Circadian Entrainment in Arabidopsis by the Sugar-Responsive Transcription Factor bZIP63

1Circadian entrainment to sugars involves the regulation of PRR7 transcription. In response to sugars, the wild-type circadian period shortens, whereas that of prr7-11 does not. Previous investigation of candidate regulators of PRR7 failed to identify candidates affecting the response of the circadian oscillator to sugars. We hypothesized that the transcription factor (TF) bZIP63 might regulate PRR7 because bZIP63 is regulated by the SnRK1 energy sensor and bZIP63 transcripts peak before PRR7 in constant light (Figure S1A). bZIP63 is a strong candidate for sugar-mediated regulation of the circadian oscillator because it binds ACGT core element motifs, and the PRR7 promoter contains five ACGT-core bZIP TF-binding motifs within 300 bp of its transcription start site, including a canonical G-box at −254 bp (Figure 1A). bZIP63 binds a PRR7 promoter region spanning −276 to −182 (Figures 1A and 1B; primer pair 1). PRR7 transcripts were downregulated in T-DNA insertion mutants and RNAi lines of bZIP63 (Figures 1C, S1B, and S1C) and upregulated in bZIP63 overexpressors (Figures 1C and S1B). We measured PRR7 transcript abundance under normal and low light, which mimics starvation, demonstrated by accumulation of the marker transcript DARK INDUCIBLE6 (DIN6) (Figure S1D). Under low-light conditions that deplete endogenous sugars, PRR7 transcripts accumulated in the wild-type before dawn (Figure 1D), whereas this was attenuated in bzip63 mutants (Figure 1D). Under high light, which elevates endogenous sugars (Figure S1D), bzip63 mutations had little effect on PRR7 transcript abundance; bzip63-1 was without effect and bzip63-2 reduced PRR7 transcript abundance slightly (Figure 1D). Sucrose supplementation and high light both suppress PRR7 transcript accumulation (Figure 1D), and bzip63 mutants prevent PRR7 upregulation under low light (Figures 1D and 1E). These data suggest that bZIP63 upregulates PRR7 in low-energy conditions and that bzip63-2 is a stronger allele (Figure 1D). At the end of the photoperiod, the bzip63 mutations did not affect PRR7 transcript abundance, consistent with the effects of sugar on PRR7 being restricted to the early photoperiod (Figure 1E). PRR7 transcripts decreased in bzip63-2 only at the night end, whereas CIRCADIAN CLOCK ASSOCIATED1 (CCA1) and its target GBS1 were upregulated at ZT20-24 (Figure S1E). Upregulation of CCA1 in bzip63-2 might be due to downregulation of PRR7. Therefore, bZIP63 upregulates PRR7 in response to low energy. This is suppressed by sugars because both sucrose supplementation and bzip63 mutants inhibit PRR7 transcript accumulation.

2Because we found a PRR7-mediated circadian system-wide effect of bZIP63, we investigated whether bZIP63 underlies a response of the circadian oscillator to sugars. Unlike the wild-type, bzip63-1 circadian period was unaffected by sucrose under low light (Figure 1F), indicating that the circadian oscillator in bzip63-1 is sugar unresponsive. This suggests the circadian oscillator did not respond to sugars in bzip63-1 because PRR7 was not upregulated by low-energy conditions.

3We investigated how regulators of bZIP63 influence the response of the circadian oscillator to sugars. KIN10 (AKIN10/SnRK1.1), an α subunit of the sugar sensor SnRK1, regulates bZIP63 activity in response to starvation. KIN10 overexpression (KIN10-ox; Figure S1F) further increased the long circadian period of the wild-type occurring under low-energy conditions (Figures 2A and 2C). The circadian system in KIN10-ox remained sugar sensitive because sucrose supplementation shortened its period (Figure 2C). This could be because overexpressed KIN10 is inhibited post-translationally by sugars. Constitutive KIN10 overexpression under high light, when sugars are high, did not lengthen the period (Figure S2A).

