During development, the animal body plan is built through the coordinated timing of cell divisions, cell differentiation, and cell movements. Considerable progress has been made in identifying the endogenous mechanisms that orchestrate these basic processes. However, less well studied is the potential influence of timing cues from the environment. Thus is early embryonic development completely “uncoupled” from the environment or are certain environmental timing cues fed into the developmental program?
The day–night cycle represents one of the most important, regular environmental changes that animals are exposed to. Most animals live in an environment where sunlight intensity, temperature, and consequently other factors such as food availability and risk from predation vary considerably with a regular 24-hr cycle. Thus the ability to modify physiology and behavior appropriately at certain times during the day–night cycle is essential for survival. A key adaptation that allows animals to anticipate environmental changes has been the development of a self-sustaining 24-hr timing mechanism or “clock” (Pittendrigh, 1993).
These 24-hr biological clocks characteristically continue to function even under artificial conditions where animals are deprived of environmental changes. In these conditions, the period length of the clock-generated rhythms is not precisely 24 hr. For this reason, they are termed “circadian” (“circa”, around; “diem”, one day) clocks. Under normal day–night conditions, the phase of the clock is reset on a daily basis by environmental signals such as changes in light and temperature (so called “zeitgebers,” time givers). This property is essential to ensure that the clock does not progressively drift out of phase (“free-run”) with respect to the environment (Roenneberg et al., 2003).
At its simplest level, the clock can be considered as a three part system: a pacemaker that generates the circadian rhythm; an input pathway whereby zeitgebers are perceived and adjust the phase of the pacemaker; and finally an output pathway through which the pacemaker regulates a wide diversity of physiological processes and behavior (Menaker et al., 1978). The discovery of circadian clock mutants has led to the identification of core clock genes in a wide range of model organisms, and we now have detailed models to explain how they interact to generate clock function. Among the most important conclusions that have emerged from molecular analysis is that the circadian clock represents a cell autonomous mechanism shared by most cell types (Schibler and Sassone-Corsi, 2002). Thus in contrast with earlier beliefs, the clock is not a property restricted to neuronal networks, but ticks in every cell and might, therefore, operate well before the establishment of a functional nervous system.
The aim of this review is to address two basic issues: (1) When do the circadian clock and clock input pathways first emerge during development? And (2) What are the regulatory targets of the clock at this stage? Specifically, do circadian clocks regulate early developmental processes? Might there be an advantage to restrict certain developmental processes to certain times of day? The concept of clocks controlling early developmental processes is certainly not new. A molecular clock that operates in presomitic mesoderm controls somite segmentation in vertebrates (Pourquie, 2003). Most of the gene components of this clock are members of the Notch pathway (e.g., the hairy family of Notch target genes) that are organized in negative feedback loops. Delays in transcription and translation within these loops have been predicted to determine the period length of the oscillation, which is as short as 30 min in the chick and 90 min in the mouse (Giudicelli and Lewis, 2004). This clock mechanism ensures that body segments are laid down sequentially in a spatially periodic pattern.
In this review, we will focus our attention on Drosophila and vertebrates. Here, both circadian clock and developmental control mechanisms have been studied extensively. Furthermore, many similarities between the fly and vertebrate systems at the molecular level have made it possible to pinpoint highly conserved core mechanisms (Wager-Smith and Kay, 2000; Panda et al., 2002b; Gilbert, 2003). We will start with an overview of the Drosophila and vertebrate circadian clocks and how they are regulated by light. We will then explore the mechanisms whereby these clocks regulate the timing of physiological processes. Subsequently we will go on to review our current understanding of the emergence and function of the circadian clock during Drosophila and vertebrate development.
DROSOPHILA CIRCADIAN CLOCKS
Several lines of evidence point to a small number of clustered neurons in the brain as being the site of a central pacemaker responsible for directing adult locomotor activity rhythms (Handler and Konopka, 1979). These neurons are located bilaterally between the optic lobes and the central brain. In the adult, they consist of three groups of dorsal neurons DN1, DN2, and DN3 and two groups of lateral neurons on each side of the brain. One group (five to eight neurons) lies dorsolaterally, the so-called LNd neurons, while the other group, the LNv neurons, lies ventrolaterally consisting of five small (s-LNv) and four large (l-LNv; Kaneko and Hall, 2000) neurons. The LNv neurons are characterized by their expression of the neurotransmitter pigment-dispersing factor (PDF; Renn et al., 1999). Recent studies have determined that the integrity of the LNv neurons is crucial for the characteristic morning bout of locomotor activity, while the LNd neurons are responsible for generating the evening locomotor activity bout (Grima et al., 2004; Stoleru et al., 2004). However, not all clock output is governed by the pacemaker neurons. Molecular studies have identified circadian rhythms of clock gene expression in most peripheral tissues which persist even in cultured Drosophila dissociated body segments (head, thorax, or abdomen) as well as in the proboscis, antennae, anterior wing margins, and legs (Plautz et al., 1997; Krishnan et al., 1999; Tanoue et al., 2004). One of the most intriguing properties of these “peripheral” clocks is that they can be entrained by direct exposure to light–dark (LD) cycles, even in vitro, predicting the widespread expression of a photopigment (Plautz et al., 1997).
There are several dedicated photoreceptive organs that contribute to entrainment of the adult circadian system of D. melanogaster: the compound eyes, the ocelli, and a pair of extra-retinal photoreceptors called the Hofbauer–Buchner (H-B) eyelets, located beneath the compound eye (Rieger et al., 2003). However, in Drosophila, neither the genetic ablation of eyes nor mutations in the well-characterized visual phototransduction pathways (e.g., norpA, a mutation affecting the phospholipase C [PLC] signaling pathway) completely block entrainment of the circadian clock by light. The explanation for this finding is that cryptochrome (Cry) expressed in the clock neurons also plays a crucial role in photic entrainment of the fly clock (Stanewsky et al., 1998; see next section). Thus the circadian clock is only “blind” in flies that lack all external and internal eye structures as well as Cry expression (Helfrich-Forster et al., 2001).
