As suites of transcripts are clock regulated in Arabidopsis 10, we hypothesized that this was due to rhythmic accumulation of transcription factors. MYB, bHLH, and bZIP are the predominant factor families previously implicated in light- and clock-regulated accumulation of targets 252627, so we decided to test how prevalent individual rhythms are within these families. Previously, we have collectively reported no less than 368 genes predicted to encode transcription factors in the MYB (131), bHLH (162), and bZIP (75) transcription factor families in Arabidopsis 252627; these were the target pools queried. We accessed expression profiles of circadian experiments, NASCArrays Experiment Reference Number: NASCARRAYS-108 28, using Affymetrix ATH1 arrays containing 22,746 probe sets in the public microarray database, GENEVESTIGATOR 2829. This probe set represents 122 MYB, 111 bHLH, and 67 bZIP genes. In these circadian datasets, a total of 185 of these genes (51 MYB, 81 bHLH, and 53 bZIP) were found to be expressed at least one time point (at p < 0.06). We noticed that the expression of more than half of MYB genes was below detection level on these hybridization samples. Perhaps this implies that many MYB transcripts in this experimental protocol were tissue or growth-stage specific. In contrast, most bHLH and bZIP genes were detected in this array experiment. The expression data we processed from these 185 genes was sufficient for further in silico analyses.

We scored the expression values of the 185 genes using the modified Cosinor analysis 30. This analysis was used successfully in previous experiments to score the circadian expression for genes in Drosophila, mouse, and Arabidopsis 103128. A previous study using this approach in Arabidopsis employed three threshold scales of significance to assess probability (pMMC-β): < 0.02, 0.05, and 0.1 28. We used the same confidence cut-offs to define rhythmic genes (Table 1). A total of 42 transcription factor genes with a pMMC-β value of 0.05, which reflected 9 MYB, 19 bHLH, and 14 bZIP transcription factors, were scored as rhythmic (Table 1, Additional file 1). The percentages of rhythmic bHLH and bZIP genes within each respective family were similar to, or even slightly higher than, that of the set of "all" genes (Table 1). The percentage of rhythmic MYB genes was less than that of the bHLH, bZIP, and the set of all genes; however, noteworthy is the percentage of expressed-MYB on the array that was, itself, lower than other sets. A graphic representation of the expression patterns illustrates when a given peak occurred during transcription factor oscillation (in (h) hours relative to zeitgeber time, which is the time of the last external temporal cue such as the dawn signal of lights-on), (Figure 1). It was noted that many bZIP genes were transcribed during the photophase of the day, whereas many MYB genes peaked during the skotophase (Figure 1). These collective results highlight that, as expected, many transcription factors oscillate, and do so at many discreet phases of the daily cycle.

In the in silico analysis (Figure 1), we found several night-expressed transcription factors. We wondered if such transcription factors could be the as of yet unidentified clock components proposed by current models 1617. To generically test this hypothesis, we surveyed T-DNA tagged lines targeting these transcription factors, obtained from the public stock center, and measured circadian rhythms of these mutants in our system. However, no strong alternations of the circadian rhythm were observed in any of these lines under our assay condition (data not shown). Our observation suggested that none of these factors are the predicted clock element, and thus alternative approaches must be used to define such circadian mutants.

To identify transcription factors that function within the circadian system, as part of control or slave oscillator, we assessed rhythmic output of lines targeted to misexpress a given transcription factor. These experiments, because they are overexpression studies, allowed us to circumvent genetic redundancy. We feel that this is a particularly important consideration in the Arabidopsis clock, as the Mybrelated sequences CCA1 and LHY have strong circadian defects when overexpressed 1213, but have only mild phenotypes as single gene loss-of-function alleles because of the redundancy inherent in the system, 14. Our hypothesis was that members of the MYB, bHLH, and bZIP families are as of yet uncharacterized components of the oscillator and/or are slave components. As a test of this hypothesis, we made use of 198 plants over-expressing 39 MYB genes, 29 bHLH genes, and 4 bZIP genes (Table 2, Additional file 2). Overexpression of each transcription factor was confirmed by RT-PCR on rosette leaf cDNA from T1 plants (data not shown, e.g. Additional file 5).