4Under low-energy conditions, such as low light, KIN10-ox caused a longer period relative to the wild-type (Figure 2C). Under high-energy conditions (either low light plus sucrose or high light), KIN10-ox had less effect (Figures 2C and S2A). Therefore, sugar levels affect the circadian phenotype in KIN10-ox. This is consistent with KIN10 regulating the circadian clock in response to energy status because low-energy conditions cause a longer period, KIN10 is activated by low energy, and sucrose rescues light-intensity-dependent effects of KIN10 upon the circadian oscillator.

5TREHALOSE-6-PHOSPHATE SYNTHASE1 (TPS1) synthesizes the signaling sugar trehalose-6-phosphate (Tre6P). Tre6P concentration tracks sucrose and negatively regulates SnRK1 activity, bZIP63, and other sugar-sensing targets. We investigated the circadian phenotype of three hypomorphic tps1 TILLING mutants. Sucrose had no effect on period in tps1-11 and tps1-13 (Figures 2B and 2D). Sucrose shortened the period of tps1-12, the weakest allele for metabolite alterations, but less than the wild-type (Figure 2D). In the presence of sucrose, all three tps1 alleles had longer periods than the wild-type (Figure 2D), presumably because low Tre6P mimics starvation. In the absence of sucrose under low-light conditions, tps1 mutants did not have a longer period than the wild-type (Figure 2D). This might be because these seedlings were already in a low-sugar state, so disrupting this pathway caused no further change.

6Energy changes might arise from fluctuations in sucrose availability. We investigated this using DARK INDUCIBLE6 (DIN6/ASN1), which is upregulated by starvation, downregulated by sugars, and downstream of the regulation of bZIP63 by KIN10. DIN6:luc had a diel rhythm, with promoter activity (2.4-fold; Figure 3A) and transcript abundance (2.7-fold; Figure S2B) increasing after dusk. DIN6 promoter activity reduced through the night, presumably as sugars became available from starch breakdown. After dawn, DIN6:luc activity decreased rapidly, suggesting the accumulation of photosynthetic sugars (Figure 3A). DIN6 promoter dynamics in light/dark cycles arose from sugar status alterations, because sucrose supplementation attenuated DIN6:luc rhythms (Figure 3A). This is also supported by increased DIN6 transcript abundance under low-light/dark cycles at pre-dawn and dusk compared with high-light controls (Figure S1D). DIN6 promoter activity and transcript accumulation are also circadian regulated, peaking in the middle of the subjective day in constant high light (LL) (Figures 3A and S2C). As with CCA1:luc, DIN6:luc was phase-advanced by sucrose under constant low light but not higher light intensity. Therefore, under light/dark cycles there are sugar-dependent cycles in a readout of KIN10- and bZIP63-mediated sugar signaling. This indicates that starvation pathways are upregulated during each night of light-dark cycles (Figures 3A and S2B). Extensive integration of circadian regulation with energy signaling is corroborated by statistically significant overlaps between KIN10- and [Tre6P]-regulated transcripts and five sets of circadian- and diel-regulated transcripts (Figures 3B and S2D–S2F).

7Diel cycles of DIN6:luc activity are likely mediated by Tre6P. DIN6:luc activity was increased in two tps1 alleles (tps1-11 and tps1-12), particularly at night when low sugar availability combines with the low TPS1 activity in these mutants to activate KIN10 and the DIN6 promoter (Figure 3C). DIN6:luc was unaltered in tps1-13, which is a weaker allele for some physiological traits. This demonstrates that Tre6P transmits diel changes in sugar status to the bZIP63-responsive promoter DIN6 (Figure 3C).

