Abstract

   Protein phosphorylation is a key regulatory mechanism in circadian
   systems. TIMING OF CAB EXPRESSION 1 (TOC1) is a core transcriptional
   repressor in the plant circadian system that is phosphorylated near its
   N terminus. Phenotype testing of TOC1 phosphosite mutants shows
   incomplete rescue of the short period toc1 mutant. We establish that
   TOC1 phosphorylation (particularly at S175) is necessary for optimal
   interaction with FAR-RED ELONGATED HYPOCOTYL3 (FHY3) and PHYTOCHROME
   INTERACTING FACTOR 5 (PIF5) at the CIRCADIAN CLOCK-ASSOCIATED 1 (CCA1)
   promoter to down-regulate CCA1 expression. Expression of the closely
   related LATE ELONGATED HYPOCOTYL (LHY) also requires TOC1 but is
   independent of TOC1 phosphorylation, suggesting different
   TOC1-dependent gene repression mechanisms. We additionally show that
   TOC1 phosphorylation–dependent interactions at specific clock gene
   promoters selectively regulate these circadian system components more
   acutely than nonrhythmic genes. Our genome-wide analysis demonstrates
   that the TOC1 phosphostate is important for optimal chromatin presence
   and robust rhythmic gene expression.
     __________________________________________________________________

   TOC1 phosphorylation discriminates between and specifies for rhythmic
   gene promoters.

INTRODUCTION

   Adapting to the day and night cycles that arise from the rotation of
   the earth, most organisms have developed an endogenous oscillator, the
   circadian clock, to synchronize their biological processes with a
   rhythm of ~24 hours ([40]1–[41]3). Plants, as sessile organisms,
   particularly rely on this internal clock to anticipate the external
   cyclic changes and adjust their growth, physiology, and development in
   advance ([42]4, [43]5).

   The molecular mechanism of circadian clock in plants was described as
   an autoregulatory transcriptional feedback network ([44]6). The central
   oscillator of Arabidopsis clock is composed of two core groups of clock
   components. One group contains two morning-expressed MYB-like
   transcription factors (TFs), CIRCADIAN CLOCK-ASSOCIATED 1 (CCA1) and
   LATE ELONGATED HYPOCOTYL (LHY). These two closely related
   morning-phased TFs bind the evening element (EE) ([45]7) of
   evening-phased clock genes, such as TIMING OF CAB EXPRESSION 1 (TOC1),
   LUX ARRHYTHMO (LUX), and EARLY FLOWERING 3 (ELF3) and ELF4 and repress
   their transcription ([46]7–[47]12). PSEUDO-RESPONSE REGULATOR (PRR)
   proteins, including TOC1, PRR3, PRR5, PRR7, and PRR9, comprise another
   core group of clock components in Arabidopsis, which sequentially
   repress CCA1 and LHY expression and thus form transcriptional feedback
   loops ([48]13, [49]14) that sustain the pace of the central oscillator.

   A key feature of circadian systems is synchronization with the
   surrounding environment through informational input pathways. Light is
   one of the major signals shaping the rhythmic expression of core clock
   genes and thus resets the clock ([50]15, [51]16). FAR-RED ELONGATED
   HYPOCOTYL3 (FHY3) acts as a phytochrome signaling integrator and plays
   an important role in gating red light to the clock system in the
   morning ([52]17). FHY3 and its paralog FAR-RED IMPAIRED RESPONSE 1
   (FAR1) are required for the light induction and normal rhythmic
   expression of CCA1 through direct promoter binding ([53]18). In
   contrast, PHYTOCHROME INTERACTING FACTOR 5 (PIF5) and TOC1 antagonize
   FHY3 activity through physical interactions and repress CCA1 expression
   from predusk to midnight ([54]18). Blue light signaling also plays a
   role in CCA1 regulation. Cryptochrome 2 (CRY2) was found to form
   photobodies with TEOSINTE BRANCHED 1-CYCLOIDEA-PCF 22 (TCP22) and
   activates CCA1 expression through LIGHT REGULATED WDs (LWDs) and the
   TBS motif ([55]19) in a blue light–dependent manner. In addition,
   PHOTOREGULATORY PROTEIN KINASES 1 (PPK1) phosphorylates TCP22 and
   enhances the formation of CRY2-TCP22 photobodies to promote the
   expression of CCA1 ([56]19). These findings indicate a close
   relationship between light signaling components and clock proteins in
   coordinating robust rhythmic expression of CCA1.

   Numerous studies in mammals, Drosophila, and Neurospora have shown the
   fundamental importance of posttranslational regulation, particularly
   phosphorylation, in the circadian clock ([57]20, [58]21). In
   Arabidopsis, phosphorylation also comprises an essential integral part
   of the circadian regulatory system ([59]22, [60]23). For example, all
   PRR proteins are phosphorylated, and phosphorylation enhances their
   heterodimerization and possibly interactions with other proteins as
   well ([61]24). A chemical screen of the altered circadian period
   identified CKL4, a member of CASEIN KINASE 1-LIKE (CKL) family, that is
   possibly involved in phosphorylation of TOC1 and PRR5 ([62]25). TOC1
   and PRR5 are the most closely related members of the PRR family, and
   PRR5 promotes TOC1 phosphorylation and nuclear transport through
   heterodimerization with the TOC1 N terminus ([63]26). In addition,
   specific TOC1 N-terminal phosphosites are essential for NF-Y-TOC1
   complex formation and its regulated photoperiodic hypocotyl elongation
   ([64]27). However, how TOC1 phosphorylation regulates its role in the
   central oscillator is still unknown.

   In this study, we establish that N-terminal phosphorylation of TOC1,
   particularly serine-175 (S175), is required for sustaining both the
   wild-type (WT) circadian period and the robustness of circadian
   oscillations. FHY3 interacts strongly with phosphorylated TOC1 and
   selectively shapes CCA1 expression rhythm, but not LHY, through the FBS
   element. Loss of either S175 phosphorylation or FHY3 results in
   shortened period and decreased amplitude, whereas the absence of both
   leads to arrhythmicity.

   Genome-wide transcriptome analyses identified global changes in
   amplitude and phase of rhythmically expressed genes between WT TOC1 and
   the alanine-substituted phosphomutant line (5X). Our findings suggest
   that these changes in circadian oscillations of gene expression are
   likely due to diminished 5X binding to central oscillator clock genes
   and a disproportionally greater level of phosphorylation-dependent
   occupancy of TOC1 on cycling targets genes relative to noncycling
   target genes. In addition, we identify a very different category of
   noncycling TOC1 target genes that point to possible noncircadian roles
   for TOC1. Together, our results expand our understanding of
   posttranslational regulation of the circadian clock, providing a
   comprehensive analysis of how TOC1 phosphorylation affects genome
   rhythmicity and interesting insights into the selective regulation of
   CCA1 by TOC1 interacting with different cofactors.

RESULTS

TOC1^S175 is essential for its function in the central oscillator

   Our previous study identified five phosphosites at the N terminus of
   TOC1, mutations at all of which (5X) result in poor interactions with
   NUCLEAR FACTOR-Y (NF-Y) transcription cofactors and increased hypocotyl
   growth under short days ([65]27). In addition to the growth phenotype,
   5X plants also showed a significantly shortened circadian period
   compared to WT ([66]27). To investigate the role of TOC1
   phosphorylation and NF-Ys in the context of the circadian oscillator,
   we first examined the clock-related phenotypes in various TOC1
   phosphosite substitutions in either toc1 single or nf-yc3/4/9 toc1
   quadruple mutant backgrounds under constant white light. We found 5X
   and TOC1^S175A both exhibit a 1.5-hour shorter period ([67]Fig. 1A and
   fig. S1), increased relative amplitude error (RAE) (fig. S2A), and
   CCA1-LUC oscillation that dampens rapidly relative to WT ([68]Fig. 1B
   and fig. S2C), indicating that phosphorylation at TOC1^S175 is
   essential for normal circadian oscillations, as we previously reported
   ([69]27). In contrast, mutation of the closely positioned T135 residue
   to alanine (TOC1^T135A) ([70]27) had no effect on period. Consistent
   with this result, the protein oscillations of only 5X and TOC1^S175A
   are phase advanced under 12-hour light/12-hour dark cycles (fig. S3, A
   and B).

Fig. 1. Shorter period and dampened CCA1-LUC oscillations in 5X and S175A
mutants.