We constructed a high-throughput time-lapse imaging system, similar to one previously reported 32. With this system, we measured the circadian rhythms of leaf movement in each transgenic line. For this, the seedlings were entrained under 24-hour light-dark cycles for ~10 days, and then the leaf positions of individual plants were imaged under constant light (LL) for an additional 7 days. The circadian parameters of the change in leaf position were analyzed (Additional files 2, 3, 4, 5 &6). We found 33 misexpressors with circadian phenotypes, and 13 out of 33 misexpressor lines targeted the three genes MYB3R2, bHLH69, and bHLH92 (Figure 2; Tables 2 and 3). MYB3R2-ox and bHLH69-ox displayed a 4–8 hour delayed phase of leaf movement rhythms (Figure 2A and 2C). Statistical analysis indicated no significant differences in circadian periodicity (Table 3; Additional files 2 and 3), while the phase difference was significant (Figure 2A–D). bHLH92-ox plants exhibited a 0.5~2 hour lengthened periodicity phenotype, compared to the wild type (Figure 2E). The periodicity differences were statistically significant; WT, 24.16 ± 0.45; bHLH92-ox line A, 26.32 ± 0.77; line B, 25.09 ± 0.32; line D, 25.16 ± 0.36; line E, 24.76 ± 0.41 (p < 0.05, R. A. E. < 0.4) (Table 3). Thus, MYB3R2 and bHLH69 controls circadian phase, and bHLH92 contributes to the regulation of circadian periodicity.

To confirm the clock phenotypes observed in MYB3R2, bHLH69, and bHLH92 over-expression lines, we employed the promoter:LUCIFERASE (LUC) system as an assay that here is used to detect rhythmic patterns of gene expression 33. LUC fusions to the well-characterized circadian-regulated promoter, CCA1 and to the COLD-AND CIRCADIAN-REGULATED 2 (CCR2, also termed AtGRP7) promoter were separately introduced into these MYB3R2-ox, bHLH69-ox and bHLH92-ox plants via fertilization. If these transcription factors act upstream of the clock, both CCA1 and CCR2 oscillations would be altered. Alternatively, if these genes act downstream, controlling the output pathway that regulates leaf movement, the CCA1 rhythm intimately associated with clock function would not be affected. MYB3R2-ox exhibited delayed phase phenotypes of both CCA1 and CCR2 rhythms under LL (Figure 3A and 3B). bHLH69-ox also delayed the phase of the CCA1 and CCR2 rhythm (Figure 3C and 3D). In our mathematical analysis, we could not find any significant effects on the circadian periodicity (Additional file 7A). The second peak positions of CCA1 rhythms in control, MYB3R2-ox, bHLH69-ox and bHLH92-ox were 51.44 ± 0.44, 54.22 ± 0.71, 53.18 ± 0.45 and 53.49 ± 1.03, respectively (Additional file 7B). P-values for MYB3R2-ox and bHLH69-ox were less than 0.01, while p-value for bHLH92-ox was 0.07. In the second peaks of CCR2 rhythm, the values in control, MYB3R2-ox and bHLH69-ox were 37.01 ± 0.23, 39.37 ± 0.63 and 39.81 ± 0.89 (p < 0.01). Thus, we found that overexpression of MYB3R2 and bHLH69 altered both CCA1 and CCR2 rhythms, suggesting that these genes can control core-clock functions, rather than being specific to the leaf-movement-output pathway.

We measured CCA1 and CCR2 rhythms in MYB3R2-ox, bHLH69-ox, and bHLH92-ox plants in constant darkness (DD) (Figure 4). This allows us to compare their behavior to the LL phenotypes. Interestingly, MYB3R2-ox plants had an advanced phased CCR2 rhythm and a delayed phase of CCA1 expression in DD (Figure 4A and 4B). The peak positions in control and MYB3R2-ox were 37.44 ± 0.44 and 35.86 ± 0.46 in the CCR2 rhythm, and 51.23 ± 0.61 and 53.92 ± 0.69 in the CCA1 rhythm (p < 0.01) (Additional file 7D). A delayed phase of CCA1 in DD was seen in bHLH69-ox lines (Control = 51.23 ± 0.61, bHLH69-ox = 54.05 ± 0.34; p < 0.01) (Figure 4C and 4D, Additional file 7D). bHLH92-ox plants also exhibited a clock phenotype in DD. Here an effect on CCA1 phase was detected (Additional file 6B). The values were 53.94 ± 0.57 (p < 0.01). We thus concluded that misexpression of any of these three transcription factors could alter clock parameters. However, the specific nature of the phenotypic effects depended on the light conditions and the output measured.