8bZIP TFs undergo regulatory interactions with many other proteins. We investigated whether these might contribute to circadian entrainment by sugars. Using a yeast two-hybrid (Y2H) screen, we identified interaction between bZIP63 and the circadian oscillator component CCA1 HIKING EXPEDITION (CHE/TCP21) (Figure S3A). Like PRR7, CHE is a transcriptional repressor of CCA1. No other known regulators of the circadian system interacted with bZIP63. We confirmed that bZIP63 interacts with CHE in planta using bimolecular fluorescence complementation (Figure S3B). We hypothesized that CHE might contribute to a response of the circadian oscillator to sugars. We found that sucrose induces CHE transcripts in the wild-type, but this was attenuated in KIN10-ox and somewhat in tps1-12 (Figure S3C). In light/dark cycles without sucrose supplementation, CHE overexpression and che loss-of-function mutants suppressed and increased the amplitude of daily CCA1:luc fluctuations, respectively (Figures S3D and S3E). When diel changes in sugar status were eliminated by cultivation on 90 mM sucrose (Figure 3A), the amplitude difference between che mutants and the wild-type was abolished (Figures S3D and S3E). This suggests that CHE might not suppress CCA1 under sugar-replete conditions, and that CHE regulates a response of CCA1 to sugars. This occurred independently from CHE binding to the TCP-binding site (TBS) within the CCA1 promoter, because mutation of the TBS did not alter the response of CCA1 to a morning sugar pulse (Figure S3F).

9To test the potential involvement of bZIP63, KIN10, Tre6P, and CHE in circadian entrainment, we measured the time-dependent adjustment of circadian phase in response to a sucrose pulse. This tests the response of the circadian oscillator to a transient stimulus, as opposed to prolonged sucrose treatments (Figures 1F and 2). A morning sugar pulse (ZT0–ZT6) advanced the wild-type circadian phase (Figures 4A and 4B). In contrast, morning sugar pulses did not advance the circadian phase in bzip63-1 mutants (Figures 4A and 4C). This demonstrates that sucrose acts as a type 1 (weak) zeitgeber in the wild-type, resulting in a circadian phase advance. This phase advance was absent in bzip63-1. There was also no phase advance in tps1-11, tps1-13, and KIN10-ox in response to morning sucrose pulses (ZT3–ZT7.5, Figures 4D–4G). tps1-12 had a phase advance in response to sucrose pulses at this time (Figure 4E), which is consistent with its weaker metabolic phenotypes. These data suggest that TPS1, KIN10, and bZIP63 might be positioned within a pathway by which sugars entrain the circadian oscillator (Figures 4A–4G).

10Two che mutations had little effect upon the morning sucrose-induced phase advance (ZT0–ZT7.5; Figures 4H and 4I), and a sucrose-induced phase delay occurred in che during the subjective night (ZT15–ZT19.5; Figures 4H and 4I). CHE-ox increased the magnitude of the sucrose-induced phase advance of CCA1 at most times of day (Figure 4J). Therefore, CHE is not required for the sugar-induced circadian phase advance in the morning but might be associated with other sugar responses. This is consistent with our hypothesis that circadian entrainment to sugars depends upon PRR7, with the alteration of CCA1 transcription occurring in response PRR7 expression dynamics.

11Lastly, we examined whether PRR7 and bZIP63 are required for correct circadian function under day/night cycles. prr7-11 and bzip63-1 have a late phase of CCA1 expression (Figures 4K and 4L) demonstrating that bZIP63 and PRR7 are required for correct oscillator phase under light/dark cycles (Figures 4K and 4L). These late-phase phenotypes (Figures 4K and 4L) suggest defective entrainment. Considering that bZIP63 participates in the regulation of the circadian oscillator in response to sugars (Figure 1F), this sugar-responsive entrainment pathway is required to ensure correct circadian phase under light/dark cycles.

