JG98

SIRT7 couples light-driven body temperature cues to hepatic circadian phase coherence and gluconeogenesis

The central pacemaker in the hypothalamic suprachiasmatic nucleus (SCN) synchronizes peripheral oscillators to coordinate physiological and behavioural activities throughout the body. How circadian phase coherence between the SCN and the periph- ery is controlled is not well understood. Here, we identify hepatic SIRT7 as an early responsive element to light that ensures circadian phase coherence in the mouse liver. The SCN-driven body temperature (BT) oscillation induces rhythmic expres- sion of HSP70, which promotes SIRT7 ubiquitination and proteasomal degradation. Acute temperature challenge dampens the BT oscillation and causes an advanced liver circadian phase. Further, hepatic SIRT7 deacetylates CRY1, promotes its FBXL3- mediated degradation and regulates the hepatic clock and glucose homeostasis. Loss of Sirt7 in mice leads to an advanced liver circadian phase and rapid entrainment of the hepatic clock upon daytime-restricted feeding. These data identify a BT–HSP70– SIRT7–CRY1 axis that couples the mouse hepatic clock to the central pacemaker and ensures circadian phase coherence and glucose homeostasis.

Physiological and behavioural activities are coordinated by an endogenous biological clock that is synchronized by a central pacemaker in the SCN1,2. At the molecular level, circadian rhythms are sustained by an autoregulatory transcriptional–transla- tional feedback loop, in which CLOCK and BMAL1 heterodimerize to promote Cryptochrome (Cry1/2) and Period (Per1/2) transcription; accumulating CRY and PER proteins dimerize and translocate to the nucleus, where they repress the transcriptional activity of the CLOCK–BMAL1 complex3–9. The CLOCK–BMAL1 complex also activates Rev-erbα/β and retinoic acid receptor-related orphan recep- tor (Ror) transcription; in turn, ROR stimulates whereas REV-ERBα/βsuppresses Bmal1 transcription via ROR elements (RORE)10–12.The endogenous biological clock is entrained by a variety of environmental signals, known as zeitgebers. Light is a predominant zeitgeber that entrains the central clock in the SCN via the retino– hypothalamic tract; the SCN then synchronizes subsidiary oscilla- tors in the periphery13. Although still unclear at the molecular level, increasing evidence suggests that the synchronization is mediated by systemic cues, such as feeding activity, BT and hormones14. In mice, BT declines in the light, resting phase, but increases in the dark, active phase15,16; the daily variation in BT integrates the central clock to the hepatic clock via heat shock factor (HSF1)-mediated stress response17–19. The light–dark cycle also regulates food intake behaviour, thus coupling the central clock to metabolic pro- cesses.

Notably, feeding serves as an independent zeitgeber of the hepatic clock, but not the central clock13,14.NAD+-dependent sirtuins are crucial nutrient sensors that cou- ple metabolism to peripheral clocks. SIRT1 rhythmically deacety- lates BMAL1 and histone H3 at circadian promoters to regulate the expression of clock-controlled genes (CCGs)23. SIRT1 also deacety- lates and thus destabilizes PER2 to modulate circadian clocks24. SIRT6 instead regulates the expression of CCGs rather than core transcriptional–translational feedback loop elements in the liver25. SIRT7 localizes in the nucleolus, where it deacetylates H3K18 and desuccinylates H3K122 to modulate chromatin remodelling, gene transcription and DNA repair26,27. Whereas loss of Sirt7 causes genomic instability and accelerated aging in mice28, a biological function for SIRT7 in circadian rhythms is unclear.All environmental zeitgebers and central and peripheral oscil- lators must be finely aligned to ensure circadian phase coherence and avoid disrupted rhythmicity, which might lead to metabolic diseases and/or accelerated aging. For example, constant light expo- sure alters feeding activity and causes obesity in mice20. Ablating the Clock gene disrupts feeding rhythms and metabolism, and accel- erates aging in mice.