Forward genetic analysis has been crucial in discovering the core molecular mechanism of the clock itself, a group of genes encoding activator and repressor transcription regulatory factors: period (per), timeless (tim), clock (clk), cycle (cyc), vrille (vri), and par domain protein 1ϵ (pdp1ϵ). These genes and their products are organized into two interconnecting transcription–translation feedback loops: a per/tim and a clk loop (Fig. 1; reviewed in Hardin, 2005). In the per/tim loop, the bHLH PAS domain factors Clk and Cyc function as transcriptional activators. They bind as heterodimers to E box enhancer elements in the promoters of the per and tim genes and thereby up-regulate their transcription. The Per and Tim proteins subsequently negatively regulate their own expression by inhibiting the Clk–Cyc heterodimer. In the clk loop, the Clk–Cyc heterodimer instead activates expression of the vri and pdp1ϵ genes again acting by means of E-box enhancer elements. Vri binds to V/P promoter elements in the clk promoter and thereby down-regulates clk expression. Subsequently a delayed accumulation of PDP1ϵ displaces Vri from the V/P elements and so derepresses clk expression. By driving a rhythm of Clk expression, this loop is thought to confer stability and robustness on the core per/tim loop. One cycle involving both loops takes approximately 24 hr to be completed and is thought to constitute the core of the clock mechanism: the relative levels of the various cycling components provide a cellular read-out of time-of-day.
The clk loop also controls rhythmic transcription of the cry gene, which we have already encountered as a circadian photoreceptor (see above). In fly heads, Cry abundance changes with the light–dark cycle, with the Cry protein accumulating in the dark and declining in the light. Once Cry is exposed to light, it binds to Tim and triggers tyrosine phosphorylation, thereby targeting Tim for degradation by means of the proteasome (Ceriani et al., 1999; Naidoo et al., 1999). These light-dependent levels of Tim are thought to contribute to the phase-shifting effect on the clock. While the function of Cry as a photoreceptor has been studied in detail, its role in peripheral clock function is not yet fully understood and may involve participation in the core feedback loop (Krishnan et al., 2001).
Yet other clock proteins seem to be involved in regulating the protein stability and subcellular localization of these clock transcriptional regulators. Thus the Double time (Dbt) kinase (also known as Discs overgrown, Dco) phosphorylates and destabilizes Per (Kloss et al., 1998; Price et al., 1998), Casein kinase 2 (CK2, α and β subunits) also phosphorylates and destabilizes Per as well as affecting its nuclear localization (Lin et al., 2002; Akten et al., 2003), and the glucose synthase kinase 3 (GSK3) homolog Shaggy (Sgg) phosphorylates Tim and thereby promotes nuclear localization of the Per–Tim heterodimer (Martinek et al., 2001). In concert with these kinases, phosphatases have also been shown to play an important role in regulating phosphorylation levels of clock proteins. Thus protein phosphatase 2a (PP2a) comprising the regulatory subunits Twins (Tws) and Widerborst (Wdb) stabilizes Per by means of dephosphorylation (Sathyanarayanan et al., 2004). Additional factors appear to play a role in mediating proteasomal degradation of PER such as the supernumary limbs (slmb) gene product, a member of the F-box/WD40 family, which functions as an E3-ligase in the ubiquitin–proteasome pathway (Grima et al., 2002; Ko et al., 2002).
THE VERTEBRATE CLOCK
As in Drosophila, a wealth of evidence points to the existence of central pacemakers in vertebrates. The best studied example is the suprachiasmatic nucleus (SCN) of mammals, which is responsible for the generation of rhythms in behavior, hormonal secretion, and various physiological functions (Klein et al., 1991). This small, paired nucleus in the anterior hypothalamus consists of around 20,000 cells and lies dorsal to the optic chiasma and lateral to the third ventricle. Each nucleus can be subdivided into a “core” and “shell” structure and, furthermore, shows a characteristic anterior–posterior polarity (Moore and Silver, 1998). Various SCN neuronal cell types have been revealed by their differential expression of certain neural peptides (e.g., arginine vasopressin [AVP] and vasoactive intestinal polypeptide; Klein et al., 1991). Recent studies documenting clock gene expression suggest that the SCN can also be subdivided into different regions in terms of light-responsive gene expression and various clock functions (de la Iglesia et al., 2000, 2004). Independent SCN neurons have been shown to function as circadian oscillators in vitro (Liu et al., 1997b); however, cell–cell contacts are thought to serve to synchronize individual cell oscillators within the context of the SCN.
In vertebrates, peripheral clocks also exist that possess a molecular mechanism equivalent to that in the central SCN pacemaker. Thus the expression of clock genes oscillates in a variety of organs, including liver, lung, and blood vessels (Yamazaki et al., 2000; Yoo et al., 2004). Additionally even mammalian fibroblast cell lines can be induced to exhibit a circadian oscillation of clock gene expression by a serum shock treatment (Balsalobre et al., 1998). The current models for the circadian timing system predict that the SCN clock acts as a “master” pacemaker that coordinates the action of the multiple peripheral clocks (Schibler and Sassone-Corsi, 2002). Several direct and indirect mechanisms have been implicated in conveying timing information from the master to the peripheral pacemakers in mammals, including glucocorticoid levels and temperature cycles (Balsalobre et al., 2000; Brown et al., 2002). Also, temporal feeding restriction is able to uncouple peripheral oscillators from the central SCN (Damiola et al., 2000). This general organization appears to vary between vertebrate groups. Thus for example, in zebrafish, direct exposure to light is able to entrain peripheral clocks even in explanted organ and cell cultures, a situation comparable to that in Drosophila (Whitmore et al., 2000). Also, although in mammals the SCN appears to function as the principal central pacemaker, in nonmammalian vertebrates such as fish, reptiles, and amphibia, the pineal complex, a photosensitive structure that is the site for nocturnal synthesis of melatonin, also appears to serve as a central pacemaker (Falcon, 1999).
