Circadian oscillations of cytosolic free calcium regulate the Arabidopsis circadian clock

Abstract

In the last decade, the view of circadian oscillators has expanded from transcriptional feedback to incorporate post-transcriptional, post-translational, metabolic processes and ionic signalling. In plants and animals, there are circadian oscillations in the concentration of cytosolic-free Ca²⁺ ([Ca²⁺]_cyt), though their purpose has not been fully characterised. We investigated whether circadian oscillations of [Ca²⁺]_cyt regulate the circadian oscillator of Arabidopsis thaliana. We report that in Arabidopsis, [Ca²⁺]_cyt circadian oscillations can regulate circadian clock function through the Ca²⁺-dependent action of CALMODULIN-LIKE24 (CML24). Genetic analyses demonstrate a linkage between CML24 and the circadian oscillator, through pathways involving the circadian oscillator gene TIMING OF CAB2 EXPRESSION1 (TOC1).

Circadian oscillators confer competitive advantage by modulating physiology and development^{1, 2}. In Eukaryotes, circadian oscillators are comprised of feedback loops of transcriptional regulators, however, the oscillator genes differ between the Kingdoms^{2, 3}. In Arabidopsis thaliana, a morning loop is formed of CIRCADIAN CLOCK ASSOCIATED1 (CCA1)⁴ and LATE ELONGATED HYPOCOTYL (LHY)⁵, PSEUDO RESPONSE REGULATOR 7 (PRR7) and PRR9⁶. The evening feedback loop involves TIMING OF CAB2 EXPRESSION1 (TOC1)⁷ and GIGANTEA (GI)⁸. These loops are connected through CCA1/LHY mediated repression of TOC1 expression⁹ and TOC1-mediated repression of CCA1 involving CCA1 HIKING EXPEDITION (CHE)¹⁰.

Circadian oscillators also incorporate post-transcriptional, post-translational and metabolic processes^11–16. In Arabidopsis, these include the F-box protein ZEITLUPE (ZTL)¹⁷ regulated blue-light dependent degradation of TOC1^{15, 18, 19} and photosynthetic sugars affect entrainment¹¹. Several studies have revealed the importance of ionic signalling for circadian timekeeping^20–23. In Drosophila, circadian oscillations of intracellular Ca²⁺ ([Ca²⁺]) regulate cellular oscillations in vivo²³ and temperature-induced increases in cytosolic [Ca²⁺] are involved in entrainment²¹.

In plants, like in the mammalian suprachiasmatic nucleus, there are circadian oscillations of [Ca²⁺]_cyt ²⁴. They are driven by cyclic adenosine diphosphate ribose-mediated Ca²⁺-release from internal stores^{16, 25–28} to encode information about photoperiod, timing and light intensity^{29, 30}. However, the functions regulated by circadian oscillations of [Ca²⁺]_cyt have not been identified. Therefore, it has been conjecture whether circadian oscillations of [Ca²⁺]_cyt represent an input to the oscillator or part of the timing mechanism, in addition to being an output²⁴.

We show that circadian oscillations of [Ca²⁺]_cyt affect the abundance of CHE and affect circadian period through a Ca²⁺-dependent regulatory protein of the plant specific CALMODULIN-LIKE (CML) family. We conclude that CML24 is part of the Arabidopsis circadian system, acting through a Ca²⁺-dependent pathway to regulate TOC1.

Results

Circadian oscillator gene expression can be altered by [Ca²⁺]_cyt signals

We identified potential targets for [Ca²⁺]_cyt, by examining by microarray the response of circadian oscillator transcripts to a single 24 h artificial oscillation of [Ca²⁺]_cyt in plants in which circadian oscillations of [Ca²⁺]_cyt were abolished, and later artificially induced (Fig. 1a and Supplementary Fig. 1). [Ca²⁺] signals did not restore high amplitude oscillations of clock transcripts like those in light and dark cycles³¹. CCA1 HIKING EXPEDITION (CHE) was the only clock transcript whose abundance correlated with the [Ca²⁺]_cyt signal, having a dynamic opposite to the imposed [Ca²⁺]_cyt oscillation (maximum repression 5.2 fold at 12 h, 4.5 fold at 16 h and 3.1 fold at 8 h) (Fig.1b). The CCA1 dynamic was modestly altered (1.9 fold activation at 20 h and 24 h) (Fig. 1b). This later increase in CCA1 transcript abundance might have been due to the earlier large repression of CHE. Artificial [Ca²⁺] oscillation had smaller effects on other circadian oscillator transcript abundance, the largest being a 2 fold reduction of PRR3 at 16 h and a 1.9 fold increase of LHY at 20 h (Fig. 1b).