12Our finding that bZIP63 regulates the circadian oscillator allowed us to investigate entrainment of the Arabidopsis circadian oscillator by sugars. We propose that daily fluctuations in sugar availability might be signaled by Tre6P to effect entrainment. Mutants impaired in Tre6P production had a reduced response of circadian period to sucrose (Figures 2B and 2D), and the circadian oscillator of tps1 mutants was not entrained by sucrose (Figures 4D–4F). bZIP63 homo- and heterodimerization is regulated by KIN10-mediated phosphorylation, so transcriptional regulation of PRR7 by bZIP63 might arise from bZIP63 phosphorylation dynamics. Mutants of both bzip63 and its negative regulator TPS1 have similar effects on the circadian oscillator because PRR7 cannot respond to sugar dynamics in both sets of mutants. KIN10-ox was insensitive to entraining sugar pulses, suggesting that SnRK1 participates in entrainment to transient sugar fluctuations (Figure 4G). Morning sucrose pulses delayed the phase in bzip63-1 (Figure 4A), whereas the phase in tps1-11 and tps1-13 was sucrose insensitive (Figures 4D and 4F). This difference might reflect additional effects of the tps1 TILLING alleles, which influence a range of phenotypes. Other kinases and pathways might be involved because KIN10-ox seedlings retained a shorter period, like the wild-type, during long-term sucrose supplementation (Figure 2C). bzip63-1 was unresponsive to prolonged sucrose stimulation (Figure 1F), whereas it retained some sucrose-induced phase changes within the phase response curve (Figure 4A). This is similar to the lip1-1 mutant, which has a very small period response to varying intensity of continuous light, but retains a phase delay response to pulsed light in phase response curves. In continuous darkness, the period of an evening reporter of the circadian oscillator (GIGANTEA) is unaltered in KIN10-ox. This might be because, in continuous darkness, evening components of the circadian oscillator appear to uncouple from morning components such as CCA1 and PRR7 investigated here.

13We found that bZIP63 binds the PRR7 promoter in a region containing a canonical G-box motif. While bZIP63 might bind to other cis elements within this region, for two reasons it is possible that bZIP63 binds to this −254 bp G-box. First, mutating the G-box within the promoter of another bZIP63- and KIN10-regulated gene, DIN6, abolishes the regulation of DIN6 by KIN10. Second, bZIP1 binding to G-box motifs is enhanced by heterodimerization with bZIP63.

14Our data suggest that the response of bZIP63 and PRR7 to sugar entrains the Arabidopsis circadian clock to dynamic energy signals during light/dark cycles. This will allow identification of other network components, since bZIP63 and other bZIPs are phosphorylated by protein kinases other than KIN10, potentially including KIN11 and casein kinase II. It will be informative to determine whether other bZIP TFs dimerize with bZIP63 to regulate PRR7 transcription and whether binding and/or activity of bZIP63 is regulated by energy status.

15bZIP63 works through protein-protein interactions in addition to protein-DNA interactions. We identified potential interactions with CHE, a regulator of CCA1. This suggests several modes of regulation of bZIP63 activity and that CHE might affect the circadian oscillator through multiple mechanisms. bZIP63 interacts with a further TCP TF, TCP2, suggesting an interacting network of bZIP and TCP TFs. This bZIP63-CHE interaction provides further evidence that TCP TFs are important regulators within the plant circadian system. The bZIP63-CHE interaction might have mechanistic similarities with the coincident binding of interacting TOC1 and PIF3 to promoters of growth-regulating genes. PIF4 is proposed as an additional regulator of circadian entrainment to sugars, although PIF4 does not bind the PRR7 promoter and roles for PIF4 within metabolic entrainment remain untested.

16Our identification of key molecular components that permit the circadian oscillator to respond to sugars brings a new dimension to the study of plant circadian systems, by identifying a TF that adjusts circadian phase to entrain the oscillator. Our finding that bZIP63 upregulates PRR7 promoter activity in response to low energy suggests that sugars regulate circadian period and phase through a signaling pathway rather than indirect metabolic changes. This underlies the plasticity of the circadian period to sugars, with this plasticity absent from prr7-11 and bzip63 mutants (Figure 1F). We propose that the dynamic sensitivity of the circadian system to cellular energy, through bZIP63 regulation of PRR7 expression, permits its continuous metabolic adjustment to contribute to energy homeostasis. This is important because circadian systems provide a selective advantage through their phase relationship with the environment. Here, we identified a mechanism that establishes that phase relationship.

17Further information and requests for resources and reagents should be directed to and will be fulfilled by Antony Dodd (antony.dodd@bristol.ac.uk).

