Research Article

Interacting Molecular Loops in the Mammalian Circadian Clock

See allHide authors and affiliations

Science  12 May 2000:
Vol. 288, Issue 5468, pp. 1013-1019
DOI: 10.1126/science.288.5468.1013

Abstract

We show that, in the mouse, the core mechanism for the master circadian clock consists of interacting positive and negative transcription and translation feedback loops. Analysis ofClock/Clock mutant mice, homozygousPeriod2Brdm1 mutants, andCryptochrome-deficient mice reveals substantially alteredBmal1 rhythms, consistent with a dominant role of PERIOD2 in the positive regulation of the Bmal1 loop. In vitro analysis of CRYPTOCHROME inhibition of CLOCK: BMAL1-mediated transcription shows that the inhibition is through direct protein:protein interactions, independent of the PERIOD and TIMELESS proteins. PERIOD2 is a positive regulator of the Bmal1 loop, and CRYPTOCHROMES are the negative regulators of the Period andCryptochrome cycles.

Circadian clock–controlled rhythms provide an orchestrated temporal program that allows for the appropriate timing of physiology and behavior, optimizing the efficiency of biological systems. In mammals, a master clock generating circadian rhythms is located in the suprachiasmatic nuclei (SCN) of the hypothalamus (1,2). Synchronization of the multiple, cell-autonomous circadian clocks within the SCN leads to coordinated circadian outputs that regulate expressed rhythms (3–7).

In its simplest form, the molecular clockwork consists of autoregulatory transcriptional and translational feedback loops that have both positive and negative elements (8). The positive components are two basic helix-loop-helix, PAS domain–containing transcription factors, CLOCK and BMAL1 (9–11). When these transcription factors heterodimerize, they drive the transcription of three Period genes (in the mouse, designatedmPer1, mPer2, and mPer3) and twoCryptochrome genes (mCry1 and mCry2) (9, 12, 13). The mPER and mCRY proteins appear to act as negative components of the feedback loop. By in vitro luciferase reporter gene assays, each protein can negatively regulate CLOCK:BMAL1-mediated transcription through CACGTG E box enhancer elements (12–14).

The mouse Timeless (mTim) protein, which can also negatively regulate CLOCK:BMAL1-mediated transcription in vitro (12, 14), is constitutively expressed in the nucleus of most SCN neurons (15). Unlike TIM inDrosophila, however, mTIM does not translocate the mPER proteins from cytoplasm to nucleus (13), nor is it light responsive (16). In fact, recent database entries show that the mTim is not the true mammalian homolog ofDrosophila tim (17).

Mutations of Clock (clk, a 51–amino acid deletion from the putative transcriptional activation domain) cause abnormally long circadian periods in behavior, with homozygous mutant mice eventually becoming arrhythmic in constant darkness (DD) (18, 19). The dominant negative, mutant CLOCK protein can still form heterodimers with BMAL1 that bind to DNA but cannot activate transcription (9). Consistent with this deficiency, the RNA rhythms of mPer1, mPer2,mPer3, and mCry1 are all significantly blunted in the SCN of homozygous Clk mutant mice (12,13, 20). In addition, mCry2 RNA levels, which are not expressed in a circadian rhythm in the SCN in our hands (21), are significantly depressed inClk/Clk mice (13), as are the RNA rhythms of two clock-controlled genes (CCGs), vasopressin prepropressophysin and albumin D-element binding protein (12, 20,22, 23).

Mice lacking both mCry1 and mCry2, however, exhibit a complete loss of circadian rhythmicity in wheel running behavior immediately upon placement in DD (24,25). mPer1 and mPer2 RNA levels are arrhythmic and expressed at tonic mid-to-high values in the SCN, consistent with a nonfunctioning circadian clock (21,25). This finding is also consonant with in vitro studies showing that the mCRY proteins are potent inhibitors of CLOCK:BMAL1-mediated transcription (13), so that in their absence, expression of genes driven by CLOCK:BMAL1 is enhanced.

