diurnal cycles with a peak expression often lagging 4 -8 h after the expression peaks of the respective message levels (Fig. 1). Notably, PRR9 and PRR3 persist at high levels well past the peak of mRNA expression and diminish rapidly upon transition to dark or light, respectively (Fig. 1, A and D) (45). To determine whether these declines are due to light/dark transitions, we tested expression under constant white light (LL) (Fig. 1E). In LL all five proteins continue to cycle very robustly, consistent with previous reports of circadian expression of PRR message levels (7). These results imply a rapid turnover of the five PRR proteins and confirm and extend previous reports (44,45).
Nuclear Localization of PRR Proteins-TOC1 (PRR1) has been shown in transient assays to be nuclear-localized (46). We tested whether other PRR proteins were similarly present in the nucleus. Representative lines expressing PRR3-GFP, PRR5-GFP, and PRR7-GFP under the respective endogenous promoters and grown under light/dark cycles were harvested and fixed at times of strong abundance (between ZT 8 and 12) and immunolocalized through the GFP moiety (Fig. 2). All three proteins are distinctly and strongly nuclear-localized in the shoot meristem (Fig. 2, D-M). In all other tissues tested (leaf mesophyll and vasculature, and root meristems) nuclear localization was also distinct (Fig. 2, N-R). These findings are consistent with transient localization assays using strong ectopic promoters (35S cauliflower mosaic virus) (47), and suggest a nuclear function for all four of the PRR proteins tested here.
TOC1 and PRR5 Are the Sole Proteolytic Targets of ZTL within the PRR Family-To determine whether ZTL might target members of the PRR family beside TOC1, independent transformations of the each of the PRR::PRR-GFP constructs were made into the ztl-1 background. The single missense mutation in the kelch domain in ztl-1 is functionally equivalent to a ztl null mutation, but has WT levels of the ZTL protein (39,48). One line showing a representative expression pattern was chosen for each PRR construct for further analysis.
All PRR9::PRR9-GFP lines showed little to no expression of PRR9-GFP protein (data not shown). We examined mRNA expression patterns of the transgene and of endogenous PRR9 in ztl-1 and found very low expression of both in all lines tested (Fig. 3A). Because enhanced accumulation of TOC1 occurs in the absence of ZTL (33), this result is consistent with a previous report showing that strong expression of TOC1 severely depresses PRR9 message levels (49). Due to this strong negative regulation of PRR9 expression in ztl mutants we were unable to examine the direct effect of reduced ZTL on PRR9 protein. This result, together with the anti-phase expression of PRR9 mRNA and ZTL protein strongly suggests that PRR9 is not a substrate of ZTL.
Of the remaining four PRRs, PRR3 and PRR7 protein showed similar amplitude and patterns of rhythmic expression in the ztl-1 and WT backgrounds (compare Fig. 1, B and D, with Fig. 3B). We tested whether PRR3 or PRR7 interact with ZTL in vivo through co-immunoprecipitation tests. Using could we detect ZTL in co-immunoprecipitates with PRR3-GFP, or even in plants expressing PRR3 at very high levels (35S cauliflower mosaic virus::GFP-PRR3) (supplemental Fig. S1). We found similar results using PRR7::PRR7-GFP plants (data not shown). Additionally, the free-running period of the prr5 prr3 double mutant is not additive but is indistinguishable from the short period prr5 mutant, suggesting PRR3 acts upstream of PRR5, or it may act in support of PRR5 activity, but not independently of it (supplemental Fig. S2C).
In contrast, we confirm that TOC1 cycling is largely eliminated in the ztl-1 mutant, due to the stabilization of the TOC1 protein (Fig. 3B) (33). Additionally, PRR5 shows strongly reduced rhythmic amplitude in ztl-1, with protein levels damping to high or intermediate levels, relative to WT (Fig. 3B). These results suggested that PRR5 may be an additional proteolytic target of ZTL, and we performed similar co-immunoprecipitation experiments using PRR5::PRR5-GFP expressing plants. Fig. 3C shows ZTL present in the immunoprecipitates at ZT9 and ZT17, indicating an interaction between the two at native expression levels, and that this interaction is stronger in the dark. The ztl-1 mutation abrogates this interaction (Fig. 3C), similar to its effect on TOC1 (33). Taken together, these data strongly support full-length PRR5 as an endogenous interactor with ZTL, confirming and extending similar results using strong ectopic expressors or truncated versions of PRR5 (35).
