Research Article

Role of the CLOCK Protein in the Mammalian Circadian Mechanism

See allHide authors and affiliations

Science  05 Jun 1998:
Vol. 280, Issue 5369, pp. 1564-1569
DOI: 10.1126/science.280.5369.1564


The mouse Clock gene encodes a bHLH-PAS protein that regulates circadian rhythms and is related to transcription factors that act as heterodimers. Potential partners of CLOCK were isolated in a two-hybrid screen, and one, BMAL1, was coexpressed with CLOCK and PER1 at known circadian clock sites in brain and retina. CLOCK-BMAL1 heterodimers activated transcription from E-box elements, a type of transcription factor–binding site, found adjacent to the mouseper1 gene and from an identical E-box known to be important for per gene expression in Drosophila. Mutant CLOCK from the dominant-negative Clock allele and BMAL1 formed heterodimers that bound DNA but failed to activate transcription. Thus, CLOCK-BMAL1 heterodimers appear to drive the positive component of per transcriptional oscillations, which are thought to underlie circadian rhythmicity.

Circadian clocks are endogenous oscillators that control daily rhythms in physiology and behavior (1). Such clocks are phylogenetically widespread (2) and are likely to reflect evolutionarily ancient, fundamental mechanisms of timekeeping important for the anticipation of daily variations in environmental conditions (3). In mammals, the circadian clock driving metabolic and behavioral rhythms is located in the suprachiasmatic nucleus (SCN) of the hypothalamus (4). Mammals and other vertebrates also have an autonomous circadian clock in each retina (5) driving rhythms in local physiology that are likely to anticipate the transitions between daytime and nighttime viewing conditions.

The starting point for a molecular analysis of the mammalian circadian mechanism was the identification of a mouse mutant, Clock, which has a phenotype affecting both the periodicity and persistence of circadian rhythms (6). CLOCK, the predicted protein product of the mutated gene (7, 8), is a member of the bHLH-PAS family, some members of which are known to function as transcription factors. The mutant Clock allele acts genetically in a dominant-negative fashion (7, 9) and encodes a protein with a 51–amino acid deletion in its putative transcriptional regulatory domain (CLOCK-Δ19). How CLOCK controls the periodicity and persistence of circadian rhythms is unknown.

Although not formally demonstrated to encode circadian clock components, three mammalian orthologs of the Drosophilaclock gene per, mper1 (10),mper2 (11), and mper3 (12), have been identified. All three are expressed in the SCN and retina, and, like Drosophila per, the levels of their transcripts exhibit a circadian oscillation. Fly and mammalian circadian clocks are thus likely to share a conserved molecular mechanism.

In Drosophila, the clock mechanism is constituted in part by a negative feedback loop in which the PER protein directly or indirectly represses transcription of its own gene (13, 14). Constitutive per mRNA expression has been observed in mutants lacking functional PER protein (14, 15), indicating that there is PER-independent positive regulation of pertranscription. A 69–base pair (bp) “clock control region” located upstream of the per gene confers circadian cycling on reporter genes that is dependent on a functional PER protein (16). The 69-bp clock control region thus includes sequences sufficient for both PER-dependent negative feedback and PER-independent positive transcriptional regulation. Within this sequence, an E-box element (CACGTG), a binding site for certain transcription factors, is required for the positive component of the transcriptional regulation (16).

Precedents for heterodimerization between bHLH-PAS proteins have suggested that CLOCK could dimerize with a bHLH-PAS partner to regulate circadian rhythms. Because E-box elements are recognition sites for bHLH DNA-binding domains (17), we hypothesized that CLOCK and its predicted partner might drive expression of one or moreper genes in mammals and that homologous proteins might do the same in Drosophila.

Screen for CLOCK-interacting proteins. To identify CLOCK-interacting proteins (CIPs), including the predicted bHLH-PAS partner of CLOCK, we performed a yeast two-hybrid screen (18) using a LEXA-CLOCK hybrid as bait (19). In our initial control experiments, CLOCK showed clear evidence of interaction with ARNT, a bHLH-PAS protein known to heterodimerize widely (20), but no evidence of CLOCK homodimerization was detected (Fig. 1A), implying that CLOCK acts in vivo with a heterodimeric partner. We then carried out a two-hybrid screen of a cDNA library constructed from hamster hypothalamus (19). Of 2 × 107 yeast transformants, 154 formed colonies positive for both HIS3and LacZ reporter genes. After accounting for multiple isolations, we ultimately identified 22 distinct CIPs that gave two-hybrid signals specific for CLOCK, and two proved to be bHLH-PAS proteins (Fig. 1B). One, isolated five times and giving an interaction signal stronger than the ARNT positive control, was BMAL1, a protein of unknown function expressed in brain and muscle (21); the other, isolated once and giving a weaker interaction signal, was ARNT2 (22), also of unknown function.