How does light entrain the vertebrate circadian clock? In mammals, a wide range of experimental evidence points to the integrity of the retina as being crucial for the entrainment of the mammalian clock by light (Foster, 1998). A subset of intrinsically photoreceptive retinal ganglion cells (ipRGCs) that directly project to the SCN by means of the RHT serve as the principal circadian photoreceptors (Berson et al., 2002; Hattar et al., 2002). These cells constitute a non–image-forming pathway and express the photopigment melanopsin, an opsin-like protein (Opn4). However, also the “image-forming” photoreceptors contribute to circadian light sensing (Hattar et al., 2003; Panda et al., 2003). In nonmammalian vertebrates, photoreceptors in extraretinal sites, such as the pineal complex or “deep brain photoreceptors” in the lining of the third ventricle of the diencephalon also appear to contribute to entraining the circadian timing system (Menaker et al., 1997). Such receptors have also been found in the zebrafish (Robinson et al., 1995; Mano et al., 1999; Kojima et al., 2000). The identity of the widely expressed photoreceptor molecule underlying the entrainment of zebrafish peripheral clocks still remains uncertain. Candidates include cryptochromes, based on the apparent involvement of blue wavelengths of light and on the photoreceptor role of Cry in Drosophila (Cermakian et al., 2002). Alternatively, novel opsins such as TMT opsin (Teleost multiple tissue opsin) that are widely expressed in zebrafish tissues are also strong candidates to act as the circadian photoreceptor (Moutsaki et al., 2003).
At the molecular level, homologues of many Drosophila melanogaster clock genes have been cloned in vertebrates and the actual core mechanism of interacting transcriptional feedback loops is very similar in flies and mammals (Fig. 1; Albrecht, 2004). Differences between the two systems include the role of Cryptochrome, a photoreceptive molecule in flies and a component of the negative limb of the feedback loop in mammals. Despite the important role of tim in Drosophila, the function of a mammalian homolog remains controversial with some claiming that mammalian mTim is more likely to represent a homolog of the Drosophila time-out gene, that has not been implicated in the circadian clock (Gotter et al., 2000). Others claim an involvement of a long isoform of mTim in generating circadian rhythmicity in the SCN and place this on the negative limb of the molecular feedback loop (Barnes et al., 2003). In the mammalian counterpart of the Drosophila clk loop, the Rev-erbα and Rora orphan nuclear receptors act in a similar manner to PDP1ϵ and Vri in Drosophila, but instead regulate the bmal1 promoter (the vertebrate cyc homolog; Albrecht, 2004). Rora and Rev-erbα expression is up-regulated by the CLOCK–BMAL1 heterodimer. Rev-erbα in turn represses expression of Bmal1, while the Rora protein subsequently competes with Rev-erbα for the same binding sites and functions as an activator of the Bmal1 promoter. As in the case of Drosophila, posttranslational modifications of vertebrate clock proteins also play a vital role in the function of the clock machinery. For example, the tau mutation in the Syrian hamster that causes shortening of the free-running period results from a point mutation in the casein kinase Iϵ gene (Lowrey et al., 2000).
The extent of circadian clock regulation has been revealed by recent DNA microarray studies. Between 1 and 5% of transcripts in Drosophila heads, and 8–10 % of those in mouse tissues are expressed with a circadian rhythm (Claridge-Chang et al., 2001; McDonald and Rosbash, 2001; Storch et al., 2002). Of interest, there is only limited overlap in cycling transcripts between different tissues, with many of the genes cycling in one tissue being expressed in other tissues at constant levels (Panda et al., 2002a; Storch et al., 2002). Despite this tissue specificity, cycling genes can be mapped to a broad range of functions in all the tissues studied. The functions include peptide synthesis, processing, and secretion; vesicle trafficking; nutrient and intermediate metabolism; detoxification; and cytoskeleton and synaptic function (Claridge-Chang et al., 2001; McDonald and Rosbash, 2001; Panda et al., 2002a; Storch et al., 2002). Furthermore, the circadian clock also plays an important role in regulating systemic functions ranging from locomotor activity rhythms and the sleep wake cycle to endocrine activity (Pittendrigh, 1993). Before exploring the function of the clock during early development, it is essential to understand the nature of the output pathways that link the circadian clock with its regulatory targets.
In the case of peripheral clocks, direct regulation of cycling transcripts by core clock components has been described for several genes. However, this only accounts for a subset of cycling transcripts (McDonald and Rosbash, 2001; Panda et al., 2002a). It has been proposed that conserved, core clock–regulated elements interact with tissue-specific transcription factor networks to generate the observed tissue-specific cycling gene profiles. For central clocks, neuronal connections that connect pacemaker neurons with other parts of the central nervous and neuroendocrine systems represent key clock outputs. Thus in Drosophila, PDF appears to function as the key circadian neurotransmitter linking the lateral neurons with other areas of the brain by means of binding to its specific receptor (Renn et al., 1999; Hyun et al., 2005; Lear et al., 2005; Mertens et al., 2005). In mammals, neural connections between SCN pacemaker neurons and other brain regions mainly within the hypothalamus also play an important role in conveying timing information (Klein et al., 1991; Kalsbeek and Buijs, 2002). Among the key regulatory targets of the SCN is the nighttime synthesis of melatonin by the pineal gland (Korf et al., 1998). Whereas in nonmammalian vertebrates, pinealocytes express photopigments and function as photoreceptors, in mammals, with the possible exception of neonates, these cells have no direct photoreceptive function (Tosini et al., 2000). The rate-limiting step in melatonin synthesis, the N-acetylation of serotonin, is catalyzed by arylalkylamine-N-acetyl transferase (AA-NAT). Regulation of the activity of this enzyme at the transcriptional and posttranscriptional levels appears to be the principal control point for melatonin synthesis. High-affinity membrane receptors for melatonin are expressed in various structures within the brain, including the SCN and pars tuberalis. In the SCN, they are predicted to mediate feedback by melatonin on the clock (Liu et al., 1997a), while in the pars tuberalis, they have been implicated in regulating pituitary function in response to changes in photoperiod (Ross and Morgan, 2002; von Gall et al., 2002).
TIMING CELL PROLIFERATION
One particular circadian clock output deserves more attention for its potential impact on development: the regulation of cell proliferation. From cyanobacteria to higher vertebrates, there are many examples of the circadian clock “gating” S-phase and mitosis of the cell cycle to occur during the night period (Bjarnason and Jordan, 2000; Mori and Johnson, 2000). Thus circadian rhythms of cell cycle have been reported in mammalian proliferative tissues such as oral and intestinal epithelia, bone marrow, and pancreas (Bjarnason and Jordan, 2000). In addition, circadian rhythms in the number of cells in S-phase are encountered in zebrafish from the larval to adult stages (Fig. 2). Most zebrafish tissues show robust, light-entrainable rhythms of S-phase that persist in constant darkness (Dekens et al., 2003). These rhythms are also observed in zebrafish cell culture cells exposed to LD cycles, implicating a cell-autonomous, peripheral circadian clock function in this clock output (Dekens et al., 2003).