18Arabidopsis thaliana (L.) Heynh. background lines Columbia-0 (Col-0), Landsberg erecta (L. er.) and Wassilewskija (Ws) were used for experimentation, with mutants and transgenic lines in these backgrounds as detailed in the Key Resources Table. Arabidopsis seedlings were cultivated at 19°C, under light conditions required by each experiment and described in the Results. Nicotiana benthemiana was cultivated at 25°C (both for growth and bimolecular fluorescence complementation analysis). Saccharomyces cerevisiae was cultured at 30°C for all assays.

19Seeds were surface-sterilized with 10% v/v sodium hypochlorite (Fisher Scientific, Loughborough, UK) and 0.02% (v/v) Triton X-100 (Fisher Scientific, Loughborough, UK) for 5 min, washed three times with sterile deionised water and sown on half-strength Murashige & Skoog media (Duchefa, Netherlands), pH = 5.7 with 0.8% w/v agar (Bactoagar, BD). Where specified, media was supplemented with 90 mM sucrose or 90 mM sorbitol as an osmotic control. This concentration of sucrose is appropriate for our experiments because it saturates the sugar response of the circadian oscillator, is the standard concentration of sucrose used for experimentation with Arabidopsis, and there is no dose-dependent effect of sucrose upon circadian entrainment. Seeds were stratified at 4°C for 2 or 3 days in darkness, then transferred into 50 μmol m-2 s-1 (before starting low light experiments) or 80 – 100 μmol m-2 s-1 (for standard light experiments) photon flux of cool white fluorescent light, at 19°C, with cycles of 12 hr light and darkness (MLR-350/352 growth chamber, Sanyo/Panasonic, Japan). Background lines for the tps mutants were derived originally from mutagenesis of Col-0 and backcrossed three times with Landsberg erecta (L. er.). We have, therefore, used L. er. as a control for experiments with tps lines. KIN10-ox was as described elsewhere. T-DNA insertion mutants bzip63-1 (SALK_006531, Col-0 background), bzip63-2 (FLAG_610A08, Ws-2 background) and bzip63-3 (FLAG_532A10, Ws-2 background) were obtained from the Arabidopsis Biological Resource Center (ABRC). Homozygous bzip63-3 was isolated using kanamycin-resistance segregation analysis, and sequencing of the flanking regions revealed that the T-DNA is inserted in the sixth exon of the bZIP63 gene. For experiments using mature plants, Arabidopsis was grown under 12 hr photoperiods of 100 μmol m-2 s-1 for 30 days before experimentation.

20To make HA-bZIP63-ox and bZIP63 RNAi lines, both the bZIP63 coding sequence (CDS) and a 350 bp fragment for RNAi were amplified using PCR primers that incorporated restriction sites (indicated in lower case); overexpressor: tctaga ATGGAAAAAGTTTTCTCC (FP); ggatccCTACTGATCCCCAACGCT (RP); RNAi: aagcttggtaccTCACTGGTCGGTTAATGG (FP); tctagactcgagCACTTGTTATAGCACTGC (RP). The bZIP63 CDS was cloned into the pFP101 vector, fused to 3xHA tag and the VP16 activation domain for overexpression by the CaMV 35S promoter. The 350 bp bZIP63 fragment was cloned antisense and sense into pHANNIBAL, then transferred to pFP100-Lacz with the CaMV 35S promoter driving antisense-sense hairpin expression.

21To produce the DIN6:luciferase construct, 2659 bp of genomic DNA upstream of the DIN6 start codon was isolated by PCR, using primers that introduced the KpnI (5′) and HindIII (3′) restriction sites (FP: CGTGGTACCTGGACATGAGTGCATGAC; RP: GCGAAGCTTGAAGAAAGTGAAAAAGATCACG). The fragment was ligated into a modified pGreenII0179 binary vector containing the LUCIFERASE+ coding sequence. The CCA1:luciferase construct used is described elsewhere. Reporter lines with CCA1(TBSm):luc and control CCA1:luc used constructs in the pPZP binary vector described elsewhere. All constructs were transformed into Arabidopsis by the floral dip method, using Agrobacterium tumefaciens strain GV3101. Transformants were identified with hygromycin (pGreenII0179) or gentamycin (pPZP) selection, or from GFP fluorescence in the seed coat (pFP101 and pFP100-LacZ), using a GFP2-equipped SMZ1000 stereomicroscope. Lines were validated by PCR and RT-PCR, and selected for similar levels of luciferase activity. Homozygous third or fourth generation seed lines were used for experiments.