The behavioral and molecular phenotypes of mPer2mutant animals are more complex. In the protein coded by the mutantmPer2 allele (mPer2Brdm1 ), 87 residues are missing from the carboxyl portion of the PAS dimerization domain (26). HomozygousmPer2Brdm1 mutant mice show a shorter circadian period followed by loss of circadian rhythmicity in DD.mPer1 and mPer2 RNA levels are still rhythmic in the SCN of homozygous mutants on initial placement in DD, but the rhythms are severely reduced in amplitude because of reduced peak RNA levels (26). The molecular phenotype of the mutant mice is surprising, because reduced RNA levels are not consistent with a potent role of mPER2 in its own negative regulation or the negative regulation of the mPer1 RNA oscillation. Instead, the molecular phenotype of the mPer2Brdm1 mutation is consistent with strong positive regulatory function of mPER2 in the core clock mechanism (26).

The discrepancy between the molecular alterations inmCry-deficient mice and homozygousmPer2Brdm1 mutant animals could be explained by the existence of another interdigitating clock feedback loop, similar to a scheme proposed in Drosophila(27). In the mouse, this other loop would involve the rhythmic regulation of Bmal1, because Clk RNA levels do not oscillate (28, 29).Bmal1 RNA is indeed rhythmic in the SCN of rats and mice with a phase opposite that of Per1 and Per2(20, 30, 31). Because mPER1, mPER2, mCRY1, and mCRY2 exhibit synchronous oscillations of nuclear localization in the SCN (13, 16), the proteins could be simultaneously negatively regulating their own transcription and positively regulating the transcription ofBmal1.

We now provide evidence for an interdigitating Bmal1feedback loop in the mouse SCN and suggest that the dynamically regulated mPER and mCRY proteins have differential effects, with mCRYs being predominantly negative regulators of the mPer andmCry cycles and the mPER2 protein being a dominant positive regulator of the Bmal1 loop.

Bmal1 RNA rhythm in Clk/Clk mice. We first documented a Bmal1 RNA rhythm in mouse SCN using in situ hybridization (32) with an antisense riboprobe to the two major Bmal1 transcripts in the SCN (33,34). Wild-type mice exhibited a robust circadian rhythm in Bmal1 RNA levels (Fig. 1A), with low levels from circadian time (CT) 6 to 9 and peak levels from CT 15 to 18.

Figure 1

Bmal1 and Clock RNA levels are differentially altered in Clock/Clock mutant (Clk/Clk) mice. (A) Bmal1RNA levels are reduced in the SCN of Clk/Clk mice [significant difference between genotypes; analysis of variance (ANOVA), P < 0.001]. (Top) Representative autoradiograms of coronal brain sections at the level of the SCN from wild-type (+/+) and Clk/Clk mice. In situ hybridization with an antisense cRNA probe was used to detect Bmal1 RNA. Magnification, ×9.4. (Bottom) Temporal profiles of Bmal1RNA levels in the SCN of wild-type (solid) and Clk/Clk(dashed) mice are depicted. Each value is the mean ± SEM of five to nine animals. Data at CT 2, 3, 22, and 24 are double-plotted. Gray bar, subjective day; black bar, subjective night. (B)Clk RNA levels are not altered in the SCN ofClk/Clk mice (ANOVA, P > 0.05). Expression of Clk mRNA was not rhythmic in the SCN of wild-type (solid line) or Clk/Clk (dashed line) mice. In situ hybridization with an antisense cRNA probe was used to detectClk RNA (nucleotides 1343 to 1831 of accession numberAF000998). Each value is the mean ± SEM of five to nine animals. Double plotting and lighting cycle as in (A).

The phase of the Bmal1 rhythm is opposite that of themPer1, mPer2, and mPer3 RNA rhythms (35), as reported (20, 30,31). In addition to driving rhythmic transcription of themPer and mCry genes (12,13), it seemed possible that CLOCK:BMAL1 heterodimers might simultaneously negatively regulate Bmal1 gene expression, as proposed for ClK regulation in Drosophila(27). If CLOCK:BMAL1 heterodimers negatively regulate Bmal1 gene expression and if the mutant CLOCK protein is ineffective in this activity, then Bmal1 RNA levels should be elevated and less rhythmic in homozygousClock mutant (Clk/Clk) mice (36). Compared with wild types, however,Clk/Clk animals expressed a severely dampened circadian rhythm of Bmal1 RNA levels in the SCN (Fig. 1A), as reported previously (20). Trough Bmal1 RNA levels did not differ between Clk/Clk mice and wild types. The peak level of the RNA rhythm in homozygous Clk mutant mice was only ≈30% of the peak value in wild types.

