cAMP-Dependent Signaling as a Core Component of the Mammalian Circadian Pacemaker

See allHide authors and affiliations

Science  16 May 2008:
Vol. 320, Issue 5878, pp. 949-953
DOI: 10.1126/science.1152506


The mammalian circadian clockwork is modeled as transcriptional and posttranslational feedback loops, whereby circadian genes are periodically suppressed by their protein products. We show that adenosine 3′,5′-monophosphate (cAMP) signaling constitutes an additional, bona fide component of the oscillatory network. cAMP signaling is rhythmic and sustains the transcriptional loop of the suprachiasmatic nucleus, determining canonical pacemaker properties of amplitude, phase, and period. This role is general and is evident in peripheral mammalian tissues and cell lines, which reveals an unanticipated point of circadian regulation in mammals qualitatively different from the existing transcriptional feedback model. We propose that daily activation of cAMP signaling, driven by the transcriptional oscillator, in turn sustains progression of transcriptional rhythms. In this way, clock output constitutes an input to subsequent cycles.

The suprachiasmatic nuclei (SCN) of the hypothalamus are the principal circadian pacemaker in mammals, driving the sleep-wake cycle and coordinating subordinate clocks in other tissues (1). Disturbed circadian timing can have a major negative impact on human health (2). The molecular clockwork within the SCN has been modeled as a combination of transcriptional and posttranslational negative-feedback loops (3), whereby protein products of Period and Cryptochrome genes periodically suppress their own expression (4). It is unclear how long-term, high-amplitude oscillations with a daily period are maintained, not least because transcriptional feedback loops are typically less precise than the oscillation of the circadian clock and oscillate at a higher frequency than one cycle per day (5). Moreover, recombinant cyanobacterial proteins can sustain circadian cycles of autophosphorylation in vitro, in the absence of transcription (6), and intracellular signaling molecules cyclic adenosine diphosphate–ribose (cADPR) and Ca2+ are essential regulators of circadian oscillation in Arabidopsis and Drosophila (7, 8). This indicates that transcriptional mechanisms may not be the sole, or principal, arbiter of circadian pacemaking (9, 10). We show that the transcriptional feedback loops of the SCN are sustained by cytoplasmic adenosine 3′,5′-monophosphate (cAMP) signaling, which determines their canonical properties of amplitude, phase, and period. This extends the concept of the mammalian pacemaker beyond transcriptional feedback to incorporate its integration with rhythmic cAMP-mediated cytoplasmic signaling.

We tracked the molecular oscillations of the SCN as circadian emission of bioluminescence by organotypical slices from transgenic mouse brain. Rhythmic luciferase activity controlled by the Per1 promoter (Per1::luciferase) revealed circadian transcription, and a fusion protein of mPER2 and LUCIFERASE (mPER2::LUC) reported circadian protein synthesis rhythms. Under these conditions, the cAMP content of the SCN was circadian (Fig. 1A) and accompanied by a circadian cycle in activity of cAMP response element (CRE) sequences reported by a CRE::luciferase adenovirus (Fig. 1B). In molluscs, birds, and the mammalian SCN, cAMP is implicated in entrainment or maintenance of clocks, or both, or mediation of clock output (1113). It has not been considered as part of the core oscillator (14). If the cAMP-mediated rhythm of CRE activity is necessary for SCN pacemaking, its suppression should compromise circadian gene expression. We treated SCN slices with MDL-12,330A (MDL), a potent, irreversible inhibitor of adenylyl cyclase (AC) (15) to reduce concentrations of cAMP to basal levels (Fig. 1A). MDL rapidly suppressed circadian CRE::luciferase activity, presumably through loss of cAMP-dependent activation of CRE sequences (fig. S1A), and caused a dose-dependent decrease in the amplitude of cycles of circadian transcription and protein synthesis observed with mPer1::luciferase and mPER2::LUC (Fig. 1C and fig. S1, B and C). Damping was reversible over several days and not an artifact of the bioluminescent reporters, because MDL also suppressed mPer1-dependent circadian transcription reported by green fluorescent protein (fig. S1D). Video imaging of mPER2::LUC expression showed that MDL (2.5 μM) rapidly suppressed cellular circadian gene expression to barely detectable levels (Fig. 1D). Prolonged exposure to MDL (1.0 μM) suppressed and desynchronized the transcriptional cycles of SCN cells (Fig. 1E). Pharmacological inactivation of AC therefore mimicked the effect of pertussis toxin (16) and loss of the gene encoding the vasoactive intestinal peptide receptor 2 (Vip2r), an activator of AC within the SCN (17). MDL also reduced cAMP to undetectable levels in NIH 3T3 fibroblast cultures (fig. S2, A and B) and suppressed circadian transcriptional cycling revealed by a Bmal1::luc reporter (3) (fig. S2, C to E). MDL had no effect, however, on luciferase expression from NIH 3T3 cells transfected with a control, noncircadian (CMV, cytomegalovirus) promoter (fig. S2F).

