The mammalian hypothalamic suprachiasmatic nucleus (SCN) acts as a master pacemaker, aligning behavioral and physiological rhythms to light-dark cycles (1
). Initially, the SCN was thought to be the only site of self-sustaining molecular pacemakers in mammals, but subsequent reports have shown such clocks to be ubiquitous (2
). Unlike those in the SCN, clocks in non–light-sensitive organs are entrained by daily feeding (2
), which theoretically allows peripheral tissues to anticipate food consumption and to optimize the timing of metabolic processes. Although nuclear receptors and other pathways have been suggested to provide input (6
), the biochemical basis for non–light-dependent clock entrainment remains unclear.
The adenosine monophosphate (AMP)–activated protein kinase (AMPK) is a central mediator of metabolic signals (9
). Nutrient-regulated diurnal phosphorylation of AMPK substrates in rat livers (10
) makes AMPK an attractive candidate contributor to peripheral clock entrainment. AMPK is a heterotrimeric protein kinase consisting of a catalytic (α) subunit and two regulatory (β, γ) subunits. Activation of AMPK by liver kinase B1 (LKB1)–mediated (11
) or calcium-calmodulin–dependent protein kinase kinase β (CAMKKβ)–mediated (12
) phosphorylation is increased in the presence of high ratios of AMP to ATP (adenosine triphosphate) or elevated intracellular calcium, respectively. Biochemical and bioinformatic studies have established the optimal amino acid sequence context in which phosphorylation by AMPK is likely (13
), which has facilitated prediction of novel substrates.
Mammalian circadian clocks require alternating actions of activators and repressors of transcription (15
). The transcriptional regulators BMAL1 (brain and muscle ARNT-like protein 1) and CLOCK (circadian locomotor output cycles kaput) activate expression of many transcripts including the period (Per1, Per2
, and Per3
) and cryptochrome (Cry1
) genes, whose protein products inhibit CLOCK and BMAL1, which results in rhythmic expression. Posttranslational modification is required for resetting the clock mechanism (15
), including ubiquitination of cryptochromes by F-box and leucine-rich repeat protein 3 (FBXL3) (16
), which leads to their subsequent degradation.
To explore the role of posttranslational modifications in nutrient-dependent signaling to mammalian peripheral clocks, we used mass spectrometry and bioinformatics to identify likely sites of regulated phosphorylation in mouse CRY1 (fig. S1
, A to C). Preliminary biochemical characterization of several observed and predicted sites suggested that modification of serine 71 (S71) and, to a lesser extent, serine 280 (S280) alters the stability of CRY1, and the other mutants examined had less or no effect on stability (fig. S1
). Both CRY1 S71 and S280 conform well to the optimal sequence phosphorylated by AMPK (fig. S2
).Mutation of either S71 or S280 to a nonphosphorylatable amino acid (alanine) was sufficient to stabilize CRY1, whereas mutation of either S71 or S280 to a phosphomimetic amino acid (aspartic acid) was sufficient to destabilize CRY1 (). Mutation of S71, which is conserved in all non–light-sensitive insect cryptochromes (6
) and higher organisms (fig. S2
), to either alanine or to aspartic acid, had a stronger effect than the analogous mutation of S280 ().
Fig. 1 Phosphorylation of S71 or S280 by AMPK destabilizes CRY1. (A) Human embryonic kidney (HEK) 293 cells expressing WT or mutant FLAG-CRY1 were treated with cycloheximide (CHX) as indicated. AA denotes CRY1 with both S71A and S280A mutations. (B) FBXL3-v5 (more ...)