We also examined the temporal profile of Clk RNA levels in the SCN of Clk/Clk mutant animals, because it has been reported that Clk RNA levels (assessed by Northern blot analysis) are decreased in the eye and hypothalamus ofClk/Clk mutant mice (19). Consistent with previous reports (28, 29), Clk RNA levels did not manifest a circadian oscillation in mouse SCN (Fig. 1B).Clk RNA levels in the SCN of Clk/Clk mutant mice were not significantly different from those in the SCN of wild-type animals (Fig. 1B). Thus, the Clk mutation appears to alter regulation of Bmal1 gene expression in SCN, but not the regulation of the Clk gene itself. Clk expression may be decreased in other hypothalamic regions.

The low levels of Bmal1 RNA in the SCN of homozygousClk mutant animals show that CLOCK is not required for the negative regulation of Bmal1. Instead, these data indicate that CLOCK is actually necessary for the positive regulation ofBmal1. The positive effect of CLOCK on Bmal1levels is probably indirect and may occur through the mPER and/or mCRY proteins, which are expressed in the nucleus of SCN neurons at the appropriate circadian time to enhance Bmal1 gene expression (13, 15, 16). In addition, themPer1, mPer2, mPer3, mCry1, and mCry2 RNA oscillations are all down-regulated inClk/Clk mutant mice (12, 13). Reduced levels of the protein products of one or more of these genes may lead to the reduced levels of Bmal1 in the mutant mice, through loss of a positive drive on Bmal1 transcription.

Bmal1 and mCry1 RNA rhythms inmPer2Brdm1 mutant mice. HomozygousmPer2Brdm1 mutant animals have depressedmPer1 and mPer2 RNA rhythms (26). We thus examined the Bmal1 rhythm in homozygous mPer2Brdm1 mutants to determine whether the positive drive on the Bmal1 feedback loop might come from the mPER2 protein. We also examined the effects of this mutation on the mCry1 RNA rhythm.

The temporal profiles of gene expression were analyzed at six time points over the first day in DD in homozygousmPer2Brdm1 mutant mice and wild-type littermates (37). The Bmal1 RNA rhythm in the SCN of wild-type animals was substantially different than that of mutant mice (Fig. 2A). Trough RNA levels did not differ between wild-type and mutant animals, but the increase inBmal1 RNA levels was advanced and truncated in the mutants, compared with the wild-type rhythm.

Figure 2

Bmal1 and mCry1 RNA rhythms are blunted in homozygous mPer2Brdm1 mutant mice. (A) Bmal1 RNA levels in the SCN ofmPer2Brdm1 mutant mice. (Top) Representative autoradiograms from wild-type (+/+) and homozygousmPer2Brdm1 (mPer2 mutant) mice. Magnification, ×9.4. (Bottom) Temporal profiles ofBmal1 RNA levels in the SCN of wild-type (solid line) andmPer2 mutant (dashed line) mice, as in Fig. 1 (significant difference between genotypes; ANOVA, P < 0.05). Each value is the mean ± SEM of four animals. (B)mCry1 RNA levels in the SCN ofmPer2Brdm1 mutant mice. (Top) Representative autoradiographs from wild-type (+/+) and homozygousmPer2Brdm1 mice. In situ hybridization with an antisense cRNA probe was used to detectmCry1 RNA (nucleotides 1081 to 1793 of accession numberAB000777). Magnification, ×9.4. (Bottom) Temporal profiles ofmCry1 RNA levels in the SCN of wild-type (solid line) andmPer2 mutant (dashed line) mice are shown, as in Fig. 1(significant difference between genotypes; ANOVA, P ≤ 0.0001). Each value is the mean ± SEM of four animals.

The mCry1 RNA rhythm was also significantly altered. In the SCN of mPer2Brdm1 mutant mice (Fig. 2B), the peak levels of the mCry1 RNA rhythm were suppressed by ≈50%, as reported for mPer1 and mPer2 RNA rhythms in this mouse line (26).