We next tested whether PRR5 is also a substrate of the SCF ZTL ubiquitin ligase. Cycloheximide (CHX) treatment of WT and ztl-1 seedlings at two time points in constant light (subjective day, LL 7 and subjective night, LL 19) was used to determine whether PRR5-GFP undergoes differential degradation rates over the circadian cycle. PRR5-GFP is more rapidly degraded during the subjective night than during subjective day, with a half-life at LL19 of about 2.5 h compared with almost 9 h at LL7 (Fig. 3D). The greater stability early in the day is consistent with the very low levels of ZTL during this time (31). In the ztl mutant PRR5-GFP is significantly more stable during the subjective day and night, relative to the WT (Fig. 3E). These results, together with the physical interaction with ZTL, demonstrate that PRR5 is a proteolytic target of ZTL in lightgrown plants, consistent with similar results reported for darkgrown etiolated seedlings or strong ectopic expressors of PRR5 (35). These findings are further supported by genetics, where the period of the double mutant ztl-3 prr5 is intermediate between that of the ztl and prr5 single mutants, consistent with PRR5 being one of at least two ZTL substrates (supplemental Fig. S2A). Similarly, if PRR5 and TOC1 are co-targets of the same SCF ZTL complex then the prr5 toc1 double mutant should be additively more severe than either single mutant alone. Con- sistent with this notion, the free-running period in the prr5 toc1 background is significantly shorter (18.4 h) than that of the toc1 (19.9 h) and prr5 (23.2 h) hours for the toc1 single mutants (supplemental Fig. S2B) (50).
Blue Light Stabilizes PRR and ZTL Protein Levels-ZTL-mediated TOC1 degradation is thought to be facilitated by the dark, implying a potential protective role by light (33). We tested whether this is also true for PRR5, and whether proteolysis differs under different light qualities. Light-grown entrained plants were shifted to continuous dark, blue or red light and harvested at different times after CHX application at CT 19. In blue light, PRR5 was significantly more stable, relative to red light and darkness (Fig. 4A) (35). Interestingly, ZTL levels were also more stable under blue light, relative to red light, under the same conditions (Fig. 4B). These results support the notion that under blue light, degradation of PRR5 is inhibited, possibly through a light-mediated disruption of the ZTL/PRR5 interaction.
Phosphorylation of the PRR proteins varies throughout the diurnal cycle. In other systems, phosphorylation of F-box protein targets is required to facilitate interaction (51,52). To further dissect the nature of the ZTL/PRR5 interaction we first examined whether PRR5 is phosphorylated at any time during the circadian cycle. Immunoblots detecting PRR5-GFP showed at least two bands of distinctly different mobilities with a similar relative abundance of both present at all time points over a diurnal cycle (Fig. 5A). We next determined that both the upper and lower forms of PRR5-GFP are the result of greater and lesser degrees of protein phosphorylation, respectively. Extracts treated with -PP shifted the PRR5-GFP mobility to lower than either of the two normally observable bands, indicating that both forms of PRR5-GFP are phosphorylated to some extent at nearly all times during the circadian cycle (Figs. 5C and 6B). Occasionally we observed a third, very rapidly migrating form at ZT21 that corresponds to the most dephosphorylated form of the protein (Fig. 5A).
Similar experiments were performed with TOC1-YFP extracts and we resolved two bands during the photoperiod (Fig. 5B). During the dark period, however, only a single, more slowly migrating band was detectable. Phosphatase treatment of TOC1-YFP extracts at ZT1 and ZT13 shows that the slower migration of the upper band results from phosphorylation (Fig.   5C). We also tested the PRR3-GFP and PRR7-GFP proteins for phosphorylation status. The multiple, ladder-like array of PRR3-GFP bands detectable during the dark period collapsed to essentially a single rapidly migrating band upon phosphatase treatment (Fig. 5C). Similarly, PRR7-GFP migration is increased in the presence of -phosphatase (Fig. 5C). Hence, all four PRR proteins undergo diurnal variation in both protein abundance and phosphorylation state.