Fig. 1. Transcripts Abundance of Circadian Clock Genes is Modulated by [Ca²⁺]_cyt.

a, Imposing oscillations of external CaCl₂ restores circadian oscillations in [Ca²⁺]_cyt in unentrained seedlings. See also Supplementary Fig. 1c for calibrated data. Results represent the mean ± S.D. (n=12 biological replicates) for one experiment. Experiments were repeated three times.

b, Effect of the imposition of a 24 h oscillation of [Ca²⁺]_cyt on the transcript abundance (expressed as log2) of the circadian clock genes. Closed circles indicate unentrained water-treated samples, green circles the unentrained CaCl₂-treated plants. To generate the oscillation, CaCl₂ was applied as described by the shaded areas and in Supplementary Table 1. Results represent the mean (n=2 biological replicates).

c, Plants treated at ZT36 and ZT48 with a solution containing 660 μM W7 and 50 mM CaCl₂ for 2 h, were assayed for changes in the abundance of circadian clock transcripts by qPCR. Dots represent each measurement and the black bars the mean ± S.D. (n= 3 biological replicates). See also Supplementary Fig. 1. Single, double or triple asterisk indicate significance of ≤ 0.05, ≤ 0.01 and ≤ 0.001, respectively after two-tailed Student´s t test analysis or two-sided Mann-Whitney Rank Sum Test (ZT36 CCA1, PRR7 and ELF4; ZT48 ELF4 and LHY).

To test further the effects of [Ca²⁺]_cyt on CHE transcript abundance, we screened a number of [Ca²⁺]_cyt agonists to identify those that had profound and persistent effects on [Ca²⁺]_cyt. N-(6-Aminohexyl)-5-chloro-1-naphthalenesulfonamide hydrochloride (W7) in combination with CaCl₂ evoked sustained increases in [Ca²⁺]_cyt³². This Ca²⁺ agonist caused a transient and large rise of [Ca²⁺]_cyt peaking at 848.5 ± 156.0 nM, followed by a sustained elevation for at least 2 h (Supplementary Fig. 1). Induction of large, sustained [Ca²⁺]_cyt increases at ZT36 and ZT48 by 2 h treatment with W7 and CaCl₂, confirmed that the major transcriptional response of the circadian oscillator was a repression of CHE abundance (Fig. 1c). Similar to the experiment with a single artificial 24 h [Ca²⁺]_cyt oscillation, the abundance of morning transcripts CCA1 and LHY did not alter immediately after the [Ca²⁺]_cyt increase (Fig. 1c). This increase in [Ca²⁺]_cyt significantly altered the abundance of transcripts expressed later in the day, being an activator of PRR9 and a repressor of PRR7, PRR3 and CHE (fold changes: PRR9 3.2 (ZT48), PRR7 2.0 (ZT48), PRR3 1.7 (ZT36) and 2.1 (ZT48), CHE 3.4 (ZT36)) (Fig. 1c). Finally, the transcript levels of the evening genes were not affected (Fig. 1c). These Ca²⁺-induced changes in CHE transcript did not alter the free running period of the oscillator, because treating plants with the W7 solution at ZT0 and ZT12 did not alter the period of CCA1:LUC (ZT0 Control 24.4 ± 0.5 h, W7 solution 24.2 ± 0.2 h; ZT12 Control 24.4 ± 0.2 h, W7 solution 24.4 ± 0.5 h; p>0.05). Thus, manipulation of [Ca²⁺]_cyt demonstrated that both 24 h [Ca²⁺]_cyt oscillations and shorter-term [Ca²⁺]_cyt signals can regulate CHE transcript abundance (Fig. 1).

A reverse genetic screen identified Calmodulin-Like 24 (CML24) as a regulator of circadian period

Because transient increases in [Ca²⁺]_cyt affected circadian oscillator gene abundance but not period, we wished to determine whether Ca²⁺ signalling has a role in regulating the clock in planta. We performed a screen of 75 well characterised Ca²⁺ signalling mutants of transporters, transducers and sensors. Five lines had significantly increased circadian period of leaf movement compared to wildtype (Fig. 2 and Supplementary Table 2); vitamin C (vtc) 1-1³³ (Col-0 23.8 ± 0.1 h, vtc1-1 24.7 ± 0.2 h, p=0.001), calmodulin-like (cml) 24-1 and the double mutant cml23-2 cml24-1(Col-0 24.1 ± 0.1 h, cml24-1 24.6 ± 0.1 h, cml23-2 cml24-1 25.0 ± 0.1 h, p<0.001 for both) and also cml24-4 and the double mutant cml23-2 cml24-4^{34, 35} (Col-0 23.9 ± 0.1 h, cml24-4 25.1 ± 0.1 h, cml23-2 cml24-4 25.6 ± 0.1 h, p<0.001 for both). cml23-2 only had a phenotype when in combination with alleles of CML24 (Fig. 2), suggesting that CML24 and CML23 could be redundant in the regulation of the clock. None of the overexpressing (CML24-OX1 and CML24-OX2) or underexpressing lines (CML24-U1 and CML24-U1) affected circadian period (Supplementary Table 2). As CML24 (previously called TCH2)^{36, 37} encodes a CALMODULIN-LIKE Ca²⁺-sensor^{34, 35} and two different alleles, cml24-1 and cml24-4 had a significantly longer period than Col-0 for leaf movements (Fig. 2), we decided to further characterize whether CML24 is involved in the regulation of the circadian oscillator. The long circadian period of cml23-2 cml24-4 double mutants was confirmed by measuring promoter activity of CCA1 fused to luciferase (CCA1:LUC) and circadian oscillations of [Ca²⁺]_cyt (Fig. 3a-3d) (Col-0 23.7 ± 0.1 h, cml23-2 cml24-4 26.1 ± 0.4 h for CCA1:LUC, p<0.001; Col-0 23.7 ± 0.2 h, cml23-2 cml24-4 25.1 ± 0.1 h for 35S:AEQ, p<0.001). To investigate the effect of CML24 on the central oscillator in more detail, we analyzed CCA1, PRR7 and TOC1 transcript abundance in Col-0 and cml23-2 cml24-4. In the third day in LL, there was a substantial delay of 4 h in the phase of CCA1, TOC1 and PRR7 transcript abundance in the mutant plants (Fig. 3e), consistent with the lengthening of circadian period by ∼1.5 h described in Fig. 3a-d. The transcript levels of the clock components were unaffected.