   [71]Fig. 1.
   [72]Open in a new tab

   Free-running period (A) and amplitude (B) of Col-0, toc1, nf-yc3/4/9
   (c349), nf-yc3/4/9 toc1 (c349t1), and TOC1 native promoter lines in
   toc1 (t1) single or nf-yc3/4/9 toc1 quadruple mutant backgrounds. #1
   and #2 indicate the two independent lines of the corresponding
   transgene. In the boxplots (A) and (B), the middle line of the box
   represents the median, x represents the mean, the bottom line and the
   top line indicate the first and third quartile, respectively. The
   whiskers extend from the ends of the box to the minimum value and
   maximum value. Circles out of the box indicate outliers. Seedlings were
   entrained in 12-hour/12-hour light/dark cycles for 7 days and then
   transferred to constant white light at ZT2 for image acquisition at
   2-hour intervals for 1 week. Forty-five seedlings from three
   independent trials were averaged (n = 45). Different letters indicate
   statistically significant differences [P < 0.01, one-way analysis of
   variance (ANOVA) followed by the Tukey-Kramer post hoc test].

   The period of the nf-yc3/4/9 triple mutant is not different from Col-0,
   and the period of the nf-yc3/4/9 toc1 quadruple mutant (table S1) is
   similar to toc1 ([73]Figs. 1, A and C, and [74]2A), supporting the
   notion of a TOC1-anchored trimeric complex at the chromatin, composed
   of NF-YB/C factors as previously reported for hypocotyl length control
   ([75]27). 5X in nf-yc3/4/9 toc1 was unable to restore clock-related
   phenotypes and showed a short period (table S1) ([76]Fig. 1A), high RAE
   (fig. S2B), and strongly dampened CCA1-LUC oscillations, similar to the
   nf-yc3/4/9 toc1 quadruple mutant (fig. S2D).

Fig. 2. Robust TOC1 chromatin residence depends on TOC1 phosphostate and
NF-YC proteins.

   [77]Fig. 2.
   [78]Open in a new tab

   ChIP-qPCR of TOC1 binding to the promoter region of core clock genes in
   Col-0, WT TOC1, and 5X in toc1 (t1) single (A) or nf-yc3/4/9 toc1
   (c349t1) quadruple (B) mutant backgrounds. Data were averaged from
   three independent trials, each trial with two technical repeats. Error
   bars indicate SEM. Asterisks indicate significant differences
   (*P < 0.05; **P < 0.001; ***P < 0.0001, one-way ANOVA followed by
   two-tail Student’s t test), n.s., not significant.

Phosphorylation stabilizes TOC1 chromatin residence at clock genes

   To investigate how TOC1 phosphorylation affects the control of central
   oscillator genes, we assayed the residence of both TOC1 and 5X at the
   promoters of several key clock genes previously reported as binding
   TOC1 ([79]28). We chose ZT14 for chromatin immunoprecipitation (ChIP)
   assays because TOC1 showed equal protein levels in both pTOC1:TOC1-GFP
   and pTOC1:5X-GFP at this time point under 12-hour/12-hour light/dark
   cycles (fig. S3, A and B).

   In most cases, the 5X mutations diminished TOC1 presence at the genes
   tested. PRR7, PRR9, and GI showed 30 to 40% less TOC1 at their
   promoters and an almost complete loss of TOC1 at CCA1 in the
   pTOC1:5X-GFP line was observed ([80]Fig. 2A). However, there were no
   differences between TOC1 and 5X binding at the LHY promoter ([81]Fig.
   2A). These results suggest that phosphorylation enhances TOC1 chromatin
   binding activity at a subset of target clock genes, such as CCA1, PRR7,
   PRR9, and GI.

   We next examined whether NF-YCs play a role in TOC1 presence at target
   clock genes. These experiments were conducted at ZT13 to ensure similar
   TOC1 levels in the respective genotypes (fig. S4). ChIP of TOC1 in an
   nf-yc3/4/9 background showed a 20 to 40% reduction in chromatin
   occupancy, relative to WT, at all tested clock promoters. The
   phosphorylation defects in 5X decreased chromatin binding further,
   except for LHY ([82]Fig. 2B). Chromatin residence of TOC1 and 5X in the
   nf-yc3/4/9 toc1 background at the LHY promoter were very similar,
   whereas 5X binding to CCA1 promoter diminished to almost background
   levels ([83]Fig. 2B). These results suggest that NF-YCs are not the
   factors responsible for the notable difference between TOC1 5X and TOC1
   WT binding at the CCA1 and LHY promoters. Together with the marginal
   period phenotype of nf-yc3/4/9, these results exclude the three NF-YC
   genes as major determinants in regulating phosphorylation-dependent
   functions of TOC1 in the circadian system.

Phosphorylation of TOC1^S175 promotes CCA1 oscillation

   To test whether TOC1^S175 is the most essential TOC1 phosphosite, we
   next performed ChIP with pTOC1:T135A-GFP and pTOC1:S175A-GFP. Similar
   to 5X, S175A showed notably reduced residence at CCA1, whereas T135A
   binding was only marginally less at CCA1 relative to WT ([84]Fig. 3A).
   The differences in TOC1 chromatin residence were not due to differences
   in protein abundance between plant lines (fig. S3, A and B). In
   contrast to CCA1, all four different TOC1 phosphovariants showed a
   similar presence at the LHY promoter ([85]Fig. 3A).

Fig. 3. TOC1 phosphomutants show reduced binding to CCA1 and diminished CCA1
oscillations.

   [86]Fig. 3.
   [87]Open in a new tab

   (A) ChIP-qPCR of TOC1 binding to the promoter region of CCA1 (-730 to
   -648) and LHY (-1062 to -958) at ZT14. Different letters indicate
   significant differences (P < 0.05, one-way ANOVA followed by two-tail
   Student’s t test), n.s., not significant. (B to E) Time-series
   expression of CCA1 (B and D) and LHY (C and E) in the indicated plant
   lines. For [(B) and (C)], insets show magnified views of the trough
   region, highlighting differences between Col-0 and phosphomutants. Data
   were averaged from three independent trials, each trial with two
   technical repeats. Error bars indicate SEM. Asterisks indicate
   significant differences of 5X and S175A compared to WT TOC1 at each
   time point (*P < 0.05; **P < 0.001; ***P < 0.0001, Student’s t test),
   n.s., not significant. For [(D) and (E)], two trials were conducted,
   each trial with two technical repeats.

   To determine the consequences of reduced 5X binding on gene expression,
   we next examined transcript levels of CCA1 and LHY in Col-0, toc1, and
   the four TOC1 phosphovariant lines over a 12-hour/12-hour light/dark
   cycle. CCA1 displayed significantly lower amplitude oscillations in 5X
   and TOC1^S175A plants than in Col-0, toc1, and TOC1^T135A lines
   ([88]Fig. 3B). In contrast, LHY expression was similar for all
   genotypes ([89]Fig. 3C). These results indicate that these
   phosphorylation defective mutants do not simply phenocopy the toc1 null
   mutant but rather are neomorphic in the context of CCA1 transcription.
   We hypothesize that 5X and TOC1^S175A may act as dominant negative
   mutants by interfering with the binding of other PRR repressors to the
   CCA1 promoter. This could occur through TOC1-PRR5 heterodimerization,
   which has been previously reported ([90]26). Consistent with this
   hypothesis, the prr5 prr7 double mutant (prr5/7) showed a 5- to
   200-fold higher trough of CCA1 expression, particularly from ZT7 to
   ZT16 ([91]Fig. 3, D and E, and fig. S3C), similar to results reported
   previously ([92]14, [93]29). Sequestration of endogenous PRR5 and/or
   PRR7 by 5X or TOC1^S175A heterodimerization could account for the
   similar high trough phenotypes of the 5X and prr5/7 mutants
   particularly between ZT7 and ZT10, where CCA1 expression is high but
   LHY is not.

Phosphorylation of TOC1^S175 enhances binding affinity with FHY3 and PIF5 but
not TCPs

   To identify factors that work together with TOC1 to cause the
   differential transcriptional regulation between CCA1 and LHY, we
   examine the promoter sequence of CCA1 and LHY and identified two
   cis-elements that are present in CCA1 but not in LHY: an FHY3/FAR1
   binding sequence (FBS) and a TCP (TEOSINTE BRANCHED 1, CYCLOIDEA, and
   PCF1) binding site (TBS) ([94]Fig. 4A).

Fig. 4. TOC1 phosphomutants interact more poorly with FHY3 and PIF5.

