The Circadian Clock. A Plant's Best Friend in a Spinning World¹

THE RHYTHM SECTION. WHICH PLANT PROCESSES ARE CLOCK-REGULATED?

Microarray experiments indicate that at least 6% of Arabidopsis genes are rhythmically expressed, with expression peaks at all phases throughout the day and night (Harmer et al., 2000; Schaffer et al., 2001). This circadian gene expression produces the rhythms that pervade plant physiology, some of which are obvious (such as the “sleep movements” of legume leaves, noted since classical times), others less so. In several cases, genes that affect a common pathway or process are expressed at the same phase, suggesting that the phase might be important in the function of that process. Many genes encoding enzymes of phenylpropanoid biosynthesis have peak RNA levels before dawn, perhaps because it is advantageous to accumulate photoprotective flavonoids before the sun rises (Harmer et al., 2000). A large proportion (68%) of the rhythmically regulated genes also directly respond to environmental stress (Kreps et al., 2002), so rhythmic expression of these genes in anticipation of predictable environmental changes might prepare the plant to withstand a stress (or make best use of a resource). Thus, circadian regulation would complement the plant's subsequent response to the stress. Recent experimental evidence shows a fitness detriment in some Arabidopsis clock mutants (Green et al., 2002), and more is likely to follow from natural variants and further physiological studies. Photoperiodism is a special case in which a circadian rhythm is combined with light signaling. The photoperiod sensor allows plants to respond to the annual cycle of day length by making flowers, tubers, or frost-tolerant buds at appropriate seasons. The selective advantages of correct seasonality are very clear; recent reports have significantly enhanced our understanding of this mechanism (see below).

The circadian clock is itself a target of regulation by environmental response pathways, so it stands at the interface between external and endogenous regulators. Light-signaling pathways from both phytochromes (phys) and cryptochromes (crys) regulate clock components to achieve entrainment, for example (for review, see Fankhauser and Staiger, 2002). There are also reports of hormonal effects upon circadian timing from several plant species. Perhaps surprisingly, circadian timing is not coordinated among cells, so it is possible to set circadian rhythms of gene expression to several different phases in different parts of a single plant or even of a single leaf (Thain et al., 2000). Moreover, the circadian system appears to differ slightly among cell types. The evidence for this is that circadian rhythms in different cell types can have different circadian periods in wild-type plants tested under identical conditions. A recent example was provided by the expression rhythm of a LUC reporter that is expressed in the leaf epidermis compared with one that is expressed in the mesophyll (Hall et al., 2002, and refs. therein). A possible advantage is that the rhythms controlled by one clock can alter their phase relative to the rhythms controlled by another clock. The biochemical difference in the clocks between cell types is unknown. It might be relatively minor because all the circadian rhythms tested in Arabidopsis depend upon a common set of genes. These “clock genes” or “clock-associated genes” function within, or close to, each cell's circadian clock: Their products produce and maintain the oscillation that drives all other circadian rhythms.

PORTRAIT OF THE ARABIDOPSIS CIRCADIAN CLOCK

The known clock mechanisms in all organisms include a gene circuit with negative feedback, involving 24-h rhythms in the levels of positively and negatively acting transcriptional regulators, in all organisms (Harmer et al., 2001; Young and Kay, 2001). These rhythmic feedback loops probably have arisen by convergent evolution because the sequences of the proteins involved share little or no homology across taxa, despite the overt similarities among the circadian rhythms that they control. The genes that we now review represent the plant kingdom's solution to the problems posed by the rhythmic environment (Staiger, 2002; Hayama and Coupland, 2003); we restrict our discussion to work on Arabidopsis, the species in which these genes were first identified (Fig. 1). The first consistent model of the Arabidopsis circadian oscillator (Alabadi et al., 2001) suggested that it is comprised of three main players, the genes encoding Myb-related transcription factors CCA1 (CIRCADIAN CLOCK ASSOCIATED 1) and LHY (LATE ELONGATED HYPOCOTYL) and a pseudoresponse regulator TOC1 (TIMING OF CAB EXPRESSION 1; also known as Arabidopsis PRR1 [PSEUDORESPONSE REGULATOR 1]). Strong loss-of-function alleles of TOC1, a double mutant of LHY and CCA1, or constitutive overexpression of any of these genes cause arrhythmia under prolonged constant light and darkness (Schaffer et al., 1998; Wang and Tobin, 1998; Alabadi et al., 2002; Makino et al., 2002; Mizoguchi et al., 2002; Mas et al., 2003). It is too early to speculate on the origin of the rhythms that remain in these plants under light/dark cycles and in some cases persist transiently in constant conditions.