Fig. 2. CML24 Regulates Circadian Period in Arabidopsis.

Average normalized traces of leaf positions (left panels). FFT-NLLS analysis of the circadian period for leaf movement experiments: dots indicate individual samples and black bars mean period ± S.E.M (right panels). a, Col-0 n=70, cml23-2 n=70, cml24-1 n=97, double mutant n=94; b, Col-0 n=63, cml23-2 n=69, cml24-4 n=49, double mutant n=110, rhythmic leaves. a shows the results of cml23-2 and cml24-1 single and double mutants, b shows the results of cml23-2 and cml24-4 single and double mutants. Red lines indicate Col-0, grey lines indicate cml23cml24 double mutants, and light blue cml24 single mutants. cml23-2 traces for leaf position were removed for clarity. All plants were grown under 12 h L: 12 h D cycles before the experiments. Data represent three (a) or two (b) independent experiments. Single or triple asterisk indicate significance of ≤ 0.05 and ≤ 0.001, respectively, after two-tailed Student's t test or two-sided Mann-Whitney Rank Sum test (b, Col-0 vs. double mutant).

Fig. 3. CML24 has profound effect on the regulation of the Arabidopsis circadian clock.

The cml23-2 cml24-4 mutant has a long circadian period of 35S:AEQ and CCA1:LUC luminescence. Mean normalized luminescence ± S.D. of 35S:AEQ (a) and CCA1:LUC (b) for wildtype (red circles) and cml23-2 cml24-4 (grey circles) from two independent experiments (a, n=8 biological replicates; b, Col-0 n=23 and cml23-2 cml24-4 n=24, biological replicates). c and d show the FFT-NLLS analysis of the samples used in a and b, respectively. Triple asterisk indicate significance of ≤ 0.001, after two-tailed Student´s t test analysis (c) and two-sided Mann-Whitney Rank Sum Test (d). e, CCA1, TOC1 and PRR7 transcripts abundance were analysed at the time point indicated in plants grown for 12 days in 12h:12h light:dark cycles and transferred to continuous light at ZT0. qPCR results represent the mean ± S.D. (n=3 biological replicates). Single, double or triple asterisk indicate significance of ≤ 0.05, ≤ 0.01 and ≤ 0.001, respectively after two-tailed Student´s t test analysis or Mann-Whitney Rank Sum Test (CCA1 ZT60 and PRR7 ZT48).

CML24 regulates circadian period in a [Ca²⁺]_cyt-dependent manner

Because circadian rhythms persist in conditions where the circadian oscillations of [Ca²⁺]_cyt are abolished, such as nicotinamide¹⁶, sucrose²⁴ or monochromatic red light²⁹, circadian rhythms of [Ca²⁺]_cyt are not necessary for a rhythmic oscillator. However, we could test for necessity for oscillations in [Ca²⁺]_cyt in the correct regulation of circadian period by determining whether the action of CML24 in the circadian clock depends upon Ca²⁺. If the effects of the CMLs are independent of Ca²⁺, then the effects of mutation on circadian period should be additive to treatments that abolish circadian [Ca²⁺]_cyt oscillations. Whereas, if the action of the CMLs is dependent on Ca²⁺, then mutation should have no further effect in the presence of Ca²⁺ antagonists. 50 mM nicotinamide increased the circadian period of leaf movement¹⁶ but in its presence, cml23-2 cml24-4 was indistinguishable from the wild type (Col-0 27.7 ± 0.2 h, cml23-2 cml24-4 28.1 ± 0.2 h, p=0.490) (Fig. 4a). Sucrose abolishes [Ca²⁺]_cyt oscillations²⁴, however under high and low light conditions it has different effects on circadian period¹¹. There was no significant difference between cml23-2 cml24-4 and wild type on 3% sucrose (high light Col-0 24.7 ± 0.1 h, cml23-2 cml24-4 25.0 ± 0.1 h, p = 0.051; low light Col-0 26.4 ± 0.2 h, cml23-2 cml24-4 26.9 ± 0.2 h, p=0.061) (Figs. 4b and 4c). Lastly, in monochromatic red light, the circadian period was not statistically different in the cml23-2 cml24-4 mutant compared to Col-0 (Col-0 26.3 ± 0.4 h, cml23-2 cml24-4 26.1 ± 0.2 h p=0.410) (Fig. 4d). The lack of an effect of cml23-2 cml24-4 in conditions that abolish circadian oscillations of [Ca²⁺]_cyt is consistent for loss-of-function mutations in putative Ca²⁺-sensor proteins and CML24 acting downstream of [Ca²⁺]_cyt. Additionally, we found that CML23 and CML24 transcript abundance (Fig. 4e) was increased in response to the artificially imposed [Ca²⁺]_cyt oscillation (Fig. 1a), with their peaks in phase with the imposed [Ca²⁺]_cyt rhythm, suggesting [Ca²⁺]_cyt positively regulates CML abundance. This could be explained by the presence in their promoters of the Ca²⁺-Responsive cis element CAM box (ACGCGT)³².

