Arabidopsis circadian clock protein, TOC1, is a DNA-binding transcription factor

Previously it was shown that TOC1 occupies a region of the CCA1 promoter that contains a TCP binding site (TBS) (17), presumably through interaction with the TCP transcription factor CHE. To explore whether TOC1 associates with other regions of the CCA1 genomic region, we performed ChIP using the previously described TOC1 minigene (TOC1p::TOC1:YFP) transgenic line (15). Seedlings were grown under 12-h light:12-h dark (LD) cycles and cross-linked at zeitgeber time (ZT) 14 when the peak of TOC1 protein accumulation occurs (15). Enrichment of CCA1 genomic regions after immunoprecipitation with an anti-GFP antibody was determined by quantitative real-time PCR (qRT-PCR) using primers that cover the CCA1 promoter and 5′ UTR regions (Fig. 1A). Our results indicate that TOC1 occupies three regions of the CCA1 promoter and CCA1 coding sequence outside of the TBS region (E and F region) (Fig. 1A). Regions correspond as follows: A = −866/−786, B = −668/−589, C = −472/−365, D = −364/−274, E = −264/−171, F = −172/−85, G = −58/+69, H = +144/+235, I = +322/+413, or J = +537/+622 base pairs from the transcriptional start site. TOC1 is associated with a region upstream of the TBS that contains a G-Box (CACGTG) (C and D region), a region 5′ of the G-box region (A region), and a region near the beginning of the CCA1 coding sequence (I and J regions) (Fig. 1A). It is possible that TOC1 is recruited to the site containing the G-box through its interaction with phytochrome interacting factor (PIFs) or phytochrome interacting factor-like (PILs) (20), but the I/J and A regions have no canonical cis-elements for transcription factors that are known to interact with TOC1 (G-box or TBS). This finding suggests that TOC1 either interacts with unknown DNA-binding factors that recruit it to these regions or that it can bind DNA directly. Previous work showed that the LHY promoter does not contain a TBS and is not bound by CHE in yeast one-hybrid assays (17). However, genetic evidence suggests that TOC1 also regulates the expression of LHY (4). To test whether TOC1 can associate with the LHY promoter, we used ChIP–qRT-PCR with primers designed to two regions of the LHY promoter (Fig. 1B), one containing a G-box and another without G-box or TBS that is located 5′ of the G-box in the LHY promoter (Fig. 1B). TOC1 associates at both of these locations in the LHY promoter. These experiments provide evidence that TOC1 can occupy the LHY promoter and indicates that TOC1 is directly associated to multiple regions in the promoters of CCA1 and LHY.

To investigate whether TOC1 has DNA-binding activity, we examined the conserved regions of TOC1 and the PRRs and found, as expected, a high degree of similarity within the CCT domain (Fig. 2A). The N terminus of the CCT domains is more divergent than the C terminus (Fig. S1) and is known, in the HAP2 protein from yeast, to bridge interactions with other proteins (12, 21); the C-terminal portion of all CCT domains contain most of the amino acids necessary for DNA binding previously reported for CO and the DNA-binding protein AtHAP2 (Fig. 2A) (16, 21). To test whether TOC1 can bind DNA, purified full-length TOC1 protein was obtained using a baculovirus-insect cell purification system and used to perform EMSAs. As shown in Fig. 2B, TOC1 binds to a 60-bp probe from the J region of CCA1 (Fig. 1A) that contains no TBS or G-box. The CO-response element was shown to be TGTG(N2-N3)ATG (16), but the probe contains only a TGTG motif, so we performed a competition EMSA with both wild-type and mutant (TGTG changed to AAAA) unlabeled J probes (Fig. 2B). Fig. 2B and Fig. S2 show that the wild-type probe competes with the labeled probe but the mutant probe does not, indicating that TOC1 binds to the TGTG motif. Interestingly, this TGTG sequence (underlined) that we termed T1ME is part of the previously defined morning element (ME) (CCACAC or GTGTGG) (22) and the hormone up-regulated at dawn motif (HUD) (CACATG or CATGTG) (23), which are enriched in the promoters of morning-phased genes. As shown in Fig. 2C, we performed EMSAs that confirmed that TOC1 binds to 60-bp synthetic probes containing the T1ME, ME, or HUD elements. It is known that the LHY and CCA1 promoter regions can confer cycling (24), and the J region of the CCA1 promoter is not necessary for the cyclic behavior of CCA1 expression. To test whether TOC1 can bind directly to the promoters of CCA1/LHY, probes that contain the T1ME in the CCA1 and LHY promoters were made. TOC1 is able to bind both probes (Fig. 2C), showing that TOC1 alone directly interacts with the promoter sequence in vitro. To test whether the CCT of TOC1 is the domain that mediates direct DNA binding, the interaction between a purified recombinant GST-tagged TOC1-CCT (TOC1-CCT) and the J probe were tested by EMSA (Fig. 2D). The TOC1-CCT variant binds DNA and the GST and no-protein controls showed no DNA-binding activity. To determine if binding is specific to the CCT domain and whether conserved amino acids are necessary for DNA binding, two CCT-domain mutant variants were generated: a single alanine to valine (A mutation), which is found in the toc1-1 semidominant mutant (6), and a quadruple alanine, arginine, glycine, and phenylalanine to valine, isoleucine, valine, and leucine (ARGF mutation), that has been shown to disrupt DNA binding in the HAP2 DNA-binding domain (Fig. 2A) (21). Both the A and ARGF mutant versions of the CCT domain lost affinity for the J probe (Fig. 2E), confirming that the CCT domain is sufficient for DNA binding and is dependent on the ARGF conserved amino acids.