Fig. 1.

Damped and desynchronized cellular circadian gene expression in SCN after suppression of cAMP and CRE rhythms. (A) Circadian oscillation of cAMP concentration (blue) and PER2::LUC bioluminescence (red), as well as cAMP concentration in SCN slices treated with MDL-12,330A (MDL) or with forskolin plus IBMX. [**P < 0.01 versus other samples, by analysis of variance (ANOVA) and post hoc Duncan's multiple-range test.] (B) Circadian oscillation of CRE activity in two representative SCN slices reported by CRE::luciferase adenovirus. (C) Reversible, dose-dependent suppression of SCN PER2::LUC expression by MDL. Arrows indicate medium changes. (D) Effect of MDL on PER2::LUC expression in individual SCN neurons (n = 20), bioluminescence expressed as relative gray-scale units (0 to 255 pixel intensity). Videomicrographs illustrate distribution of PER2::LUC expression before (left) and during (right) treatment with MDL. V, third ventricle. Scale bar, 500 μm. (E) Desynchronization of cellular pacemakers of SCN revealed by (top) raster plots and (bottom) Rayleigh plots of PER2::LUC oscillations of 20 representative SCN neurons before and after addition of MDL. Data are representative of three or more slices.

If cAMP sustains the clock, interference with cAMP effectors should compromise pacemaking. Treatment of brain slices with inhibitors of cAMP-dependent protein kinase had no effect, however, on circadian gene expression in the SCN (fig. S3). cAMP also acts through hyperpolarizing cyclic nucleotide–gated ion (HCN) channels (18) and through the guanine nucleotide–exchange factors Epac1 and Epac2 (Epac, exchange protein directly activated by cAMP) (19). The irreversible HCN channel blocker ZD7288, which would be expected to hyperpolarize the neuronal membrane, dose-dependently damped circadian gene expression in the SCN (Fig. 2, A and B). This is consistent with disruption of transcriptional feedback rhythms by other manipulations that hyperpolarize clock neurons (17, 20, 21). Brefeldin A, applied at a dose that antagonizes Epac but does not affect synaptic transmission (22), also rapidly and chronically suppressed SCN pacemaking (Fig. 2C and fig. S4A). Thus, circadian pacemaking is sustained by cAMP effectors, as well as by AC activity. Direct activation of the effectors might compensate, therefore, for inactivation of AC by MDL. A hydrolysis-resistant Epac agonist [8-(4-chlorophenylthio)-2′-O-methyladenosine-3′,5′-cyclic monophosphorothioate, Sp-isomer, Sp-8-CPT-2′-O-Me-cAMPS] (Fig. 2D and fig. S4B) transiently activated oscillations in transcriptional activity in SCN treated with MDL. An agonist active on both Epac and cAMP-dependent protein kinase (PKA), namely, Sp-8-CPT-cAMPS, also transiently activated circadian gene expression, whereas an agonist specific for PKA (6-Bnz-cAMP) had no effect (fig. S4B). Video imaging showed that Epac agonist synchronously activated circadian gene expression in individual SCN cells (Fig. 2E). The transcriptional cycles induced by Epac agonism damped rapidly in the presence of MDL, however, which demonstrated that, when cAMP concentrations were permanently suppressed, the reactivated transcriptional feedback loops were not self-sustaining.