   [95]Fig. 4.
   [96]Open in a new tab

   (A) Schematic representation of the positions of various cis-elements
   in the CCA1 and LHY promoter. FBS, FHY3/FAR1 binding sequence; TBS, TCP
   binding sequence; G-box, PIF binding motif; LBS, LUX binding sequence;
   5′UTR, 5′ untranslated region. (B and C) Co-IP of TAP-tagged CHE/TCP
   (B) and FHY3 or PIF5 (C) with TOC1-GFP or TOC1 carrying alanine
   substitutions in N. benthamiana. TAP-tagged proteins were detected by
   anti-MYC, and TOC1 was detected by anti-GFP. Asterisks in (C) indicate
   PIF5 that was unspecifically recognized by the secondary α-rabbit. (D)
   Quantitation of protein interactions in (B) and fig. S4 [(A) and (B)].
   (E) Quantitation of protein interactions in (C) and fig. S5A. For [(D)
   and (E)], results from three independent experiments were averaged
   (except CHE with two trials). Error bars indicate SEM, and different
   letters indicate significant difference (P < 0.05, one-way ANOVA
   followed by the Tukey post hoc test). (F) Co-IP of TAP-tagged FHY3,
   FAR1, or PIF5 with TOC1-GFP and TOC1 carrying alanine or aspartate
   substitution in N. benthamiana. (G) Quantitation of protein
   interactions in (F) and fig. S5B. Results from three independent trials
   were averaged. Error bars indicate SEM, and different letters indicate
   significant difference (P < 0.05, one-way ANOVA followed by the Tukey
   post hoc test).

   Because phosphorylation affects TOC1 binding affinity with the NF-Y
   cofactors ([97]27), we tested the interactions of TOC1 phosphovariants
   with FHY3/FAR1 and TCPs. CHE, TCP20, and TCP22 are members of the TCP
   family of TFs ([98]30). CHE interacts with and recruits TOC1 to the
   CCA1 promoter, whereas TCP20 and 22 form tetramers with LWD1/2 to
   activate CCA1 expression ([99]31, [100]32). However, none of the three
   TCP proteins show significant differences from WT in their interactions
   with TOC1 phosphovariants ([101]Fig. 4, B and D, and fig. S5).

   We next tested FHY3 and PIF5 which interact with TOC1 and coordinately
   regulate CCA1 expression under light/dark cycles ([102]18). In these
   cases, FHY3 and PIF5 both showed a 60 and 40% reduced interaction with
   5X and S175A, respectively ([103]Fig. 4, C and E, and fig. S6A). In
   contrast, the phosphomimetic mutant S175D and WT TOC1 interacted
   similarly with FHY3 and PIF5 ([104]Fig. 4, F and G, and fig. S6B),
   suggesting that the differential binding affinity between TOC1
   phosphovariants and FHY3/PIF5 results from phosphorylation, not protein
   conformation. Our results are also consistent with a previous report
   mapping the TOC1-FHY3 interaction domain to a region that includes S175
   (amino acids 141 to 520) ([105]18). In contrast to FHY3, FAR1
   interactions with TOC1 were negligible ([106]Fig. 4, F and G, and fig.
   S6B).

   Despite the diminished binding of 5X and S175A to PIF5 ([107]Fig. 4, C,
   E, F, and G, and fig. S6), chromatin residence of TOC1 at the CCA1
   promoter is unaffected in a pif3/4/5 mutant (fig. S7A), similar to
   results we and others have reported for hypocotyl growth–related genes
   ([108]27, [109]33). These results indicate that PIF5 is not involved in
   anchoring or recruiting TOC1 to the CCA1 promoter.

   Notably, ChIP–quantitative polymerase chain reaction (qPCR) at the
   G-box, a DNA motif common to both CCA1 and LHY ([110]Fig. 4A), showed a
   loss of 5X chromatin occupancy only at the CCA1 promoter, similar to
   that seen for the FBS and TBS DNA elements (fig. S7B). This probably
   resulted from the close proximity of FBS, TBS, and G-box sequences
   [within 140 base pairs (bp); [111]Fig. 4A] and the low resolution of
   ChIP.

FHY3 absence strongly diminishes TOC1 presence at CCA1

   To investigate whether FHY3 affects chromatin occupancy of TOC1
   phosphovariants at CCA1 and LHY, we examined chromatin binding of TOC1
   and 5X in the fhy3 toc1 background. The absence of FHY3 resulted in 87%
   lower TOC1 binding to the CCA1 FBS site, in contrast to no effect on
   TOC1 binding to the LHY promoter. Similar to 5X in the toc1 single
   mutant, 5X binding in the absence of FHY3 was also nearly absent at the
   CCA1 promoter ([112]Fig. 5A), indicating that FHY3 anchors TOC1 to the
   CCA1 promoter through the FBS element. These results, together with
   [113]Fig. 4E, show that FHY3-TOC1 binding affinity is enhanced by TOC1
   phosphorylation, leading to an increased presence of TOC1 at the CCA1
   promoter.

Fig. 5. FHY3 is required for TOC1 binding to the CCA1 promoter and robust
CCA1 oscillation.

   [114]Fig. 5.
   [115]Open in a new tab

   (A) ChIP-qPCR of TOC1-GFP and 5X-GFP’s residence at target promoters at
   ZT14 in toc1 and fhy3 toc1 mutant backgrounds. Data were averaged from
   three independent trials, each trial with two technical repeats. Error
   bars indicate SEM. Asterisks indicate significant differences
   (**P < 0.001; ***P < 0.0001, one-way ANOVA followed by two-tail
   Student’s t test), n.s., not significant. (B and C) Time-series
   expression of CCA1 and LHY in Col-0, fhy3, fhy3 toc1, and phosphomutant
   lines of TOC1 in the fhy3 toc1 mutant background. Shown is a
   representative trial from two independent experiments. Expression was
   normalized to Col-0 at ZT1. Insets show magnified views of the trough
   region, highlighting differences between Col-0 and mutants.

   Consistent with the absence of FBS element in the LHY promoter, TOC1
   occupancy at the LHY promoter was the same in the fhy3 toc1 and toc1
   backgrounds ([116]Fig. 5A). 5X in fhy3 toc1 showed significantly
   reduced binding to LHY than 5X in toc1, but this likely resulted from
   the lower TOC1 protein abundance when pTOC1:5X-GFP was crossed into
   fhy3 toc1 (fig. S8).

   Together, these results suggest that the selective regulation of CCA1
   by TOC1 phosphorylation results from the anchoring of TOC1 by FHY3
   through its FBS element. In contrast, LHY lacks FBS and TOC1 residence
   is likely through the G-box and independent of its phosphorylation
   state. However, we also performed protoplast dual-luc experiments with
   a mutated CCA1 FBS motif (CCA1mFBS) ([117]18). Constitutive expression
   of either TOC1 or 5X results in strong repression of CCA1-LUC, with no
   prominent differences between the two TOC1 forms with either the WT or
   mFBS form of the promoter (fig. S9). The ability of strong expression
   of TOC1 to repress CCA1mFBS in the absence of the FBS implies TOC1
   binding to other promoter elements, such as the G-box and TBS. This
   result also highlights the sensitivity of promoter activation to TF
   dosage.

fhy3 mutant derepresses CCA1 expression, similar to 5X and S175A

   We next examined the consequences of FHY3 absence and TOC1
   phosphorylation on CCA1 transcript oscillations under light/dark
   cycles. fhy3 showed a prominently higher, derepressed expression of
   CCA1 from ZT4 to ZT19 ([118]Fig. 5B and fig. S10), consistent with
   reduced TOC1 presence at CCA1 in the absence of FHY3 ([119]Fig. 5A).
   The expression phase of CCA1 in fhy3 toc1 was slightly earlier than in
   fhy3, like that seen in the toc1 mutant ([120]Fig. 3B), indicating that
   TOC1 drives phase relationships even under light/dark cycles.
   pTOC1:TOC1-GFP overcomplemented fhy3 toc1 to a Col-0–like CCA1
   expression pattern possibly due to higher than endogenous TOC1 levels
   in pTOC1:TOC1-GFP; both pTOC1:TOC1-GFP lines showed a 1-hour longer
   period than Col-0 ([121]Fig. 1A). pTOC1:5X-GFP failed to rescue toc1
   and displayed similar CCA1 expression pattern as fhy3 toc1 (compare
   [122]Figs. 3B and [123]5B and fig. S10).