The PRR proteins have been shown to be transcriptional repressors that associate with the CCA1/LHY promoters in ChIP experiments (19). Based on the results obtained with the TOC1-CCT domain, we investigated whether PRR-CCT domains encompass a similar DNA-binding activity. The four PRR CCT domains bound to the J probe, with PRR3 showing weak binding (Fig. 2F). This finding may coincide with a divergence in the PRR3 protein sequence at the fourth conserved arginine that is important for DNA binding in HAP2 (Fig. 2A). Taken together, these results indicate that TOC1 and the PRRs are a family of DNA-binding proteins and that the DNA-binding activity relies on the CCT domain shared by these proteins.

Because a toc1 loss-of-function mutant has low CCA1/LHY expression, it was proposed that TOC1 is an activator of CCA1/LHY (4), but conversely TOC1 overexpression results in low CCA1/LHY expression (8, 9). To address this discrepancy and test whether TOC1 activates or represses CCA1/LHY, we used a published ethanol-inducible TOC1 transgenic line (ALC::TOC1) (25). We treated both the transgenic line and wild-type with either a mock treatment or 2.5% ethanol for 3 h starting at ZT22 in LD and collected samples each hour. As shown in the upper left panel of Fig. 3A, TOC1 is induced approximately sixfold by the ethanol treatment. At the same time points, CCA1 and LHY are repressed up to twofold, suggesting that TOC1 acts as a repressor of these genes (Fig. 3A). We also performed the same experiments starting at ZT6 and ZT18 in constant light conditions (LL) and completed longer time courses (Fig. S3). TOC1 induction represses CCA1 and LHY expression under all cases except for the 1- and 7-h time points collected at ZT19 and ZT25 (Fig. S3 B and C).

To confirm that TOC1 represses CCA1 expression, a full-length TOC1 overexpression construct was cobombarded with a CCA1::LUC⁺ reporter (CCA1 promoter driving expression of firely luciferase) into Arabidopsis seedlings. Luciferase levels were threefold lower than the control when TOC1 is overexpressed (Fig. 3B and Fig. S3A), confirming that TOC1 can repress CCA1 expression. To test if this result is dependent on a functional CCT DNA-binding domain, we introduced the A and ARGF mutations (Fig. 2 A and F) into the CCT domain of the full-length TOC1 and cobombarded these with the CCA1::LUC⁺ reporter. Without a functional DNA-binding domain, TOC1 is not able to repress CCA1 at both ZT1 and ZT13 (Fig. 3B and Fig. S4A). The expression levels of wild-type and mutant TOC1 were quantified by qRT-PCR and confirmed that the effects were not because of differences in expression level (Fig. S4B).

