The mammalian molecular oscillator

The first mammalian clock gene was identified in 1994 by using a N-ethyl-nitrosurea (ENU) mutagenesis screen to look for circadian rhythm mutants in the mouse. With this approach, the laboratory of Takahashi identified and characterized a semi dominant mutation, Clock (for circadian locomotor output cycles kaput), which lengthens circadian period and abolishes persistence of rhythmicity (Antoch et al., 1997; King et al., 1997; Vitaterna et al., 1994).

Isolated SCN neurons from Clock mutant mice exhibit arrhythmic patterns of electrical firing, paralleling the effects on locomotor activity in the animal in prolonged DD (Herzog et al., 1998). Other clock genes have been found mostly by their homology to Drosophila counterparts. These include three period homologs (mPer1, 2, and 3), two Clock homologs (mClock and Npas2, for neuronal PAS domain protein 2), and Bmal1, a gene homologous to cycle. BMAL1 was identified by virtue of its interaction with CLOCK (Gekakis et al., 1998).

Mice homozygous for a Bmal1 null allele have severely disrupted behavioral and molecular rhythms in LD and are completely arrhythmic in DD (Bunger et al., 2000). Two cryptochrome homologs (mCry1 and 2) also exist and represent core pacemaker components.

This is in contrast to Drosophila cry, which is involved in photoreception and appears to be required for clock function only in some peripheral tissues. Two homologs of the regulatory kinase dbt are also found in mouse (Ck1ε/Tau and Ck1δ). A tim homolog also exists in the mouse genome, but the implication of this gene in circadian rhythm generation is somewhat controversial (Benna et al., 2000; Gotter et al., 2000; Zylka et al., 1998a). Rev-erbα and Ror (members of the retinoic acid-related orphan receptor (ROR) family), and nono (non-POU domain-containing octamer-binding protein) were identified by biochemical approaches and contribute to the robustness of the molecular clockwork circuitry. (For review, see (Ko and Takahashi, 2006))

The current model of mammalian circadian oscillator includes two interconnected feedback loops of gene expression: a CLOCK/BMAL1-PER/CRY loop and a CLOCK/BMAL1-REV-ERB/ROR loop (Figure 15). In the first feedback loop, the positive elements CLOCK and BMAL1 heterodimerize via their helix-loop-helix PAS domains and bind E-box cis-regulatory enhancer DNA sequences via their bHLH domain, similarly to the orthologous fly’s dCLK/CYC bHLH/PAS transcription factors. This results in the transcriptional activation of target genes including Per1, Per2, Per3, Cry1 and Cry2. The mutant CLOCK∆19 protein leads to the formation of functionally defective CLOCK/BMAL1 heterodimers and, as a consequence, exhibits markedly blunted molecular rhythms (Gekakis et al., 1998). However in Clock-/- mice, circadian rhythm generation is still observed (Debruyne et al., 2006), evoking that Clock can have one or more paralogs with redundant function. Npas2 was found to be the prime candidate for such a role (DeBruyne et al., 2007).

PER proteins contain PAS domains. They accumulate in the cytosol and form multimeric complexes with CRY proteins that promote their entry into the nucleus, where the PER-CRY

complexes repress their own transcription by acting on the CLOCK/BMAL1 complexes (Kume et al., 1999). As a consequence Per and Cry mRNAs and proteins decrease in concentration, and once the nuclear levels of the PER/CRY complexes are insufficient for auto-repression, a new cycle of Per and Cry transcription can start (Figure 15). PERs and CRYs thus represent the negative elements of the loop, and CRYs seem to have taken the role of Drosophila TIM. The expression of all three Per paralogs cycles at both the mRNA and protein levels, with steady-state message levels peaking at slightly different times during the day (Shearman et al., 1997; Sun et al., 1997; Takumi et al., 1998a; Takumi et al., 1998b; Tei et al., 1997; Zylka et al., 1998b). Cry1 and Cry2 message levels and CRY1 and CRY2 proteins cycle coordinately with the Per transcripts and proteins (Kume et al., 1999). The clock continues to oscillate with single mutations in either Per or Cry genes, suggesting a partial functional redundancy of the various isoforms. Nevertheless the period length in DD is longer (1h) in Cry2 and shorter (0.5-1.5h) in Per1, 2, 3, and Cry1 null mice than in wild type mice. After prolonged exposure to DD, some Per1 and Per2 null mice can become arrhythmic. The difference in period length suggests that the Per isoforms also underwent a functional diversification (van der Horst et al., 1999; Zheng et al., 2001). Although Per1 is dispensable for the rhythmic transcription of Per, Cry and Bmal1 genes, Per2 is required since circadian transcription of these genes is low and flat in Per2 mutants or Per1Per2 double mutants. This observation indicates that PER2 is involved in the positive regulation of Per, Cry and Bmal1 expression (Bae et al., 2001). Per1Per2 or Cry1Cry2 double knock out animals immediately loose rhythmicity under free-running conditions, reflecting their essential function in the core clock mechanism. In Cry double mutants, Per1 and Per2 mRNA levels are constitutively high, while the levels of Bmal1 remain low, supporting the idea that CRYs negatively regulate Per genes expression and like PER2, participate in the positive regulation of Bmal1 expression (Shearman et al., 2000). While CRYs seem clearly to act in the negative limb of the feedback loop, Per2 participates in a mechanism whereby PER and

CRY proteins are involved in the positive regulation of Bmal1 transcription (see below).

Bmal1 is the strongly oscillating component in the mammalian positive limb of the feedback loop, while Clock expression exhibits only weak cycle; this is in contrast to Drosophila, in which the rhythmically transcribed positive element is dClk.