   In contrast, loss of FHY3 and/or TOC1 phosphorylation did not change
   LHY expression ([124]Fig. 5C and fig. S10). Together with the ChIP
   results, these findings show that FHY3 and TOC1 phosphorylation have no
   effect on LHY transcription due to the absence of FBS site on the LHY
   promoter.

FHY3 and TOC1^S175 phosphorylation are required for CCA1 rhythmicity

   We next examined how phosphorylation affects the coordination between
   TOC1 and FHY3 in oscillator function. fhy3 single mutants display a
   much shorter circadian period (21.9 hours) and notably dampened
   CCA1-LUC oscillation compared to Col-0 under constant white light
   ([125]Fig. 6, A, B, and E) (table S2). CCA1-LUC oscillations in fhy3
   toc1 double mutants are further dampened and largely arrhythmic
   ([126]Fig. 6, A to E). In contrast, fhy3 and toc1 single mutants
   displayed an RAE within a range (<0.6) indicative of robust rhythmicity
   ([127]Fig. 6C). These results show that FHY3 and TOC1 act
   synergistically in sustaining a robust circadian period and
   oscillation.

Fig. 6. TOC1 phosphorylation and FHY3 are required for robust CCA1 circadian
oscillations.

   [128]Fig. 6.
   [129]Open in a new tab

   Free-running period (A), amplitude (B), RAE (C and D), and average
   CCA1-LUC bioluminescence (E and F) of the indicated plant lines. The
   boxplots were generated as explained in the [130]Fig. 1 caption. Dashed
   lines in [(C) and (D)] indicate the RAE cutoff for the robustness of
   circadian clock. Thirty seedlings were averaged, and different letters
   indicate statistically significant differences (P < 0.01, one-way ANOVA
   followed by the Tukey-Kramer post hoc test).

   WT TOC1 and TOC1^T135A overcomplemented fhy3 toc1 and exhibited a
   WT-like circadian period and amplitude ([131]Fig. 6, A and B), probably
   due the higher than endogenous levels of TOC1-GFP expression,
   consistent with the slightly longer than WT (Col-0) period of
   pTOC1:TOC1-GFP/toc1 ([132]Fig. 1A). However, 5X and S175A were unable
   to complement toc1 in the fhy3 toc1 mutant, displaying arrhythmic and
   dampened CCA1-LUC phenotype, similar to fhy3 toc1 ([133]Fig. 6, A and
   F). Furthermore, the oscillations in 5X/fhy3 toc1 and S175A/fhy3 toc1
   were much less robust, with half of the seedlings at an RAE higher than
   0.6 ([134]Fig. 6D). This suggests that TOC1 phosphorylation is required
   for oscillator function and S175 is the most essential residue.

   We further examined the changes in the relative phosphorylation state
   of TOC1 over the period of its highest expression level, ZT10 to ZT22.
   This addressed the question of whether changes in TOC1 phosphorylation
   might play a regulatory role. Using a phospho-specific TOC1 antibody
   (fig. S11A), we found the relative phosphorylation state of TOC1
   changed very little between ZT10 and ZT22 (~2-fold) and not at all from
   ZT13 to ZT19 (fig. S11, B to D). These findings suggest that
   phosphorylation at S175 is essential for full functionality throughout
   the normal window of TOC1 expression.

TOC1 and 5X lines show genome-wide circadian rhythmicity effects

   The above findings demonstrate the very specific role FHY3 and its
   partnership with phosphorylated TOC1 plays in the regulation of CCA1
   expression. However, to gain a broader insight into the genome-wide
   effects of TOC1 phosphorylation on the rhythmicity of gene expression,
   we next performed RNA sequencing (RNA-seq) analyses over a 24-hour time
   course using pTOC1:TOC1-GFP and pTOC1:5X-GFP lines. The experiments
   were conducted under 12-hour light/12-hour dark conditions to better
   observe the effects of 5X on both the phase and amplitude of global
   gene expression under entrainment and closer to a natural light/dark
   environment. In the TOC1 and 5X lines, similar numbers of expressed
   genes (23,393 and 23,312, respectively) and rhythmically expressed
   genes [10,206 (43.6%) and 9604 (41.2%)] were observed ([135]Fig. 7A)
   (tables S3 and S4).

Fig. 7. TOC1 phosphostate alters the phase and amplitude of rhythmic genes.

   [136]Fig. 7.
   [137]Open in a new tab

   (A) Number and percentage of rhythmic and nonrhythmic genes in TOC1 and
   5X lines identified by time-series RNA-seq under 12-hour light/12-hour
   dark cycles. (B) Venn diagram showing the overlap of rhythmically
   expressed genes in TOC1 and 5X lines. (C) Heatmap of the expression
   pattern of 5680 common rhythmic genes sorted by the oscillation phases
   in TOC1 and 5X lines. (D) Phase distribution density curves of rhythmic
   transcripts in TOC1 and 5X plants. (E) Amplitude density curves of
   rhythmically expressed genes in TOC1 and 5X lines. P values in [(D) and
   (E)] indicate significant difference by the Kolmogorov-Smirnov test
   (two-sided). (F) Amplitude reduction and phase shift shown in the core
   clock genes due to impaired phosphorylation of TOC1. The angular
   coordinates represent phase 0 to 24 hours. The log[10] transformed
   values (peak-trough) represent the amplitude of the gene oscillation
   and are plotted as radial coordinates. Circles represent 5X lines, and
   squares represent TOC1 lines. (G) Sinusoidal model fitting traces for
   the expression patterns of CCA1 and LHY in TOC1 versus 5X lines that
   were determined by RNA-seq. The actual data were plotted as
   means ± SEM, and the confidence interval was plotted as the shade of
   each curve. r^2 of the fit to the sine model for CCA1 are
   r^2(TOC1) = 0.8056 and r^2(5X) = 0.9917. r2 of the fit for LHY are
   r^2(TOC1) = 0.9304 and r^2(5X) = 0.9202. (H) Venn diagram and
   subsequent GO analysis of CCA1 target genes [ChIP-seq; Nagel et al.
   ([138]34)] that have altered phase and/or amplitude (phase change > 2
   hours; amplitude change > 1.5-fold) in 5X relative to TOC1.

   A total of 5680 genes were identified to commonly oscillate in both
   TOC1 and 5X lines, whereas 4526 genes and 3924 genes were oscillating
   only in TOC1 or 5X plants, respectively ([139]Fig. 7B) (table S5). Gene
   Ontology (GO) terms of circadian rhythm and carbohydrate biosynthetic
   process were two of the most strongly enriched among the common cycling
   genes, indicating that these processes continue to be rhythmic even in
   the absence of the phosphosites mutated in 5X. TOC1-specific categories
   of protein glycosylation, cell wall biogenesis, and response to stress
   were specifically strongly enriched, whereas categories related to root
   development were among the highly enriched 5X-specific rhythmic genes
   (fig. S12).

   The heatmaps of common rhythmic transcripts between TOC1 and 5X
   revealed differential expression patterns, with the 5X exhibiting
   dampened expression peaks in genes phasing at ZT14 but increased
   expression peaks in genes phasing at ZT22 ([140]Fig. 7C). The density
   curves of phase distribution of the common rhythmic genes in TOC1 and
   5X plants showed an ~4 hours advanced phase of the most accumulated
   transcripts in the 5X versus TOC1 line ([141]Fig. 7D and fig. S13). As
   indicated with the heatmaps, the TOC1 line showed the highest density
   of genes phasing at 14.89, whereas the 5X line had the highest number
   of genes phasing at 10.87 ([142]Fig. 7D). These results are consistent
   with the shorter period of the 5X line ([143]Fig. 1), manifesting as a
   global gene expression phase advance.

   In addition, the amplitude distribution showed a forward-shifted peak
   in the 5X line indicative of a more dampened gene oscillation in 5X
   than in TOC1 plants. The most enriched rhythmic transcripts in TOC1 and
   5X plants had an amplitude at 151.36 (log[10] 151.36 = 2.18) and 83.18
   (log[10] 83.18 = 1.92), respectively ([144]Fig. 7E).

   Further analyses of only core clock genes showed alterations in their
   rhythmicity even under 12-hour light/12-hour dark cycles, with most
   genes exhibiting slightly advanced phase and dampened amplitude in the
   5X background ([145]Fig. 7F and fig. S14) (table S6). In support of our
   targeted qPCR results ([146]Fig. 3, B to F), genome-wide RNA-seq data
   showed a nearly sixfold reduction in CCA1 amplitude in the 5X
   background ([147]Fig. 7G). In contrast, LHY expression did not show a
   prominent expression difference regardless of TOC1 phosphorylation
   status ([148]Fig. 7G), consistent with our qPCR results ([149]Fig. 3, B
   to E). These results further support our notion of the selective
   regulation of targets through TOC1 interactions with different
   cofactors.