22Imaging of luciferase bioluminescence was performed as described previously. Briefly, circular clusters of 9 day old seedlings were supplied with 2 mM (two doses) or 5 mM (single dose) of D-luciferin potassium salt (Melford Labs, UK, or Biosynth AG, Switzerland) between 1 hr and 24 hr prior to commencing imaging. Luciferase bioluminescence was integrated for 800 s each hour using either a Photek HRPCS intensified CCD camera (Photek, Hastings, UK) or LB 982 NightSHADE (Berthold Technologies, Bad Wildbad, Germany). Light was controlled automatically to provide the stated photon irradiance and LD cycles or constant (LL) light from a red/blue LED mix (wavelengths 660 nm and 470 nm). Circadian oscillation parameters were calculated from four 24 hr cycles, excluding the first 24 hr of data, using the Fast Fourier Transform Non-Linear Least-squares method within the BRASS software (http://millar.bio.ed.ac.uk/). The mean peak height parameter was calculated as an average of the differences between peak (maximum value) and trough (minimal value) measured on days 2, 3, 4, 5 and 6 of LL (thus excluding first 24 hr), and for LD was calculated by averaging the differences between peaks and troughs measured for each day under light/dark conditions. Phase response curves were produced using the same method as described previously.

23Sampling and RNA isolation for real-time PCR was performed as described previously. Primers were PRR7: TTCCGAAAGAAGGTACGATAC (FP); GCTATCCTCAATGTTTTTTATGT (RP); PP2AA3 reference transcript: CATGTTCCAAACTCTTACCTG (FP); GTTCTCCACAACCGCTTGGT (RP) (for Figure 1C); CHE: TAATGGGTGGTGGTGGTTCTG (FP); GCAAAGCTCCAGACTTGTCC (RP); Figure S3C); DIN6: TTCACCTTTCGGCCTACGAT (FP); ATCGGCATGTTGTCAATTGC (RP); ACT2 reference (TGAGAGATTCAGATGCCCAGAA (FP); TGGATTCCAGCAGCTTCCAT (RP) (for Figure S3C).

24Aerial parts of 12-day old seedlings were vacuum-infiltrated with 1% v/v formaldehyde solution and incubated at room temperature for 20 min to crosslink DNA-protein complexes. After cross-linking, tissues were ground in liquid nitrogen using mortar and pestle. Chromatin extraction was performed as described elsewhere. The isolated chromatin was re-suspended in 300 μL of nuclear lysis buffer (50 mM Tris-HCL, pH 8; 10 mM EDTA; 1% w/v SDS; 1X Pierce protease inhibitor cocktail #88265 (Thermo Scientific, Massachusetts, USA). Resuspended chromatin was sonicated to achieve DNA fragments ranging 0.3 - 1 Kb. Immunoprecipitation was performed using EpiQuik Plant ChIP kit (Epigentek Group, Farmingdale, USA) following manufacturer’s instructions. DNA-protein complexes were immuno-precipitated using a monoclonal anti-HA antibody (Santa Cruz Biotechnology, Texas, USA). Analysis of enrichment of target genes was performed by qPCR (ABI 7500 Fast Real-time PCR System). For PRR7, primer pair 1 was GACGTTTTCCTTACCCACCA (FP), ATTGGCGAGGATTAGTGACG (RP), and primer pair 2 TGCTTTTGTATGGTTGGATTTTT (FP), TGAAGAACGACGAATTCTCAAA (RP) (Figures 1A and 1B). Data were normalized using cycle thresholds from immunoprecipitated and non-immunoprecipitated samples as elsewhere.