These data suggest that maintenance of a normal Bmal1 RNA rhythm is important for the positive transcriptional regulation of themPer and mCry feedback loops. Thus, rhythmicBmal1 RNA levels may drive rhythmic BMAL1 levels, which, in turn, regulate CLOCK:BMAL1-mediated transcriptional enhancement in the master clock. Indeed, mPer1, mPer2, andmCry1 RNA rhythms are all blunted in the SCN ofmPer2Brdm1 mutant mice, in which theBmal1 rhythm is also blunted [(26) and present data]. In addition, the homozygousmPer2Brdm1 mutation is associated with a shortened circadian period and ensuing arrhythmicity in DD (26).

These data, along with the fact that Clk RNA levels are unaltered in the SCN of homozygous mPer2Brdm1 mutants (26), also provide evidence that mPER2 is a positive regulator of the Bmal1 RNA rhythm. This effect may be unique to mPER2. For example, the diurnal oscillation inmPer2 RNA is not altered in the SCN ofmPer1-deficient mice (38), and mPer1,mPer2, and Bmal1 RNA circadian rhythms are not altered in the SCN of mPer3-deficient mice (39). Moreover, circadian rhythms in behavior are sustained in mice deficient in either mPer1 or mPer3 (38,39).

mCRY-mediated nuclear translocation of mPER2. There are at least two ways that the mPer2Brdm1 mutation could alter the positive drive of the clock feedback loops. The mutation could disrupt mPER:mCRY interactions important for the synchronous oscillations of their nuclear localization and/or alter the protein's ability to interact with other proteins (e.g., transcription factors). We examined whether the PAS domain is necessary for functionally relevant mPER2:mCRY interactions using immunofluorescence of epitope-tagged proteins in COS-7 cells (40).

Coexpression of mPER1 or mPER2 with either mCRY1 or mCRY2 in COS-7 cells translocates >90% of mPER1 and mPER2 into the nucleus (13). To determine whether the PAS domain of mPER2 is required for this translocation, residues 1 to 434 (mPER21-434), which include the PAS domain, were examined in COS-7 cells. mPER21-434 was localized to both cytoplasm and nucleus (89% of transfected cells) (Fig. 3), and the localization was not changed by coexpression with mCRY1. Coexpression of mPER2435-1257 with mCRY1, however, markedly changed the cellular location of the mPER2 fragment from cytoplasm only (84%) to nucleus only (85%). Coexpression of mPER2Brdm1 (missing residues 348 to 434) with mCRY1 also moved mutant mPER2 into the nucleus, from cytoplasm only (100% when expressed alone) to predominantly both cytoplasm and nucleus (81%) (Fig. 3). The same patterns of cellular localization were found when mCRY2 was coexpressed with these mPER2 constructs. Thus, functional mPER2:mCRY interactions are not mediated through the PAS domain. Similarly, the PAS domain was not important for the mCRY-mediated nuclear translocation of mPER1 in COS-7 cells (41).

Figure 3

mPER2:mCRY interactions occur independently of the PAS domain. Location of immunofluorescence of V5-tagged mPER2 constructs expressed in COS-7 cells with (+) or without (−) mCRY1. The mPER2Brdm1 construct was generated by overlap extension polymerase chain reaction. All constructs were confirmed by sequence analysis. The V5 epitope was at the COOH-terminus. The cellular location of immunofluorescence was scored as cytoplasm only (C), both cytoplasm and nucleus (B), or nucleus only (N). Values shown are the mean percentages from two experiments; all values were within 17% of the mean. Gray bars, PAS domain.

Because the mPER:mCRY interactions necessary for nuclear transport occur outside the PAS region, the PAS domain of an mPER2:mCRY heterodimer might be free to bind to an activator (e.g., transcription factor) and shuttle it into the nucleus to activate Bmal1transcription. Alternatively, once in the nucleus, mPER2:mCRY heterodimers or mPER2 monomers could coactivate Bmal1transcription through a PAS-mediated interaction with a transcription factor (27). mPER2 itself does not have a DNA-binding motif (42).

Bmal1 RNA levels in mice lacking mCry1 andmCry2. The tonic mid-to-high mPer1 andmPer2 RNA levels in mCry-deficient mice (21) suggest that CLOCK:BMAL1 heterodimers might be constantly driving mPer1 and mPer2 gene expression in the absence of transcriptional inhibition by the mCRY proteins. To examine whether Bmal1 RNA levels would also be modestly elevated, we compared Bmal1 RNA levels in the SCN of mCry-deficient mice with those in the SCN of wild-type mice of the same genetic background at CT 6 and at CT 18 (43). We also examined Clk RNA levels in these animals.