   We next considered that the altered rhythmicity of the cycling genes in
   5X plants partially resulted from the altered waveform (i.e., higher
   trough and earlier phase) of CCA1. When we compared the differentially
   altered (phase or amplitude differences) oscillating genes in 5X with
   known CCA1 target genes ([150]34), 352 genes identified as direct
   targets of CCA1 were enriched (tables S7 and S8). These genes are
   predominately involved in circadian rhythm, cold and stress responses,
   starch biosynthesis, flowering, and light signaling, among which are
   well-known TFs involved in these processes (e.g., PIFs, BBXs, PRRs,
   CBF1, and CORs) ([151]Fig. 7H). TOC1 has been previously linked to
   control of drought responses through direct binding to the ABA-related
   gene, ABAR, and to light signaling responses, through interaction with
   PIF proteins ([152]27, [153]33, [154]35). Thus, through our analysis of
   the TOC1 phosphorylation state, we find support for the notion that
   TOC1 is both directly and indirectly involved in the transcriptional
   control of these responses to abiotic signals.

TOC1 occupancy at rhythmically expressed genes is enriched by phosphorylation

   We next sought to determine whether the rhythmic divergence between
   TOC1 and 5X can be accounted for by differences in chromatin occupancy
   at target genes. Genome-wide ChIP sequencing (ChIP-seq) analyses of
   TOC1 and 5X chromatin binding intensity at ZT14 (the time point of
   equivalent TOC1 and 5X protein levels) was conducted using
   pTOC1:TOC1-GFP and pTOC1:5X-GFP lines. A total of 4846 and 5148 target
   peaks were identified in TOC1 and 5X lines, respectively, of which 3773
   were in common ([155]Fig. 8A) (tables S9 to S11). GO pathway analyses
   revealed that responses to hormone and abiotic stimulus were highly and
   commonly enriched regardless of TOC1 phosphorylation state (fig. S15C).
   Of the target genes common between TOC1 and 5X, 5X showed similar
   chromatin binding intensity, but the presence of small peaks at gene
   bodies and distant promoter regions ([156]Fig. 8B) suggesting that 5X
   target binding is less stringent than TOC1. More notably, examination
   of the chromatin binding profiles using only rhythmically expressed
   target genes shows that the intensity of 5X was prominently lower than
   TOC1 ([157]Fig. 8C), suggesting that phosphorylation of TOC1 promotes
   chromatin presence more robustly on rhythmic genes.

Fig. 8. TOC1 phosphostate is central to promoter binding to rhythmically
expressed genes.

   [158]Fig. 8.
   [159]Open in a new tab

   (A) Venn diagram showing the overlap of targets in TOC1 and 5X lines.
   (B and C) Binding profiles of TOC1 and 5X at the transcription start
   site (TSS) ± 1 kb on all target genes (B) and rhythmic target genes (C)
   at ZT14. TOC1 binding profile at ZT2 was used as control. (D)
   Quantitative profiling of rhythmic TOC1 target genes identified through
   time-course RNA-seq. Genes exhibiting rhythmic expression patterns
   (P < 0.05) were classified into six phase-specific clusters using the
   Bio-Cycle algorithm, with cluster distribution frequencies annotated
   above corresponding histogram bars. Relative TOC1-GFP protein abundance
   driven by the native TOC1 promoter is also shown, demonstrating
   antiphase oscillation relative to gene cluster distribution of TOC1
   target genes (red line). (E) GO analysis of the rhythmic and
   nonrhythmic genes bound by TOC1. The enrichment score is the negative
   natural logarithm of the enrichment P value derived from the Fisher’s
   exact test. Count indicates that the number of genes has been
   identified in each pathway. (F) Profiles of TOC1 binding to rhythmic
   target genes versus nonrhythmic target genes. TES, transcription end
   site.

   To better compare the nature of rhythmic TOC1 target genes with
   noncycling TOC1 target genes, we first clustered the cycling genes by
   time of day (phase) (fig. S16) and then plotted them across six
   phase-specific time points superimposed with TOC1 protein abundance
   ([160]Fig. 8D). Peak phase expression of the TOC1 target genes is
   generally antiphase to TOC1 protein abundance, supporting previous
   evidence that TOC1 acts as a transcriptional repressor. Divergence from
   a perfectly antiphase relationship between TOC1 abundance and peak
   phase of gene expression may arise from TOC1 cofactors modifying peak
   TOC1 efficacy.

   Comparative GO analysis between cycling and noncycling TOC1 genes
   showed a strong divergence in the nature of the targets ([161]Fig. 8E).
   Expectedly, circadian rhythm and response to temperature and cold were
   the most strongly enriched cycling TOC1 target genes. In contrast,
   general gene expression and genes involved in translational elongation
   were the most commonly enriched noncycling TOC1 target genes ([162]Fig.
   8E). However, the TOC1 binding profile of the noncycling genes showed a
   much greater presence of TOC1 throughout the gene body ([163]Fig. 8F).
   This suggests that some noncycling gene residence of TOC1 may not be
   functionally significant.

FHY3 and phosphoTOC1 shape cyclic gene rhythmicity

   Consistent with our ChIP-qPCR results ([164]Fig. 2A), our genome-wide
   ChIP-seq results confirm that 5X is nearly absent at the CCA1 promoter,
   resulting in notably elevated CCA1 expression at ZT14 ([165]Fig. 9A).
   In contrast, LHY is similarly bound by TOC1 and 5X, reflecting a
   similarly repressed expression at night ([166]Figs. 2A and [167]9A).
   Other core clock genes targeted by TOC1, such as PRR9, PRR7, GI, and
   ELF4, are bound more weakly (~30%) by 5X than TOC1, similar to our
   ChIP-qPCR results ([168]Fig. 2A), leading to enhanced expression of
   these genes in 5X plants (fig. S17). These results show that
   phosphorylation is crucial to TOC1 chromatin binding specificity and
   strength, particularly on cycling genes, leading to altered rhythmicity
   of their expression.

Fig. 9. FHY3 and TOC1 phosphostate shape robustness and period of
rhythmically expressed genes.

   [169]Fig. 9.
   [170]Open in a new tab

   (A) IGV view of ChIP-seq profiles and RNA-seq data at CCA1 and LHY
   locus in TOC1 and 5X lines. Up (CCA1) and down (LHY) trace in RNA-seq
   track indicate forward- and reverse-strand RNA reads, respectively. (B)
   Motifs enriched in differentially oscillating genes bound by TOC1,
   along with their enrichment P values and candidate TFs shown in the
   table. (C) Venn diagram showing the overlap of differentially
   oscillating genes bound by FHY3 and TOC1. (D and E) Comparison of the
   phase (D) and amplitude (E) of differentially oscillating transcripts
   (phase change > 2 hours; amplitude change > 1.5-fold) bound by both
   FHY3 and TOC1 (120 genes) or only by FHY3 (115 genes) between the TOC1
   and 5X ChIP-seq datasets. (F) GO analysis of the 120 differentially
   oscillating genes bound by FHY3/TOC1. The enrichment score is the
   negative natural logarithm of the enrichment P value derived from the
   Fisher’s exact test. Count indicates that the number of genes has been
   identified in each pathway. (G) Proposed mechanism for the role of TOC1
   phosphorylation in the control of CCA1 gene expression. (H)
   Differential roles of TOC1 and 5X in regulating rhythmic target genes
   versus nonrhythmic target genes.

   When TOC1 target genes from our ChIP-seq are compared to the targets
   identified from Huang et al. ([171]28), we identified 64.2% of the
   target genes from this earlier study. This comparison indicates a good
   degree of consistency of our ChIP-seq with previous reports (fig.
   S15A). Motif analysis of TOC1-bound sequences identified significantly
   enriched G-box that are bound by the basic helix-loop-helix (bHLH) and
   basic Leu zipper (bZIP) TFs, such as ABFs and PIFs (fig. S15B). HY5,
   FHY3, and TCP binding motifs were also identified among TOC1 targeting
   sequences ([172]Fig. 9B), consistent with our prior ChIP-qPCR results.