This brings us to the second regulatory feedback loop, which is responsible of the high amplitude of Bmal1 expression. It also involves CLOCK/BMAL1 heterodimers that, in parallel of increasing Per and Cry transcription, induce the transcription of retinoic acid-related orphan nuclear receptors, Rev-erbα and Rorα through E-box elements in their promoters. REV-ERB and ROR proteins are both able to bind retinoic acid-related orphan receptor response elements (ROREs) present in Bmal1 promoter. REV-ERBα (and probably REV-ERBβ) represses Bmal1 transcription, whereas RORα (and probably RORβ and RORγ)

Figure 15. A network of transcriptional–translational feedback loops constitutes the mammalian circadian clock. See text for details. This Figure is adapted from (Ko and Takahashi, 2006).

is an activator of Bmal1 transcription. The combination of repression and activation by REV-ERBs and RORs, respectively, renders Bmal1 transcription (and to a lesser extent Clock transcription) circadian, with a phase opposite to that of Rev-erbα, Per and Cry transcription (Figure 15) (Akashi and Takumi, 2005; Guillaumond et al., 2005; Preitner et al., 2002; Sato et al., 2004). Thus, as evoked above, PERs and CRYs indirectly regulate Bmal1 expression positively by repressing CLOCK/BMAL1-dependent transcription of Rev-erbα. Again this shows that the mammalian oscillator functions in an analogous manner to the Drosophila oscillator, exploiting nuclear orphan receptors of the REV-ERB and ROR family to modulate Bmal1 expression, in lieu of VRI and PDP1ε that modulate dClk transcription, in a second interlocking loop. While the feedback loop within the positive limb is not absolutely necessary for circadian rhythm generation, it ensures a more robust oscillator function (Preitner et al., 2002). As expected, Rev-erbα overexpression is sufficient to abolish the expression of core clock controlled genes, since it clamps BMAL1 levels to nadir values (Kornmann et al., 2007a).

Similarly to Cyanobacteria, N.crassa, Arabidopsis and Drosophila, post-transcriptional regulations are essential for oscillator function in mammals. In particular, reversible phosphorylation provides plausible mechanisms for the regulated formation of protein complexes, their nuclear entry, and their ultimate degradation, each step introducing delays into the feedback loop. The phosphorylation status of PER1, PER2, CLOCK and BMAL1 proteins varies dramatically during the circadian cycle (Lee et al., 2001; Sahar et al., 2010). The first evidence for the involvement of casein kinase 1ε (CK1ε) in the generation of the circadian rhythm comes from the genetic analysis of tau mutant hamsters, which display a dramatically shortened period (20h in homozygous animals) in wheel running activity (Ralph and Menaker, 1988). Subsequent mapping studies revealed that this mutation resided within casein kinase 1ε locus (Lowrey et al., 2000). By creating null and tau mutations of CK1ε in mice, a recent study shows that CK1ε knockout animals have a relatively mild phenotype,

likely because of a redundant function of the highly related CK1δ. Indeed, the tau mutation was revealed to be a gain-of-function mutation of CK1ε that targets PER, but not CRY proteins, and is responsible for the accelerated degradation of PERs by the ubiquitin-proteasome pathway (Akashi et al., 2002; Meng et al., 2008). So CK1ε /CK1δ have a similar role as the Drosophila DBT kinase, likely participating in the generation of the delay between Per mRNA and protein accumulation by controlling PER protein stability and subcellular localization. PER2 phosphorylation appears to enhance the dependency on CRYs for protein stability, probably through direct PER2/CRYs association. The physical interaction between PER and CRY proteins drives their subsequent translocation into the nucleus. Their interaction seems to protect phosphorylated PER2 from ubiquitination and subsequent degradation by the proteosome and thus promotes PER2 nuclear accumulation (Lee et al., 2001; Yagita et al., 2002). In addition to phosphorylation, the protein of the circadian clock can be subjected to acetylation. It has indeed been shown that PER2 and BMAL1 proteins can be acetylated, and that the level of this acetylation varies over the 24h period. The CLOCK protein itself seems to possess an acetyl-transferase enzymatic activity that is responsible for BMAL1 acetylation. The acetylated form of BMAL1 has been suggested to facilitate the recruitment of the repressor CRY1 to the CLOCK/BMAL1 complex, thereby promoting transcriptional repression (Hirayama et al., 2007). Less is known about PER2 acetylation.

However, its deacetylation is mediated by SIRT1, an NAD⁺-dependent deacetylase. SIRT1 binds CLOCK/BMAL1 in a circadian manner and promotes the deacetylation and degradation of PER2. In accordance with this scenario, circadian gene expression is affected in Sirt1 knockout fibroblasts (Asher et al., 2008). BMAL1 stability is also modulated by SIRT1 (Nakahata et al., 2008). The NAD+ dependence of SIRT1 deacetylase activity suggests that SIRT1 may connect cellular metabolism to the circadian core clockwork circuitry.

SUMOylation may also affect clock protein stability, representing another level of control within the core circadian clock (Cardone et al., 2005). CRY protein turnover is controlled by

the F-box protein FBXL3, a component of the SKP1-CUL1-F-box-protein (SCF) E3 ubiquitin ligase complex. FBXL3 loss-of-function animals exhibit a long period phenotype (~26h).

FBXL3 targets CRY proteins for ubiquitination, which is important for the modulation of CLOCK/BMAL1 transcriptional activity over the circadian period (Siepka et al., 2007). Thus, in some cases post-transcriptional modifications are essential for clock function and in others it provides layers of regulatory fine-tuning. The prevalence of these modifications reveals a degree of complexity greater than initially imagined, where protein production, interactions, and post-transcriptional modifications go well beyond the rhythmic dynamics of RNA levels.