Supplemental Data

Full Text
Orchestrated Transcription of Key Pathways in Arabidopsis by the Circadian Clock
Stacey L. Harmer, John B. Hogenesch, Marty Straume, Hur-Song Chang, Bin Han, Tong Zhu, Xun Wang, Joel A. Kreps, and Steve A. Kay

Supplementary Material

Supplemental Table 2. Genes implicated in phenylpropanoid biosynthesis are under clock control, peaking before subjective dawn.
Gene nameAccession number
Phenylalanine ammonia-lyase (PAL1)AAD18156
Phenylalanine ammonia-lyase (PAL2)P45724
Cinnamate 4-hydroxylaseAAC99993
4-Coumarate:CoA ligase 3AAD47194
Cinnamyl-alcohol dehydrogenase (ELI3-1)CAA48027
Cinnamyl-alcohol dehydrogenase (CAD1)S71179
Putative cinnamoyl CoA reductaseAAC63661
Probable caffeoyl-CoA O-methyltransferaseT05431
Chalcone synthaseP13114
Flavanone 3-hydroxylaseAAC49176
Flavonol synthaseQ96330
Leucoanthocyanidin dioxygenaseT05119
Dihydroflavonol 4-reductaseP51102
UDP rhamnose-anthocyanidin-3-glucoside rhamnosyltransferase-like proteinCAB81407
Similar to glutathione S-transferaseAAD39312
Glutathione S-transferaseAAD32887
Glutathione S-transferaseAAA74019
Glutathione S-transferase CAA74639
Glutathione S-transferaseAAD32888
Glutathione-conjugate transporter ( AtMRP2)AAC16268
Glutathione-conjugate transporter (AtMRP4)CAA05625
Phenylcoumaran benzylic ether reductase-likeZ49777
Sinapoylglucose:malate sinapoyltransferase-likeAC004401

Supplemental Table 3. A 9-base pair (bp) motif was found in the promoters of many evening-phased genes. Genomic DNA corresponding to individual genes on the Affymetrix Arabidopsis GeneChip was identified using the N-terminal 300 bp of each gene as subject and the BLASTN algorithm with the following parameters: expect = 1E-80, number of sequences to report = 10, number of alignments to report = 10 (1). This file was analyzed for subject sequences of length > 10,000 bp, then parsed for 1500-bp sequences upstream of the start methionine from each gene. This process resulted in the identification of promoter sequences corresponding to 56% of all genes represented on the array. Sequences 1500 bp upstream from the ATG corresponding to each cycling gene were searched using AlignACE and ScanACE programs with the following parameters: expect = 10, word size = 10, iterations = 5 (2). Consensus sequences corresponding to conserved motifs were analyzed against Transfac (3). The sequence AAAATATCT was found 46 times in the promoters of the 31 genes listed below.
Gene nameAccession numberNumber of evening elements
Unknown proteinAAA502351
Unknown proteinAAB630761
Similar to extracellular dermal glycoproteinAAC721201
Putative DnaJ proteinAAD035701
Unknown proteinAAD181021
Putative zinc finger proteinAAD305761
Unknown proteinAAD313731
Unknown proteinAAD414211
Auxin transport protein (PIN3)AAD526951
GTP cyclohydrolase II CAA038841
Photosystem II oxygen-evolving complex proteinCAA756291
2-Dehydro-3-deoxyphosphoheptonate aldolase CAB389111
Unknown proteinH71403 1
Beta-9 tubulin (TUB9) M847061
1-Aminocyclopropane-1-carboxylate oxidase Q065881
1Proteinase inhibitor IIS305781
Probable hexose transporter (STP14)T004501
Unknown proteinT045481
Aldose 1-epimerase homolog T077191
Unknown proteinT077251
Unknown proteinT102391
CDP-diacylglycerol synthetase homolog X943061
Unknown proteinAAD328702
Fructose-bisphosphate aldolase P221972
Cysteine protease RD19A P432962
Fatty acid elongase homologT047712
Unknown proteinAAC394683
R1-like protein / starch kinase AAD313373
Unknown proteinAAD419773
Unknown proteinT052264

Supplemental Figure 1. Assessment of microarray reproducibility. Biotin-labeled cRNA prepared from plants harvested at CT0 was hybridized sequentially to two Affymetrix GeneChips manufactured on the same silicon wafer. The solid line indicates a difference of a factor of 2, and the dashed lines a factor of 3, between the two hybridizations. Circadian-regulated genes are colored red; all other genes are gray.