25Chromatin immunoprecipitation (ChIP)-qPCR was performed using two transgenic lines overexpressing HA-tagged bZIP63 (HA-bZIP63-ox1 and HA-bZIP63-ox2 in Figure 1B), using an anti-HA antibody (+) and an anti-mouse IgG antibody as a negative control. ChIP using overexpressor lines is a common approach (e.g.,), with the bZIP63 overexpressors accumulating approximately double the transcript as the wild-type under high light at ZT0 (Figure 1C). Primer pair 1 amplifies the promoter region containing G-box motif and primer pair 2 as a control amplifies a region without any putative bZIP binding site. Data in Figure 1B compare the cycle threshold (Ct) of the anti-HA antibody IP samples (adjusted relative to Ct of input DNA) and Ct of mock IP (adjusted relative to Ct of the input DNA); t tests compared control and anti-HA treated samples (HA-bZIP63-ox1 p = 0.011; HA-bZIP63-ox2 p = 0.014).

26Full-length coding regions of bZIP63 (AT5G28770.2) and CHE (AT5G08330.1) were cloned using specific primers (CHE: TAAGCAGGATCCATGGCCGACAACGACGGAGC (FP); TAAGCAGGTACCACGTGGTTCGTGGTCGTC (RP); bZIP63: TAAGCAGGATCCATGGAAAAAGTTTTCTCCG (FP); TAAGCAGGTACCCTGATCCCCAACGCTTC (RP) into binary vectors pSPYNE-35S (YFPN terminus; aa 1-155) or pSPYCE-35S (YFPC terminus; aa 156-239) to generate the fusion proteins bZIP63-YFPN, bZIP63-YFPC and CHE-YFPC. The binary vectors containing the constructs bZIP63-YFPN, bZIP63-YFPC and CHE-YFPC were inserted into Agrobacterium tumefaciens strain C58C1, which was grown in Luria-Bertani (LB) liquid medium containing appropriate antibiotics (100 μg ml-1 rifampicin, 50 μg ml-1 kanamycin, 10 μg ml-1 tetracycline) at 28°C for 16 hr and 200 RPM agitation. Bacterial cells were harvested by centrifugation and the pellet resuspended in 5 mL of infiltration buffer (10 mM MgCl2, 10 mM MES-KOH pH 5.7 and 200 μM acetosyringone (#D134406, Sigma-Aldrich). Agrobacterium strains carrying the constructs containing putative interacting proteins were mixed and co-infiltrated in the abaxial surface of four-week-old Nicotiana benthamiana leaves at a final OD600 = 0.5 each. To enhance expression of fusion proteins, Agrobacterium C58C1 carrying the pCH32 helper plasmid that suppress gene silencing was co-infiltrated in all experiments. After 3-4 days, infiltrated regions of Nicotiana leaves were excised and pavement cells visualized in a confocal laser scanning microscope (LSM 510 Meta, Carl Zeiss, Thornwood, NY, U.S.A.) with an argon laser (excitation = 488 nm, emission = 524 nm).

27Full-length coding regions of bZIP63 (AT5G28770.2) and CHE (AT5G08330.1) were cloned using specific primers (CHE: caccATGGCCGACAACGACGGAGC (FP); TCAACGTGGTTCGTGGTCGTC (RP); bZIP63: caccATGGAAAAAGTTTTCTCCGAC (FP); CTACTGATCCCCAACGCTTC (RP) into pENTR-D-TOPO (Thermo Scientific) following the manufacturer’s instructions, to generate pENTR-D-TOPO:bZIP63 and pENTR-D-TOPO:CHE intermediary entry vectors. The sequence indicated in lowercase in the bZIP63 and CHE forward primers (FP) were incorporated by PCR into the amplicon for directional cloning in pENTR-D-TOPO. Subsequently, the intermediary entry vectors were recombined into the yeast expression vectors pDEST32 and pDEST22 of the ProQuest Two-Hybrid System with Gateway Technology (Thermo Scientific) following manufacturer’s instructions. The resulting yeast expression vectors pDEST32:bZIP63; pDEST32:CHE; and pDEST22:bZIP63, as well as the empty vectors pDEST32 and pDEST22 (used here as negative controls) were transformed into Saccharomyces cerevisiae strain PJ69-4a (MATa trp1-901 leu2-3,122 ura3-52 his3-200 gal4Δ gal80Δ LYS2::GAL1-HIS3 GAL2-ADE2::GAL7-lacZ). Y2H assays were performed using yeast double-transformants (pDEST32:bait + pDEST22:prey) lines. The yeast lines obtained were grown for 16 hr (overnight) at 30°C and 200 RPM agitation in minimal SD (Synthetic Defined) liquid medium lacking leucine and tryptophan (SD/-Leu/-Trp), auxotrophic markers for pDEST32 and pDEST22 vector selection, respectively. To verify yeast double transformation, yeast cells carrying the vector combinations shown in the figure were grown in solid SD medium lacking both leucine and tryptophan (SD/-Leu/-Trp. Protein-protein interactions were evaluated by dropping 5 μL of overnight yeast cultures (106 CFU.mL-1) in SD/-Leu/-Trp/-His, where histidine (His) is the auxotrophic marker for protein-protein interaction, and grown at 30°C for 3 days. To reduce background growth due to residual activity of HIS3 gene, 3 mM of 3-Amino-1,2,4-Triazol (3AT), a competitive inhibitor of the HIS3 reporter enzyme, was added to test plates.

