Multiple feedback loops of the Arabidopsis circadian clock provide rhythmic robustness across environmental conditions

Modeling the Effects of Altered Levels of Activators and Repressors on Plant Clock Function.

The opposite circadian phenotypes seen in Arabidopsis plants mutant for LHY-type or RVE8-type Myb-like transcription factors (14⇓⇓⇓–18) suggest that they have antagonistic functions in the clock network. Previous in silico analysis of mammalian clock components has suggested that the coordinate action of repressors and activators helps control the circadian phase and amplitude of common targets genes (19). We hypothesized that LHY-type and RVE8-type factors, acting as repressors and activators, respectively, of shared targets, might play a similar role in the plant circadian network. Since nonclock-regulated activators of gene expression are thought to play an important role in the plant clock network (20), we modified the mathematical expression developed by Ueda et al. (19) to include a noncycling activator of gene expression in addition to the cycling activator and repressor. In our model, we combined CCA1 and LHY activity as one cycling repressor component (termed “LHY”) and members of the RVE8 subclade as one cycling activator component (termed “RVE8”) (Fig. 1A), and we assumed equal rhythmic amplitude and regulatory potency for both. Times of peak regulatory activities were assigned based on the reported peak phases of LHY and RVE8 protein accumulation (17, 21, 22) (Fig. 1B). This simple model predicts that genes coordinately regulated by LHY- and RVE8-type factors would have peak expression near dusk (Fig. 1C), consistent with the phases of the transcripts of most genes identified as RVE8 or CCA1 targets in genome-wide studies (18, 23, 24).

We next simulated an RVE8 loss-of-function mutant by reducing the strength of the cycling activator in the model, which caused a later peak and lower trough for the hypothetical target gene (Fig. 1D), consistent with published expression patterns of RVE8 targets, such as TOC1, in the rve468 mutant (18). Simulation of an LHY reduction-of-function mutant resulted in an earlier peak phase and higher trough level (Fig. 1E), recapitulating experimental results observed in cca1 lhy double mutants (15). Finally, simulation of a cca1 lhy rve468 mutant by reducing the strength of both the cycling activator and the repressor resulted in output gene expression with a near–wild-type peak phase but reduced amplitude (Fig. 1F). This simulation suggests that near-normal circadian phase can be achieved in Myb mutants as long as the activity of the positive and negative factors are balanced but that lower overall expression levels of the Mybs might compromise rhythmic amplitude.

Removal of both Repressors and Activators Rescues Growth and Flowering Phenotypes.

To test the hypothesis that simultaneous reduction of function of LHY-clade and RVE8-clade genes might rescue phenotypes seen in plants mutant for one or the other type of factor, we compared the phenotypes of cca1-1 lhy-20 (25) (hereafter cca1lhy), rve4-1 rve6-1 rve8-1 (18) (hereafter rve468), and cca1-1 lhy-20 rve4-1 rve6-1 rve8-1 (hereafter quint). qRT-PCR analysis of the quintuple mutant revealed that no CCA1 or RVE4 transcripts could be detected but that RVE8 is expressed at low levels and RVE6 is expressed at about one-half the level seen in the wild type (SI Appendix, Fig. S1). LHY is also detectably expressed in both the cca1lhy and quint mutants, with peak levels ∼30% of the wild type (SI Appendix, Fig. S1). Therefore, these mutants represent loss of function but not complete null alleles for LHY-type and RVE8-type genes.

Previous studies reported numerous phenotypes for cca1lhy mutants, including early flowering, and short hypocotyls (15, 26). In contrast, rve468 triple-mutant plants display delayed flowering, elongated hypocotyls, and larger rosettes (due to both longer petioles and larger leaf blades) than the wild type (27). We found that these growth and flowering phenotypes are largely alleviated in the quint mutants. While the petioles of cca1lhy plants are shorter and those of rve468 are longer than the wild type, the petioles of quint mutant plants are not significantly different from the wild type in long days (LDs; 16 h light, 8 h dark) and are closer to the wild type than either parental mutant in short days (8 h light, 16 h dark) (Fig. 2A and SI Appendix, Fig. S2A). Similarly, while cca1lhy has shorter (26) and rve468 has longer hypocotyls (27), hypocotyl length in the quint mutant is not distinguishable from the wild type (Fig. 2B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Simultaneous reduction of activator and repressor functions rescues developmental and circadian period and phase phenotypes. (A) Petiole lengths of plants grown in short days (8 h light:16 h dark; n = 12–14). (B) Hypocotyl lengths of seedlings grown in continuous monochromatic red light (25 μmol m⁻² s⁻¹; n = 24–30). (C) Number of leaves produced before bolting in plants grown in LDs (16 h light:8 h dark; n = 26–27). (D) Free-running period of CCR2::LUC+ activity (plants maintained in constant 35 μmol m⁻² s⁻¹ red + 35 μmol m⁻² s⁻¹ blue light; n = 24–36). Groups that do not share a letter have significantly different mean values (P < 0.01, one-way ANOVA with Tukey’s post hoc test). (E–H) Normalized CCR2::LUC+ activity of plants maintained in (E–G) LDs or (H) continuous light. Col, the wild-type genetic background for all mutants (Col-0). Error bars ± SEM; note that error bars are smaller than symbols in E–H.