To test whether TOC1 acts as a general transcriptional repressor, we used the Gal4-LexA/UAS system, which tests proteins for positive or negative transcriptional potential (26). TOC1 represses the activity of VP16 and did not activate the reporter on its own (Fig. 3C), showing that TOC1 is a transcriptional repressor. PRR5, -7, and -9 contain a repression motif in the IR of the protein that is not present in TOC1, so we also tested the three domains of TOC1 (PR, IR, and CCT) (Fig. 3C, cartoon under plot) for transcriptional potential. The results in Fig. 3C show that the TOC1 PR domain contains the transcriptional repression potential, supporting the idea that TOC1 does not contain the transcriptional repression motif in the IR as was seen for PRR5, -7, and -9 (19).

To determine whether the PR domain can repress CCA1 expression in the absence of the CCT and IR, we overexpressed the PR domain in a transgenic line carrying a CCA1::LUC⁺ reporter construct and monitored oscillations in luciferase activity in LL (Fig. 3D). We averaged the luciferase activity of 32 T1 lines where either TOC1 or TOC1-PR are overexpressed, and 16 CCA1::LUC⁺ control seedlings. Overexpression of the full-length TOC1 caused dampening of CCA1 expression and arrhythmia after 2 d in LL (Fig. 3D and Fig. S5), as expected (8, 9). However, TOC1-PR overexpression caused a shortened period in luciferase activity (Fig. 3D and Fig. S5). This finding corresponds with previous findings that a nonfunctional, or in this case absent, CCT domain prevents TOC1 or the PR repression domain from negatively regulating CCA1 expression (4, 6). Additionally, the short period caused by PR-constitutive expression is the same as seen in the toc1-1 semidominant mutant and toc1 loss-of-function mutants (4), suggesting that overexpression of the PR domain alone has the same effect on the clock as a loss-of-function mutant.

We next reversed the transcriptional polarity of TOC1 by making a translational fusion with the strong transcriptional activation domain, VP64. We overexpressed the full-length TOC1 with (TOC1-VP64-ox) and without (TOC1-ox) the VP64 in a CCA1::LUC⁺ genetic background and tracked luciferase activity in 18 T1 generation TOC1-ox or TOC1-VP64-ox lines and 12 CCA1::LUC⁺ control lines. As was seen previously, most of the TOC1 overexpression lines are arrhythmic [11 of 18 with a relative amplitude error (RAE) > 0.6] (Fig. 3E) or have a short period call when rhythmic (RAE < 0.6), whereas the TOC1-VP64-ox lines are mostly rhythmic (16 of 18 with RAE < 0.6) with a short period (Fig. 3E). This finding shows that reversing the transcriptional potential of TOC1 makes it behave like a loss-of-function mutant and confirms the importance of TOC1 repression activity to proper clock function.

To identify potential targets and output pathways controlled by TOC1, we performed microarray analysis on RNA collected from ALC::TOC1 and wild-type Arabidopsis mock-treated or ethanol-treated for 3 h and collected at ZT1 in either LD or LL. We treated all samples with 20 μM MG132 to block proteasome function and increase TOC1 protein levels after induction. Differentially expressed genes specific to ALC::TOC1 and ethanol treatment and not wild-type or mock-treated were identified using LIMMA (27). This approach controls for the general effects of ethanol on Arabidopsis and for any gene-expression changes caused by the ALC::TOC1 transgene before treatment.

In total, TOC1 induction caused misexpression of 2,566 genes, with 1,254 genes up-regulated and 1,312 genes down-regulated (Fig. 4A). Of the up-regulated genes, 570 were misregulated in LD and 860 were misregulated in LL, with an overlap of 176 genes. Of the down-regulated genes, 580 were misregulated in LD and 876 were misregulated in LL. There was only one gene that overlapped between the genes that were down-regulated in LD and those that were up-regulated in LL, and no genes were shared between the genes that were down-regulated in LL and up-regulated in LD, showing that there was good agreement between the up-regulated and down-regulated genes from all datasets. The groups of genes were then compared with genes that were called cycling from four different microarray datasets, two collected from plants growing under LD conditions [LDHH_SM (28) and LDHH_ST (29)] and two collected from plants growing under LL conditions [LL12_LDHH (5) and LL23_LDHH (30)]. The genes that change by TOC1 induction in LD and LL had more cycling genes in all four of the tested datasets than would be expected by chance, showing that TOC1 regulates expression of genes with cyclic behavior (Fig. 4B).