Fig. 2.

Influence of effectors of cAMP signaling on SCN circadian pacemaking. (A) Effect of HCN channel blocker ZD7288 (arrow) on SCN mPER2::LUC circadian gene expression. (B) Damping of peak bioluminescence in SCN slices treated with vehicle or ZD7288 (Pre, pretreatment; means ± SEM, n ≥ 4). (C) Brefeldin A suppresses circadian gene expression in PER2::LUC SCN. (D) Transient reactivation of MDL-suppressed circadian PER2::LUC expression in SCN slices by Sp-8-CPT-2′-O-Me-cAMPS. (E) Acute activation of cellular circadian gene expression (expressed as relative gray-scale units) by Epac agonist in presence of MDL, illustrated by raster (top) and graphical plots (bottom) of 20 representative cells. (F) Phase shifts of SCN circadian PER2::LUC bioluminescence rhythm by Epac agonist (Sp-8-CPT-2′-O-Me-cAMP, red) but not vehicle (black) (means ± SEM, n ≥ 3 per time point; **P < 0.01 versus vehicle, by ANOVA and Bonferroni test).

Epac can lead to activation of the transcription factor CRE-binding protein (CREB) by phosphorylation (23), and so CRE sequences in Per1 and Per2 are likely points of integration between Epac and the core loop. An Epac agonist acutely triggered CREB phosphorylation (fig. S4C) and CRE::luciferase activity (increase ± SEM: vehicle, 1.9% ± 0.7%; Epac agonist, 38.4% ± 13.0%; n = 4) in SCN slices treated with MDL. If Epac activity were rate-limiting during the normal circadian cycle, acute activation should reset the oscillator, and indeed, a short-acting, hydrolyzable Epac agonist (Sp-8-CPT-2′-O-Me-cAMP) phase-shifted SCN slices (Fig. 2F). As with cAMP agonists (13), treatment of slices with Epac agonist during the circadian day advanced the SCN. The dependence of circadian gene expression on cAMP mediators Epac1 and Epac2 and HCN confirms the necessary contribution of cAMP signaling in sustaining the SCN pacemaker.

If circadian cAMP signaling is an intrinsic part of the pacemaker, feeding back into the transcriptional loops rather than being solely an output, it should determine their temporally specific parameters of phase and period. To test this, we decoupled cAMP concentrations from the transcriptional oscillator by treating SCN slices with forskolin (For), the activator of AC, and the cAMP phosphodiestease inhibitor 3-isobutyl-1-methylxanthine (IBMX). This chronically elevated cAMP levels (Fig. 1A) and acutely increased mPER2::LUC activity (Fig. 3A). Previously asynchronous SCN slices were resynchronized, an effect inconsistent with cAMP signaling acting solely as an output (fig. S5A). Continued exposure of slices to For-IBMX elevated the circadian nadir of mPER2::LUC expression, damping amplitude and definition of the circadian profile of the slice and of individual cells across the SCN (Fig. 3A and fig. S5, B and C). With sustained elevation of cAMP concentrations, the imposed synchrony between free-running slices dissipated (fig. S5D).

Fig. 3.

Alterations in the phase of the SCN oscillator after acute transitions in cAMP concentrations. (A) PER2::LUC bioluminescence rhythms from SCN treated with vehicle or forskolin and IBMX (green arrowhead), followed by washout (blue arrowheads). Dotted lines highlight synchrony of For-IBMX–treated slices, but not control slices, after washout. (B) Phases of PER2::LUC rhythms in individual SCN immediately before (pre-) and 4 days after (post-) washout of vehicle or For-IBMX. Removal of For-IBMX caused resynchronization, driving slices to a common phase, regardless of phase before washout.