Fig. 4. Circadian oscillations of [Ca²⁺]_cyt are necessary for the correct function of the Circadian Oscillator.

a, Circadian period estimates of leaf movement in continuous high light (80 μmol m^-2 s^-1) of Col-0 and cml23-2 cml24-4 plants treated with either 50 mM nicotinamide (Col-0 n=15, cml23-2 cml24-4 n=33, biological replicates) or water (Col-0 n=16, cml23-2 cml24-4 n=29, biological replicates). b, Circadian period estimates of CCA1:LUC rhythms in continuous high light (80 μmol m^-2 s^-1) or c, continuous low light (10 μmol m^-2 s^-1) of Col-0 and cml23-2 cml24-4 plants grown in the presence of either water (high light Col-0 n=7, cml23-2 cml24-4 n=8, low light n=16, biological replicates), 90 mM sucrose (high light n=8, low light n=16, biological replicates) or 90 mM mannitol (Col-0 n=8, cml23-2 cml24-4 n=7, biological replicates). d, Circadian period estimates of CCA1:LUC rhythms in Col-0 and cml23-2 cml24-4 plants under continuous high mixed red (660 nm) and blue (470 nm) light (80 μmol m^-2 s^-1) (n=4, biological replicates) and continuous monochromatic blue or red light (40 μmol m^-2 s^-1) (blue n=7, red Col-0 n=11 and cml23-2 cml24-4 n=10, biological replicates). e, Effect of the imposition of a ramp of external CaCl₂ (Fig. 1a) on the expression (log2) of CML24 and CML23. CaCl₂ was applied as described by the shaded areas and in Supplementary Fig. 1c. Plant material was harvested from the onset of treatment every 4 h for 24 h to extract RNA for probing with microarray. Results represent the mean (n=2 biological replicates). f, Circadian period estimates of CCA1:LUC rhythms in continuous high light (80 μmol m^-2 s^-1) of Col-0 (n=7 biological replicates) and cml23-2 cml24-4 (n=8 biological replicates) plants treated from the day before going into continuous high light either with water or 200 μM cPTIO. Period estimates were obtained by BRASS and are shown as mean ± S.E.M. Data were obtained from 1 independent experiment. Experiments were repeated at least twice. Single, double or triple asterisk indicate significance of ≤ 0.05, ≤ 0.01 and ≤ 0.001, respectively after two-tailed Student´s t test analysis or two-sided Mann-Whitney Rank Sum test (a (water), b (mannitol) and f).

CML24 not only senses Ca²⁺, it also regulates NO³⁵ [35] as shown by high levels of NO in cml24 mutants³⁵. We therefore, tested whether the high NO in the mutants could be the cause of the long period by two experiments. We investigated the effect of NO on circadian regulation of [Ca²⁺]_cyt and CHLOROPHYLL A/B BINDING PROTEIN2:LUC (CAB2:LUC) in wild type plants and found no evidence for a role for NO, because the NO donor, SNAP, and the scavenger, cPTIO, were without effect on circadian rhythms (Supplementary Fig. 2). The high NO levels found in cml23-2 cml24-4, were not responsible for the long period phenotype, because the mutant long period phenotype persisted in the presence of cPTIO (Fig. 4f)³⁸.

We conclude that the effect of the cml23-2 cml24-4 mutations on the circadian clock requires [Ca²⁺]_cyt and is independent of the effects of CML24 on NO generation.

CML24 regulates circadian period by a pathway involving TOC1 and possibly CHE

To investigate how CML24 affects circadian period, we tested whether it is involved in the [Ca²⁺]_cyt-mediated transcriptional regulation of circadian clock genes. The clock transcripts regulated by [Ca²⁺]_cyt in Col-0 (Fig. 1c) were also regulated by [Ca²⁺]_cyt in cml23-2 cml24-4 (Supplementary Fig. 3), suggesting that CML24 is not involved in the transcriptional regulation by [Ca²⁺]_cyt.