To determine whether the instability of CRY1 phosphomimetic mutants reflects altered interaction with known binding partners, we expressed FLAG-tagged wild-type (WT) or mutant CRY1 with v5-tagged FBXL3 or untagged PER2 and analyzed their interactions by immunoprecipitation of the FLAG-tagged CRY1, followed by immunoblot. Mutation of S71 to aspartic acid (S71D) greatly increased the affinity ofCRY1 for FBXL3, in concert with destabilizing CRY1 in the context of FBXL3 overexpression ( and fig. S3A
), which suggested that phosphorylation of S71 increases the interaction betweenCRY1 and FBXL3. CRY1 harboring an S280D mutation coprecipitated double the amount of FBXL3 relative to WT CRY1. In contrast, CRY1-S71D lost the ability to bind PER2, whereas CRY1-S280D retained PER2 binding, as did the other mutants examined ( and fig. S3B
). Thus, CRY1-S71D exhibited decreased binding to PER2 and increased binding to FBXL3, each of which is expected to destabilize CRY1 and which together likely account for its instability.
Using purified components, we found that AMPK directly phosphorylates CRY1 in vitro (). To examine CRY1 S71 phosphorylation in vivo, we tested several phospho-specific antibodies for the ability to recognize WT but not S71A CRY1 phosphorylated by AMPK. We found that an antibody against acetyl coenzyme A carboxylase 1 (ACC1) phosphorylated on S79 indeed specifically recognized CRY1 phosphorylated on S71 (). To determine whether endogenous AMPK mediates this phosphorylation, we used retroviruses to stably express FLAG-CRY1 in mouse embryonic fibroblasts (MEFs) that were genetically wild-type (WT) or null (ampkα1−/−;ampkα2−/−) for the catalytic subunits of AMPK (AMPK−/−).We detected phosphorylation of purified FLAG-CRY1 after treatment with the AMPK agonist aminoimidazole carboxamide ribonucleotide (AICAR) only in the WT cells, which demonstrated that in vivo phosphorylation of S71 is mediated by AMPK ().
To investigate whether AMPK regulates cryptochrome stability, we generated WT and AMPK−/−
cells stably expressing FLAG-tagged WT or doubly nonphosphorylatable (AA) CRY1 and found that WT, but not AA, CRY1 is degraded after treatment with AICAR ( and figs. S3C and S4
). The regulation of CRY1 stability by AMPK phosphorylation of S71 and S280 was confirmed by subjecting these cells to a time course of cyclo-heximide treatment in the presence or absence of AICAR ( and fig. S5
). AICAR treatment of WT cells reduced the half-life of WT CRY1 from 45 min to less than 30 min, whereas the half-life of AA CRY1 exceeded 2 hours under all conditions. Genetic loss of AMPK stabilized WT CRY1 but did not affect the stability of the non-phosphorylatable AA mutant, which was more stable than WT CRY1 regardless of the presence or absence of AMPK.
Given the importance of feeding-derived signals for circadian clock resetting, the regulation of AMPK by glucose availability, and the destabilization of cryptochromes by AMPK, we examined the effects of AMPK expression and glucose availability on circadian rhythmicity in fibroblasts (6
). When WT fibroblasts were cultured in medium containing either limiting glucose or AICAR, the amplitude of circadian expression of REV-ERBalpha (reverb
α, also known as nuclear receptor subfamily 1, group D, member 1, Nr1d1
) and D-site albumin-binding protein (dbp
) was enhanced ( and fig. S6
), consistent with a model in which glucose deprivation activates AMPK and reduces CRY1 stability, leading to de-repression of CLOCK:BMAL1 targets. Notably, neither glucose deprivation nor AICAR affected the expression of reverb
α and dbp
in MEFs lacking either LKB1 or AMPK( and fig. S6
, B to D),which indicates that these effects of glucose limitation are mediated by AMPK, although other glucose-dependent signaling pathways may also affect circadian clocks (18
Fig. 2 Disruption of AMPK signaling alters circadian rhythms in MEFs. (A) MEFs were synchronized by serum shock and transferred to media containing glucose and AICAR as indicated. Quantitative PCR (QPCR) was performed by using cDNA from samples collected at (more ...)