Figure 1

Yeast two-hybrid screen for CLOCK-interacting proteins. Shown are triplicate yeast patches expressing the indicated LEXA protein (rows) and the indicated VP16 protein (columns). Blue precipitate indicates cumulative β-Gal activity, resulting from activation of the LacZ reporter gene by protein-protein interaction. Each triplicate represents three independent transformants. (A) Control two-hybrid assays showing interaction of LEXA-CLOCK with VP16-ARNT but not with VP16-CLOCK. Background is defined by the signal from LEXA-CLOCK with VP16. Yeast expressing the LEXA-CLOCK(1-580) bait were transformed with VP16 (negative control), VP16-CLOCK(1-389), or VP16-ARNT(70-474) expression plasmids. The latter two include a complete bHLH-PAS region; VP16-CLOCK(1-389) shows a robust interaction with LEXA-ARNT(70-474) (25). (B) Two-hybrid assays showing characterization of several CLOCK-interacting proteins obtained in the two-hybrid screen. Yeast expressing the LEXA-CLOCK(1-580) bait or the LEXA-p65 (synaptotagmin) control bait were transformed with the indicated VP16 expression plasmids. VP16 and VP16-ARNT plasmids are the negative and positive controls, respectively, for LEXA-CLOCK, as in (A). VP16-ARNT2, VP16-BMAL1, and VP16-CIP8 are library plasmids isolated from yeast colonies positive in the two-hybrid screen.

Expression at known circadian clock sites. Next we examinedarnt, arnt2, and bmal1 transcripts for coexpression with Clock and mper1 transcripts in the mouse brain by in situ hybridization (23). Of the three potential CLOCK partners, only bmal1 transcripts showed a clear pattern of expression like Clock (7) andmper1 transcripts (10) in coronal brain sections (24), including robust expression in the SCN (Fig.2A) and pars tuberalis (25).bmal1 transcripts were coexpressed with bothClock and mper1 transcripts throughout the rostrocaudal extent of the mouse SCN, although the hybridization signal with the bmal1 riboprobe was somewhat lower than that withClock or mper1 (Fig. 2A).

Figure 2

Localization by in situ hybridization of bmal1 transcripts as compared withClock and mper1 transcripts at the two known circadian clock sites in mammals. (A) Coexpression throughout the rostrocaudal extent of the mouse SCN. Sets of three neighboring coronal brain sections from a single mouse, each set taken from a different rostrocaudal level. Within each set, the three sections were hybridized to mper1, Clock, orbmal1 antisense riboprobes, respectively, as indicated. Rostrocaudal level of each set within the SCN is indicated at left. Only faint background hybridization was observed with sense control riboprobes (25). (B) Coexpression in the mouse retina. Three neighboring parasaggital eye sections hybridized tomper1, Clock, or bmal1 antisense riboprobes, respectively, as indicated. S, sclera; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Only faint background hybridization was observed with sense control riboprobes, except for the sclera, which showed robust hybridization to all three sense riboprobes (25).

We also tested for coexpression in retina, the other known site of an independent circadian clock in mammals. Clock,bmal1, and mper1 transcripts showed coexpression in the mouse retina in a characteristic pattern (Fig. 2B), with strong expression in the inner nuclear layer (INL), the outer nuclear (photoreceptor) layer (ONL), and a discrete subset of cells in the ganglion cell layer (GCL). bmal1 appeared to show somewhat higher expression in the ONL relative to the INL than didClock and mper1. Only about 1 in 10 cells in the GCL showed a hybridization signal for the three riboprobes (25), and these hybridizing cells could correspond to a subset of ganglion cells or they could be displaced amacrine cells (26). Taken together, the coexpression of these transcripts in the SCN and retina strongly suggest that CLOCK, BMAL1, and mPER1 are colocalized within circadian clock cells.