28Seedling cultivation and sampling occurred as for RNA sampling. Total protein was isolated in 1.1 M glycerol, 5 M Tris-MES (pH 7.6), 1 mM EGTA and 2 mM dithiothreitol with protease inhibitor cocktail P9599 (Sigma). Protein concentrations were quantified with Bradford reagent (Bio-Rad). Proteins were separated on 10% polyacrylamide gels and transferred to nitrocellulose membranes (Bio-Rad), which were subsequently stained with Ponceau Red to verify equal protein loading. Membranes were incubated with KIN10 antiserum (AKIN10/SNF1-related protein kinase catalytic subunit alpha antibody, Agrisera) at 1:1000 dilution with 1% w/v fat-free milk powder and 0.1% v/v Tween 20, and incubated subsequently with goat anti-rabbit IgG HRP conjugate (GtxRb-003-DHRPX from ImmunoReagents, Raleigh, NC) at 1:2000 dilution. Blots were developed using Pierce ECL-2 reagent (Thermo Scientific). Two independent biological repeats were performed of each experiment (data from one repeat shown in Figure S1F).

29Lists of rhythmic genes and lists of genes that are regulated by KIN10 were compared and their overlaps analyzed for significance. Rhythmic genes were selected from two nycthemeral experiments (light/dark) and three circadian experiments (constant light). Lists of genes with all the rhythmic transcripts in the nycthemeral experiments or all the circadian genes in the circadian experiments were also analyzed. The KIN10-regulated genes were selected from genes having altered expression in KIN10-ox or in lines overexpressing the E. coli trehalose 6-phosphate synthase (OtsA), which elevates Tre6P in planta. Statistical significance of overlaps and the representation factor were estimated using web-based software designed by Jim Lund (University of Kentucky), with statistical significance quantified using a hypergeometric test (http://elegans.uky.edu/MA/progs/overlap_stats.html). 25% - 40% of transcripts upregulated by KIN10-ox and 19% - 34% of transcripts downregulated by otsA-ox oscillate under light/dark cycles. 49% - 59% of transcripts upregulated by KIN10-ox and 37% - 41% of transcripts downregulated by otsA-ox oscillate under constant conditions. Each overlap was calculated as the proportion of the total number of rhythmic transcripts in each phase bin. Only overlaps with p < 0.01, and representation factor < 0.5 (#) or > 2 (∗) and were considered significant.

30Statistical analysis was performed using Sigmaplot 13.0. Details of statistical tests used, replication levels, and nature of error bars are provided in figure legends. Circadian rhythm parameters were determined using the Biological Rhythms Analysis Software System (BRASS) (University of Edinburgh; millar.bio.ed.ac.uk). Statistical significance of intersections between transcriptomes was calculated with a hypergeometric test. No data were excluded from analysis. Statistical significance is indicated by the p value, where ∗ = p < 0.05, ∗∗ = p < 0.01 and ∗∗∗ = p < 0.001.