Bmal1 knockout mice are arrhythmic and short lived7,31. Intriguingly, only marginal effects were observed on Per and Cry rhythmicity in the liver when Bmal1 expression was attenuated32,33. Restricted feeding (RF) entrains the hepatic clock more rapidly in mice when the SCN is ablated34. These suggest that the synchronizing cues from the central pacemaker counteract the phase entrainment of liver clock induced by RF, likely via a Bmal1- independent, yet unknown mechanism.Here we present SIRT7 as an early responsive element to light- driven timing cues in the mouse liver. We found that BT oscillations induce rhythmic heat shock protein 70 (Hsp70) transcription, and HSP70 interacts with SIRT7 to promote its ubiquitination and pro- teasomal degradation. SIRT7 deacetylates CRY1 and promotes its degradation mediated by FBXL3, coupling the hepatic clock to the central pacemaker. Hepatic SIRT7 thus regulates circadian glucose homeostasis and hepatic gluconeogenesis. The data highlight a BT–HSP70–SIRT7–CRY1 axis in regulating circadian phase coher- ence and glucose homeostasis.

Results
SIRT7 protein rhythm is clock controlled. NAD+-dependent sir- tuins are essential nutrient sensors that bridge circadian rhythms and metabolic processes in peripheral tissues, such as mouse liver35, whereas their nutrient-independent roles in the circadian clock are limited. We first examined sirtuin oscillations across a light–dark cycle in the mouse liver. We observed prominent BMAL1 and CRY1 rhythmicity, as well as SIRT1, SIRT3, SIRT5, SIRT6 and SIRT7 (Fig. 1a,b). Of interest, SIRT7 and CRY1 protein levels were tightly correlated across the light–dark cycle: SIRT7 accumulated in the daytime but dropped at night (around zeitgeber time (ZT) 12), whereas CRY1 exhibited the inverse pattern. Notably, Sirt7 messenger RNA levels in mouse liver lacked an obvious oscil- lation (Extended Data Fig. 1a), and Sirt7 mRNA and SIRT7 protein levels were almost undetectable in the hypothalamus (Extended Data Fig. 1b,c). Light represents the most prominent zeitgeber of the circadian clock. To further decipher the light–dark effect on SIRT7 protein expression, we examined the level of hepatic SIRT7 at ZT4 (day) and ZT16 (night) with or without a 2-h light pulse from ZT14 to ZT16 in the dark period in mice fed ad libitum. We observed a dra- matic reduction in Sirt7 expression at ZT16 compared with ZT4. Light exposure in the dark period elevated the level of SIRT7 but decreased that of CRY1 (Fig. 1c,d), suggesting that the light–dark cycle modulates SIRT7 levels. To examine whether SIRT7 rhyth- micity is clock controlled or induced by physiological changes upon light exposure, we switched the mice from a light–dark to a dark– dark condition. Consistent with the self-sustained property of the circadian clock, SIRT7 and CRY1 levels oscillated in a circadian manner in the dark–dark condition (Fig. 1e,f). These data suggest that SIRT7 protein oscillation is clock controlled.

The peripheral clock is sensitive to feeding signals controlled by the SCN13. Given the lack of SIRT7 in the hypothalamus, we rea- soned that light might regulate hepatic SIRT7 via systemic cues, such as a feeding–fasting cycle. To test this possibility, we examined SIRT7 levels at ZT4 and ZT16 under feeding or fasting conditions (no food available from ZT0 to ZT16), with a normal light–dark cycle or a 2-h light exposure from ZT14 to ZT16. As shown, whereas fasting had little effect on hepatic SIRT7 levels in the dark period (ZT16NF versus ZT16), SIRT7 levels markedly increased upon a 2-h light exposure (ZT16NF + light exposure versus ZT16NF) (Fig. 1g,h). This suggests that light resets the SIRT7 expression pat- tern, independent of a feeding–fasting cycle. In line with rhythmic SIRT7 expression, the level of H3K18ac, a direct target of SIRT7 deacetylase activity27, exhibited a totally inversed oscillating pattern. By contrast, the protein levels of SIRT1, SIRT6 and their deacety- lating target H3K9ac minimally changed between ZT4 and ZT16, regardless of feeding, fasting or light exposure. Although SIRT3, SIRT5 and BMAL1 oscillations were obvious between ZT4 and ZT16 in mice fed ad libitum, little change was observed during light exposure with or without fasting. Notably, the mRNA levels of Sirt7, Cry1, Bmal1 and Per2 in livers were hardly affected by light expo- sure in the dark period (Extended Data Fig. 1d). Collectively, these data indicate that the rhythmicity of hepatic SIRT7 is entrained independent of feeding and fasting.