A permutation-based z-score was calculated to compare the phase distribution of the cycling genes that change by TOC1 induction (Fig. 4C). The genes up-regulated and down-regulated by TOC1 induction in LDHH entrainment show a peak of enrichment at the morning and evening light transitions for both the LL and LD (Fig. 4C) induction conditions, with the highest peaks before dawn or subjective dawn. There is also a peak at dusk (LD) or subjective dusk (LL). The peaks seen during the dawn transition may represent the direct targets of TOC1 because those targets are expected to be phased to the morning, and the peak in the middle of the day may represent downstream targets of the morning-expressed transcription factors that are regulated by TOC1. The LL12 dataset shows a similar pattern of peaks as LDHH (Fig. S6A), and the peaks in the LL23 dataset are shifted (Fig. S6A). When all datasets are combined (Fig. S6B), the peaks follow the same trend as the LDHH (Fig. 4C).

To look for enrichment of cis-elements we took 500 bp of the promoter sequence from the shared up- (176 from Fig. 4A) and down- (144 from Fig. 4A) regulated genes from our dataset. These sequences were analyzed by three different motif-finding software packages [MEME (31), WEEDER (32, 33), and the R package rGADEM (34)]. Three elements were found to be enriched in the promoters of these genes using this stringent approach. In the down-regulated genes, a variant of the TBS (GGGCCAT) (35) was found (Fig. 4D). In the up-regulated genes a G-box variant (CACGTGKM) and a plant-specific GA motif (AGARRGARRRAGADR) are enriched. The GA motif is the cis-element for a family of DNA-binding proteins called BASIC PENTACYSTEINES (BPCs) (36). Interestingly, three of the four regions that we show to be occupied by TOC1 in the CCA1 promoter contain the enriched elements. The E/F regions contain a TBS, the C/D regions contain a G-box, the A region contains a site similar to the AGARRGARRRAGADR element (AAGGAGGAAGAAG or ARRGARRRAGADR) (Fig. 1A), and we have shown direct binding through the T1ME for the I/J region. These results suggest that there are complex transcriptional mechanisms underlying TOC1 regulation of gene expression. The T1ME was not found to be enriched in this dataset, but this may be expected because we have not defined the full cis-element sequence using an unbiased method, such as protein-binding microarray or systematic evolution of ligands by exponential enrichment. We have listed all genes that contain a T1ME, HUD, or ME in Dataset S1, and despite not being enriched in the dataset, the presence of these elements in a promoter may indicate direct binding by TOC1.

To find clock output pathways we examined genes differentially regulated by TOC1 induction at least 1.8-fold for gene ontology (GO) enrichment in biological processes. Flow charts of these results are presented to visualize the hierarchy of GO categorization (Figs. S7 and S8). The down-regulated genes in LL were enriched in two main categories: response to stimulus, which includes stress, and biotic stimulus and metabolic process (Fig. S7). This finding coincides with the circadian clock being a major regulator of daily metabolic processes (37) and it is likely that TOC1 will be a hub that regulates genes involved in metabolism directly. Up-regulated genes in the LD experiment showed enrichment of categories of abiotic stimuli—such as light, temperature, and oxidative stress—all categories that coincide with the dark-to-light transition in the morning. This finding is in contrast to the down-regulated genes in LD, which show an enrichment for genes involved in biotic responses, mainly defense responses, which has recently been shown to be circadian-regulated (38, 39). These experiments show that TOC1 is involved in the regulation of circadian-regulated pathways that are critical to plant function.

Arabidopsis ecotype Colombia (Col-0) was used for all experiments. Growth conditions and transgenic lines are described in SI Materials and Methods.

ChIP was done as described previously (17). Primer details are described in SI Materials and Methods.

EMSAs were done with standard conditions provided in the Lightshift Chemiluminescent EMSA kit (Thermo Scientific). Protein purification conditions, EMSA buffers, and probe sequences are provided in SI Materials and Methods.

An established TOC1 inducible transgenic line (ALC::TOC1) (25) and Col-0 (wild-type) were used in all experiments. Growth and induction conditions, qRT-PCR primers, and microarray hybridization are described in SI Materials and Methods.

The Gal4/UAS system has been described previously (26). Experimental details are provided in SI Materials and Methods.

Details of microarray analysis are described in SI Materials and Methods.