   To identify potential transcriptional coregulators of TOC1 target genes
   and the effect of phosphorylation, we compared oscillating genes
   showing the altered phase or amplitude in 5X lines with genes bound
   differentially by 5X. Motif analysis of the overlapped genes identified
   the FHY3 binding element among those significantly enriched in the
   rhythmic genes bound by TOC1 ([173]Fig. 9B), consistent with the
   essential role of FHY3 in maintaining the robustness of circadian
   rhythms together with TOC1. To identify additional genes that are
   coregulated by TOC1 and FHY3, we overlapped the phase- and
   amplitude-altered genes from the 5X dataset with genes from FHY3
   ([174]36) and TOC1 ChIP-seq datasets. From this three-way comparison,
   120 genes altered in 5X plants were bound by both FHY3 and TOC1 (table
   S12), whereas 115 genes were bound only by FHY3 (table S13) ([175]Fig.
   9C). The 120 overlapping genes were phase advanced by more than 3
   hours, and the amplitude of these genes decreased by ~31% in 5X plants
   ([176]Fig. 9, D and E). In contrast, the 115 genes that were only bound
   by FHY3 exhibited much reduced amplitude compared to the 120 genes,
   with similar phase and amplitude between TOC1 and 5X plants ([177]Fig.
   9, D and E). The 120 overlapped genes were primarily enriched for GO
   terms describing rhythmic processes, response to stress, signal
   transduction, and regulation of circadian rhythm, consistent with known
   targets of TOC1 and FHY3 ([178]Fig. 9F). However, for the 115
   nonoverlapped genes, no GO terms were enriched. Together, these results
   suggest that FHY3 and phosphorylated TOC1 are essential to the general
   maintenance of robust gene oscillations, as we have specifically shown
   here for CCA1.

DISCUSSION

FHY3 is key cofactor with phosphoTOC1 regulating CCA1 expression

   TOC1 is a core transcriptional repressor in the plant circadian clock.
   With peak expression in the early evening, genome-wide studies have
   shown a wide-ranging effect of TOC1 on gene expression well beyond the
   core elements of the circadian oscillator ([179]13, [180]28, [181]35)
   (this study). In addition, although TOC1 is phosphorylated ([182]24,
   [183]27), the role of this modification in the context of the circadian
   system has been unclear. Here, we show that TOC1 repression of FHY3
   activation of CCA1 depends on the phosphorylation of TOC1 N-terminal
   residues, particularly serine-175.

   The failure of the S175A mutation to rescue the short period toc1
   mutation ([184]Fig. 1) suggests a deficiency in the transcriptional
   repressor functions of TOC1^S175A. We show that two aspects of TOC1
   function in the circadian system are affected by phosphorylation:
   protein-protein interactions and chromatin residence. TOC1 binds DNA
   via the C-terminal CCT domain ([185]37) but requires NF-YB and NF-YC
   cofactors for robust presence at the promoters of hypocotyl
   growth–related genes and N-terminal TOC1 phosphorylation to promote
   formation of the NF-TOC1 trimeric complex ([186]27). Here, we have
   extended these findings to show that full TOC1 chromatin binding at
   core circadian clock genes similarly requires members of the NF-YC
   family and these same phosphosites ([187]Fig. 2). Notably, our previous
   results, under short days and red light, showed that the nf-yc3/4/9
   mutant had a slightly longer hypocotyl than WT but shorter than toc1
   ([188]27). Similarly, the free-running period of nf-yc3/4/9 in red
   light was slightly shorter than WT but much longer than toc1. The
   quadruple mutant was additively longer hypocotyl/shorter period than
   either toc1 or nf-yc3/4/9 alone. [189]Figure 1A reports a similar
   genotype/period relationship under white light, although the slightly
   shorter period of nf-yc3/4/9 was not statistically significant under
   the stronger white light conditions. Hence, despite the different light
   and growth conditions, the formation of an NF-TOC1 complex ([190]27)
   appears critical to TOC1 repressor activity.

   The phosphostate-dependent binding of TOC1 to the CCA1 promoter
   ([191]Fig. 2A) prompted a more detailed investigation into the cause.
   Using the relative insensitivity of the TOC1 phosphostate to LHY
   promoter binding for comparison, we identified the FHY3-TOC1 protein
   interaction as the primary determinant of the TOC1
   phosphostate-dependent difference between the CCA1 and LHY TOC1
   chromatin presence and the cause for the increased CCA1 expression in
   the 5X, S175A, and fhy3 mutant backgrounds ([192]Fig. 3). A previous
   work established TOC1 amino acid residues 141 to 520 as necessary for
   FHY3 interaction ([193]18). Within this region are four of five
   serine/threonine residues mutated in 5X (T135, S175, S194, S201, and
   S204) including the single most effective mutation, S175A. Alanine
   substitution of T135 that was not within the interaction region
   exhibited trivial changes in the circadian rhythm, FHY3 interactions,
   and chromatin residence ([194]Figs. 1, [195]3, and [196]4). Our
   findings now confirm and expand on that work to show that
   phosphorylation of TOC1 at S175 is crucial to the interaction. Notably,
   at this residue TOC1 phosphorylation state appears largely constitutive
   during the times of maximal TOC1 expression (ZT13 to ZT19; fig. S11D),
   suggesting that this posttranslational modification is a necessary, but
   not regulatory, component of TOC1 activity.

   Previous works have implicated both FAR1 and FHY3 as important
   regulators of circadian function ([197]18, [198]38). However, roles for
   these two closely related TFs in the activation of ELF4 and CCA1/LHY
   through promoter binding have tended to emphasize FHY3 as the more
   important of the pair. In most cases, fhy3 mutants are more severe than
   far1 lines, with the double mutant often additive. Here, we show a
   strong difference in the relative roles of the two factors in how TOC1
   controls CCA1 expression, consistent with these previous reports.
   Although other work has shown a TOC1-FAR1 interaction in yeast two
   hybrid and bifluorescence complementation assays ([199]18), we were
   unable to demonstrate measurable interactions. Our in planta
   coimmunoprecipitations (Co-IPs) between TOC1 and FAR1 showed very weak
   to no interaction, in strong contrast with the significant
   phosphostate-dependent interaction between TOC1 and FHY3 ([200]Fig. 4
   and fig. S6A). Different assays and experimental conditions may account
   for the different results, but our findings are consistent with most
   reports emphasizing the primary role for FHY3 over FAR1 in light
   signaling and circadian systems ([201]18, [202]39, [203]40). Together,
   our results demonstrate that phosphorylated TOC1 is recruited by FHY3
   via the FBS to maintain both the amplitude and phase of CCA1 expression
   ([204]Fig. 9G).

   PIF5 has been implicated to act with TOC1 to repress FHY3 activation of
   CCA1 transcription ([205]18). Our results show TOC1 binding to CCA1 and
   LHY promoters is independent of PIF5 presence, although TOC1
   interaction with PIF5 is phosphostate dependent ([206]Fig. 4E and fig.
   S7A). PIF5 and PIFs, in general, are associated with circadian
   regulation ([207]41). pif5 single mutants have no effect on CCA1
   expression, although PIF5 can be found at the CCA1 promoter and strong
   ectopic PIF5 expression can repress CCA1 expression ([208]18). Only
   combinatorial stacking of multiple pif mutants alters the circadian
   period, suggesting genetic redundancy among the PIF family ([209]41).
   Together with our findings, these results suggest that phosphorylated
   TOC1 is important in the recruit of one or more PIF species to the
   FBS/G-box vicinity of the CCA1 promoter to control phase and amplitude
   under light/dark cycles ([210]Fig. 9G). The role of PIF5 may be related
   to sucrose sensing as Shor et al. ([211]41) showed enhanced binding to
   the CCA1 and LHY promoters at higher sucrose levels. A PIF5-TOC1
   complex might then be a point of confluence where circadian and
   metabolic signals converge to regulate CCA1 and LHY expression.

   LHY transcription is repressed by TOC1 ([212]13, [213]28), consistent
   with TOC1 chromatin presence ([214]Figs. 2 and [215]3). In vitro
   binding of an NF-TOC1 trimer at the LHY promoter supports our findings
   ([216]37). The toc1 null shows an increase in LHY mRNA during the time
   of maximum TOC1 protein presence ([217]Fig. 3C) ([218]24), but the
   waveform of the mutant is unaffected by the presence or absence (fhy3
   toc1) of FHY3 ([219]Fig. 5C). Hence, the NF-TOC1 trimeric complex
   ([220]27) is either acting alone [possibly with a histone deacetylase,
   like HDA15; ([221]27)] or with one or more unknown cofactors to repress
   LHY expression via the G-box or related sequences. The additional
   FHY3-dependent component of CCA1 regulation, not present for LHY,
   raises the interesting question of what FHY3 contributes to LHY
   expression, particularly because its phase of circadian expression is
   so similar to LHY. In addition, paradoxically, the transcriptional
   activator FHY3 is also necessary for TOC1 recruitment to the promoter
   ([222]Fig. 5), which then acts to repress CCA1 expression.