Detailed insight into the mechanisms whereby clock components interact with the cell cycle regulatory machinery has come from recent mouse studies. For example, it has been demonstrated that the circadian clock times the extensive hepatocyte proliferation that occurs during liver regeneration. Specifically, the delay between removal of liver tissue (partial hepatectomy) and the subsequent first wave of mitosis depends upon the time of day that the surgery was performed (Matsuo et al., 2003). Transcription of the wee1 kinase G2/M checkpoint control gene appears to represent a key clock regulatory target in this process. Thus expression of the wee1 transcript and consequently levels of the WEE1 kinase protein and phosphorylation of its target, the Tyr-15 residue of Cdc2, show pronounced circadian rhythms. Three E-box elements within the wee1 gene promoter serve as direct binding sites for CLOCK–BMAL1 heterodimers and, thereby, confer regulation by the core clock molecular machinery (Matsuo et al., 2003). Circadian clock components have also been implicated in regulating the proliferation of osteoblasts during bone formation. Knockout mice strains where either both per1 and per2 genes or both cry genes have been inactivated show a high bone mass phenotype as well as a severely disrupted circadian clock. This bone phenotype results from abnormally high levels of osteoblast proliferation (Fu et al., 2005). It appears that in osteoblasts, the CLOCK–BMAL1 heterodimer (presumably as a complex with the negative clock elements PER and CRY) down-regulates expression of c-myc that in turn suppresses expression of the G1 cyclin cyclin D1. By means of this mechanism, clock genes mediate the leptin-dependent sympathetic inhibition of bone formation (Fu et al., 2005).
More recently, intriguing studies have implicated both PER1 and PER2 proteins as tumor supressors. Thus the PER1 protein has been reported to physically interact with and activate components of the ATM pathway (Gery et al., 2006). The ATM kinase and its downstream effector CHK2 are activated by double-strand breaks in DNA and are involved in initiating DNA repair and cell cycle arrest or apoptosis. In addition, per2 knockout mice develop cancer at a significantly higher rate than their wild-type siblings after exposure to γ-radiation. In wild-type mice, clock gene expression is up-regulated and c-myc expression is down-regulated by γ-radiation; however, in the per2 null mutant mice, the clock gene induction response is not observed and c-myc expression is de-repressed, ultimately leading to tumor formation (Fu et al., 2002). This tumor-suppressor role of the per genes may be independent of their function within the circadian clock mechanism. This conclusion is based on the observation that knockout mice where both cry genes are inactivated (showing a severely disrupted circadian clock phenotype) and their wild-type siblings display an equal sensitivity to spontaneous and irradiation-induced cancer (Gauger and Sancar, 2005).
The presence of circadian cell cycle rhythms in unicellular organisms has led to the theory that they may represent an adaptation to minimize the damaging effects of ultraviolet (UV) radiation on sensitive steps of the cell cycle such as DNA replication (Nikaido and Johnson, 2000). The clock restricts these stages of the cell cycle to times of day when there is little or no UV light (Roenneberg and Foster, 1997). However, whether this is also the case in multicellular organisms such as vertebrates is more questionable because here the penetration of UV light to internal tissues may well be limited.
As we have seen, the wide range of physiological processes under circadian control may certainly have direct relevance for development—but is the clock already working at the earlier stages of embryogenesis? In the next sections, we will first explore how the clock matures during development before going on to discuss potential functions of the clock in embryogenesis.
WHEN AND HOW DOES THE CIRCADIAN TIMING SYSTEM DEVELOP?
Basic issues that have been investigated in both Drosophila and vertebrates include defining when clock gene expression rhythms first emerge, when circadian clock outputs mature, and also if clocks start alone or whether they require zeitgeber input. In Drosophila, the first clock-driven circadian rhythm detected during development is the sensitivity to light of the larval light avoidance behavior (the so-called photophobic response; Mazzoni et al., 2005). Subsequently, the clock regulates the timing of eclosion of the adult from the pupal stage (see later; Konopka and Benzer, 1971). Light exposure is able to entrain the adult Drosophila clock as early as the onset of larval development. Remarkably, a single light pulse given at the first larval stage just after hatching from the egg followed by a return to constant darkness alters the phase of adult eclosion as well as locomotor activity rhythms (Sehgal et al., 1992). These results predict that the clock is already running from the first larval stage and that the phase of the clock-generated rhythms persists during the considerable rearrangement of the body plan that constitutes metamorphosis in the pupa. Indeed, lateral neurons expressing PDF are already present in the first-stage larval brain and exhibit circadian rhythms of per and tim gene expression (Helfrich-Forster, 1997; Kaneko et al., 1997). These neurons persist during the pupal stage and seem to subsequently represent the s-LNv neurons of the adult brain. The l-LNv lateral neurons only differentiate later in the pupa. So, it seems that central pacemaker neurons are already functional as the larva hatches from the egg. The larval visual system, the Bolwig's organ, projects and signals to these neurons that in turn transmit rapid photophobic signals (see next section; Mazzoni et al., 2005). The LNs thereby confer a circadian rhythm in the sensitivity of the photophobic response to light input. Interestingly, this photophobic function of the LNs does not rely upon PDF as a neurotransmitter (Mazzoni et al., 2005). How are the clock rhythms in the larval LNs first established? Drosophila raised from the egg until the adult stage under constant darkness and constant temperature conditions still develop robust circadian rhythms of locomotor activity (Sehgal et al., 1992). However, the phases of the rhythms observed in a population of flies raised in constant darkness conditions are random when comparing individuals. Thus the Drosophila clock is able to mature independently of exposure to zeitgebers. However, light or temperature cycles clearly do play an essential role in synchronizing the phase of the emerging clock with the environment.