Binding of CLOCK-BMAL1 heterodimer to per gene E-box. On the grounds that the circadian mechanism is likely to be conserved between Drosophila and mammals (10-12), we tested CLOCK-ARNT, CLOCK-ARNT2, and CLOCK-BMAL1 heterodimers for binding to the E-box element within theDrosophila per clock control region, known to be important for regulation of per gene expression. In effect, we asked the Drosophila E-box to select the heterodimer likely to be relevant to the mammalian circadian mechanism.

To test DNA binding, we performed yeast one-hybrid assays (27), in which the binding of test proteins to a given DNA fragment is signaled by activation of the LacZ gene, resulting in an increase in β-galactosidase (β-Gal) activity. The one-hybrid strains we constructed are virtually identical to the two-hybrid strain used above in genotype and reporter gene, differing from the two-hybrid strain in the short DNA sequences inserted adjacent to the LacZ reporter gene (27).

CLOCK and BMAL1 together exhibited robust DNA binding to fragments containing the E-box but not to identical fragments with mutated E-box sequences (Fig. 3, A and B). In contrast, CLOCK, ARNT, ARNT2, or BMAL1 alone did not detectably bind to theper clock control region, nor did CLOCK and ARNT together or CLOCK and ARNT2 together (Fig. 3A). Failure to detect a signal in the one-hybrid assay could reflect poor binding of the bHLH-PAS dimers to the DNA sequence, poor dimerization (only the dimers bind DNA), or both. Detection of CLOCK-ARNT and CLOCK-ARNT2 heterodimerization in two-hybrid assays with the same reporter gene and the identical expression plasmids (Fig. 1) argues that poor binding of the heterodimers to this particular E-box is the most likely explanation.

Figure 3

Yeast one-hybrid DNA-binding assays showing binding of CLOCK-BMAL1 heterodimer to the E-box site from theDrosophila per gene clock control region. Shown are triplicate yeast patches expressing the indicated proteins. DNA binding is indicated by β-Gal activity, resulting from activation of theLacZ reporter gene (27). Blue precipitate indicates cumulative β-Gal activity, and each triplicate represents three independent transformants. (A) 69-bp clock control region. Reporter yeast transformed with full-lengthClock cDNA expression plasmid (+CLOCK) or nonrecombinant expression plasmid (−CLOCK) in combination with the indicated VP16 plasmid. (B) Full-length CLOCK and VP16-BMAL1 expressed in four reporter strains. +, reporter gene includes 21-bp sequence. −, reporter gene lacking 21-bp sequence. wt, 21-bp sequence, including E-box, from clock control region. mut1 and mut2, same 21-bp sequence, but with E-box sequence scrambled two different ways (27).

Of the candidates tested, only CLOCK and BMAL1 formed heterodimers that robustly bound to the E-box sequence known to mediate positive regulation of the Drosophila per gene. In conjunction with the coexpression of bmal1, Clock, andmper1 transcripts in the SCN and retina (Fig. 2), these results implicate CLOCK-BMAL1 heterodimers in circadian clock function and suggest that CLOCK-BMAL1 heterodimers drive mammalianper gene expression.

Cloning and analysis of mper1 5′ flanking region. To test the hypothesis that CLOCK-BMAL1 heterodimers drive expression of one or more mammalian per genes, we isolated overlapping genomic clones encompassing ∼35 kilobases (kb) frommper1 and its 5′ flanking region (28) on the hypothesis that one or more E-box elements identical to that of theDrosophila per clock control region would be found near the transcription start site. Restriction mapping and sequencing indicated that one clone included the 5′ untranslated region, a putative transcription start site, and 3.2 kb of predicted 5′ flanking sequence (28). The complete double-stranded sequence of this 3.2-kb region revealed three E-box sites within 1.2 kb of the putative transcription start site, all identical to that from theDrosophila per clock control region (Fig.4A).