There is good evidence that CCA1/LHY proteins act redundantly in the late night and early day, binding to an evening element (EE; AAATATCT) in the promoter of TOC1 and repressing its expression (Harmer et al., 2000; Alabadi et al., 2001, 2002; Mizoguchi et al., 2002). When CCA1/LHY levels fall late in the day, the TOC1 protein is proposed to activate the transcription of CCA1/LHY, thus forming the outline of a transcriptional feedback loop (Alabadi et al., 2001). The accumulation pattern of TOC1 protein in plants has not been reported yet. This activation is probably indirect because it requires at least three other genes that are co-expressed with TOC1 in the evening: ELF3 (EARLY FLOWERING 3; Schaffer et al., 1998), GI (GIGANTEA; Fowler et al., 1999), and ELF4 (EARLY-FLOWERING 4; see below, Doyle et al., 2002). The biochemical functions of the cognate proteins are unclear. Paradoxically, CCA1 expression is somewhat reduced rather than activated by overexpression of TOC1 (Makino et al., 2002; Mas et al., 2003). Thus, the current portrait of the Arabidopsis clock amounts to a silhouette; we discern the familiar negative feedback loop but, as in other species, evolution has endowed the Arabidopsis clock with more than the minimum mathematical requirement for an oscillation.

LIGHT INPUT PATHWAYS IN CHIAROSCURO

The light input to the oscillator is provided by the photoreceptor families, crys and phys, which mediate signaling by high-intensity blue light and lowintensity blue light or red light, respectively. This area of the plant circadian system has been reviewed recently (Devlin, 2002; Fankhauser and Staiger, 2002). One aspect of light input deserves further mention, namely the gating mechanism or “zeitnehmer” (time bringer), by which the clock rhythmically regulates its input signals. In the elf3 mutant, unregulated light input stops the clock 8 to 10 h after dawn; therefore, timing breaks down in long days (McWatters et al., 2000; Covington et al., 2001). ELF3 is, therefore, an essential part of the plant circadian system, though it may not be central to rhythm generation. ELF3 protein can bind to PHYB; thus, it may inhibit light signaling in the evening (Covington et al., 2001). This zeitnehmer mechanism is analogous to the rhythm of light responsiveness that creates the photoperiod sensor (see below).

In contrast to the photoreceptors, their signaling pathway(s) to the clock components are not yet well defined. In etiolated seedlings, PIF3 (PHYTOCHROME INTERACTING FACTOR 3) binding to a G-box sequence (CACGTG) in the CCA1 promoter might mediate CCA1 activation by light, when activated phyA and/or phyB bind to PIF3 at the promoter (for review, see Quail, 2002). In green tissues, Carré and coworkers recently showed that light regulates LHY protein levels rather than RNA abundance, which can account for driven rhythms of LHY overexpressors in light/dark cycles (Kim et al., 2003). However, the weak rhythmicity that remains in lhy; cca1 double mutant plants is still entrained by light/dark cycles, so it is unclear whether these genes are essential for entrainment of the clock (Alabadi et al., 2002; Mizoguchi et al., 2002).

