A circadian rhythm is an endogenously generated rhythm with a period of about 24 h, approximating the period of the rotation of the earth on its axis. These rhythms provide temporal organization of biological processes from cyanobacteria to mammals [3]. In plants, circadian rhythmicity is widespread and pervasive [4, 5]. Approximately one-third of the Arabidopsis transcriptome shows circadian oscillations in abundance in continuous conditions [6], but if one looks under a variety of light and temperature cycles that proportion grows to an astonishing ~90% [7], underlining the probable importance of circadian rhythm to overall fitness [4, 5].

Circadian clocks of taxonomic groups as diverse as plants, fungi and animals are composed of multiple interlocked feedback loops with positive and negative components [3] and many of the components of these clocks are encoded by members of gene families. The Arabidopsis circadian clock, an example of this common design principle, is composed of at least four interlocked feedback loops (Figure 1). In the central loop (blue in Figure 1), TIMING OF CAB EXPRESSION 1 (TOC1), the founding member of a family of five PSEUDO-RESPONSE REGULATOR (PRR) genes, is a positive regulator of CIRCADIAN AND CLOCK ASSOCIATED 1 (CCA1) and LATE ELONGATED HYPOCOTYL (LHY). CCA1 and LHY are members of a small family of REVEILLE genes that encode single Myb domain transcription factors. Others members of this family have been shown to play roles in clock function as well as in regulation of clock output pathways [4, 5]. To complete the first loop, CCA1 and LHY bind to the TOC1 promoter to inhibit its expression. In a second loop (green in Figure 1) within the central loop, CCA1 and LHY also repress expression of CCA1 HIKING EXPEDITION (CHE), which encodes a TCP transcription factor that binds to and represses expression from the CCA1 promoter [1]. In the third loop (yellow in Figure 1), termed the 'morning' loop based on the time of peak mRNA accumulation of its constituents, CCA1 and LHY are positive regulators of two TOC1 relatives, PRR7 and PRR9, that are negative regulators of CCA1 and LHY [4, 5]. In a fourth loop (gray in Figure 1), termed the 'evening' loop based again on the time of peak mRNA accumulation, TOC1 represses a component, 'Y', that includes GIGANTEA (GI) and possibly PRR5. This component in turn positively regulates TOC1 expression at least in part through modulation by GI of proteasomal degradation of TOC1 mediated by the F-box protein ZEITLUPE (ZTL) [4, 5]. In addition, proper regulation of CCA1 and LHY requires other clock genes, including EARLY FLOWERING 4 (ELF4), which encodes a protein of unknown function, and LUX ARRHYTHMO/PHYTOCLOCK1 (LUX/PCL), which encodes a Myb domain transcription factor; these and other clock components have yet to be fully incorporated into current clock models [4, 5]. The number of interlocked feedback loops will undoubtedly increase as the regulatory relationships among clock components are more fully described.

The value of model organisms such as Arabidopsis stems from the generalization of knowledge acquired in the model to all flowering plants, and especially to those of agricultural significance. The increasing availability of genomic sequences from multiple plants is now permitting our first insights into this issue.

Molecular phylogenetic analysis of the PRR genes indicates that the common ancestor of the monocots and eudicots had three PRR gene clades [2]. Since the divergence of the monocots and eudicots, the clades corresponding to PRR3/PRR7 and PRR5/PRR9 have expanded independently in both lineages as a result of genome duplications [2]. Within the eudicots, or 'true dicots', a subset of the former broad classification of dicots that includes more than half of extant plant species, two further genome duplications occurred in Arabidopsis following its divergence from papaya (Carica papaya) but, after each duplication, one of the paralogs was lost. In contrast, poplar has retained the duplicate copies of PRR5, PRR7 and PRR9, which originated in a genome duplication, termed the Salicoid duplication, that occurred in the poplar lineage after its separation from the papaya-Arabidopsis lineage. PRR3 has been completely lost from the poplar genome, although it is unclear whether this loss predated or followed the Salicoid duplication. The Brassica rapa genome has triploidized since its divergence form Arabidopsis approximately 14.5 million years ago, yet for no members of the B. rapa TOC1/PRR gene family have all three copies persisted, making it clear that differential PRR gene loss has occurred [8].

Takata et al. [9] have conducted a parallel analysis of angiosperm CCA1/LHY genes, and their observations are consistent with those obtained in their analysis of the PRR genes; the common ancestor of monocots and eudicots had one CCA1/LHY gene and there has been independent duplication of the LHY/CCA1 genes in the monocots and eudicots. Within the eudicots, there has been independent duplication in poplar and Arabidopsis.