Figure 4

Analysis of mper 1gene 5′ flanking region. (A) Location of E-box sites within 5′ flanking region. Numbered axis represents distance in base pairs from the putative transcription start site, marked as +1 (28). Filled boxes represent the locations of the three E-boxes; the sequence of each E-box with 6 bp of flanking sequence (lowercase letters) on each side is shown at the top. (B toE) Transactivation from E-box sites by CLOCK-BMAL1 heterodimer. Transcriptional activation in mammalian cells of luciferase reporter from sequences derived from the 5′ flanking region of the mper1 gene (B to D) or from the Drosophila per gene (E). (B) 2.0-kb fragment including all three E-boxes. (C) 54-bp fragment consisting of the three E-boxes and their immediate flanking sequences linked together (A). 54-bp mut, 54-bp fragment in which all three E-boxes were scrambled (30). (D) Each individual E-box with immediate flanking sequences, labeled as in (A). dist., distal; mid., middle; prox., proximal. (E) 69-bpDrosophila per clock control region. (B and E) + or −, presence or absence, respectively, of indicated insert in luciferase reporter plasmid. (C and D) + or −, presence or absence, respectively, of indicated luciferase reporter plasmid in cell transfection. (B to E) Expression plasmid with (+) or without (−) the indicated CLOCK, BMAL1, and CLOCK-Δ19 full-length cDNA inserts. Shown are the mean are SEM of ≥6 independent experiments (C and E) or of ≥3 independent experiments (B and D). Some of the standard error bars are too small to be seen at this scale.

Transactivation by CLOCK-BMAL1 heterodimers frommper1 E-boxes. Our hypothesis that CLOCK-BMAL1 heterodimers drive per gene expression requires that they function to activate rather than suppress transcription, suppressor activity having been documented in the bHLH-PAS family (29). To test this prediction, we cloned a 2.0-kb fragment of the mper1 5′ flanking region that includes all three E-boxes (30) for use in luciferase reporter gene assays in mammalian cells (31). CLOCK and BMAL1 together, but not either alone, produced a substantial increase in transcriptional activity (9.3-fold; P < 0.005) (Fig. 4B), indicating that recognition sites for the CLOCK-BMAL1 heterodimer are contained within the mper1 5′ flanking region. We conclude that CLOCK acts as a transcriptional activator.

To determine if one or more of the mper1 E-boxes were sites of action of CLOCK-BMAL1 heterodimers, we examined CLOCK and BMAL1 transcriptional activity from a 54-mer in which the threemper1 E-boxes and their immediate flanking sequences (Fig.4A) were linked together (30). CLOCK alone produced a small increase in transcriptional activity (∼twofold), BMAL1 alone produced only a neglible increase in activity, but CLOCK and BMAL1 together produced a substantial transcriptional activation (14.1-fold, P < 0.0005) (Fig. 4C). No transcriptional activity of CLOCK and BMAL1 together was detected with a 54-mer in which the three E-box sequences were mutated (30) (Fig. 4C), indicating that transcriptional activation by CLOCK-BMAL1 heterodimers requires at least one of the E-box sites.

We next tested each of the three mper1 E-box sites independently (30). CLOCK and BMAL1 together produced a small increase in transcriptional activity from the distal E-box (2.1-fold), a moderate increase from the middle E-box (3.6-fold), and a substantial increase from the proximal E-box (8.5-fold) (Fig. 4D). Thus, all three E-boxes are potential sites of regulation ofmper1 by CLOCK-BMAL1 heterodimers. Together with the coexpression of Clock, bmal1, andmper1 transcripts in the SCN and retina, these results strongly suggest that CLOCK-BMAL1 heterodimers drive mper1gene expression.

A similar set of results for CLOCK and BMAL1 was obtained with the 69-bp Drosophila per clock control region as the DNA target site (Fig. 4E). Thus, per gene regulatory mechanisms have been conserved between Drosophila and mammals, and a homologous heterodimer is likely to exist in Drosophila. We have identified a Drosophila homolog of BMAL1 (dBMAL1), and experiments addressing its function in conjunction with aDrosophila CLOCK homolog (dCLOCK) are reported separately (32).

CLOCK-Δ19 and BMAL1 together showed no transactivation activity frommper1 E-boxes (Fig. 4C), the amount of activity being significantly below that of BMAL1 alone (P < 0.0005). A similar result was observed with CLOCK-Δ19 and BMAL1 together from the Drosophila per E-box (Fig. 4E). We conclude that CLOCK-Δ19 is defective in transcriptional activation activity.

Mechanism of the Clock dominant-negative mutation. The Clock mutant allele acts as a dominant negative in mice (7, 9), and CLOCK-Δ19 and BMAL1 together showed no transcriptional activation from mper1 E-box elements (Fig. 4C). To determine the mechanism of the dominant-negativeClock mutation, we further characterized CLOCK-Δ19 with regard to heterodimerization and DNA binding. In two-hybrid assays, CLOCK-Δ19 showed interactions with ARNT, ARNT2, and BMAL1 that were similar in signal strength and identical in rank order to those shown by wild-type CLOCK (Fig.5A; compare with Fig. 1B). Thus, CLOCK-Δ19 heterodimerizes, apparently normally, with BMAL1 and other bHLH-PAS proteins. CLOCK-Δ19 is not, however, equivalent to the wild type in regard to all protein-protein interactions, because several CIPs we isolated apparently interacted only with wild-type CLOCK (for example, CIP8 in Fig. 5A; compare with Fig. 1B).