Other potential mechanisms of light input have been suggested. Previous studies of the toc1-1 allele revealed no light-dependent effects of TOC1; however, toc1-1 is now classed as a weak allele because it maintains circadian rhythms under all lighting conditions, albeit with a 21-h period. The more recent work used RNAi transgenes and a stronger allele, toc1-2, that severely reduce TOC1 RNA levels and result in arrhythmia specifically under red light (Mas et al., 2003). The strong alleles also reduce the responsiveness of hypocotyl elongation and CCA1 gene expression to red light, suggesting that TOC1 might be involved in phy signaling to targets other than the clock. The only hint of a biochemical function for TOC1 protein is its potential to bind to PIF3 and a related protein, PIL1 (PIF3-LIKE 1; Makino et al., 2002), but this at least provides a potential link to photoreceptor signaling. Under blue or white light, the severe toc1 mutants remain rhythmic with a short period, so signaling from phy requires more TOC1 than cry signaling, though both are affected in the mutants (Mas et al., 2003).

An intriguing gene family is made up of ZTL (ZEITLUPE), FKF (FLAVIN-BINDING KELCH REPEAT F-BOX), and LKP2 (LOV DOMAIN KELCH PROTEIN 2). Mutations or misexpression of each gene can affect circadian rhythms, but most interestingly, the ztl mutant's effect on circadian period varies depending on the fluence rate of ambient light (Devlin, 2002). Their protein sequences share a PER-ARNTSIM domain, multiple kelch domains, and an F box. Similar PER-ARNT-SIM domains of other proteins bind a flavin chromophore, as in the plant phototropin photoreceptors (Briggs and Christie, 2002), so a similar cofactor might confer light dependence on ZTL function. Kelch domains are typically involved in protein interactions, and other F-box proteins recruit target proteins to E3 ubiquitination complexes, marking the target proteins for degradation. This suggests a function in the light-dependent ubiquitination of a clock component(s), which might provide light input independently of LHY/CCA1 (for review, see Fankhauser and Staiger, 2002).

NEWCOMERS AMONG THE CLOCK-ASSOCIATED GENES

System identification (finding all the relevant components) is an important step in understanding biological regulation. However, finding the primary function of a gene starting from a mutant phenotype is never trivial. The intimate connections between light signaling and circadian regulation often result in overlapping phenotypes: The srr1 mutant (sensitivity to red light reduced) is a recent example that shows defects both in phyB signaling and in circadian rhythms (Staiger et al., 2003). The challenge is exacerbated by gene duplications: Additional members of the CCA1/LHY protein family have been identified, for example (for review, see Carre and Kim, 2002), so we eagerly await their functional characterization.

ELF4 IS NECESSARY FOR CIRCADIAN RHYTHMS AND PHOTOPERIODISM

The elf4 mutant was identified by its early flowering in short photoperiods (Doyle et al., 2002). It carries a T-DNA insertion in the ELF4 gene, which encodes a predicted protein of 111 amino acids without identifiable protein signatures. Its small size alone might lead to speculation that it functions as a posttranslational protein modifier or a secreted signal. The ELF4 transcript shows robust, circadian expression in wild-type plants, with a peak in the evening. The elf4 mutation affects the rhythms of the circadian-regulated genes CAB (CHLOROPHYLL A/B BINDING PROTEIN, also known as LHCB) and CCR2 (COLD-CIRCADIAN RHYTHM-RNA BINDING 2). Their expression in a population of seedlings (studied using LUC [luciferase] reporter gene fusions) rapidly became arrhythmic under constant conditions, but individual seedlings were transiently rhythmic with widely varying periods. Circadian leaf movements were similarly affected, suggesting a role for ELF4 in the accuracy and persistence of circadian rhythms. Overall, elf4's rhythmic defects resembled those of lhy;cca1 double mutant plants; consistent with this interpretation, the elf4 mutation drastically reduces expression of CCA1 (Doyle et al., 2002). A striking feature of elf4 plants is their very early flowering in short photoperiods, whereas under long-day conditions, they flowered at about the same time as wild type. A high expression of CO (CONSTANS) in the mutant during short days is the likely cause of the early flowering seen under these conditions. ELF4 appears to be closely linked with the circadian oscillator; whether its circadian defect is the sole cause of its early flowering (see below) remains to be investigated.