After 5 to 7 days of treatment with For-IBMX, we acutely reduced cAMP concentrations by transferring slices to fresh medium. This was done as a “wedge” experiment (24, 25), such that the reduction of cAMP concentrations was imposed at different phases of the ongoing oscillations of different slices. If cAMP signals constitute part of the pacemaker and not solely its output, the enforced decline in cAMP concentrations would set the transcriptional oscillator to a new unique phase. Consequently, the gene expression rhythms of all slices would be synchronized, regardless of their phase before washout. Washout did not synchronize vehicle-treated SCN (Fig. 3B and fig. S5D). In contrast, washout resynchronized SCN previously treated with For-IBMX to a common phase distinct from that of control slices. Note that the extrapolated phase of peak PER2::LUC activity occurred about 39 ± 11 min after the time of washout, which is consistent with the delay of about 1 hour between the circadian minimum of cAMP content and peak PER2::LUC activity in SCN slices (Fig. 1A). Hence, the behavior of the transcriptional loop was determined by acute changes in cAMP signaling, decoupled pharmacologically from that loop. As with protein synthesis (25), these results identify cAMP as a component of the SCN oscillator.

Finally, if cAMP signaling is an integral component of the SCN pacemaker, altering the rate of cAMP synthesis should affect circadian period. 9-(Tetrahydro-2-furyl)-adenine (THFA) is a noncompetitive AC inhibitor (15) that slows the rate of Gs-stimulated cAMP synthesis, which attenuates peak concentrations (fig. S2, A and B). THFA dose-dependently increased the period of circadian pacemaking in the SCN, from 24 to 31 hours (Fig. 4, A and B), with rapid reversal upon washout. Rhythm amplitude decreased at higher concentrations of THFA (fig. S6A). Imaging of individual cells revealed that THFA increased period in neurons across the SCN (fig. S6B). Other noncompetitive inhibitors also lengthened SCN period (fig. S6C). The effect of THFA was additive to that of the Clock mutation (Fig. 4C), which suggests THFA acts in addition to, and independently of, E-box–mediated transactivation by CLOCK and BMAL1. Further, THFA acted additively with inhibition of c-Jun N-terminal kinase (JNK), by generating unusually long periods of 36 hours (fig. S6, D and E). THFA also lengthened the period of circadian oscillators in peripheral tissues from mPER2::LUC mice and fibroblasts transfected with Bmal1::luc reporter (fig. S7, A to D). Note that THFA lengthened the circadian period of wheel-running when delivered continuously and directly to the SCN of mice via intracerebral cannulae (Fig. 4D and fig. S7E). The differential circadian effects of AC inhibitors, damping versus period-lengthening by MDL and THFA, respectively, reflect their particular actions on cAMP kinetics. The current results therefore suggest that noncompetitive inhibitors, such as THFA, might be of therapeutic value in patients with acute (jet lag, shift work) or maintained [familial advanced sleep phase syndrome (26)] acceleration of circadian period.

Fig. 4.

Prolonged SCN circadian period in vitro and in vivo after inhibition of AC by THFA. (A) Effect of THFA on circadian period of representative SCN slices, reported by mPer1::luciferase. (B) Sigmoidal curve fit to one-site inhibition model with 95% confidence limits. Red data point indicates period after washout. (C) Circadian period in SCN from wild-type and Clock mutant mice, before or during treatment with THFA (**P < 0.01 versus pretreatment, by ANOVA and Bonferroni test). All data plotted as means ± SEM, n ≥ 3. (D) Representative double-plotted, wheel-running records of mice treated with (left) vehicle and (right) THFA (delivered to SCN via osmotic minipump). Mice entrained to 12 hours light (shaded) and 12 hours dim red light were released into continuous dim red light. Asterisk indicates day of surgery and commencement of infusion.

We conclude that circadian pacemaking in mammals is sustained, and its canonical properties of amplitude, phase, and period are determined by a reciprocal interplay in which transcriptional and posttranslational feedback loops drive rhythms of cAMP signaling and that dynamic changes in cAMP signaling, in turn, regulate transcriptional cycles. Thus, output from the current cycle constitutes an input into subsequent cycles. The interdependence between nuclear and cytoplasmic oscillator elements we describe for cAMP also occurs in the case of Ca2+ and cADPR (7, 8), which highlights an important newly recognized common logic to circadian pacemaking in widely divergent taxa.

Supporting Online Material

Materials and Methods

Figs. S1 to S7


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article