Figure 5

Properties of CLOCK-Δ19. (A) Two-hybrid assay showing interaction of CLOCK-Δ19 with bHLH-PAS proteins but not with CIP8. Arranged as in Fig. 1. (B) One-hybrid assay showing binding of CLOCK-Δ19–BMAL1 heterodimer to E-box site. Arranged as in Fig. 3A. The three proteins indicated at left were LEXA hybrids (18) in this experiment, but the LEXA domain is not relevant to the results in the one-hybrid reporter strain, and it does not affect the function of CLOCK in this assay.

To assess DNA-binding activity, we performed one-hybrid assays using a yeast reporter strain carrying the per clock control region. When CLOCK or CLOCK-Δ19 was paired with BMAL1, robust DNA-binding activity was detected, but it was not detected when either was paired with negative controls (Fig. 5B). Thus, the E-box–binding activity of the CLOCK-Δ19–BMAL1 heterodimer is intact. Together these studies indicate that CLOCK-Δ19 acts genetically in a dominant-negative fashion because it forms a DNA-binding heterodimer, its partner thereby sequestered in a protein-DNA complex deficient in transactivation activity. We conclude that the abnormal circadian rhythms ofClock mutant mice are caused by this defect, which we predict would lead to decreased transcription of mper1 and possibly other circadian clock genes. Strong support for this conclusion comes from the observation that in Clock mutant mice the peak levels of mper1 transcript in the SCN are significantly reduced, compared with those of wild-type mice (33).

Role of CLOCK and BMAL1 in circadian clock mechanism. We propose that CLOCK-BMAL1 heterodimers directly activate transcription of mper1 and possibly other circadian clock genes. As inDrosophila, it is likely that the mammalian clock mechanism is constituted in part by the direct or indirect repression ofper genes by PER proteins. Central to the mechanism of circadian oscillations would therefore be the establishment of an alternating regime of per gene activation by CLOCK-BMAL1 heterodimers and PER-dependent inhibition of this activation, responsible in turn for the rising and falling phases of the circadian oscillation in the levels of per transcripts. In the fungusNeurospora, a similar positive role in a circadian transcriptional feedback loop has been proposed for the WC-1 and WC-2 proteins, which are required for normal expression of the circadian clock gene frequency (frq) and consequently for the circadian oscillation of frq expression (34). Although shown genetically to be required for this regulation, it is not yet known if WC-1 and WC-2 act directly as transactivators offrq expression.

How might PER-dependent negative feedback inhibit or overrideper gene transcriptional activation by the CLOCK-BMAL1 heterodimer? One possibility would be PER-dependent sequestration of CLOCK, BMAL1, or both, leading to a loss of CLOCK-BMAL1 DNA-binding activity, similar to the mechanism by which the Id protein inhibits myogenic differentiation (35). Other possibilities would include a PER-dependent repressor that binds either to the E-box or to a nearby site within the per gene upstream region. It is not known if PER (including any other associated proteins) acts directly to achieve negative feedback, but the observation that coexpression of Drosophila PER and its partner TIM is sufficient to inhibit dCLOCK-dependent per gene activation in cultured cells suggests a direct mechanism (32).

We have identified BMAL1 as a probable component of the circadian clock, defined a biochemical function for CLOCK, and implicated a mammalian per homolog as a target gene regulated by CLOCK-BMAL1 heterodimers. These findings place CLOCK in a specific role within the circadian oscillator, and they mechanistically tie together CLOCK, a genetically demonstrated component of the circadian clock, and a mammalian per homolog, likely to be a component of the clock. How the circadian transcriptional feedback loop is generated and regulated will provide the basis for understanding how biological clocks evolved, how they operate, and how they control behavioral and physiological programs.

Note added in proof: CLOCK-BMAL1 (MOP3) heterodimerization and transactivation from an E-box site have been reported independently (36).

  • * These authors contributed equally to this work.


View Abstract

Navigate This Article