Medium version | Full size version

Supplemental Figure 2. All phases of gene expression are represented. cRNA samples were prepared as described and hybridized to oligonucleotide-based microarrays (4). Genes determined to be clock-regulated (5) were sorted according to their peak phase of expression. Number of genes peaking at each phase is indicated.

Medium version | Full size version

Supplemental Figure 3. Clock control of light signaling genes. Data were normalized such that the median signal strength for each gene over all time points was 1. The average signal strength at each time point was then graphed as a ratio relative to the median signal strength of that gene. (A) Photoreceptor genes are clock-controlled. PHYB (P14713) is in red, CRY1 (CAB78016) and CRY2 (AAD09837) are in light blue, and NPH1 (AF030864) is in dark blue. PHYB has been previously reported to cycle (6). (B) Downstream signaling component genes are clock-controlled. SPA1 (AAD30124) is in pink and RPT2 (AAF33112) is in green. (C) The phytochrome and cryptochrome signaling pathways mediate input to the circadian clock and are also clock outputs. In contrast, there is no perturbation of circadian function in nph1 mutants (7), indicating that NPH1 is solely an output component.

Medium version | Full size version

Supplemental Figure 4. Clock control of genes implicated in stress responses. (A) Cold- and stress-induced genes (in light blue) are clock-regulated. These genes encode putative LEA protein (CAB42908), ribosomal-protein S6 kinase homolog (ATPK19) (D42061), related to CIC protein (AAB82643), CCR1 (Q03251), CCR2 (Q03250), SEN1 (AAA80303), ERD15 (T02438), cysteine protease (RD19A) (P43296), DC 1.2 homolog (X80342), COR15b (S43320), and COR6.6 (CAA38894). Cycling genes encoding lipid modifying enzymes are in pink: fatty acid elongase homolog (T04771), delta(9)-desaturase (BAA25181), sphingolipid desaturase (AAC62885), putative UDP-glucose:sterol glucosyltransferase (CAB06081), CDP-diacylglycerol synthetase homolog (X94306), and putative lecithin-cholesterol acyl transferase (AAF99739). The gene encoding transcription factor DREB1A/CBF3 (T05799) is indicated in purple. (B) Possible mechanism for clock-mediated control of chilling resistance. Regulation of lipid-modifying genes by DREB1a/CBF3 has not been demonstrated and is speculative.

Medium version | Full size version

Supplemental Figure 5. Genes implicated in sugar metabolism peak in the subjective afternoon. (A) Genes involved in glucose oxidation. Genes encoding glycolytic enzymes are in pink: fructose-bisphosphate aldolase (P22197), putative fructose bisphosphate aldolase (CAB86897), and aldose 1-epimerase homolog (T07719). Genes encoding oxidative pentose phosphate enzymes are in blue: probable phosphoriboisomerase (Q9ZU38), cytosolic glucose-6-phosphate 1-dehydrogenase (AJ010970), and possible glucose-6-phosphate dehydrogenase (AAD09232). (B) Genes encoding enzymes involved in galactinol synthesis: myo-inositol 1-phosphate synthase isozyme-2 (AAC49172) (in brown), and putative galactinol synthase (AAB63818) and UDPglucose 4-epimerase (S62783) (in gold). Myo-inositol 1-phosphate synthase peaks earlier in the day than the other genes in this figure; we surmise this is because its product is required in a number of other pathways such as lipid synthesis. (C) Genes encoding hexose transporters: putative sugar transporter (AAB70415), putative sugar transporter (AAB70414), glucose transporter STP1 (P23586), and putative sugar transporter (STP14) (T00450) (all in red). (D) Genes encoding trehalose 6-phosphate synthases: putative trehalose-6-phosphate synthase (AAF16560), trehalose-6-phosphate synthase homolog (T02267), and putative trehalose-6-phosphate synthase (AAD08939) (all in green). (E) Model depicting roles of gene products in allocation of assimilated carbon. Colored arrows indicate the known or predicted function of the enzyme encoded by the corresponding clock-controlled gene. Not all intermediate steps in each pathway are shown.