TEJ IMPLICATES PROTEIN POLY(ADPRIBOSYL)ATION IN MAINTAINING PERIOD LENGTH

The tej mutant (Panda et al., 2002) was identified in the screen for mutants with altered rhythms of CAB: LUC expression, which also identified the first toc1 and ztl mutants. tej mutant plants showed a light-independent period lengthening of 2 h in CAB, CCA1, and CCR2 expression; leaf movement rhythms were similarly affected. Flowering time was slightly earlier than wild type under both long and short days. The recessive mutation results from an amino acid substitution in the TEJ protein, which functions as a poly (ADP-Rib) glycohydrolase. Poly(ADP-ribosyl)ation is a posttranslational protein modification, which is conferred by poly (ADP-Rib) polymerase and removed by poly (ADP-Rib) glycohydrolase; in other species, it has been implicated in DNA repair, DNA damage signaling, and the regulation of transcription and proteasome function (refs. in Panda et al., 2002). Applying an inhibitor of poly (ADP-Rib) polymerase, 3-aminobenzamide, rescued the phenotype of tej. This strongly suggests that the mutant phenotypes were due to excessive poly(ADP-ribosyl)ation of a clock-related protein, the identity of which is unknown (Panda et al., 2002).

THE APRR/TOC1 QUINTET. IS THEIR DAILY ROUND IMPORTANT FOR CIRCADIAN TIMING?

TOC1 shares sequence homology with a set of pseudoresponse regulator genes, APRR9, 7, 5, and 3, which have been termed the “APRR quintet” (Matsushika et al., 2000; Strayer et al., 2000; Suzuki et al., 2001). APRR9, APRR7, APRR5, and APRR3 and TOC1 (APRR1) are expressed sequentially every 2 to 3 h, starting soon after dawn with the expression of APRR9, until the evening when TOC1 is expressed. APRR9 expression is also phy activated (Matsushika et al., 2000). The interactions among these proteins are unknown, but recent studies of transgenic plants overexpressing APRR9, APRR5, and TOC1 show that their expression is interrelated. TOC1 overexpression, for example, abolishes APRR9 expression under constant light and damps rhythmic APRR7-APRR3 expression to low levels (Makino et al., 2002). Plants that overexpress APRR9 or APRR5 affect flowering time and show a red light-dependent short-hypocotyl phenotype (also found for TOC1 overexpressors). APRR9 overexpression confers an early phase and/or a short period on many rhythmic genes under constant white light (Matsushika et al., 2002). In APRR5 overexpressors, APRR9 and APRR7 gene expression was reduced toward the trough level, whereas expression of APRR3 and TOC1 was increased toward the peak level (Sato et al., 2002). This is consistent with a cascade mechanism, in which regulation proceeds along the quintet from APRR9 to TOC1. Studies of null mutants in these genes are now required to understand their function in the circadian system. The quintet might provide a flexible output mechanism that can regulate a gene at any desired phase from dawn to dusk. They might thus participate in TOC1 activation toward the end of the day, counteracting its repression by CCA1/LHY.

PHOTOPERIODIC REGULATION OF FLOWERING TIME

FUTURE PROSPECTS

The identification of the molecular components of the plant circadian system and its associated photo-period sensor will accelerate with the wider application of genome-wide data collection, quantitative biochemical studies of individual components, and mathematical modeling. Methods of regulating gene expression in specific cell types will be applied to manipulate the clock, matching the reporter gene methods for monitoring rhythms in specific tissues. Some classic questions will be re-visited using these new tools: Circadian timing can differ among cells, for example, so which leaf cells are responsible for the CO expression rhythm (Salisbury and Denney, 1971)? We look forward to understanding how the Arabidopsis model is altered in plant species with other adaptations of timing. To name but a few, the phase jump from diurnal to nocturnal gas exchange, when Mesembryanthemum crystallinum switches from C3 to CAM metabolism; the photoperiodic responses of some crop species, which contribute to determining harvest times; and the short-day response of many trees, which must trigger the formation of dormant buds before temperatures drop below freezing in the winter.