Light-entrained BT oscillations regulate SIRT7 rhythmicity in the mouse liver. BT is a pivotal systemic entrainment cue used by the SCN to synchronize peripheral clocks14,19. Murine BT can be elevated when environmental temperatures exceed the ther- moneutral zone (30 °C)36. Altering ambient temperature (AT) imparts changes in circadian gene expression in peripheral organs18. Because BT oscillations were maintained during dark–dark condi- tions (Extended Data Fig. 2a), we asked whether SIRT7 oscillation is regulated by BT. We subjected mice to a high AT (32 °C) at dif- ferent phases of the circadian cycle and measured BT. As expected, BT was high at ZT16 and low at ZT4, when mice were maintained at room temperature (Fig. 1i). High AT increased BT at ZT4, but not at ZT16, which correlates with down-regulated SIRT7 expression but increased CRY1 expression at ZT4 (Fig. 1j,k). By contrast, the cold exposure (4 °C) at ZT16 decreased BT (Fig. 1l) and increased SIRT7, but down-regulated CRY1 expression (Fig. 1m,n).

Next, we examined whether light exposure at night stabilizes SIRT7 expression via BT. BT declined upon light exposure, but this was blocked by a concomitant high AT (Fig. 1o). Importantly, the light-induced increase in SIRT7 level at ZT16 was abolished by con- comitant high AT challenge (Fig. 1p,q). Given that BT rhythmic- ity is regulated by feeding and locomotor activity14, we examined the effects of fasting on BT oscillation. Indeed, the BT exhibited only a slight decline during fasting, but a dramatic decrease upon light exposure at ZT16 in fasted mice (Extended Data Fig. 2b). Furthermore, we examined SIRT7 protein levels in fasted and refed mice after 24-h fasting starting from ZT12. No significant change in SIRT7 protein levels was observed in mouse liver tissues by fast- ing or refeeding (Extended Data Fig. 2d). As a positive control, the phosphorylated AKT (p-AKT) S473 level was reduced by fasting, and refeeding increased the p-AKT S473 level to a similar extent to the feeding group. Notably, the BT was dramatically reduced during the light phase by 24-h fasting (Extended Data Fig. 2e). These find- ings support the notion that a light-driven BT oscillation regulates the rhythmicity of SIRT7 expression that is independent of a feed- ing–fasting cycle.

HSP70 mediates SIRT7 degradation in the mouse liver. BT resets peripheral clocks via transcription factor HSF1, which drives the oscillation of HSPs, including HSP70, HSPCA, HSP105 and HSPA8 (refs. 17,32). We found that the mRNA levels of Hsps were signifi- cantly enhanced in the liver by high AT challenge at ZT4 (Extended Data Fig. 2g). HSP70 is a key chaperone molecule that regulates pro- tein homeostasis37. Its protein level (low at ZT4 but high at ZT16) was negatively correlated with SIRT7 level (high at ZT4 but low at ZT16). We asked whether BT regulates SIRT7 rhythmicity via HSP70. To answer this question, we did co-immunoprecipitation (co-IP) experiments to determine any potential interaction between SIRT7 and HSP70, and found that SIRT7 interacts with HSP70 (Fig. 2a,b). Further degradation assay revealed that overexpression of HSP70 accelerated the degradation of hemagglutinin (HA)-SIRT7 (Fig. 2c,d), which was blocked by the proteasome inhibitor MG132 (Fig. 2e,f). In addition, ectopic HSP70 enhanced HA-SIRT7 polyubiquitination (Fig. 2g). Hsp70 knockdown by small interfering RNA suppressed HA-SIRT7 degradation (Fig. 2h,i) and down-regulated HA-SIRT7 polyubiquitination (Fig. 2j). These indicate that HSP70 promotes SIRT7 degradation via the ubiquitin–proteasome system. To ascertain the function of HSP70, we immunoprecipitated endogenous SIRT7 from liver lysates across a circadian cycle.