We then investigated the genetic linkage between CML24 and components of the oscillator. We studied epistatic interactions in the control of circadian rhythms of leaf movement between mutations in CML23/24 and CCA1, LHY, TOC1, ZTL, ELF3, ELF4 and LUX. Double CML23/24 mutants were used for epistasis due to having a larger measurable effect on period compared to the single CML24 mutants. We identified epistasis between mutations in CML23/CML24 and TOC1 in the regulation of the period of leaf movement (Fig. 5a, Supplementary Fig. 4 and Supplementary Table 3). toc1-2 single mutant had a short period (C24 24.3 ± 0.1 h, toc1-2 21.4 ± 0.3 h, Mann-Whitney Rank Sum Test p<0.001; Fig. 5a)²⁹. In the triple mutant toc1-2 cml23-2 cml24-4, the long period arising from mutations in CML23/CML24 was absent, having a period that was indistinguishable from the single mutant toc1-2, and significantly shorter than cml23-2 cml24-4 mutant (One-way ANOVA p=1 and p<0.001, respectively) (Fig. 5a, Supplementary Fig. 4 and Supplementary Table 3). This indicates that toc1-2 is epistatic to cml23-2 cml24-4. We did not find epistatic interactions between mutations in CML23/CML24 and CCA1, LHY or ZTL. cml23-2 cml24-4 increased period in the cca1-11, lhy-21 and ztl-3 backgrounds, resulting in additive phenotypes (Fig. 5b-5d and Supplementary Table 3). Analysis of genetic interactions between CML23/CML24 and ELF3, ELF4 and LUX is complicated by the arrhythmic phenotypes caused by loss-of-function of these evening complex genes. Triple mutants of cml23-2 cml24-4 and members of the evening complex where therefore all arrhythmic in LL (Supplementary Fig. 4). Crossing the wild type genetic backgrounds used in this study was without effect (Supplementary Table 3).

Fig. 5. Epistatic Analysis of Leaf Movements Rhythms Shows that TOC1 is Functionally Linked to CML24 to Regulate Circadian Period.

Average normalized traces of leaf positions and FFT-NLLS analysis of the circadian period for leaf movement experiments. a shows the results of cml23-2 cml24-4 with toc1-2 (Col-0 n=33, C24=31, double mutant=22, clock gene mutant=24, triple mutant=25), b with cca1-11 (Col-0 n=7, Ws-2=6, double mutant=29, clock gene mutant=21, triple mutant=19), c with lhy-21 (Col-0 n=24, Ws-2=24, double mutant=22, clock gene mutant=24, triple mutant=14) and d with ztl-3 (Col-0 n=22, double mutant=25, clock gene mutant=26, triple mutant=15). Grey lines indicate cml23-2 cml24-4, blue clock gene single mutants and black the triple mutants, respectively. Wild-type traces for leaf position were removed for clarity. All plants were grown under 12 h L: 12 h D cycles before the experiments. Data are presented from one experiment representative of two (cca1-11, lhy-21, ztl-3) or three (toc1-2) independent experiments. See also Supplementary Fig. 4 and Supplementary Table 3. Single or triple asterisk indicate significance of ≤ 0.05 and ≤ 0.001, respectively, after Kruskal-Wallis One Way Analysis of Variance on Ranks followed by Dunn´s method was used to compare the triple mutant to the single and cml23-2 cml24-4 double mutant.

TOC1-mediated repression of CCA1 involves CHE ¹⁰. Therefore, the regulation of CHE by [Ca²⁺]_cyt and interaction between mutations in CML23/CML24 and TOC1, prompted investigation of whether CHE is also part of the genetic pathway by which CML24 regulates circadian period. We identified epistatic interaction between mutations in CML23/CML24 and CHE in the regulation of circadian leaf movements. As previously reported for CCA1:LUC⁺ rhythms, the circadian period of leaf movement in che-1 and che-2 single mutants was indistinguishable from the background (two-tailed Student´s t-test p=0.461 and p=0.681, respectively; Fig. 6a and 6b)¹⁰. In the triple mutant che-2 cml23-2 cml24-4, the long period arising from mutations in CML23/CML24 was absent, being indistinguishable from the single mutant che-2 and significantly shorter than the cml23-2 cml24-4 mutant (One-way ANOVA p>0.05 and p<0.05, respectively) (Fig. 6a, Supplementary Fig. 4 and Supplementary Table 3). However, we found no evidence that che-1 was epistatic to cml23-2 cml24-4. che-1 cml23-2 cml24-4 triple mutants had a significantly longer period relative to che-1 and similar to cml23-2 cml24-4 (One-way ANOVA, p<0.05 and p>0.05, respectively) (Fig. 6b, Supplementary Fig.4 and Supplementary Table 3). This is consistent with che-2 being a stronger allele than che-1 in the lhy background¹⁰, explaining why there is an epistatic interaction with cml23-2 cml24-4 and che-2 but not with che-1.

Fig. 6. Epistatic Analyses of Leaf Movements Rhythms and Flowering Time Shows that CHE is Functionally Linked to CML24.