In wild-type animals, the typical circadian variation inBmal1 RNA levels was apparent with high levels at CT 18 and low levels at CT 6 (Fig. 4A). InmCry-deficient mice, on the other hand, Bmal1 RNA levels were low at both circadian times (Fig. 4A). Clk RNA levels did not differ as a function of circadian time or genotype (Fig. 4B).

Figure 4

Bmal1 and Clk RNA levels are differentially regulated in mCry-deficient mice. (A) Attenuated peak levels of Bmal1 RNA inmCry-deficient mice. Quantitation of Bmal1 RNA levels in the SCN of wild-type (solid bars) andmCry-deficient (open bars) mice. Values are the mean ± SEM of five animals. Mice were studied on the first day in DD. *, significant difference in wild-type mice, P < 0.001; there was no significant difference in mCry-deficient mice,P > 0.05. (B) Clk gene expression in the SCN of mCry-deficient mice. Quantitation of Clk RNA levels in the SCN of wild-type (solid bars) andmCry-deficient (open bars) mice showed no differences (P > 0.05). Values are the mean ± SEM of five animals.

The unexpected low Bmal1 gene expression in the SCN ofmCry-deficient mice suggests that the Bmal1feedback loop is disrupted in the mutant animals, with a resultant nonfunctional circadian clock. Nevertheless, enough Bmal1gene expression and protein synthesis occur for heterodimerization with CLOCK so that, without the strong negative feedback normally exerted by the mCRY proteins, mPer1 and mPer2 gene expression is driven sufficiently by the heterodimer to give intermediate to high RNA values (depending on RNA stability).

mPER1 and mPER2 localization in mCry-deficient mice. The mid-to-high mPer1 and mPer2 RNA levels in the SCN of mCry-deficient mice and simultaneous low Bmal1 levels suggest that mPER1 and mPER2 proteins may not be exerting much positive or negative influence on the core feedback loops. To test this, we determined whether mPER1 and mPER2 were tonically expressed in the nuclei of SCN cells inmCry-deficient mice, because nuclear location appears necessary for action on transcription (13,16).

mPER1 immunoreactivity exhibited a robust rhythm of nuclear staining in the SCN of wild-type mice, with high values at CT 12 (328 ± 3.5, mean ± SEM of positive nuclei per 30 μm section,n = 3) (Fig. 5A) and significantly lower values at CT 24 (54 ± 5, n = 3; P <0.01) (Fig. 5B), as reported in other strains of mice (15, 16).

Figure 5

The intracellular location of mPER1 and mPER2 is altered in the SCN of mCry-deficient mice. (Top panels) mPER1 immunoreactivity (mPER1-IR) in the SCN and piriform cortex of wild-type (+/+) and mCry-deficient mice. (A and B) Representative coronal sections of the SCN in wild-type mice at CT12 (A) and CT 24 (B). (C andD) Expression in mCry-deficient mice at CT12 (C) and CT24 (D). OC, optic chiasm. (E and F) High-power contrast interference views of SCN from wild-type (E) andmCry-deficient (F) mice. (G and H) Nuclear mPER1-IR in the piriform cortex of wild-type (G) and mutant (H) mice. (Bottom panels) mPER2 immunoreactivity (mPER2-IR) in the SCN and piriform cortex of wild-type (+/+) and mCry-deficient mice. (A′ and B′) Representative coronal sections in the SCN of wild-type mice at CT12 (A′) and CT 24 (B′). (C′ and D′) Expression in mCry-deficient mice at CT12 (C′) and CT24 (D′). (E′ and F′) High-power contrast interference views of SCN from wild-type mice (E′) andmCry-deficient mice (F′). (G′ and H′) mPER2-IR in the piriform cortex in wild-type (G′) and mutant (H′) mice.

In mCry-deficient mice, however, mPER1 immunoreactivity was detected in the nucleus of a similar number of SCN neurons at each of the two circadian times (CT 12, 140 ± 9, n = 3; CT 24, 152 ± 21, n = 3) (Fig. 5, C and D), and the values at each time were at ≈40% of those seen at peak (CT 12) in wild-type animals.