A rhythmic interaction between HSP70 and SIRT7 was observed, with maximum binding at ZT12 and ZT16, consistent with high HSP70 levels but low SIRT7 in the dark period (Fig. 2k). We next tested whether HSP70–SIRT7 interaction is modulated by BT via examin- ing the interaction at different ATs. As shown, whereas a high AT increased HSP70 binding to SIRT7 at ZT4, a low AT reduced this interaction at ZT16 (Fig. 2l) in a. c, Representative immunoblots showing SIRT7 and CRY1 protein levels in mouse livers from mice maintained under normal feeding conditions with or without a 2-h light pulse from ZT14 to ZT16. n = 3 per time point. d, Quantification of band intensities of three independent blots shown in c, unpaired two- tailed Student’s t-test. e, Representative immunoblots showing SIRT7 and CRY1 levels in the livers of mice maintained under a 12 h:12 h light–dark or dark–dark cycle. The dark–dark group was placed in darkness for 24 h, and liver tissues were collected every 4 h for 24 h. n = 3 per time point. f, Quantification of band intensities of three independent blots shown in e. g, Representative immunoblots showing the indicated protein levels under normal feeding or fasting (no food available from ZT0) conditions with or without 2-h light exposure.

Mice were fasted from ZT0 to ZT16 under a normal light–dark cycle or exposed to a 2-h light pulse from ZT14 to ZT16. n = 3 per time point. h, Quantification of band intensities of three independent blots shown in g, unpaired two-tailed Student’s t-test. i, Mouse rectal temperature under room or high AT at ZT4 and ZT16. n = 6 per time point, unpaired two-tailed Student’s t-test. j, Representative immunoblots showing SIRT7 and CRY1 protein levels under room temperature or high AT conditions at ZT4 and ZT16. n = 3 per time point. k, Quantification of band intensities of three independent blots shown in j, unpaired two-tailed Student’s t-test. l, Mouse rectal temperature under room or cold AT at ZT16 n = 6 per time point, unpaired two-tailed Student’s t-test. m, Representative immunoblots showing protein levels of SIRT7 and CRY1 at ZT16 under room temperature or cold AT conditions at ZT16. n = 3 per time point. n, Quantification of band intensities of three independent blots shown in m, unpaired two- tailed Student’s t-test. o, Mouse rectal temperature with 2-h light exposure under room or high AT at ZT16. n = 6 per time point, unpaired two-tailed Student’s t-test. p, Representative immunoblots showing protein levels of SIRT7 and CRY1 in mouse liver with 2-h light exposure, when maintained at room temperature or at high AT conditions at ZT16. n = 3 per time point. q, Quantification of band intensities of three independent blots shown in p, unpaired two-tailed Student’s t-test. Data represent the means ± s.e.m. of three independent experiments. C, cold; DD, dark–dark; H, high temperature; L, light exposure; LD, light–dark; NF, no food available from ZT0; R, room temperature; temp, temperature. Molecular weight (kDa) is indicated on the right side of the immnoblots.

We then examined whether HSP70 is required for BT-generated SIRT7 rhythmicity. As shown, a 2-h heat shock (39.5 °C) led to reduced SIRT7 expression but elevated HSP70 and CRY1 expres- sion in mouse embryonic fibroblasts (MEFs) (Fig. 2m,n). Hsp70 knockdown by small interfering RNA (siRNA) in MEFs elevated SIRT7 but down-regulated CRY1 protein levels. The effects of heat shock on SIRT7 and CRY1 levels were totally blocked in the case of Hsp70 knockdown, suggesting that a high BT relies on HSP70 to down-regulate SIRT7 protein levels.
Next, we investigated whether BT regulates CRY1 via SIRT7, using a Sirt7−/− mouse line generated by the CRISPR–Cas9 pro- cedure. We first noted that loss of Sirt7 significantly up-regulated CRY1 protein level in MEFs (Fig. 2o,p). Interestingly, a 2-h heat shock up-regulated HSP70 expression in both wild-type (WT) and Sirt7−/− MEFs, but the increase in CRY1 expression was observed only in WT. We next applied temperature challenges to Sirt7−/− and WT mice. Again, an increase in CRY1 expression in Sirt7−/− livers was observed (Extended Data Fig. 3a,c). Consistent with the in vitro data, Sirt7 depletion completely abolished high AT challenge- induced CRY1 up-regulation at ZT4 in mouse livers (Extended Data Fig. 3a). Similarly, the cold treatment-induced CRY1 down- regulation was not observed in Sirt7−/− livers (Extended Data Fig. 3c). HSP70 expression level was comparable in WT and Sirt7−/− livers (Extended Data Fig. 3a,c). Together, these data suggest that BT regulates SIRT7 protein stability through HSP70.