We next examined the timing of the transition from vegetative to reproductive growth. As previously reported, we found a slight delay in rve468 (27) and advance in cca1lhy plants (15) grown in LD, whether measured as number of leaves or as number of days to flowering (Fig. 2C and SI Appendix, Fig. S2B). As seen for vegetative growth phenotypes, the flowering time phenotype of quint mutants is more similar to the wild type than to either mutant parent. Thus, simultaneous reduction of both LHY-clade and RVE8-clade gene function rescues multiple clock-associated developmental phenotypes.

Removal of both Repressors and Activators Rescues Circadian Phase and Period but Not Amplitude.

We next examined activity of the circadian-regulated CCR2::LUC+ reporter in these lines. As expected given previously reported period phenotypes of rve468 and cca1lhy mutants (15, 18, 28), these lines exhibit late- and early-phase phenotypes, respectively, both in LD and short day (Fig. 2 E and F and SI Appendix, Fig. S2 C, D, F, and G). In contrast, the phase of the quintuple mutant is indistinguishable from the wild type in LD and closer to it than either parental line in short day (Fig. 2G and SI Appendix, Fig. S2 E–G). This rescue of diel phase in the quint mutant is consistent with the predictions of our in silico analysis (Fig. 1F) and supports the hypothesis that LHY- and RVE8-like proteins antagonistically regulate common clock genes. Furthermore, it suggests that the period phenotype of the quint mutant is closer to the wild type than either parental line. Indeed, we found that the free-running period of luciferase activity is intermediate between the short period of cca1lhy mutants and the long period of rve468 mutants (Fig. 2 D and H). As also predicted by our in silico analysis, rhythmic amplitude of CCR2::LUC+ activity in plants deficient for both the activator and repressor Mybs is reduced relative to the wild type (Fig. 2H and SI Appendix, Fig. S2H).

We next used qRT-PCR to examine patterns of core clock gene expression in constant light and standard growth conditions. As expected based on our luciferase data, we found that all transcripts show an earlier peak in the cca1lhy mutant and a later peak in the rve468 mutant but that peak phase in the quint is close to the wild type (Fig. 3). This same pattern was seen both for EE-containing genes thought to be directly regulated by CCA1 and RVE8 [TOC1, LUX, ELF4, GI, PRR5, and PRR9 (12)] and for genes that do not have EE in their promoters (ELF3 and PRR7) that may be indirectly regulated by these antagonistic Mybs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Loss of both repressors and activators rescues circadian-phase phenotypes but reduces cyclic amplitude of most clock genes. Expression of indicated genes was determined relative to the reference gene PP2A by qRT-PCR. Plants were harvested at the indicated times after release into continuous 17.5 μmol m⁻² s⁻¹ red + 17.5 μmol m⁻² s⁻¹ blue light at 22 °C. Col, the wild-type genetic background for all mutants (Col-0). Error bars ± SEM for two biological replicates.

Although near–wild-type phases of clock gene expression are observed in the quint line, rhythmic amplitude is reduced for most transcripts (Fig. 3). While peak transcript levels of a few genes (ELF3 and LUX) are not reduced relative to the wild type, peak levels of most transcripts, especially PRR5, PRR7, and PRR9, are lower. In addition, trough levels of most transcripts, most notably ELF3, are elevated in the quint compared with the wild type. Together, these changes in waveform result in statistically significantly lower rhythmic amplitude for almost all of these clock-regulated genes in the quint mutant compared with the wild type (SI Appendix, Fig. S3). This reduction in amplitude in the quint is similar to that observed for CCR2::LUC+ luciferase activity (Fig. 2H and SI Appendix, Fig. S2H) and is consistent with the predictions of our in silico model (Fig. 1F).

The Low-Amplitude quint Oscillator Is Sensitized to Genetic and Environmental Perturbation.

We next examined the robustness of circadian rhythms in the quint background to various perturbations. Overexpressing RVE8 at high levels using the strong 35S promoter has previously been reported to cause a short-period phenotype (17). We examined the effects of driving more moderate overexpression of RVE8 using the UBQ10 promoter (SI Appendix, Fig. S4 A and B). In wild-type T1 plants, the largest effect of expression of this transgene was a 2-h decrease in free-running period (Fig. 4A). In contrast, circadian period in the quint T1 plants was 4–6 h shorter than nontransformed controls, and rhythmic amplitude was reduced relative to the wild-type T1 plants (SI Appendix, Fig. S4C). This difference in phenotype between the wild type and the quint could not be attributed to different levels of RVE8 expression (Fig. 4A and SI Appendix, Fig. S4 A and B). Thus, the circadian clock of quint plants is more susceptible to genetic perturbation than that of wild-type plants.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Loss of both repressors and activators reduces clock robustness. (A) RVE8 protein levels (normalized to actin) in T1 UBQ10p::RVE8-Ypet-3xFLAG plants (Col and quint genetic backgrounds) are plotted vs. the difference in period between each T1 plant and its background genotype. (B) Period estimates of CCR2::LUC+ activity in plants grown at 22 °C in light/dark cycles and then released into constant 35 μmol m⁻² s⁻¹ red + 35 μmol m⁻² s⁻¹ blue light at the indicated temperatures. Mean periods are only plotted for samples with RAE < 0.6 (n = 29–36, 11–18, and 5–54 for 12 °C, 22 °C, and 30 °C, respectively, except that, at 12 °C, n = 1 for 35S:RVE8). Error bars ± SEM. (C) Rhythmic amplitude (measured as RAE) at different temperatures. Plants were grown as in B. Col, the wild-type genetic background for all mutants (Col-0).