Average normalized traces of leaf positions and FFT-NLLS analysis of the circadian period for leaf movement experiments. a shows the results of cml23-2 cml24-4 with che-2 (Col-0 n=29, cml23-2 cml24-4=48, che-2=47, triple mutant=40) and b with che-1 (Col-0 n=29, cml23-2 cml24-4=48, che-1 =39, triple mutant=46). Wild-type traces for leaf position were removed for clarity. Data are presented from one experiment representative of two (che-1) or three (che-2) independent experiments (Supplementary Fig. 4 and Supplementary Table 3). Flowering time responses under long (16 h:8 h) or short days conditions (8 h:16 h) for che-2 cml23-2 cml24-4 (c) and che-1 cml23-2 cml24-4 (d) mutants. Number of leaves were recorded when the emerging bolt was 5 mm high. Dots represent the individual plants and the black bars the mean ± S.D. (n=16; in LD Col-0 n=15; in SD cml23-2 cml24-4 n=15 and che-2 triple mutant n=13). Single or triple asterisk indicate significance of ≤ 0.05 or ≤ 0.001, respectively after Kruskal-Wallis One-Way ANOVA analysis followed by Tukey test (LD) or Dunn’s method (a, b and SD), when the triple mutant was compared to the che and cml23-2 cml24-4 mutants. Flowering rate was calculated using the number of days since germination when the number of leaves was recorded. che-2 cml23-2 cml24-4 mutant used was HL18 and che-1 cml23-2 cml24-4 was HL15. An independent experiment in LD was done using three different mutants lines (Supplementary Fig. 5).

Because the circadian oscillator contributes to the photoperiodic regulation of flowering and cml23-2 cml24-4 mutants are late-flowering³⁵, we tested epistasis between CML23/CML24 and CHE by measurement of flowering time. che-2 is an early flowering mutant and, as suggested above, possibly the stronger allele (Supplementary Fig. 5). As in the circadian experiments, there is an epistatic relationship between mutations in CML23/CML24 and CHE in the regulation of flowering time (Fig. 6c and Supplementary Fig. 5) when che-2 was used. che-2 cml23-2 cml24-4 mutant flowered at the same time as che-2, and significantly earlier than cml23-2 cml24-4 (One-way ANOVA, p>0.05 and p<0.05, respectively) (Fig. 6c and Supplementary Fig. 5). However, and similarly to leaf movement, flowering time provided no evidence of epistasis between che-1 and cml23-2 cml24-4 (Fig. 6d and Supplementary Fig. 5).

The data suggest that the [Ca²⁺]_cyt-dependent regulation of circadian period by CML24 is not directly mediated by CCA1, LHY and ZTL and that CML24 might regulate TOC1 and CHE, because functional copies of these two clock genes are required to express the cml23-2 cml24-4 phenotype.

Discussion

The Ca²⁺-sensor CML24 is a regulator of Arabidopsis circadian period

We tested the hypothesis that circadian oscillations of [Ca²⁺]_cyt can feed back into the circadian oscillator. We demonstrate that [Ca²⁺]_cyt signals can regulate the expression of the Ca²⁺-binding CALMODULIN-LIKE24 (CML24) and that CML24 also regulates circadian period, with the loss-of-function phenotype being absent when [Ca²⁺]_cyt rhythms are abolished. We conclude that correct circadian period is dependent on CML24 and circadian rhythms of [Ca²⁺]_cyt. Epistatic analysis suggests that TOC1 and likely CHE genetically interact with CML24 (Supplementary Fig. 6). Additionally, we show that [Ca²⁺]_cyt signals can regulate the expression of CHE in a CML24-independent manner and transcriptional regulation of CHE is unlikely to regulate circadian period.

It was previously reported that CML24 regulates flowering time³⁵. Our new data demonstrates that CML24 also regulates circadian period in Arabidopsis because two different alleles (cml24-1 and cml24-4) alone or in combination with the null allele of CML23 (cml23-2)³⁵, had a long circadian period (Fig. 2). We found that CML24 has robust and profound effects on period (Fig. 3). The period lengthening persisted in different clock mutant backgrounds such as cca1-11, lhy-21 and ztl-3 (Fig. 5). The magnitude of the period lengthening of the CML24 mutants (from 0.6 to 2 h) is larger or similar to previously reported mutations in important circadian genes: prr7-11 and prr9-1 (0-2 h)³⁹, che-1 and che-2 (no effect on period)¹⁰, prr3-1 and prr5-3 (∼1 h)⁴⁰, lnk1-1 (no effect on period), lnk2-2 (1 h), lnk1-1 lnk2-2 (2 h)⁴¹, che-1/lhy and che-2/lhy double mutants have a significantly shorter circadian period (~ 0.5 or 1 h, respectively) compared to the lhy mutant¹⁰. CML24-OX1 and CML24-OX2 were without phenotype, which might be expected for a sensor protein whose activity depends on and might be limited by Ca²⁺ concentration, rather than abundance of the sensor protein. Meaning that the presence of 24 h [Ca²⁺]_cyt oscillations might be critical for the production of the physiological response as observed in Fig. 4a-d and that in the over-expressor lines, the Ca²⁺ signature might be still decoded. Nevertheless, the limitation of other protein targets of CML24 and activators cannot be ruled out.