SIRT7 directly binds and deacetylates CRY1. We repeatedly observed that the CRY1 protein level is inversely correlated with SIRT7, peaking in the night but dropping in the daytime, and that Sirt7 deficiency led to a dramatic increase in CRY1 protein in day- time. We speculated that SIRT7 might directly regulate CRY1. We first examined whether SIRT7 interacts with CRY1. Co-IP revealed that SIRT7 interacted with CRY1 (Fig. 3a,b). Endogenous CRY1 was found in anti-SIRT7 immunoprecipitates (Fig. 3c). Finally, a glutathione S-transferase (GST) pull-down assay supported a direct interaction between His-CRY1 and GST-SIRT7 (Fig. 3d). As a protein deacetylase38, SIRT7 might deacetylate CRY1. As shown, HA-CRY1 acetylation levels were remarkably reduced in the presence of SIRT7 but were unaffected when cotransfected with catalytically inactive SIRT7 (H187Y) (Fig. 3e,f). Further, FLAG-CRY1 deacetylation was abolished in cells treated with nicotinamide (NAM), a pan sirtuin inhibitor (Fig. 3g,h). Thus, we conclude that SIRT7 deacetylase activity is required for CRY1 deacetylation. We also found that the CRY1 acetylation level was increased in SIRT7 KO HEK293 cells (Fig. 3i,j). To test whether SIRT7 directly targets CRY1, we did an in vitro deacetylation assay. As shown, the acetylation level of FLAG-CRY1 was decreased in the presence of GST-SIRT7 and NAD+, and the decrease was blocked by NAM (Fig. 3k,l). Although CRY2 is a close homologue of CRY1, its acetylation level was hardly affected by NAM (Extended Data Fig. 4a). We further examined whether other sirtuins affected acety- lation of CRY1. As shown, only SIRT7 inhibited CRY1 acetylation (Extended Data Fig. 4b). The data suggest that SIRT7 specifically deacetylates CRY1.

SIRT7 deacetylates CRY1 at K565/579. To identify the specific residues on CRY1 targeted for SIRT7-mediated deacetylation, we performed mass spectrometry. Three lysine residues were found differentially acetylated, that is, K22/565/579 (Supplementary Fig. 1). We generated nonacetylatable CRY1 mutants by replac- ing each lysine K22, K565 or K579 in turn with arginine. Each mutant (FLAG-CRY1-KR (lysine (K) residues mutated to arginine (R))) exhibited a comparable baseline acetylation level with WT (Fig. 3m,n). The effect of NAM on the various FLAG-CRY1-KR mutants was then examined. As shown, the NAM treatment enhanced the acetylation levels of WT and K22R but had a minimal effect on K565R and only a moderate effect on K579R (Fig. 3m,n), indicating that K565 and K579 are targeted by SIRT7 for deacety- lation. We then generated FLAG-CRY1-2KR, wherein both K565 and K579 were mutated to R565 and R579, respectively. FLAG- CRY1-2KR displayed reduced acetylation levels compared with WT, and ectopic SIRT7 was unable to further down-regulate the acety- lation level (Fig. 3o,p). In addition, the acetylation level of FLAG- CRY1-2KR did not increase in SIRT7 KO cells (Fig. 3q,r) or cells treated with NAM (Extended Data Fig. 5), indicating that SIRT7 deacetylates CRY1 predominantly at K565 and K579.