How does the origin of the circadian clock in Drosophila compare with the situation in vertebrates? In the case of mammals, this issue has been addressed recently in an elegant study using clock-mutant, knock-out strains of mice (Jud and Albrecht, 2006). Female mice that lack circadian clock function as a result of double mutations in clock genes (mPer1Brdm1/Per2Brdm1 or mPer2Brdm1/Cry1−/−) were mated with wild-type males. Heterozygous pups were raised by the arrhythmic, mutant mothers in constant darkness, and then subsequently, locomotor activity patterns for the individual offspring were monitored by a running wheel assay. The results demonstrated that the pups all displayed circadian rhythms of locomotor activity. So, as in Drosophila, it seems that the mammalian clock can initiate its function independently of zeitgebers as well as independently of a functional circadian clock in the mother. However, many studies have documented the importance of a maternal contribution to the entrainment of the fetal clock by the environmental LD cycle. Thus rat pups born and reared under constant darkness display a circadian rhythm of pineal aanat expression that is in phase with the circadian time of the mother (Reppert et al., 1984). Similar results have been obtained in hamsters (Davis and Gorski, 1985). Furthermore, in experiments where the maternal SCN was lesioned, whereas maturation of the fetal clock was unaffected, synchrony between the pups was altered (Reppert and Schwartz, 1986; Davis and Gorski, 1988).
When do circadian rhythms first appear in vertebrates? In rodents, the emergence of circadian clock outputs seems to occur during the first 2 to 3 weeks after birth (Davis, 1981). Thus for example, rhythmic production of melatonin is first detected at the end of the first week of postnatal development (Tamarkin et al., 1980). However, as in the case of Drosophila, are there clues from the central pacemaker itself that the circadian clock might start to function earlier? In rats, SCN development occurs gradually, beginning at embryonic day (E) 14 and continues until E17 (with birth occurring at E21–E22; Moore, 1991). Synaptogenesis in the SCN progresses slowly during late prenatal and early postnatal development and increases notably between postnatal day (P) 4 and P10 (Moore, 1991). Various approaches have been used to measure the emergence of circadian clock function in the developing SCN. Thus SCN neuron metabolic activity as monitored by 2-deoxyglucose uptake shows day–night rhythms in the fetal rat SCN starting between E19 and E21 (Reppert and Schwartz, 1984). In addition, day–night rhythms of AVP mRNA levels and in the firing rate of the SCN neurons become apparent at E21 (Reppert and Uhl, 1987) and E22 (Shibata and Moore, 1987), respectively. Thus there seems to be a functional circadian clock operating in the SCN already just before birth, but it is not clear whether these rhythms are already able to direct clock outputs. It is possible that these activities are required for proper maturation of the clock neurons themselves or of the neural networks within the emerging SCN, similar to the role spontaneous retinal activity waves play in visual system development (Wong, 1999).
The identification of clock genes has enabled more detailed studies of the circadian clock in the developing SCN (Sladek et al., 2004; Kovacikova et al., 2006). The clock genes per1, per2, cry1, bmal1, and clock are already expressed in rat SCN neurons before birth at E19; however, there are no detectable oscillations in expression. Clock gene expression rhythms develop only gradually. Thus at E20, a rhythm of per1 expression starts to become apparent; then at P1, per1, bmal1, and per2 rhythmic expression is established. At P2, rhythmic cry1 expression emerges and only by P10 are all the clock genes expressed with a high-amplitude rhythm. Thus an intriguing conclusion is that circadian clock outputs in SCN neurons emerge before the complete maturation of the rhythmic clock gene expression patterns that constitute the core of the clock mechanism.
Studies have also explored the ontogeny of clock gene expression rhythms in a peripheral tissue clock: the heart (Sakamoto et al., 2002). Circadian expression of per1 and bmal1 were first detected in the developing rat heart between P2 and P5, whereas rhythmic per2 was only detected by P14. Adult patterns of heart clock gene expression were only encountered after stage P20. Thus it seems that, in peripheral tissues, as for the SCN, rhythmic clock gene expression appears gradually during early postnatal development. Of interest, however, not all genes cycle as they do in the adult, predicting that the precise organization of the pacemaker mechanism may well be different in these immature stages. It is tempting to speculate that this immature pattern of clock gene mRNA expression might reflect more reliance upon posttranscriptional regulatory mechanisms within the clock mechanism, early in development. Alternatively, this property may prevent coupling between the circadian clock and output genes until the clock mechanism is completely mature.
One vertebrate in which the emergence of clock function has been studied more extensively is the zebrafish (Vallone et al., 2005). The advantages offered by this species for the molecular genetic analysis of early vertebrate development makes it an ideal model to explore the origins and early functions of the circadian clock (Nüsslein-Volhard and Dahm, 2002). When are circadian rhythms first detected during zebrafish development? After hatching, robust high-amplitude circadian rhythms of locomotor activity are evident as soon as larvae start to swim actively and search for food (starting from the 5th day postfertilization [5 dpf]). The emergence of these rhythms is dependent on the developing fish being exposed for the first 4 days of development to a LD cycle (Hurd and Cahill, 2002). Circadian rhythms in the frequency of S-phase–positive nuclei in various tissues are usually first apparent 4 dpf and subsequently increase in amplitude during larval development (see later; Dekens et al., 2003). Increasing the rate of development by raising the temperature does not advance the time of appearance of this rhythm. Thus it seems that the number of LD cycles experienced rather than the precise developmental stage is the determinant for the timing of appearance of the circadian cell cycle rhythm (Dekens et al., 2003). Circadian rhythms of melatonin synthesis emerge much earlier. In the presence of an LD cycle, a significant nocturnal increase in melatonin is first detected on the second night postfertilization—a result confirmed by the mRNA expression pattern of zfaanat2 in the pineal gland (Gothilf et al., 1999; Kazimi and Cahill, 1999). A circadian rhythm of melatonin synthesis subsequently persists upon transfer to constant darkness.