promoter is repressed by REV-ERBα. Therefore, we examined the effects of reducing glucose availability on circadian rhythms using fibroblasts stably expressing luciferase under the control of a Bmal1
promoter. In a standard (high-glucose) medium, Bmal1-luciferase
exhibited a high-amplitude circadian rhythm with a period of 25.3 hours (, B and C). Decreasing the amount of glucose in the medium increased the circadian period up to 30.7 hours and decreased the amplitude of Bmal1-luciferase
expression (). AICAR treatment had similar effects on circadian period and amplitude (, B to D), which reinforced the idea that the circadian effects of glucose limitation are mediated by AMPK. We also found evidence to suggest that AMPK activation shifts the phase of entrainment in mouse fibroblasts and mouse livers (6
) (figs. S7 and S8
If AMPK-directed cryptochrome phosphorylation regulates the phase of peripheral clocks, the activity, expression, and/or localization of AMPK must be subject to diurnal regulation. Studying mouse liver, we found that the phosphorylation of AMPK substrates Raptor-Ser792 and ACC1-Ser79 was reproducibly higher during the subjective day than at night ( and fig. S9
). We also observed circadian expression of the mRNA encoding the regulatory AMPKβ2 subunit ( and fig. S10
). Because the subunit composition of AMPK complexes regulates its localization (19
), oscillating AMPKβ2 could diurnally regulate the nuclear import of AMPK. Indeed, AMPKα1 exhibited a robust circadian rhythm of nuclear localization ( and fig. S11A
), peaking synchronously with ampkβ2
expression. The time of peak AMPKα1 nuclear localization coincides with minimum nuclear CRY1 (), which suggests that rhythmic nuclear import of AMPK may contribute to the AMPK-mediated phosphorylation and degradation of cryptochromes.
Fig. 3 AMPK is rhythmic and regulates CRY1 stability in mouse livers. (A) IB for phospho-Raptor-S792 (pRaptor) and Raptor in mouse liver lysates. CT, circadian time. (B) QPCR analysis of cDNA prepared from mouse livers. (C) IB for AMPKα1, AMPKα2, (more ...)
To determine whether AMPK affects circadian clocks in vivo, we measured CRY1 protein in liver nuclei from mice injected with either saline or AICAR during the early nighttime hours. AMPK activation reduced endogenous CRY1 by an average of 67% in vivo (). To specifically examine AMPK’s role in the liver, we studied circadian proteins and transcripts in the livers of control mice (LKB1+/+
) or littermates harboring loss of lkb1
in hepatocytes (LKB1L/L
). Liver-specific deletion of lkb1
abolished AMPK activation in that organ (20
) and significantly increased the amount of CRY1 present in liver nuclei across the circadian cycle, particularly during the day-time hours when AMPK was most active in control mice ( and fig. S11B
). This increase was associated with decreased REV-ERBα and decreased amplitude of circadian transcripts throughout the circadian cycle (). Thus, loss of AMPK signaling in vivo stabilizes cryptochromes and disrupts circadian rhythms, consistent with the hypothesis that this pathway contributes to the metabolic control of light-independent peripheral circadian clocks.
In summary, we provide evidence that mammalian cryptochromes, which evolved from blue light photoreceptors, have been repurposed by AMPK to act as chemical energy sensors and can transduce nutrient signals to clocks (fig. S12
). Although our data suggest that AMPK phosphorylation of cryptochromes contributes to metabolic entrainment of peripheral clocks, we expect that the communication of nutritional status to clocks is complex and that additional pathways contribute in vivo (6
). Given that AMPK is a central regulator of metabolic processes, the rhythmic regulation of AMPK has implications for the circadian regulation of metabolism. Genetic alteration of circadian clocks either ubiquitously (21
) or in a tissue-specific manner (23
) elicits dramatic changes in feeding behavior, body weight, running endurance, and glucose homeostasis, each of which is also altered by manipulation of AMPK (25
). The abilities of AMPK to mediate circadian regulation and of CRY1 to function as a chemical energy sensor suggest a close relation between metabolic and circadian rhythms.