Phosphorylation of TOC1 and PRR5 May Enhance ZTL Binding-We next tested whether the interaction of ZTL with the two PRR substrates (TOC1 and PRR5) is affected by phosphorylation status. Because both proteins show detectable differences in mobility we first tested binding to ZTL-GST using protein extracts from time points when both forms of the respective protein are detectable. Both the slower and faster migrating forms of TOC1-YFP can bind ZTL (Fig. 6A). When we then tested phosphatase-treated extracts compared with mock treated samples, both forms of TOC1-YFP bound to ZTL, . TMG seedlings were entrained in LD and harvested at ZT13. Total protein extracts were incubated with or without -protein phosphatase 15 min at 30°C prior to incubation with ZTL-GST resin. B, binding of PRR5-GFP expressing Arabidopsis protein extracts to ZTL-GST resin. PRR5-GFP (PRR5::PRR5-GFP) expressed in wild type Columbia seedlings were entrained and harvested as in A. Total protein extract preparation, -PP treatment, incubation with ZTL-GST resin and sample separation (8% SDS-PAGE (acrylamide:bisacrylamide, 149:1)) are as described in A. PRR5-GFP was detected with an anti-GFP antibody. ϪPP and ϩPP, extracts incubated with or without -PP, respectively. ϪPP*, extracts preincubated on ice prior to ZTL-GST resin incubation. ZTL-GST, ZTL-GST resin; GST, GST resin alone. All blots are representative of three trials.
Experiments with PRR5-GFP gave similar results. Both faster and slower migrating bands bound to ZTL-GST (Fig. 6B). PRR5-GFP from phosphatase-treated extracts still associated with ZTL-GST, but with reduced affinity. Together these results indicate that phosphorylated forms of both TOC1 and PRR5 bind ZTL, and that this modification may act to modulate their respective abundances.
Phosphorylation of TOC1 and PRR3 Is Necessary for Robust Interactions-PRR3 and TOC1 can interact in yeast and in planta, and PRR3 has been implicated in the stabilization of TOC1 protein (53). We examined whether the phosphorylation state of these proteins influences their interactions. First we established that TOC1 and PRR3 interact in planta through co-infiltration of TOC1 (TAP-TOC1) and PRR3 (PRR3-GFP) in N. benthamiana (Fig. 7). Full-length and the N terminus (amino acids 1-243) of TOC1 interact well with PRR3, whereas the TOC1 C terminus (amino acids 243-648) shows no detectable affinity (Fig. 7A). Interestingly, this is the same region of TOC1 that interacts with ZTL (Fig. 7B).
We next bound TAP-TOC1 expressed in N. benthamiana to IgG resin and tested whether TOC1 phosphorylation affected in vitro binding of PRR3-GFP extracted from Arabidopsis seedlings. Pre-treatment of resin-bound TOC1 with -phosphatase resulted in a single, faster migrating TOC1 band upon proteasemediated release from the resin, indicating the effectiveness of the phosphatase treatment (Fig. 8A). Interaction between TOC1 and PRR3 is strongly diminished by this pre-treatment of TOC1 with phosphatase (Fig. 8A).
We next observed the relative binding affinity of PRR3-GFP from tissue harvested at different time points to phosphorylated TOC1-TAP. The strong diurnal and circadian-regulated progression of PRR3 phosphorylation over time (Fig. 1, D and  E) suggested a way to assess relative binding affinity within a single extract. Although the less phosphorylated forms of PRR3 predominate at ZT10 and -13, there is a disproportionately larger fraction of more highly phosphorylated PRR3 bound to TOC1 at these times (Fig. 8B). Similarly, at ZT17, when highly phosphorylated PRR3 is most abundant, this form binds extremely well. In support of this finding, CIP pretreatment of PRR3-GFP Arabidopsis extracts results in the dephosphorylation of PRR3 and the subsequent loss of TOC1 binding (Fig.  8C). Taken together these results support the notion that phosphorylation of both TOC1 and PRR3 is necessary for optimal interaction.