The key conclusion from these studies is that the common ancestor of the monocots and eudicots had the basic components necessary for the construction of a circadian clock with multiple interlocked feedback loops prior to the separation of these groups 200 million years ago [2]. This makes it very likely that the Arabidopsis clock will prove a useful model for most agricultural species. It will be interesting to determine whether the more basal angiosperms, such as the Magnoliales, also share this common clock architecture.

One consequence of gene duplication is that it allows the two copies to subdivide the functions of the ancestral copy (functional specialization or sub-functionalization), or for one copy to acquire a new function (neo-functionalization) while the other retains the original function, thus preserving fitness; but is there evidence for either functional specialization or acquisition of novel functions among PRR genes during evolution of the angiosperms? The strongest evidence comes from Arabidopsis, where clock function is best studied. TOC1 and four other PRR genes each show circadian oscillations in transcript abundance, with peak abundance occurring at intervals spanning the day starting at dawn with PRR9, followed by PRR7, PRR5, PRR3, and finally at dusk with TOC1 (PRR1) [4, 5]. As shown in Figure 2, TOC1 is recruited to the CCA1 promoter and is a positive regulator of CCA1 expression, although the molecular details remain incompletely described [1]. PRR9, PRR7, and PRR5 are recruited to the promoters of CCA1 and LHY and negatively regulate their expression [10]. It is likely that the sequential expression of PRR9, PRR7, and PRR5 contributes to sustained repression of CCA1 and LHY expression throughout the day. This indicates that, while the function of these three genes is partially redundant, with normal expression of the three genes the temporal window of CCA1/LHY repression is extended. Thus PRR9, PRR7, and PRR5 offer an example of sub-functionalization in the temporal domain. Although the function of the rice (Oryza sativa) orthologs of PRR9, PRR7, and PRR5 has not been established, there is a similar sequential pattern of expression of OsPRR73/OsPRR37 and then OsPRR95/OsPRR59, followed by OsTOC1 (OsPRR1) [11].

In Arabidopsis, the PRR3 gene offers an example of acquisition of a novel function. PRR9, PRR7, and PRR5 all have a similar role in negatively regulating CCA1 and LHY, suggesting that this represents the ancestral function (Figure 2). PRR3 appears, instead, to have acquired a novel and specialized function in the vascular tissue, where PRR3 binds to TOC1 and, in doing so, blocks the interaction of TOC1 with ZTL, the F-box protein that targets TOC1 for proteasomal degradation [12]. Thus, PRR3 exhibits a restricted domain of expression and has acquired a novel function, the regulation of TOC1 stability through protein-protein interaction (Figure 2). In Arabidopsis, loss of PRR3 function confers only a very small shortening of circadian period [13], which is consistent with the apparent loss of PRR3 in poplar, without concomitant perturbation of clock function.

There are additional suggestions of evolving function in the PRR7 lineage. In Arabidopsis, PRR7 contributes to the determination of flowering time, although the effects are not large and PRR7 is not a major determinant of flowering time among natural populations [14]. In contrast, in the monocots barley and wheat, PRR7 (Ppd-H1 and Ppd-D1, respectively) is one of the major determinants of photoperiod sensitivity and flowering time [15, 16]. Whether this represents a true acquisition of novel function in the monocots or a loss of function in the eudicots remains uncertain and will require more detailed dissection of the roles of PRR7 in the flowering pathways of monocots and eudicots.

There remains a great deal of work to achieve a mechanistic understanding of how the circadian clock keeps time. Four of the five PRR proteins are recruited to DNA yet they do not possess recognized DNA-binding domains and are not known to bind DNA directly. How are they recruited to the CCA1 and LHY promoters and what makes TOC1 a positive regulator while PRR5, PRR7, and PRR9 are repressors? Takata et al. [2, 9] establish that the common ancestor of monocots and eudicots had PRR and CCA1/LHY genes and, therefore, the materials with which to construct a functional circadian clock. How has the differential amplification of these two gene families in the angiosperm lineages allowed modulation of circadian timekeeping? How well does the outline presented in Figures 1 and 2 apply across the angiosperms and to more primitive plants? Within species, has variation among clock genes contributed to fitness? There is no shortage of questions and the increasing availability of genome sequences and tools to probe gene function in many species make this a wonderful time to study the basis of circadian timing.