As described above, we detected circadian alternations when misexpressing given transcription factors. Because the effect of bHLH92-ox was dependent on the light condition, we continued our focus on MYB3R2-ox and bHLH69-ox to further characterize the molecular basis for their phenotypes. To this end, we analyzed transcripts of the central oscillator genes CCA1, LHY, TOC1, and GI in MYB3R2-ox and bHLH69-ox (Figure 5). The MYB3R2-ox and bHLH69-ox plants were entrained under 12 hour light/12 hour dark cycles, and then transferred to constant light conditions. Replicate samples from these plants were harvested every 4 hours for RNA isolation and expression analysis using reverse transcriptase (RT)-PCR. Both MYB3R2-ox and bHLH69-ox were found to result in a repressed transcript level of LHY and TOC1 (Figures 5A and 5B). CCA1 mRNA was found to be slightly decreased in MYB3R2-ox, while this was increased in bHLH69-ox (Figure 5C). Interestingly, in bHLH69-ox, induction of GI expression was found to display a nearly opposite phase with high expression level (Figure 5D). At zeitgeber time (zt) = 4, LHY and CCA1 mRNA are highly accumulated in MYB3R2-ox. This response could be accounted for by an acute light response. Alternatively, a phase delay or defective entrainment in MYB3R2-ox might cause the high expression seen at this time-point. Interestingly, the alteration in GI expression did not result in a dramatic alteration in the timing of the floral induction (Additional file 8). We suggested that MYB3R2 functions as a regulator of CCA1, LHY, or TOC1 transcription, and suggest that bHLH69 plays a similarly important role to regulate CCA1 and GI expression.

We next investigated whether the MYB3R2, bHLH69, and bHLH92 genes were transcribed in a circadian manner. We performed this experiment as the expression profiles of these three genes were not part of the publicly available datasets described above 29. The mRNA accumulation patterns of these transcription factors were assayed by RT-PCR from RNA extracted from plants grown under light-dark cycles and then transferred to LL (Figure 6). MYB3R2 and bHLH92 were likely to be expressed in a circadian manner with a peak between late night and dawn. In contrast, the mRNA accumulation of bHLH69 was not found to oscillate.

Here we describe that MYB3R2, bHLH69, and bHLH92 can contribute to the plant-circadian system. Alongside the characterized transcription factors CCA1, LHY, EPR1, FLC, and LUX, this now adds to the list of transcriptional clock-modulators. We provide evidence that MYB3R2 and bHLH69 influences circadian phase, whereas bHLH92 influences phase and periodicity, dependent on environmental conditions. These effects were not as strong as misexpression of CCA1, LHY, or LUX. It is still unclear how MYB3R2, bHLH69, and bHLH92 function in the clock system. Their specific effects depended on the light environment and on the output measured. Expression analysis showed that misexpression of MYB3R2 and bHLH69 resulted in altered transcript levels of clock genes; however, the circadian rhythms still kept 24-h periodicity albeit with aberrant phase. Thus, transcription factors identified here may play a role in environmental input to the clock or the mediation of its effects, rather than functioning as central-clock components. Such processes are described 34. A future effort to explore a detailed analysis of these transcription factors and the identification of target DNA elements remains to be carried out.

Plant circadian systems possess interlocked feedback loops 171819. In addition, there are various junctures for signal convergence and divergence in the input and output pathways 3536. Though a set of clock regulating transcription factors has already been identified sufficient to describe much of the oscillator framework 11, many other components are believed to still be lacking 10.

We described an additional 20 misexpression lines that exhibit altered clock phenotypes. Whether these genes are components of input core, or output pathways, is as of yet unknown. This was as the phenotypes detected in these lines were not substantiated with alternative transgenic inserts. Forward-genetic analysis of loss-of-function phenotypes of these lines is worth further attention to determine their function in the clock.

A current mathematical model proposes the existence of an unknown transcription factor "X", which activates CCA1 and LHY 1617. We tested hundreds of transcription factors as a pilot study of functional phenomics within the circadian clock, but it appears that in this test, we did not find "factor X." Still, functional analysis of more than a thousand transcription factors still remains. Thus, our pilot efforts substantiate that it is worthwhile to interrogate, via further functional genomic efforts, all known Arabidopsis transcription factors and to analyze their circadian responses.

To dissect the transcriptional regulation in circadian clock, we employed a systematic analysis of global-gene expression. In our in silico analysis, up to 20% of the transcription factors assayed were clock controlled (Table 1). This percentage is slightly higher than that of "all" circadian rhythmic genes in the Arabidopsis genome 10. This implies that the circadian oscillations in the output pathways are not just regulated by a select group of transcriptional feedbacks, and actually require a large number of rhythmic transcription factors. In contrast, the mammalian system is reported to only use around 16 cycling transcription factors, which oscillate in a circadian manner, to regulate the clock 37. The evolution of transcription-factor function and recruitment of molecular targets leads to the linking of many processes of plant physiology to the circadian system. Some components of these systems direct the clock itself, whereas others are only components of rhythmic physiological outputs.