Does the clock start independently of environmental input as appears to be the case in Drosophila and the mouse? One of the first reports exploring the appearance of circadian clock function during early zebrafish development concluded that a functional circadian clock was maternally inherited with the phase of the rhythm being inherited from the mother (Delaunay et al., 2000). However, these potentially exciting conclusions based on in situ hybridization assays of zfper3 expression were inconsistent with earlier results and have not been confirmed in several subsequent studies using independent assays of clock function (Kazimi and Cahill, 1999; Hurd and Cahill, 2002; Dekens et al., 2003; Kaneko and Cahill, 2005; Lahiri et al., 2005). Thus maternal inheritance of the zebrafish clock seems a highly unlikely scenario. Instead, many independent lines of evidence point to an absolute requirement for exposure to a light–dark or temperature cycle for the appearance of rhythmic circadian clock outputs. For example, when embryos are raised under constant darkness, low-amplitude, erratic rhythms of locomotor activity (phase locked to the time of handling during the experiment) are only visible in a minority of larvae (22%) between 5 dpf and 9 dpf (Hurd and Cahill, 2002). Reducing the number of LD cycles experienced during the first 4 days following fertilization results in a corresponding decrease in the amplitude of the subsequent free-running rhythms. Larvae raised in constant darkness also fail to show rhythmic melatonin synthesis; however, exposure to a LD cycle at a later stage of development will still initiate a normal circadian rhythm of melatonin synthesis (Kazimi and Cahill, 1999). In a similar manner, zebrafish raised in constant darkness conditions also fail to establish circadian rhythms of S-phase–positive nuclei (Dekens et al., 2003). These results have been confirmed by studies of clock gene expression: embryos and larvae raised under constant darkness show constant expression levels (Lahiri et al., 2005). Only exposure to a temperature or LD cycle is sufficient to initiate circadian rhythms of clock gene expression as well as cell cycle rhythms (Lahiri et al., 2005; and unpublished results). Together, these results contrast with the situation in Drosophila and mammals, where the clock apparently starts without external cues and only requires them for phase adjustment with the environment.
An alternative explanation for the observed lack of rhythmicity in constant conditions is that individual clocks may be functioning at the single-cell level—but asynchronously. In support of this notion are observations of clock gene expression at the single-cell level made in zebrafish cell lines after long periods in constant darkness (Carr and Whitmore, 2005). The circadian rhythms of clock gene expression at the single cell level gradually drift out of synchrony and with sufficient time, no rhythm is detectable at the population level. Single-cell clocks continue to function with somewhat erratic period lengths and amplitudes. Exposure to a brief pulse of light is sufficient to resynchronize the single-cell clocks; therefore, rhythms re-emerge at the population level. If asynchrony of single-cell clocks in the zebrafish embryo does explain the apparent lack of clock rhythmicity, then it will be interesting to determine which cues are responsible for synchronizing single cell clocks during Drosophila and mammalian development under constant conditions.
Thus rhythmic clock gene expression as well as clock outputs mature extremely early during zebrafish development. This finding may represent an adaptation to the extremely rapid development and early hatching of a free-living larva (complete embryo development within 24 hr postfertilization [hpf] and hatching of free-living larval stages within 3 dpf). It is tempting to speculate that there may be a strong selective pressure for the zebrafish to develop a functional clock as early as possible, for example, to avoid predation, optimize feeding, or reduce UV damage to proliferating cells in the advancing larval stages.
THE DEVELOPMENT OF CIRCADIAN PHOTORECEPTORS
Light in both Drosophila and vertebrates appears to play an important role in entraining the developing circadian clock. How does the clock perceive light at these early stages? In Drosophila, light is able to entrain the oscillator present in the LNs from the first larval stage. At this stage of development, none of the classic adult photoreceptive structures are present. It seems likely that light perception involves Cry that is already expressed in the clock neurons at this early developmental stage (Klarsfeld et al., 2004). In addition, larvae possess simple photoreceptive structures: the Bolwig's organs (BOs). These paired structures, each consisting of 12 photoreceptor cells, have been implicated in the larva's ability to mount the photophobic response (Hassan et al., 2000; Mazzoni et al., 2005). The Bolwig nerves (BNs) project to the brain and terminate in the vicinity of the LNs. Indeed, there appear to be strong developmental interactions between the BN axons and the LNs (Malpel et al., 2002). It seems, therefore, likely that light also entrains the clock by means of the BOs. This prediction is supported by the expression of the norpA-encoded phospholipase C in the BOs and genetic evidence pointing to the norpA-dependent signaling pathway playing a more significant role in larval than in adult circadian photoreception (Kaneko et al., 2000). During pupal development, the BOs are replaced by adult photoreceptive structures: the H-B eyelet. The establishment of new visual input to the LNs from the adult eyelet also coincides with significant remodeling of the LNs (Malpel et al., 2002).
How and when during development is the mammalian circadian system able to respond to light? As described previously, ipRGCs expressing melanopsin are considered the principal circadian photoreceptors (Berson et al., 2002; Hattar et al., 2002). Melanopsin gene expression is first detected from E10.5 during mouse retinal development, which coincides with the emergence of the ipRGCs (Tarttelin et al., 2003). In the mouse, the RHT is present at P0 and there is evidence for a functional retinal output maturing relatively early, around P4 (Munoz Llamosas et al., 2000). However, using a calcium imaging approach, Sekaran and coworkers have recently shown that the ipRGCs are actually light-responsive from P0 and form functional connections with the SCN as judged by light-induced gene expression (Sekaran et al., 2005). Of interest, at P0, the density of ipRGCs is 5 times higher than in the adult retina (Sekaran et al., 2005). Why there should be so many ipRGCs during early development remains an intriguing issue. One possibility is that there is “overproduction” of ipRGCs, which then compete for the generation of functional synapses with their target neurons, a means to ensure reliable formation of the connections (Kandel et al., 2000). Taken together, these results point to the timing of the appearance of a functional light input pathway coinciding with the first emergence of circadian clock function.