CML24 binds Ca²⁺ at EF hands to cause a conformational change but has no other identified functional domains³⁵. Our data are consistent with CML23 and CML24 acting as Ca²⁺-sensors because we demonstrate that the effect of CML23 and CML24 mutations on circadian period depends on [Ca²⁺]_cyt as shown by the absence of an effect of the cml23-2 cml24-4 mutation when circadian oscillations of [Ca²⁺]_cyt are suppressed^{16, 24, 29} (Fig. 4a-4d). This also demonstrates that sucrose regulates circadian period through a pathway independent of CML23 and CML24 because in low light, added sugar shortened circadian period¹¹ which was unaffected by cm23-2 cml24-4 (Fig. 4c). In the presence of nicotinamide, the circadian period in wild type and cml23-2 cml24-4 was the same. However, the period of the double mutant was increased by nicotinamide (Fig. 4a), suggesting that nicotinamide might target both [Ca²⁺]_cyt-dependent and -independent pathways, or additional Ca²⁺-sensors might be involved⁴². The Arabidopsis genome encodes over 50 CaM and CMLs^42,43 and other Ca²⁺ sensors, which could also contribute to circadian regulation.

CML24 genetically interacts with TOC1 and possibly with CHE to regulate circadian period

The absence of the cml23-2 cml24-4 circadian phenotype in toc1-2 cml23-2 cml24-4 indicates that the CMLs proteins are unable to exert their regulator function if TOC1 is absent (Fig. 5a and Supplementary Fig. 4). CML24 and TOC1 are expressed in diverse tissues and organs⁴⁴ which is consistent with our genetic studies, but more studies are necessary to conclude how cytosolic CML24 regulates TOC1 function. CHE might have a role because there was a genetic interaction between CML23/CML24 and CHE in the regulation of circadian period when the che-2 allele was used (Fig. 6).

Because we found that Ca²⁺ alone or in combination with W7 was able to suppress CHE transcript abundance in Col-0 plants (Fig. 1) and in the cml double mutant (Supplementary Fig. 3), we suggest that the genetic interaction between CML23/CML24 and CHE is not dependent on transcriptional regulation and that the effect of W7 is not through an effect on CML23/24. Additionally, circadian period was not affected by a transient increase in [Ca²⁺]_cyt following W7 treatment. This is not surprising, because che mutants, in which CHE transcript abundance is constitutively reduced, period is unaffected¹⁰.

Whilst we do not consider that the Ca²⁺-induced transcriptional changes in CHE affect circadian period, it might be of functional significance because it is consistent with the CHE binding site, also known as Site IIb⁴⁵, being similar to the [Ca²⁺]_cyt-regulated Site II promoter element (AGGCCCAT)³². [Ca²⁺]_cyt-regulation of Site II is most likely through the TCP family of transcription factors³², of which CHE is a member. CHE binds to the class I TCP-binding site (TBS) (GGTCCCAC) in the CCA1 promoter and represses its expression¹⁰. In addition to CHE, CCA1 transcript was the only clock gene that was modestly activated around 8 h after the last pulse of CaCl₂ and CHE repression (Fig. 1b). CHE oscillates 9 h out of phase with CCA1 transcript¹⁰ and at a similar phase with [Ca²⁺]_cyt oscillations. Whilst we conclude that transcriptional changes in CHE are unlikely to mediate changes in circadian period, it raises the possibility that CHE transmits information about Ca²⁺ signals to the CCA1 promoter.

Our data identify roles for Ca²⁺ and CML24 in the circadian clock. These findings unveil for first time in plants a function for circadian oscillations of [Ca²⁺]_cyt and expand the architecture of the plant circadian oscillator.

Methods

Plant material and growth

Arabidopsis mutant lines used in the reverse genetic screen were supplied generously from laboratories working in the area of Ca²⁺ signalling in plants and are listed in Supplementary Table 2. The T-DNA insertion mutants che-1 and che-2¹⁰ and the point mutation single mutants toc1-2⁹ and lux-4⁴⁶ were provided by Steve Kay (The Scripps Research Institute, USA); the T-DNA insertion mutants cca1-11, lhy-21, elf3-4 and elf4-1⁴⁸ were donated by Seth Davis (University of York, UK); the T-DNA single mutant ztl-3 (SALK_035701)¹⁷ was obtained from Nottingham Arabidopsis Seed Centre (NASC), UK. To obtain the triple mutants of the circadian oscillator genes with the cml23-2 cml24-4 double mutant (Col-0), the single mutants che-1 (Col-0), che-2 (Col-0), cca1-11 (WS), lhy-21 (WS), toc1-2 (C24), ztl-3 (Col-0), elf3-4 (WS), elf4-1 (WS) and lux-4 (C24), were crossed independently to cml23-2 cml24-4 (Col-0) double mutants to generate triple mutants. The F2 progeny was then self-fertilized to obtain an F3 generation. The F3 and F4 generations were then genotyped to ensure all the mutant alleles were homozygous. F4 or subsequent generations were used for the epistatic study. Similarly, to the circadian clock mutants, different Arabidopsis ecotypes (WS and C24) were also crossed to Col-0.

Growth of Arabidopsis thaliana, photon-counting imaging of aequorin and luciferase luminescence and transformation techniques were as described in ²⁹ unless otherwise stated.