PRR3 as an Inhibitor of the ZTL/TOC1 Interaction-TOC1 levels are diminished in vivo in the absence of PRR3, and levels are enhanced when PRR3 levels are enhanced. Additionally, yeast three-hybrid tests suggest that PRR3 might interfere with the TOC1/ZTL interaction (53). These reports, together with our above findings, recommended an in planta test of this hypothesis. We co-expressed GFP-PRR3, TAP-TOC1, and ZTL in N. benthamiana and observed the amount of ZTL coimmunoprecipitated with TOC1 in the presence and absence of PRR3. PRR3 diminished the relative amount of ZTL found in association with TOC1. Similar tests using only the N terminus of TOC1 gave similar results (Fig. 9). Because the same N-terminal region of TOC1 interacts with ZTL and PRR3, these results are consistent with PRR3 acting to competitively inhibit the ZTL/TOC1 interaction, contributing to the phase-specific enhancement of TOC1 levels.

DISCUSSION
To determine additional targets of the F-box protein ZTL and understand their post-translational regulation, we have characterized the five PRR proteins in Arabidopsis that previously have been identified as components of the plant circadian clock (10 -12, 44, 46, 54, 55). The high amplitude cycling of the PRR proteins, which lag behind the respective message levels, demonstrates a rapid turnover of all five proteins, indicating a highly regulated control of protein abundance. High level ectopic expression of these proteins have marked effects on the circadian period but these previous studies altered both the phase and level of expression (43,49,56,57). Results presented here, together with the prr mutant phenotypes, suggest a strong significance to the high amplitude, phase-specific expression of the PRR proteins. Input of TOC1 and its deletions was detected with an anti-PAP antibody, whereas immunoprecipitated TOC1 and its deletions were detected with an anti-His antibody. ZTL was detected with anti-ZTL antibody. Ϫ, co-infiltration of p19-expressing Agrobacterium with PRR3-GFP or ZTL used as a negative controls. All blots are representative of three trials.

ZTL Targets a Subset of Late-phased PRR Proteins to Control
Circadian Period-Through combined approaches involving examination of gene expression patterns, in vivo protein-protein interactions, and genetics we have delimited the PRR target substrates of ZTL to TOC1 and PRR5, confirming and extending previous reports (33,35). Both proteins have very similar phases of expression and act similarly to lengthen the free-running period. The additivity of their mutant circadian pheno-types suggests they may be part of a common complex, which may have diminished activity in the absence of one or the other, or they may participate in separate, similarly acting complexes. Further identification of TOC1 and PRR5 interactors will be necessary to resolve these possibilities. Previous studies have demonstrated the severe loss of circadian cycling in prr5 prr7 and prr5 prr7 prr9 mutants (13,14). These results are consistent with multiloop models where early phased and late phased PRR proteins comprise two separate functional groups, implying that the simultaneous loss of single members from both groups should have a more severe reduction of circadian function than double mutants within either of the two groups (5,6,54). The different combinations of late phased toc1, prr3, and prr5 mutants examined here do not result in rapid damping but are simply a shorter period. These data are consistent with the notion that these three proteins together comprise, or are part of, a functional unit of the circadian oscillator.
Light-dependent Regulation of ZTL Targets-TOC1 and PRR5 mRNA and protein show maximal expression during the subjective night, in phase with ZTL peak expression (31). This is unexpected given that high levels of ZTL should result in concomitantly low substrate (TOC1 and PRR5) levels. The greater stability of PRR5 in blue light is consistent with the increased PRR5/ZTL interaction we observed in the dark (Fig. 3C), with both results implying an inhibition of interaction and degradation of PRR5 in the light. The greater stability of ZTL under blue light further suggests the presence of a factor that simultaneously stabilizes ZTL but restricts targeting of PRR5 for degradation. A strong candidate for this factor is GI. Recently, we have identified GI as a post-translational stabilizer of ZTL through an interaction enhanced by blue-light dependent activation of the ZTL LOV domain (42). GI binding via the LOV motif may thus simultaneously stabilize ZTL while conformationally blocking access of ZTL substrates to the kelch interaction region. The GI/ZTL interaction would allow the increase of all four proteins (GI, ZTL, TOC1, and PRR5) during the  photoperiod, followed by the dissociation of GI and ZTL in the dark, freeing ZTL to associate with and degrade the PRR substrates. Thus a combination of clock-controlled phasing of transcription (GI, TOC1, and PRR5) together with blue lightenhanced interaction and stabilization (GI and ZTL) could result in the similarly late phased, high amplitude cycling observed for these essential clock components. Indirect Transcriptional Regulation of PRR9 by ZTL-Two previous reports have described the strong suppression of PRR9 message levels when TOC1 and PRR5 are strongly overexpressed (49,56). Our results, and others, support these ectopic studies by showing that when TOC1 and PRR5 levels rise and lose their distinctly phased expression patterns in ztl mutants PRR9 expression is strongly dampened. These data demonstrate how ZTL, through its role in the control of PRR5 and TOC1 protein levels, indirectly controls the phase and amplitude of PRR9 expression. Current computational models of the plant circadian clock are comprised of three or four interlocked loops, incorporating five known components (LHY/CCA1, PRR7/PRR9, and TOC1) along with a number of hypothetical factors, and these models are largely built around transcriptional activation and repression (5,6). Our results make explicit the central role of protein stability in the control of transcription and in circadian timing.