In zebrafish, retinal photoreceptor differentiation begins at 48 hpf and the earliest visual responses are observed around 72 hpf, the time of hatching (Branchek, 1984; Branchek and Bremiller, 1984; Easter and Nicola, 1996). However, extraretinal photoreception matures much earlier. Thus in pineal photoreceptor cells, the expression of photoreceptor markers is already detected around 24 hpf (Masai et al., 1997; Falcon et al., 2003). Direct entrainment of peripheral clocks by light appears even earlier. Exposing embryos to light from as early as 10 hpf (just after gastrulation) induces the expression of a set of genes that includes clock genes (zfper2 and zfcry1a) (Ziv and Gothilf, 2006). Indeed, it seems likely that this property represents a key part of the mechanism relaying light information directly to peripheral clocks (Ziv and Gothilf, 2006). This result is particularly impressive because, at such an early stage, certainly none of the dedicated photoreceptor cells or organs have differentiated. Very recently, it has been reported that light-induced expression of the zfper2 gene is essential for the subsequent maturation of circadian expression of the zfaanat2 gene in the pineal gland (Ziv and Gothilf, 2006). A light pulse of 1 hr given as early as the blastula stage is sufficient to initiate rhythmic zfaanat2 expression in the pineal from the second night postfertilization. This effect of light is not observed when zfper2 expression has been “knocked down” by morpholino treatment, indicating that induction of this gene is crucial for maturation of rhythmic zfaanat2 expression (Ziv and Gothilf, 2006). This result represents a key piece of evidence that light-induced clock gene expression is a prerequisite for initiating clock rhythmicity during early zebrafish development. It seems likely that acute light-induced expression of negative elements of the clock mechanism resets the phase of individual cell clocks and so results in the emergence of circadian rhythms at the whole animal level.
So an additional element of the circadian timing system, the light input pathway in peripheral tissues, matures very early during zebrafish development. It seems likely that the optical transparency of embryos and larvae together with the early maturation of light-induced gene expression may be essential properties contributing to the general early maturation of the clock in this species.
DOES DEVELOPMENT AFFECT THE CLOCK?
Once the clock has been established, how does clock function persist during the cell proliferation and differentiation that accompanies the continued development and growth of the animal? The subcellular localization of core clock components as well as their posttranslational modifications and stability are key factors contribution to the accurate generation of a circadian rhythm (Albrecht, 2004; Hardin, 2005). All of these basic properties are very likely to be influenced by the passage through the cell cycle as well as cell differentiation. In the case of central pacemakers such as lateral neurons in Drosophila and the SCN in mammals, pacemaker cells represent terminally differentiated cell types that may undergo only limited proliferation. However, retaining clock function during the extensive growth and morphogenesis of peripheral tissues may be more problematic. The observation that a single light pulse delivered to the early blastula in zebrafish is able to subsequently entrain rhythms in the pineal gland in the 36-hr-old embryo suggests that the clock does continue to accurately measure time during embryogenesis (Ziv and Gothilf, 2006). This suggestion is consistent with several previous observations made with cyanobacteria and also mammalian fibroblasts: After mitosis, daughter cells inherit the clock time from the original cell (Mori et al., 1996; Kondo et al., 1997; Nagoshi et al., 2004). Although passage through mitosis has been reported to induce some phase shifts in the circadian clock rhythm (Nagoshi et al., 2004), one can anticipate the existence of mechanisms that preserve circadian timing function, despite radical changes in subcellular organization that accompany passage through the cell cycle. The precise nature of these “protective” mechanisms remains unclear.
DOES THE CIRCADIAN CLOCK TIME DEVELOPMENT?
We now turn to the second issue of this review. Once a functional circadian clock has emerged, is there any evidence that it regulates developmental processes? In addition, might individual clock components contribute to regulating development independently of their role in the circadian clock?
Before trying to answer these questions, one important observation to stress is that a lack of circadian clock function, as the result of genetic mutations is in many cases associated with grossly normal early development (Konopka and Benzer, 1971; Dolatshad et al., 2006). Thus a functional circadian clock does NOT seem to be a prerequisite for normal development. The lack of a tight tank between the circadian clock and developmental processes may make some sense: in many cases, the earliest steps of embryogenesis occur in a relatively protected microenvironment, for example inside the chorion or in utero. Thus the potential advantages offered by the clock in optimizing the timing of physiological processes with respect to the day–night cycle may not be relevant for the embryo. Alternatively, the clock might exert some functions during development that are only relevant in the natural environment. In this scenario, the clock might “gate” and thereby protect certain developmental steps from damage by environmental factors and so confer a selective advantage.
In Drosophila, the circadian clock plays an important role in timing eclosion. This key developmental event is restricted to occur around dawn, even if flies are ready to eclose hours earlier. In this way, flies are restricted to emerge from the pupa during periods of highest humidity that might facilitate processes such as expansion of the wings. Interactions between a peripheral clock in the prothoracic gland and the central clock of the LNs, acting by means of PDF, appear to underlie the circadian timing of eclosion (Myers et al., 2003). In addition, the circadian clock has been implicated in affecting more general life history traits such as preadult development time, and adult lifespan in Drosophila (Kyriacou et al., 1990; Klarsfeld and Rouyer, 1998). It is generally believed that faster clocks speed up preadult development, and shorten adult lifespan, whereas slower clocks slow down development and lengthen adult lifespan. A clear positive correlation exists between the free-running period (τ) of different Drosophila per mutants maintained under constant dim light or continuous darkness, and their preadult development time. In the case of the mouse, immediately after birth, survival may be compromised in certain circadian clock mutants. Thus for example, the ClockΔ19 mutant line has been shown to suffer from an increased incidence of perinatal delivery problems and poor survival of offspring to weaning (Dolatshad et al., 2006).
In zebrafish, circadian rhythms of S-phase of the cell cycle are first observed at 4 dpf and subsequently increase in amplitude during larval development and into adulthood (Dekens et al., 2003). Thus although the extensive cell proliferation that accompanies early embryogenesis does not seem to be timed by the circadian clock, there appears to be a clear role for the clock in regulating cell proliferation during subsequent larval development. The number of cells that enter S-phase during one 24-hr window is reduced in larvae maintained in constant darkness compared with those in LD cycles, suggesting that exposure of zebrafish to LD cycles might constitute a mitogenic stimulus (Dekens et al., 2003). Also, as discussed previously, restricting the timing of S-phase and mitosis may constitute a UV damage avoidance strategy. In this respect, it is interesting to note that the highest amplitude circadian rhythms of S-phase are encountered in the skin and gut epithelium, suggesting some preferential clock role in these tissues. Indeed, these represent sites of sustained high levels of cell proliferation originating from a stem cell population.