Gene codes of all transcription factors studied, referred to as AGI numbers, were collected within the Arabidopsis Information Resource. The gene-expression profiles in circadian experiment were available from the public microarray database Genevestigator 2928. Expression values from the data were subject to score circadian rhythms with COSOPT 30. Data was collated without non-linear regression (Additional file 1).

Full-length coding sequences (ATG-to-Stop) from MYB-, bHLH- and bZIP transcription factors were amplified from respective cDNAs by PCR using attB-sites containing gene-specific primers. Gateway Entry clones were generated via BP-reaction using the vector pDONR201 (Gateway system, Invitrogen, USA). Via LR-reaction the transcription factor cDNAs were transferred behind the double enhancer cauliflower mosaic virus 35S promoter in the plant expression vector pLEELA 38. Transgenic lines were generated by Agrobacterium tumefaciens-mediated transformation of Arabidopsis plants (Col-0) according to the floral-dip protocol 39. All transgenic lines were selected and self-fertilized. T2 plants were analyzed in this study.

Seeds were surface sterilized with a 70% ethanol rinse, immediately followed by a rinse with 33% bleach, and then a twice washed with sterile water. The seeds were then aseptically sown on Murashige-Skoog (MS) 1.5% agar medium containing 3% sucrose (pH 5.7) with suitable antibiotic (25 mg/L Kanamycin or 12 mg/L Phosphinotrycin; dependent on the transgene) followed by stratification at 4°C for 4–5 days. Seedlings were grown for 9 days at 22°C under 12 hr light/12 hr dark cycles of 75 μmol m^-2 sec^-1 cool white fluorescent light. For flowering-time measurements, experiments were as described 40.

After 9 days of entraining growth, seedlings were transferred to fresh MS 1.5% agar medium (pH 5.7) containing 3% sucrose without antibiotic, and then agar blocks harboring single seedlings were placed to 25-well square tissue culture dishes (Bibby Sterilin). A set of twenty seedlings in a set within each dish were viewed from the side from plates in a vertically placement. The seedlings were entrained as described above for another day, and then were placed into a growth chamber for imaging over 1 week under constant white light (25–40 μmol m^-2 sec^-1) at a constant 22°C. A total of 14 dishes, containing 280 seedlings, were prepared and imaged with 14 monochromatic charge coupled device video cameras per an experiment. Images of seedlings from every camera were transferred to a computer via a Flashbus card and through a custom-built parallel-port controller unit (Universal Imaging, Germany) (system development by Visitron Systems), and were captured and saved every 30-min with a computer program Metaview 4.5 (Universal Imaging) over a week. The vertical positions of primary leaves from the images were measured and analyzed using Metamorph and BRASS, the latter provided by Prof. Andrew Millar (University Edinburgh) as described 2832. Period lengths were estimated from the leaf movement data by the fast Fourier transform nonlinear least-squares method 41. Mean period estimates for each line were based on 10–20 leaf traces from two to four independent experiments analyzed.

Imaging was performed as described via established protocols, where the light was provided from red and blue light-emitting diodes at ~2 μmol m-2 s-1 54243. Period length and Relative Amplitude of error (RAE) were estimated using FFT-NLLS program 41.

Seedlings grown for 1 week under LD cycles and replicate samples were harvested every 4 hours under LL conditions. Total RNA was isolated from the seedlings using the RNeasy Plant Mini Kit (Qiagen, USA), and then was treated with DNase I before reverse transcription. Reverse transcription was performed on 1.0 μg of total RNA with SuperscriptII (Invitrogen). Quantitative PCR were performed with iQ5 real-time PCR system (BIO-RAD). Gene-specific primers were described previously: CCA1, LHY and TOC1 44, GI and TUB 15. Primers for MYB3R2, bHLH69 and bHLH92 were designed as follow:

MYB3R2-FW, 5'-CTTGGACCACAGAGGAAGAAGT-3'

MYB3R2-RV, 5'-TGTTGTTGGTGGTGGTAACCTA-3'

bHLH69-FW, 5'-CCATCCTAATGACGCTCTCTTC-3'

bHLH69-RV, 5'-ATCAGTGGCTTGACCTCTCCTA-3'

bHLH92-FW, 5'-CTGAGAAAGAATTGGGAGGAGA-3'

bHLH92-RV, 5'-GACCATCCTTTGCTGATTTTTC-3'