Medium version | Full size version

Supplemental Figure 6. Genes encoding enzymes involved in mineral assimilation are under clock control. (A) Nitrogen assimilation genes are circadian-regulated. Genes encoding peptidases are in pink [putative serine carboxypeptidase II (AAB80670) and putative prolylcarboxypeptidase (AAF18628)]; amidases are in light blue [probable formamidase (T04713), similar to arginases (AAD17371), and probable formamidase (T04712)]; a nitrate transporter (AtNTP3) (CAB38706) and an ammonium transporter (ATM1;2) (AAD38253) are in dark blue; glutamate dehydrogenase (GDH1) (S71217) is in green; and asparagine synthetase (ASN1) (P49078) is in gold. ATM1;2 is predicted to be in the thylakoid membrane by the PSORT program (, and the gene has been previously reported to cycle under diurnal conditions (8). (B) Known and predicted roles of the products of the genes depicted in (A). The mitochondrion is bounded by an orange box, the chloroplast by a green box, and the cytoplasm by a black box. (C) Sulfur assimilation genes are circadian-regulated. Genes encoding sulfate transporters are in red [sulfate transporter (ATST1) (CAB41310) and sulfate transporter (S74246)]; adenosine 5´-phosphosulfate reductase (APR1) (AAB80957) is in light blue; phosphoglycerate dehydrogenase (O04130) is in gold; and serine O-acetyltransferase, chloroplastic (AAC37474) is in green. (D) Known functions of the products of the cycling genes depicted in (C). The chloroplast is bounded by a green box and the cytoplasm by a black box. (E) Genes encoding the sulfate and nitrate transporters peak coordinately toward the end of the night. Sulfate transporters genes are in red and the nitrate transporter gene is in dark blue. (F) Assimilation of sulfate requires products of the carbon and nitrogen assimilation pathways. Nitrogen-containing compounds are in blue, sulfur-containing compounds are in red, and compounds containing both nitrogen and sulfur are in pink. Not all intermediate steps in each pathway are depicted.

Medium version | Full size version

Supplemental Figure 7. Flowering-time genes are clock-controlled. (A) Green traces represent genes encoding CCA1 (T02684) and LHY (CAA07004), orange trace GI (CAB56039), black trace FT (BAA77838), and blue trace COL1 (Y10555). All of these genes have been previously reported to cycle (9-11). (B) External coincidence model for control of flowering time. Temporal expression pattern of a hypothetical gene is shown under long-day and short-day growth conditions. The phase of peak expression differs under the two conditions such that its peak expression level coincides with periods of light input under long-day but not short-day conditions. Expression above some threshold level during daylight hours may trigger flowering (9).

Medium version | Full size version

Functional Interdependence Between Photosynthesis Gene Products The coregulation of photosynthesis genes is probably a reflection of their functional interdependence, since their protein products work as a unit to harvest light energy (12) (Fig. 1C of the article). Additional interdependence exists between the LHC proteins and their ligand chlorophyll, neither of which is stable in the unbound state. Therefore, circadian control not only allows the coordinate up-regulation of these cooperative light-harvesting genes in anticipation of dawn, but also helps synchronize the synthesis of LHC proteins and chlorophyll. These data suggest that clock regulation of the photosynthetic machinery allows plants to efficiently use each day's allotment of sunlight.

Dawn-Phasing of Uptake of Minerals May Facilitate Their Reduction Clock regulation of these genes may help temporally coordinate these interdependent metabolic pathways. In addition, the assimilation of these minerals requires large amounts of reducing power, which is produced in abundance by photosynthesis. The near-dawn phasing of the nitrate and sulfate transporters (Web fig. 6E) may help ensure that there is enough reducing power for the conversion of these minerals into biologically useful forms after their uptake.

Perspective The clock-regulated genes identified in this study are implicated in a variety of physiological functions, giving us our first glimpse of the range of processes controlled by the clock in a eukaryote. Of course, changes in steady-state mRNA levels are not always mirrored by changes in protein levels or enzyme activities. But numerous examples of such correlated changes do exist (13-16). We were able to assign many of the cycling genes identified in this experiment into functional groups. Perhaps most notable was the circadian regulation of pathways involved in plant adaptations to sunlight. A dramatic up-regulation of photosynthesis genes occurred early in the day, presumably allowing plants to coordinately assemble the photosynthetic machinery and to take full advantage of every photon of the day's sunlight. Even more striking was the synchronous up-regulation just before dawn of over 20 enzymes involved in phenylpropanoid synthesis. Products of this pathway provide UV protection for plants (17, 18), supporting the proposition that a primary driving force for circadian clock evolution is a "flight from light" (19).