[Ca²⁺]_cyt manipulation

To obtain plants with undetectable circadian [Ca²⁺]_cyt rhythms (unentrained), meaning that [Ca²⁺]_cyt remained at basal levels, 35S:AEQ WS seeds²⁹ were grown in opaque 7 mm x 9 mm plastic rings sealed at the base with 0.5 µm nylon mesh (Normesh, UK) on sucrose-free 0.5 MS agar seeds, germinated without stratification and grown in continuous white light (LL) for at least 12 days. Artificial [Ca²⁺]_cyt rhythms were induced in these plants by step-wise addition of external CaCl₂ during the subjective day, followed by removal (Supplementary Table 1). During treatment, seedlings were floating on the mesh rafts on temperature-adjusted solutions. All experiments were repeated at least twice.

Induction of a single 24 h [Ca²⁺]_cyt peak was carried essentially the same except using a FLUOstar plate reader (BMG Labtech, Germany). 200 µl of treatment solution to provide final concentrations of 0 mM to 150 mM CaCl₂ (Supplementary Table 1) was injected into wells of a 96 well plate containing individual 12 day old 35S:AEQ transformed seedlings that had previously been reconstituted with 20 μM coelenterazine (20 °C, overnight). Seedlings were washed with temperature-adjusted deionized water before measurement and luminesce was measured⁴⁸. Solutions were replaced with the successive treatment every hour. Results were assessed from three independent experiments each consisting of a minimum of 12 replicates per treatment.

For [Ca²⁺]_cyt measurements after treatment with N-(6-Aminohexyl)-5-chloro-1-naphthalenesulfonamide hydrochloride (W7) (Calbiochem) see Supplementary Methods.

Microarray analysis

Plants treated with CaCl₂ as described above to generate 24 h [Ca²⁺]_cyt oscillations were harvested every 4 h for 24 h. RNA was isolated as described in ¹⁶. RNA was hybridised to GeneChips Arabidopsis ATH1 Genome Array using NASC’s International Affymetrix service. Raw intensity data were normalized across chips using RMA (http://www.bioconductor.org) automatically log transforming the expression data. Array 1-22_Calcium_12h_Rep2_ATH1 was removed from the analysis after failing hybridization quality control.

qPCR analysis

Experiment to determine the effect of W7: Col-0 and cml23-2 cml24-4 plants were treated with W7 solution (660 μM W7 and 50 mM CaCl_2, containing a final concentration of 2.5 % (v/v) DMSO) or deionized water at ZT36 and ZT48 in constant light for 2 h and then frozen in liquid N₂.

Experiment to determine the effect of cml23-2 cml24-4 mutation on the circadian clock transcript abundance: Col-0 and cml23-2 cml24-4 mutant were grown at 20 ºC under 12 light/12 dark and then transferred into constant light conditions. After 2 days under constant light, plants were collected every 2 h from ZT48 to ZT72.

Total RNA was extracted of three biological replicates of at least four pooled plants each, using the RNeasy Plant Mini Kit (QIAGEN) and RNase-Free DNase on-column treatment (QIAGEN). cDNA was synthesized from 500 ng RNA with the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific) using oligo(dT) primers. The gene-specific products were amplified using the Rotor-Gene SYBR Green PCR Kit on a Rotor-Gene 6000 Real-Time PCR machine (QIAGEN). Primers used are detailed in Supplementary Methods. Relative transcript levels were determined by incorporating PCR efficiencies⁴⁹.

Leaf movement analysis

Analysis of circadian rhythms of the first true leaves was performed as described in ¹⁶ without experimenter knowledge of seed lines. In those experiments where nicotinamide was applied, 50 mM Nicotinamide (Sigma) or deionized water was applied once a day for 2 days before the start of imaging; 50 µl of solution was applied to the aerial parts of each seedling. The experiments were repeated twice but for the triple mutants, cml23-2 cml24-4 che-2 and cml23-2 cml24-4 toc1-2 were repeated three times. For cml23-2 cml24-4 toc1-2 two independent lines were used.

Photoperiodic flowering time screening

Plants were sown directly onto soil and grown in 20 °C and 100 µmol m^-2 s^-1 in either long day (LD) conditions (16 h L:8 h D) or short day (SD) conditions (8 h L: 16 h D). Flowering time was defined as when the emerging bolt reached a height of approximately 5 mm. Screening experiments typically had 6 to 8 plants of each line per growth condition, and confirmation experiments had 15 to16 plants of each line per growth condition.

Statistical Analysis

All statistical tests, n number, the measure of the centre and the error bars are described in figure legends when appropriate. Other statistical parameters are listed in Supplementary Statistical Parameters section. For comparison between two groups, two-tailed Student´s t-test or two-sided Mann-Whitney Rank Sum Test were used. Both test analyses were considered significant if p<0.05, p<0.01 and p<0.001. For comparison between more than two groups, one-way ANOVA followed by Holm-Sidak method or Kruskal-Wallis one-way ANOVA followed by Dunn´s method or Tukey test were used. ANOVA tests were performed with an alpha level of 0.05 or 0.001.