Roles for Phosphorylation of TOC1 and PRR5-During the photoperiod, TOC1 exists in two states of either lesser or greater phosphorylation, but appears in the more fully phosphorylated form during the dark period. In contrast, PRR5 dis-plays two phosphorylation states nearly unchanged in relative abundance during the entire diurnal cycle. One possible function of the phosphorylation is enhanced binding to ZTL, because in yeast and other systems substrate phosphorylation generally enhances interaction with the cognate F-box proteins (51). However, recent reports from plant systems indicate this is not always the case. The Arabidopsis AUX/IAA proteins do not appear to require phosphorylation for binding and subsequent ubiquitylation by the TIR1 family of F-box proteins (58). Similarly, the DELLA substrate of the rice SCF GID2 complex, SLR1, binds to GID2 in both the phosphorylated and unphosphorylated forms (59,60). Our findings support these results, as TOC1 and PRR5 interactions with ZTL are only mildly affected by their phosphorylation status, with slightly higher binding correlating with a more phosphorylated forms.
More significant is the strong effect of phosphorylation on the promotion of the PRR3 and TOC1 interaction. This finding correlates very well with the coincidence in vivo of an advanced level of PRR3 phosphorylation with the more highly phosphorylated form of TOC1 late in the subjective night. The notion that an enhanced interaction between PRR3 and TOC1 during this time results in protection of TOC1 from degradation by ZTL is consistent with a recent report demonstrating higher levels of TOC1 in a PRR3 overexpressor and reduced TOC1 in the prr3 mutant (53) (see Fig. 10). Furthermore, the N terminus of TOC1 is required for both the PRR3 and ZTL interaction, establishing this domain as critical to sustaining the appropriate TOC1 levels necessary for proper circadian function. Fur- . TOC1/ZTL interaction is diminished by PRR3. ZTL and TAP fusions of full-length or N terminus TOC1 were co-expressed with or without GFP-PRR3 in N. benthamiana using the same techniques and constructs as previously described, and protein extracts were used for immunoprecipitation (IP) by IgG-agarose. TAP immune complexes probed for TAP-TOC1 and co-immunoprecipitated (co-IP) ZTL are shown. TOC1, TAP-TOC1; TOC1NT, TAP-TOC1 N terminus; PRR3, GFP-PRR3. Representative of three trials is shown. All proteins were still present in post-binding supernatants (data not shown). to proteasome-dependent degradation through direct interaction. TOC1 and PRR3 proteins are phosphorylated and their phosphorylation state varies diurnally and circadianly. Early after lights-on, TOC1 is expressed at low levels in both a lesser (or non-) and the greater phosphorylated form, whereas the latter form is only present during the night. Similarly, PRR3 protein shows ladder-like multiply phosphorylated forms predominately at night. TOC1, PRR3, and ZTL all show diurnal protein levels peaking around ZT13. Although the ZTL/TOC1 interaction is slightly enhanced by TOC1 phosphorylation, the phosphorylated forms of TOC1 and PRR3 strongly preferentially interact. In this model, ZTL and PRR3 compete for interaction with the N terminus of TOC1, and following the enhancement of the TOC1/PRR3 interaction through phosphorylation, this pairing results in greater TOC1 stabilization by competitively blocking the ZTL/TOC1 interaction. The black bar indicates night.
ther studies will define the kinases responsible for PRR3 and TOC1 regulation and how subcellular levels of TOC1 may be controlled by phosphorylation-dependent interactions.