Genome-wide gene rhythmicity depends on TOC1 phosphorylation status

   Both in circadian studies and, in general, gene regulation, protein
   phosphorylation plays a major role in determining genome-wide gene
   expression ([223]42–[224]47). To further understand the importance of
   TOC1 phosphorylation on a broader scale, we examined the genome-wide
   effects of the 5X mutant on rhythmicity, phase, and amplitude of gene
   oscillations under the more natural conditions of light/dark cycles.
   The advanced phase of output genes ([225]Fig. 7D) and core clock genes
   ([226]Fig. 7F) expressed in the 5X background is consistent with the
   shorter period ([227]Fig. 1), but the reduced amplitude in that
   background ([228]Fig. 7E) suggests a weakening in the robustness of
   rhythmic gene expression in general. In some circadian systems,
   phosphorylation of key clock proteins alters their stability, leading
   to weakened rhythmicity ([229]46). 5X and TOC1 protein oscillations
   were similarly robust under light/dark cycles (fig. S3) and a change in
   protein abundance alone likely cannot account for lower amplitude of
   target genes. However, weaker homodimerization and weaker interaction
   with NF-YB and C components ([230]27), leading to poor chromatin
   binding and reduced gene repression, may be the most likely reason for
   low amplitude gene expression oscillations in the 5X background.
   Focusing on 5X effects on CCA1 target genes alone, rhythmic genes are
   disproportionally affected ([231]Fig. 7H), which is indicative of the
   overall phase shift and amplitude dampening on oscillating genes
   ([232]Fig. 7, D and E).

TOC1 phosphostate determines chromatin binding at rhythmic gene promoters

   As a transcriptional repressor, if phosphorylation of TOC1 alters
   chromatin presence, it might be expected to affect binding at all
   target genes. However, only rhythmically expressed genes showed a much
   lower binding profile for 5X (compare [233]Fig. 8, B and C; see
   [234]Fig. 9H), consistent with lower amplitude oscillations of the
   rhythmic genes in this background. This interesting result suggests
   that a range of different TOC1 cofactors, not just FHY3, may contribute
   to TOC1’s role in sustaining oscillations and require TOC1
   phosphorylation for optimal functionality.

   A similar enrichment of a core circadian transcriptional regulator in
   Drosophila, CLOCK (CLK), specifically at the promoters of rhythmic
   genes, has been described previously ([235]48). As in our study, these
   researchers also found CLK binding to noncycling target genes. In our
   work, nonrhythmic TOC1 target genes (fig. S16B) were enriched for
   translational elongation ([236]Fig. 8E), part of the complex process of
   mRNA translation that has been found previously to be circadian
   regulated at multiple steps, including initiation, elongation, and
   phase separation ([237]49). It is possible that the arrhythmicity or
   low rhythmicity of these gene transcripts, despite TOC1 binding, is due
   to their high mRNA stability. Alternatively, other transcriptional
   activators binding throughout the diurnal time course could override
   any phase-dependent effects of TOC1. Also, the greater degree of gene
   body binding of TOC1 on the nonrhythmic targets ([238]Fig. 8F) may
   suggest a nonproductive presence of TOC1 at some of these genes. Future
   work will be necessary to distinguish between these possibilities.

MATERIALS AND METHODS

Plant materials and growth conditions

   The WT and all mutants of Arabidopsis thaliana plants used in this
   study were of the Colombia-0 (Col-0) ecotype. The toc1-101 [toc1;
   ([239]50)], pif3-3 pif4-2 pif5-3 toc1-101 (pif345toc1), prr5-1 prr7-1
   [prr5prr7; ([240]51, [241]52)], nf-yc3-1 nf-yc4-1 nf-yc9-1 [nf-yc349;
   ([242]53)], and nf-yc3-1 nf-yc4-1 nf-yc9-1 toc1-101 [nf-yc349toc1;
   ([243]27)] mutant lines, as well as CCA1:LUC line (Salome and McClung,
   2005) were previously described. The fhy3-11 (SALK_002711) ([244]18)
   was obtained from the Arabidopsis Biological Resource Center
   ([245]http://arabidopsis.org/). The toc1 and TOC1 native promoter lines
   were crossed with fhy3-11 to produce the fhy3 toc1 double mutant and
   TOC1 phosphovariants in the double mutant background, and the CCA1:LUC
   line was crossed with various mutants ([246]27) such as fhy3 and
   pif3/4/5, respectively. The point mutations in 5X
   (T135A/S175A/S194A/S201A/S204A), 135A, and 175A lines were introduced
   by site-directed mutagenesis driven by the TOC1 native promoter as
   described previously ([247]27). Homozygous F3 progenies of each plant
   line were used for bioluminescence and ChIP assays.

   Arabidopsis seeds were surface sterilized and stratified at 4°C for 4
   days. Arabidopsis seedlings used in this study were all grown at 22°C.
   For bioluminescence acquirement and period estimate, seedlings were
   entrained under 12-hour white fluorescent light/12-hour dark cycles for
   7 days on MS (Murashige and Skoog) plates with 3% sucrose and 0.8% agar
   as previously described ([248]54). Seedlings were then subjected to
   constant white (50 μmol m^−2 s^−1) light treatment conducted in a
   Percival E30LEDL3 growth chamber (Percival Scientific, Perry, IA). For
   protein immunoblotting, transcript abundance analyses and ChIP assays
   in 12-hour white fluorescent light (50 μmol m^−2 s^−1)/12-hour dark
   cycles, for 10 days on MS plates with 3% sucrose and 0.8% agar. For
   transient expression experiments in Nicotiana benthamiana, 4-week-old
   plants were infiltrated with Agrobacteria as described earlier
   ([249]24). Tissues were harvested on the third day at ZT12 (16-hour
   light/8-hour dark).

Constructs

   To construct TAP-tagged TCPs, FHY3, FAR1, and PIF5, the full-length
   coding region was amplified and each amplicon was subcloned into
   pENTR/D-TOPO and placed upstream of the TAP tag by LR recombination
   (Invitrogen) into the binary vector pYL436 (ABRC; CD3-679). To generate
   TOC1-GFP, 5X-GFP, T135A-GFP, S175A-GFP, and S175D-GFP for Co-IP assays
   in N. benthamiana, the cDNA fragments of each gene were subcloned into
   pENTR2B or pENTR/D-TOPO and placed into the binary vector
   pCsVMV-GFP-N-1300 by LR recombination ([250]26). To generate plasmids
   for transient expression in protoplasts, the cDNA fragments of TOC1 and
   5X were transferred from pENTR/D-TOPO entry clones to pCsVMV:GFP-C-999
   destination vector by LR recombination. The point mutations in CCA1mFBS
   ([251]18) were introduced by site-directed mutagenesis on the basis of
   the CCA1:LUC reporter plasmid described before ([252]55).

Protein extraction and immunoblots

   Total protein extraction was carried out as previously described
   ([253]24, [254]56). For the immunoblot analyses, proteins were
   separated by 8 to 10% SDS–polyacrylamide gel electrophoresis (PAGE)
   (acrylamide:bisacrylamide, 37.5:1) for regular protein electrophoresis
   or 8% SDS-PAGE (acrylamide:bisacrylamide, 150:1) to differentiate WT
   and phosphosite mutants of TOC1. Immunoblotting was performed using
   1:5000 dilution of primary polyclonal anti–green fluorescent protein
   (GFP) (Abcam, ab6556), 1:10,000 dilution of polyclonal ADK primary
   antibodies (gift from D. Bisaro) followed by ECL detection using
   anti-rabbit immunoglobulin G (IgG) with horseradish peroxidase
   (HRP)–linked whole antibody (GE Healthcare, NA934V). For immunoblot of
   primary anti-Myc (Sigma-Aldrich, M4439; 1:4000 dilution), HRP-linked
   anti-mouse IgG (Sigma-Aldrich, A0198) was used as a secondary antibody.
   Chemiluminescence reactions were performed with Supersignal West Pico
   and Femto Chemiluminescent Substrates (Thermo Fisher Scientific). To
   detect N-terminal phosphorylated TOC1, immunoblotting was performed
   using 1:5000 dilution of primary polyclonal anti-GFP (Abcam, ab6556),
   1:1000 dilution of primary anti-TOC1-P (Agrisera, AS22 4701), and
   1:12,000 dilution of polyclonal ADK primary antibodies followed by ECL
   detection using anti-rabbit IgG with HRP-linked whole antibody (GE
   Healthcare, NA934V). Chemiluminescence reactions were performed as
   described above.