The double mCry mutation also altered the subcellular distribution of mPER1 staining in the SCN. In wild-type mice, mPER1 staining viewed under contrast interference was nuclear with a very condensed immunoreaction and a clear nucleolus (Fig. 5E, arrow). The neuropil of the SCN in wild types was devoid of mPER1 immunoreactivity. In the SCN of mCry-deficient animals, mPER1 staining was nuclear (Fig. 5F, solid arrow), but the nuclear profiles were less well defined and less intensely stained, and perinuclear, cytoplasmic immunoreaction could be observed (Fig. 5F, open arrow). In addition, the neuropil staining for mPER1 was higher in mCry-deficient mice, although dendritic profiles were not discernible. In the same brains, the constitutive nuclear staining for mPER1 normally seen in the piriform cortex was not altered in mCry-deficient animals (Fig. 5, G and H).

mPER2 immunoreactivity also exhibited a robust rhythm of nuclear staining in the SCN of wild-type mice, with high values at CT 12 (371 ± 11, n = 3) (Fig. 5A′) and significantly lower counts at CT 24 (31 ± 3, n = 3;P < 0.01) (Fig. 5B′), as reported in another strain (16). In contrast, the pattern of mPER2 immunoreactivity in the SCN of mCry-deficient mice was markedly altered, with few mPER2 immunoreactive cells in the SCN ofmCry-deficient animals at either circadian time (CT 12, 12 ± 1, n = 3; CT 24, 8 ± 2,n = 3; Fig. 5, C′ and D′).

In the wild-type mice, the mPER2 staining profiles were nuclear, with well-defined outlines and nucleoli devoid of reaction product (Fig. 5E′, arrow). In the few mPER2 immunoreactive cells in the SCN ofmCry-deficient mice, low-level mPER2 staining was observed in the nucleus (Fig. 5F′), but the profiles were poorly defined and low-intensity perinuclear staining could also be observed (open arrow). As for mPER1, genotype had no discernible effect on nuclear mPER2 immunoreactivity in the piriform cortex (Fig. 5, G′ and H′), although there was evidence of a low level of perinuclear immunoreactivity for mPER2 (Fig. 5H′, arrows) in piriform cortex ofmCry-deficient mice.

The marked reduction of mPER2 staining in the SCN ofmCry-deficient animals suggests that the mCRY proteins are either directly or indirectly important for mPER2 stability, asmPer2 RNA levels are at tonic intermediate to high levels inmCry-deficient mice, similar to those found formPer1 RNA levels (21). It seems unlikely that we are not detecting mPER2 in the cytoplasm of mCry mutants, because antibody to mPER2 can detect cytoplasmically localized antigen in SCN cells (16).

The low levels of mPER2 immunoreactivity in the SCN ofmCry-deficient mice, in conjunction with tonically lowBmal1 RNA levels, are consistent with an important role of mPER2 in the positive regulation of the Bmal1 loop. Because mPER1 is present in SCN nuclei in mCry-deficient mice yetBmal1 RNA is low, mPER1 likely has little effect on the positive regulation of the Bmal1 feedback loop or negative regulation of the mPer1, mPer2, andmPer3 cycles.

mPER1 and mPER2 can each enter the nucleus even in the absence of mCRY:mPER interactions. mPER1 is expressed in the nucleus of SCN neurons from mCry-deficient mice, and both mPER1 and mPER2 are constitutively expressed in the nucleus of cells in the piriform cortex of mCry-deficient animals. The phosphorylation state of mPER1 dictates its cellular location in the absence of mPER1:mCRY interactions, because its phosphorylation by casein kinase I epsilon leads to cytoplasmic retention in vitro (44). Thus, the nuclear location of both mPER1 and mPER2 in vivo may depend on several factors, including interactions with mCRY and other proteins and their phosphorylation.

mCRY-induced inhibition of transcription. The intermediate to high levels of mPer1 and mPer2gene expression throughout the circadian day inmCry-deficient mice (21, 25) are consistent with a prominent role of the mCRY proteins in negatively regulating CLOCK:BMAL1-mediated transcription, as in vitro data have suggested (13). The endogenous expression of themCry1, mCry2, mPer1, mPer2, and mPer3 genes in mammalian cell lines, however, has obscured rigorous in vitro analysis of the mechanism. Therefore, we used an insect cell line, Schneider (S2) cells, a Drosophilacell line that expresses cycle (the Drosophila Bmal1) but not per, tim, andclock (45, 46), to study the negative regulation of mCRY1 and mCRY2 on E box–mediated transcription. The luciferase reporter consists of a tandem repeat of the Drosophila per E box (CACGTG) and flanking nucleotides fused to hsp70 driving luciferase (46,47).