Nighttime production of the hormone melatonin matures very early during zebrafish development (Gothilf et al., 1999). Interestingly, melatonin has been implicated in stimulating cell proliferation in zebrafish embryos by acting through specific melatonin receptors (Danilova et al., 2004). The onset of this effect coincides with the beginning of endogenous melatonin production and with an increase in melatonin receptor expression (18.5 hpf). This action of melatonin on cell proliferation was reported to subsequently decrease around 38 hpf, coincident with a reduction in the levels of melatonin receptor expression (Danilova et al., 2004). Because melatonin is produced at night, it has been postulated that melatonin also contributes to gating cell proliferation to times of day when levels of potentially damaging UV light are reduced (Danilova et al., 2004).
NONCLOCK FUNCTIONS FOR CLOCK GENES
Historically, our earliest view of the circadian clock was built up as the result of forward genetic analysis screening for mutants with alterations in locomotor activity patterns or eclosion rhythms of adult flies (Konopka and Benzer, 1971). As a result, by definition, any clock gene mutation that also conferred lethality would not have been identified. As such, early models of the circadian clock were very much biased—being made up of elements that were not essential for normal development. However, as our biochemical analysis of the clock has become increasingly detailed and as a result of more sophisticated genetic analysis, it has been revealed that many key clock components also play an essential role in development (Fig. 1). Certain loss-of-function clock gene mutations do result in embryonic lethality.
In Drosophila, null mutations in components of the per/tim clock loop as well as cry do not appear to have a significant effect upon development up to the adult stage (Hardin, 2005). However, factors constituting the clk loop play various important roles in development. Thus for example, the Vri gene product has been implicated in dorsoventral patterning during early embryogenesis (George and Terracol, 1997). Specifically, this factor has been shown to function as a maternal enhancer of embryonic patterning defects caused by mutations in the TGFβ homolog decapentaplegic, leading to increased ventralization of embryos. In addition, PDP1 appears to function as part of a larger complex by interacting with the myogenic factor MEF2 and, thereby, regulates the transcription of muscle-specific genes (Lin et al., 1997). Its widespread expression in developing mesoderm, ectoderm, and endoderm has led to the prediction that it might also be involved in the terminal differentiation of other tissues.
Proteins that regulate the stability, turnover, or subcellular localization of core clock transcription factors perhaps not too surprisingly are not dedicated exclusively to the circadian clock. Many also play pivotal roles during development. Thus for example, null mutations in the dbt kinase gene (also identified as discs overgrown) lead to the absence or reduction in size of imaginal discs and consequent lethality in the embryo, larva, or pupa, depending on the precise mutation (Zilian et al., 1999). The GSK3 homolog shaggy plays an essential role in development by functioning as a segment polarity gene (Siegfried et al., 1990). It participates in wingless signalling and, thereby, regulates the subcellular localization of β-catenin Armadillo. The serine threonine phosphatase PP2A also plays a key role in the Wnt signalling pathway by regulating β-catenin abundance (Seeling et al., 1999). In addition, the F-box/WD40 protein slmb is known to regulate the levels of transcription factors in both the wingless and hedgehog signalling pathways (Jiang and Struhl, 1998). Slmb loss-of-function mutants typically die as early larvae.
In vertebrates, several lines of evidence have also lead to speculation that clock genes might play non–clock-related functions affecting early development. Several clock genes are initially expressed as maternally inherited transcripts, well before circadian rhythms of clock gene expression are detected. In the mouse, a systematic transcriptome analysis of the preimplantation embryo has revealed that bmal1 and Clock are maternally inherited transcripts (Hamatani et al., 2004). Their levels decrease gradually during the first stages of development before rising again with the midblastula transition. In the case of zebrafish, both zfper2 and zfper3 are maternally inherited transcripts (Delaunay et al., 2003). Furthermore, these two genes exhibit a differential tissue-specific expression pattern during early development. The zfper2 transcript is only expressed at elevated levels in the developing pituitary and olfactory bulb, while zfper3 is expressed more ubiquitously in the brain and eye (Delaunay et al., 2003). It has been suggested that this different spatiotemporal expression pattern might reflect tissue-specific differences in maturation of the embryonic circadian system or distinct nonclock functions for zfper2 and zfper3 during development (Delaunay et al., 2003). In early Xenopus development, Xclock transcripts initially accumulate in the anterior neural plate where they do not exhibit circadian cycling (Green et al., 2001). Subsequently the Xclock expression domain expands along the whole neural tube. Early in gastrulation, Xclock expression is up-regulated by treatment with the secreted polypeptide Noggin and by the transcription factor otx2 that has been implicated in the determination of anterior fate. In addition, Xclock has been shown to up-regulate otx2 expression: in theory, constituting a positive feedback loop. Indeed, expression of a dominant-negative mutant form of XClock down-regulates otx2 expression and leads to anterior defects, namely a smaller head and abnormally formed eyes (Morgan, 2002).
From these studies it is apparent that circadian clock outputs mature predominantly after the major part of embryonic development is completed. A circadian clock appears to be dispensable for normal embryogenesis and instead seems to represent an adaptation for the onset of an independent life. However, many clock genes are characteristically expressed from the earliest stages of development and seem to perform functions separate from their participation in the clock feedback loops. Notably, null mutations in several genes involved in timing posttranslational modifications within the clock mechanism result in embryonic lethality. This finding may reflect either basic roles essential for early cell survival or specific functions in patterning or morphogenesis. These nonclock functions must clearly have helped to shape the evolution of clock genes and, thereby, the clock itself. Many basic questions still remain unanswered: Were genes recruited to a function within the clock mechanism before or after their “supplementary” roles were established? Could multiple roles for certain clock genes reflect an ancient ancestral state where development processes were indeed gated to occur only at certain times of day? How is timing information from the clock mechanism protected from processes such as cell division and cell differentiation? Which autonomous mechanisms underlie the establishment of the first clock cycles during early development? Answers to at least some of these questions promise to fill many important gaps in our understanding of circadian clock biology. Furthermore, this knowledge should help us to better understand the basic properties of other timing mechanisms that regulate developmental processes.
We thank Patrick Blader for critically reading this review manuscript. D.V. and K.L. were funded by the Max-Planck-Gesellschaft, N.S.F. was funded by CNRS, and T.D. was funded by DFG and EMBO.