Co-IP assays

   Co-IP assays were conducted as previously described ([255]27). Briefly,
   Agrobacteria containing corresponding expression clones were
   coinfiltrated into 4-week-old N. benthamiana leaves, and the total
   proteins were extracted and used for immunopreciptation with Human
   IgG-Agarose (Sigma-Aldrich, A6284). The immune complex was washed and
   subsequently released from resin by HRV-3C protease digestion. The
   eluted TAP- and GFP-tagged proteins were detected by Western blotting
   using the Myc antibody (Sigma-Aldrich, M4439) and GFP antibody (Abcam,
   ab6556), respectively.

ChIP assays

   ChIP assays were conducted as previously described. Briefly, 10-day-old
   Arabidopsis seedlings grown under light/dark conditions were harvested
   and cross-linked with 1% formaldehyde by vacuum infiltration.
   Cross-link was then quenched by 0.125 M glycine. ChIP extracts were
   incubated with anti-GFP (Abcam ab6556) overnight and immunoprecipitated
   by Protein G Dynabeads (Thermo Fisher Scientific, 10004D). The DNA of
   input and reverse cross-linked IP samples was purified using the
   QIAquick PCR clean up Kit (Qiagen) and subjected to qPCR using
   promoter-specific primers ([256]27). To perform ChIP assays using
   protoplasts, pCsVMV:GFP-C-999 (express GFP alone) was used for plasmid
   construction and the ChIP assays was performed as previously described
   ([257]27).

Luminescence measurement and circadian rhythm analysis

   Luminescence acquirement was conducted as previously described
   ([258]54). Data analysis and period estimate (based on the CCA1:LUC
   reporter) were performed as in ([259]57).

RNA extraction and qPCR

   Ten-day-old seedlings grown under light/dark conditions were harvested
   at indicated time points. Total RNA extraction, DNase treatment, cDNA
   synthesis, and qPCR were done as described earlier ([260]57).

Dual bioluminescence assay

   Protoplast isolation and DNA transformation were performed as
   previously described ([261]55). The 35S:Rluc (Renilla reniformis
   luciferase) and respective effectors were cotransformed with the
   CCA1:LUC reporter, and the bioluminescence assay was performed using a
   dual-luciferase reporter assay kit as per the manufacturer’s
   instructions (Promega, E1910). Luminescence was detected using a
   96-well dual-injection luminometer (Centro LB960; Berthold
   Technologies) with a MikroWin 2000 readout system. Results were
   normalized to the GFP expression levels to balance different protoplast
   transfection effectiveness.

ChIP-seq library preparation

   Ten-day-old Arabidopsis plants were harvested at ZT14, cross-linked
   with 1% formaldehyde, and quenched with 0.2 M glycine. For each
   experiment, about 0.5 to 1 g of the sample was ground into fine powder
   in liquid nitrogen. Nuclei were lysed in buffer S (50 mM Hepes-KOH, 150
   mM NaCl,1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 1% SDS, 2
   mM NaF, 2 mM Na[3]VO[4], 50 μM MG132, 50 μM MG115, 50 μM ALLN, and
   protease inhibitor cocktail) at 4°C, and chromatin was fragmented into
   200 to 600 bp by ultrasound processing using a Bioruptor (Diagenode).
   The lysates were centrifuged at 20,000g at 4°C for 10 min. Then, 5 to
   10 μl of GFP antibody (Abcam, ab6556) and 200 μl of suspended protein G
   magnetic beads (Life Technologies, 10003D) were incubated at 4°C for 1
   to 6 hours and rotated to generate antibody-bead complexes. The
   protein-DNA complexes were subjected to immunoprecipitation by
   incubating the antibody-bead complexes with the fragmented chromatin.
   Then, the target protein-DNA complexes were eluted from the beads by
   adding 100 μl of freshly prepared ChIP elution buffer [50 mM tris-HCl
   (pH 7.5), 10 mM EDTA, and 1% SDS]. After reverse cross-linking of DNA
   and protein, ChIP-DNA was extracted using phenol:chloroform:isoamyl
   alcohol (P3803, Sigma-Aldrich), precipitated with ethanol, and
   resuspended in tris-EDTA (TE) buffer. ChIP-DNA libraries were prepared
   using an NEBNext Ultra II DNA Library Preparation Kit. AMPure XP beads
   (Beckman, A63881) were used to select 250- to 650-bp library fragments.
   Last, the DNA fragments were sequenced using an Illumina HiSeq X Ten
   system (150-bp paired-end reads).

Whole transcriptome sequencing (RNA-seq) library preparation

   RNA was isolated from 10-day-old Arabidopsis plants grown under
   light/dark conditions using the RNeasy Plant Mini Kit (QIAGEN, 74904)
   according to the manufacturer’s instructions. A 1.5-μg total RNA was
   depleted of ribosomal RNAs using TruSeq Stranded Total RNA with
   Ribo-Zero Plant (Illumina, RS-122-2401) for RNA-seq according to the
   manufacturer’s instructions. RNA-seq was performed on an Illumina HiSeq
   X Ten system (150-bp paired-end reads).

ChIP-seq data analysis

   Raw data were checked for quality and filtered by FastQC and
   Trimmomatic, respectively. Clean reads were then mapped to the tair10
   reference genome with BWA-MEM (version 0.7.17), using default
   parameters ([262]58). After removing potential PCR repeats with
   Samtools (version 1.9) ([263]59), peak calling was performed with MACS
   (version 2.1.1) (-f BAM -g 1.2e+8) ([264]60). If RSC > 1, NSC > 1, and
   FRIP > 40% of the obtained peak, it can be used for subsequent
   analysis. ChIP-seq peaks were visualized by Integrative Genomics Viewer
   (IGV) (version 2.5.0) ([265]61), and the input file is in bigWig
   format, which can be converted by bamcoverage in deeptools (version
   3.2.1) ([266]62, [267]63).

RNA-seq data analysis

   Raw data of RNA-seq were first evaluated by FASTQC, and qualified data
   were filtered out by Trimmomatic (version0.36) ([268]64). The filtered
   reads were mapped to the tair10 reference genome using Hisat2 (version
   2.1.0) ([269]65), which requires the use of the RNA-strandness RF
   parameter as the data were strand-specific library building. Gene
   expression was quantified using HTseq. If the sum of the expression of
   the same gene in all samples is greater than 5, the gene is defined as
   expressed and used for subsequent analyses ([270]63).

Rhythmicity characterization using time-series RNA-seq data

   Rhythmicity prediction was performed as previously described ([271]66).
   Briefly, RNA-seq data were normalized at all time points, and BIO_CYCLE
   was used to predict rhythmicity of the expressed genes. The cycle
   parameter is set between 20 and 28. The rhythmicity of the gene was
   determined according to the P value and relative entropy error. If the
   P value is less than 0.05 and the relative entropy error is closer to
   1, the gene was considered as rhythmic. The differentially oscillated
   genes between TOC1 and 5X plants were determined by phase change > 2
   hours and amplitude change > 1.5-fold.

   The sinusoidal model fitting for the RNA-seq traces was performed by
   using the ggplot2 package in R on the six–time-point expression data of
   RNA-seq with three replicates ([272]67). The fitting formula for the
   sinusoidal model is cos(2 × pi × Time/24) + sin(2 × pi × Time/24).
   Because the fitted sinusoidal function is cyclical, providing
   expression data at six time points allows for the plotting of diurnal
   expression curves for each gene. The data point in the figure shows the
   average of the three replicates, and the confidence interval of the
   curve was determined by the geom_smooth parameter.

KEGG pathway analysis

   The plant GeneSet Enrichment Analysis Toolkit was used in Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis.
   The FDR (false discovery rate) of <0.05 was set as the cutoff for
   statistical significance
   ([273]http://structuralbiology.cau.edu.cn/PlantGSEA/analysis.php)
   ([274]68).

Motif analysis

   Motif enrichment analysis was done by Homer software as previously
   described ([275]59, [276]69). The parameter “-mset plants” was added
   for plant motif analysis, and other parameters are default.

Acknowledgments