Because S2 cells express endogenous cyc, transfection withdclock alone caused a large increase in transcriptional activity (265-fold) (Fig. 6A), as described (46). As for dCRY (48,41), this activation was not inhibited by either mCRY1 or mCRY2. When cotransfected, mouse (m)CLOCK and syrian hamster (sh)BMAL1 heterodimers induced a large increase in transcriptional activity (1744-fold) that was reduced by >90% by mCRY1 or mCRY2 (Fig. 6A). Moreover, cotransfection of shBmal1 and human (h)Mop4 (49), but not transfection ofhMop4 alone, similarly caused a large increase in transcriptional activity in S2 cells (539-fold), like that previously found for hMOP4: shBMAL1 heterodimers in mammalian cells (10, 13). hMOP4:shBMAL1-mediated transcription was also blocked by either mCRY1 or mCRY2 (Fig. 6A). The mCLOCK: shBMAL1- and hMOP4:shBMAL1-induced transcription in S2 cells was dependent on an intact CACGTG E box, because neither heterodimer caused an increase in transcription when a mutated E box reporter was used in the transcriptional assay (41). Immunofluorescence of epitope-tagged mCRY1 or mCRY2 expressed in S2 cells showed that each was >90% nuclear in location, as in mammalian cells (13).

Figure 6

mCRY-induced inhibition of transcription occurs independently of the mPER and mTIM proteins. (A) Effects of mCRY proteins on transcriptional activation in Drosophila S2 cells. The luciferase reporter contains four copies of aDrosophila per E box and flanking nucleotides fused to hsp70 driving luciferase (48). Presence (+) or absence (−) of reporter and expression plasmids (1 ng ofdclock; 50 ng of mClock, shBmal1, andhMop4; 100 ng of mCry1 and mCry2) is denoted. Values are luciferase activity expressed as relative to the response in presence of activators (100%). Each value is the mean ± SEM of three replicates from a single assay. The results shown are representative of three independent experiments. (B) Yeast two-hybrid assays showing mCRY interactions with mCLOCK, hMOP4, and shBMAL1. Patches of yeast are shown expressing LEXA baits with VP16 transactivation hybrids. The blue color indicates β-Gal activity of protein:protein-induced activation of lacZ reporter gene. Triplicate patches are from three independent transformants. The same interactions were found in replicate experiments. (C) Effects of mCRY proteins on transcriptional activation in COS-7 cells. The luciferase reporter (pGL3-Basic) used contains a 200–base pair fragment of the 5′ flanking region of the mouse vasopressin prepropressophysin gene, which includes the endogenous promoter (12). Presence (+) or absence (−) of reporter (10 ng) and expression plasmids (0.25 μg of mClock,shBmal1, hMop4, and dclock; 0.1 μg of mCry1 and mCry2) is denoted. Values are luciferase activity expressed as relative to the response in presence of activators (100%). Each value is the mean ± SEM of three replicates from a single assay. The results shown are representative of three independent experiments.

These data indicate that mCRY1 and mCRY2 are nuclear proteins that can each inhibit mCLOCK:shBMAL1-induced transcription independent of the mPER and mTIM proteins and of each other. The results also show that the inhibitory effect is not mediated by the interaction of either mCRY1 or mCRY2 with the E box itself, because E box–mediated transcription was not blocked by the mCRY proteins when transcription was activated by dCLOCK:CYC heterodimers. It thus appears that the mCRY proteins inhibit mCLOCK: shBMAL1-mediated transcription by interacting with either or both of the transcription factors, because a similar inhibition was found with hMOP4:shBMAL1-induced transcription. Yeast two-hybrid assays (50) revealed strong interactions of each mCRY protein with mCLOCK and shBMAL1 (Fig. 6B). Weaker interactions were detected between each mCRY protein and hMOP4. This is further evidence of functionally relevant associations of each mCRY protein with each of the three transcription factors (51).