One characteristic of circadian clocks in diurnal organisms, such as plants, is that at high fluence rates the clock runs slightly faster than at low fluence rates (29). To test how the quintuple mutant responds to changes in the light environment, we examined period and rhythmicity across different fluence rates of light (SI Appendix, Fig. S5A). Period in the quint mutant is similar to the wild type at all fluence rates tested; however, the slope of the response is opposite in the two genotypes. We next examined rhythmic robustness as measured as relative amplitude error (RAE). The RAE of a perfect sine wave is defined as 0, while an RAE = 1 indicates no significant rhythm (the error in the amplitude is equal to the amplitude value itself). The RAEs of both cca1lhy and quint plants are higher than in the wild type at all light intensities, indicating lower rhythmic robustness in these mutants (SI Appendix, Fig. S5 B–I).

Another defining characteristic of circadian clocks is their ability to maintain a relatively unchanged free-running period across a range of temperatures, a phenomenon termed temperature compensation (30, 31). To test the effect of altered ambient temperatures on clock function, we entrained the control and mutant lines in light/dark cycles at 22 °C and then monitored CCR2::LUC+ activity after transfer to constant light at 12 °C, 22 °C, or 30 °C, all physiologically relevant temperatures for Arabidopsis. As expected, free-running period in the wild type is largely unchanged over this temperature range (Fig. 4B). Period in the rve468 mutants is undercompensated, with increased temperature resulting in a shortening of free-running period.Conversely, period in plants overexpressing RVE8 is overcompensated, with free-running period lengthening with increases in temperature. Although cca1lhy mutants are arrhythmic at 12 °C, as previously reported (7), their longer period at 30 °C than at 22 °C also shows overcompensation. The quint mutant has a slightly shorter period at 30 °C than at lower temperatures and thus, is slightly undercompensated relative to the wild type, although less so than rve468 plants. Together, these data indicate that an appropriate balance of expression between the activator- and repressor-type Mybs is required for temperature compensation of free-running period.

In addition to period stability at different temperatures, another important trait is rhythmic robustness across different environmental conditions. Wild-type plants have robust rhythms at all temperatures tested: 12 °C, 22 °C, and 30 °C (Fig. 4C and SI Appendix, Fig. S6). In contrast, at 12 °C, CCR2::LUC+ activity in both cca1lhy and RVE8 overexpressing plants is almost completely arrhythmic: one-half of the plants did not return any period estimate, and the ones that did had very high RAE values (SI Appendix, Fig. S6D). The quint mutant is less rhythmic than the wild type at 12 °C but much more so than cca1lhy, with only 3% of quint plants failing to return a period estimate. Of those quint that did return a period estimate, the median RAE was intermediate between that of the wild type and cca1lhy. The partial rescue of rhythmicity in quint relative to cca1lhy mutants and the arrhythmicity of 35S::RVE8 plants indicate that a balance between LHY-type and RVE8-type Mybs is essential for rhythmicity at low temperatures. Similarly, while luciferase rhythms in the wild type are robust at 30 °C, cca1lhy and quint mutants both have very poor rhythms with high and indistinguishable RAE values. In contrast, 35S::RVE8 plants have RAE values close to the wild type (Fig. 4C and SI Appendix, Fig. S6). These data suggest that LHY clade function is indispensable for rhythmicity at high temperatures but that balanced expression between activator- and repressor-type Mybs is not an important factor. Together, these data show that the Mybs play a key role in the maintenance of cyclic robustness at nonoptimal temperature conditions but are not essential under favorable conditions.

Repressor- and Activator-Type Mybs Are Deeply Conserved.

The finding that expression of both activator- and repressor-type Mybs is important for both rhythmic amplitude at favorable temperatures and rhythmic robustness under nonoptimal conditions prompted us to investigate whether these distinct factors are of ancient origin or a recent evolutionary innovation. A phylogenetic tree of homologous proteins encoded by angiosperm, nonvascular land plants, and green algae genomes revealed clades containing LHY-like and RVE8-like proteins in these highly diverse species (SI Appendix, Fig. S7). The presence of both types of proteins even in unicellular green algae indicates their ancient origin and suggests that these antagonistic transcription factors may have been part of the earliest plant clocks.