References and Notes

1. S. F. Altschul, W. Gish, W. Miller, E. W. Myers, D. J. Lipman, J. Mol. Biol.215, 403 (1990).

2. J. D. Hughes, P. W. Estep, S. Tavazoie, G. M. Church, J. Mol. Biol.296, 1205 (2000).

3. R. Knuppel, P. Dietze, W. Lehnberg, K. Frech, E. Wingender, J. Comput. Biol. 1, 191 (1994).

4. Total RNA was prepared from the staged tissue samples using the Qiagen RNeasy Plant Mini Kit. Double-stranded cDNA was synthesized from 5 ?g of total RNA using Superscript II reverse transcriptase (Gibco BRL). This cDNA was used as a template to synthesize biotin-labeled cRNA by in vitro transcription using an ENZO BioArray High Yield RNA transcript labeling kit (ENZO). Amplified cRNA was fragmented and hybridized to GeneChip microarrays (Affymetrix) overnight at 45�C. The hybridized arrays were washed and stained with biotinylated anti-streptavidin antibody and a phycoerythrin streptavidin conjugate and then scanned using a Hewlett-Packard GeneArray Scanner. Affymetrix GeneChip software was used to determine the average difference values between perfectly matched oligonucleotide probes and single base pair mismatches for each probe set. Data were then scaled globally such that the average intensity of each microarray equaled a target intensity of 200 (20). The resulting hybridization intensity values reflect the abundance of a given mRNA relative to the total mRNA population and were used in all subsequent analyses.

5. Analysis for rhythmic gene expression was performed by an algorithm (CORRCOS) developed to empirically test for statistically significant cross correlation between test cosine waves of specified period and phase and each experimentally observed 44-hour gene expression profile. Means and SEMs of duplicate pairs of gene expression profiles were first detrended by unweighted linear regression, the results of which were then converted to standard normal deviates (by dividing the detrended means and SEMs by the detrended series SD) prior to subsequent analysis. One thousand cosine test periods, equally spaced in frequency from 1/8 hour (the highest frequency; inverse of 2Δt where the sampling interval Δt = 4 hours) to 1/8000 hours in increments of (1/2Δt)/1000, were considered at a phase, ϑ, resolution of 1% of each test period, τ (i.e., 101 ϑ's per τ ranging from -τ/2 leqϑleqτ/2 at Δϑ = τ/100). Cosine acrophase (i.e., peak, max) occurs at ϑ = 0. Statistical significance was assessed at each period/phase by an empirical, resampling procedure. One thousand surrogates of each series of original standard normal deviates were created in which both (i) the data were temporally randomized (i.e., by randomly shuffling the order of their occurrence) and (ii) pseudo-Gaussian distributed random noise was added to each data point in proportion to its associated duplicate SEM estimate. The mean and SD of the cross-correlation values calculated for each of the 1000 surrogates were assessed at each period/phase. Significance probability was calculated at each period/phase based on the Z score of the difference of the original calculated cross-correlation (CCorig) minus the mean cross-correlation from the 1000 surrogates (CCsurr) relative to the SD from the 1000 surrogates (SDsurr) [i.e., Z = (CCorig - CCsurr)/SDsurr]. The probability associated with this Z score was reported as a one-sided probability that was multiple-measures corrected with respect to the number of original time series data points (N = 12). Additional analysis was performed with the GeneSpring program software package (Silicon Genetics).

6. L. K. Bognar et al., Proc. Natl. Acad. Sci. U.S.A.96, 14652 (1999).

7. D. E. Somers, S. A. Kay, unpublished data.

8. S. Gazzarrini et al., Plant Cell11, 937 (1999).

9. A. Samach, G. Coupland, Bioessays22, 38 (2000).

10. S. Ledger, C. A. Strayer, S. Kay, J. Putterill, unpublished data.

11. J. H. Ahn, D. Weigel, unpublished data.

12. J. H. Nugent, Eur. J. Biochem. 237, 519 (1996).

13. B. Piechulla, Chronobiol. Int.16, 115 (1999).

14. C. Heintzen, M. Nater, K. Apel, D. Staiger, Proc. Natl. Acad. Sci. U.S.A. 94, 8515 (1997).

15. Z. Y. Wang, E. M. Tobin, Cell93, 1207 (1998).

16. J. C. Dunlap, Cell96, 271 (1999).

17. L. G. Landry, C. C. S. Chapple, R. L. Last, Plant Physiol.109, 1159 (1995).

18. J. Li, T.-M. Ou-Lee, R. Raba, R. G. Amundson, R. L. Last, Plant Cell 5, 171 (1993).

19. C. S. Pittendrigh, Annu. Rev. Physiol. 55, 17 (1993).

20. L. Wodicka, H. Dong, M. Mittmann, M. H. Ho, D. J. Lockhart, Nature Biotechnol. 15, 1359 (1997).