We next tried to determine whether the mCRY-induced inhibition of transcription was through interaction with CLOCK and/or BMAL1. Because neither mCRY1 or mCRY2 inhibited dCLOCK:CYC-mediated transcription, we examined the ability of each to inhibit dCLOCK:shBMAL1-mediated transcription. This aspect of study could not be examined in S2 cells because of the strong activation induced by transfectingdclock alone in S2 cells where there is strong endogenouscyc expression (52). mCRY1 and mCRY2 completely inhibited mCLOCK: shBMAL1- and hMOP4:shBMAL1-induced transcription in COS-7 cells (Fig. 6C, left and middle, respectively), whereas the cryptochromes did not inhibit dCLOCK:shBMAL1-mediated transcription by more than 20% (Fig. 6C, right). Thus, mCRY inhibits mCLOCK:shBMAL1-induced transcription through interaction with either mCLOCK alone or through an association with both mCLOCK and BMAL1 in a multiprotein complex. Unfortunately, we were unable to examine inhibition of mCLOCK:CYC heterodimers, because cotransfection ofmClock and cyc did not activate transcription in either insect cells (see above) or mammalian cells, even though strong interactions between mCLOCK and CYC were detected in yeast (41).

Molecular organization of the mammalian clockwork. Our working model of the SCN clockwork proposes three types of interacting molecular loops (Fig. 7). ThemCry genes comprise one loop, with true autoregulatory, negative feedback, in which the protein products feed back to turn off their transcription. The second loop is that manifested by each of themPer genes and some CCGs (for example, vasopressin prepropressophysin). This loop type is driven by the same positive elements (CLOCK:BMAL1) as the mCry loop, but transcription is not turned off by the respective gene products. Instead, the mCRY protein acts as a negative regulator, leaving the protein products free for other actions. Thus, mPER2 positively drives transcription of theBmal1 gene (detailed below), mPER1 may function to stabilize protein components of the loop (16), and CCG products (which might include mPER3) function as output signals (12, 22, 39). The rhythmic regulation of Bmal1 comprises the third loop with rhythmicity controlled by the cycling presence and absence of a positive element dependent on mPER2. This positive feedback loop augments the positive regulation of the first two loops.

Figure 7

Model of circadian clockwork within an individual SCN neuron. Gold, mCRYs; blue, mPER1; red, mPER2; green, mPER3; black, CCG; gray, hypothetical activator of Bmal1transcription. C, CLOCK; B, BMAL1. CLOCK:BMAL1 heterodimers activate (+) rhythmic transcription of mCry, mPer, and clock-controlled genes (CCG). The mCRY and mPER proteins form complexes important for nuclear translocation of the mPER proteins (13). The phosphorylation (P) state of the mPER proteins may also regulate their cellular location and stability (44, 53, 54). mPER2 positively regulates Bmal1 transcription by translocating an unidentified activator into the nucleus and/or acting as a coactivator (?). The nuclear-localized mCRY proteins directly interact with CLOCK and BMAL1 to negatively regulate (−) CLOCK:BMAL1-mediated transcription. The functions of nuclear mPER1 and mPER3 have not yet been clearly defined (?). CCGs transduce downstream effects.

This model of interacting loops proposes that at the start of the circadian day, mPer and mCry transcription are driven by accumulating CLOCK:BMAL1 heterodimers acting through E box enhancers. After a delay, the mPER and mCRY proteins are synchronously expressed in the nucleus where the mCRY proteins shut off CLOCK:BMAL1-mediated transcription by directly interacting with these transcription factors. At the same time that the mCRY proteins are inhibiting CLOCK:BMAL1-mediated transcription, mPER2 either shuttles a transcriptional activator into the nucleus or coactivates a transcriptional complex to enhance Bmal1 transcription (27) (Fig. 7). The importance of theBmal1 RNA rhythm is to drive a BMAL1 rhythm after a 4- to 6-hour delay. This delay in the protein rhythm would provide increasingly available CLOCK:BMAL1 heterodimers at the appropriate circadian time to drive mPer/mCry transcription, thereby restarting the cycle. We thus predict that BMAL1 availability is rate limiting for heterodimer formation and critical for restarting the loops. Delineating factors that regulate clock protein stability and interactions (phosphorylation and proteolysis) are important next steps for defining how a 24-hour time constant is built into the clockwork (44, 53,54).

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: reppert{at}helix.mgh.harvard.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article