SIRT7-mediated deacetylation destabilizes CRY1. In unsynchro- nized SIRT7 KO HEK293 cells, the mRNA level of CRY1 was slightly decreased, whereas the protein level was significantly increased (Extended Data Fig. 6a–e). By contrast, CRY1 protein level was decreased in SIRT7-overexpressing cells (Extended Data Fig. 6d,f). These findings indicate that SIRT7 most likely regulates CRY1 protein stability. Indeed, the FLAG-CRY1 degradation rate was accelerated in the presence of HA-SIRT7 compared with control (Fig. 4a,b), whereas the degradation of mutant CRY1 (FLAG-CRY1- 2KR) was inhibited (Fig. 4c,d). Similar findings were observed for the endogenous CRY1 (Extended Data Fig. 7a–d). Moreover, the CRY1 degradation rate was not affected in the presence of cata- lytically inactive mutant SIRT7 (H187Y) (Extended Data Fig. 7e). As such, we propose that SIRT7 regulates CRY1 protein stability. Because CRY1 degradation was blocked by MG132, we reasoned that SIRT7 regulates CRY1 degradation via the ubiquitination–pro- teasome pathway. Indeed, ectopic expression of SIRT7 promoted polyubiquitination of CRY1 (Fig. 4e). Lysine acetylation enhances protein stability by blocking ubiquitination on the same residue39. K565/579 were previously identified as target ubiquitination resi- dues of F-box and leucine-rich repeat protein 3 (FBXL3), a subunit of ubiquitin protein ligase complex (SKP1–cullin–F-box)40–42. We speculated that SIRT7 might deacetylate CRY1 K565/579 to facili- tate the ubiquitination of CRY1 by FBXL3 and subsequent degra- dation. Indeed, the ubiquitination level of FLAG-CRY1-2KR was dramatically reduced compared with WT but remained unchanged upon SIRT7 overexpression (Fig. 4e).

To determine whether acetylation stabilizes CRY1, we investi- gated the ubiquitination and degradation rate of CRY1 in SIRT7 KO HEK293 cells. As shown, the degradation of FLAG-CRY1 was significantly suppressed in SIRT7 KO cells compared with control (Fig. 4f,g). By contrast, the degradation rate of FLAG-CRY1-2KR was comparable between SIRT7 KO and control cells. Moreover, the polyubiquitination level of CRY1 was reduced in SIRT7 KO cells (Fig. 4h). We asked whether the increased acetylation pro- tects CRY1 from FBXL3-mediated degradation in SIRT7 KO cells. As expected, FLAG-FBXL3 accelerated HA-CRY1 degradation, which was significantly attenuated in SIRT7 KO cells (Fig. 4i,j). We further examined FBXL3-mediated CRY1 polyubiquitination in SIRT7 KO cells. Although the polyubiquitination level of CRY1 was slightly increased upon FBXL3 overexpression in SIRT7 KO cells, which is likely attributable to other lysine residues, the overall ubiq- uitination level of CRY1 was significantly lower in SIRT7 KO cells compared with controls (Fig. 4k). These results demonstrate that increased CRY1 protein stability in SIRT7 KO cells is due to elevated K565/579 acetylation, which prevents FBXL3-mediated ubiquitina- tion and subsequent proteasomal degradation. Finally, because adenosine monophosphate–activated protein kinase (AMPK) also regulates the hepatic clock via phosphoryla- tion-induced CRY1 degradation43, we examined whether K565/579 deacetylation affects AMPK-mediated CRY1 degradation. As shown, the glucose starvation activated AMPK, which led to reduced protein levels of FLAG-CRY1 and FLAG-CRY1-2KR to a similar extent (Supplementary Fig. 2), indicating independent func- tion of SIRT7 and AMPK in the regulation of CRY1 stability.

SIRT7 regulates the circadian phase of hepatic clocks. The data thus far suggest an essential role for SIRT7 in hepatic clocks, as confirmed by that deletion of Sirt7 led to a dramatic increase in CRY1 expres- sion at ZT0 and ZT6 in mouse livers (Fig. 5a,b). We confirmed that hepatic CRY1 was acetylated in a cyclic manner, peaking at ZT12– ZT18 in WT mice; this oscillation was, however, disrupted in Sirt7−/− mice with elevated acetylation at ZT0–ZT6 (Fig. 5c,d). We asked whether SIRT7 regulates core clocks in a cell-autonomous manner. Indeed, a reduced amplitude of Bmal1, Cry1, Dbp, Rev-erbα and Rev- erbβ mRNA levels was observed in Sirt7−/− MEFs compared with WT (Extended Data Fig. 8a). Moreover, the protein level of CRY1 was significantly elevated, whereas that of BMAL1 was reduced in Sirt7−/− cells (Extended Data Fig. 8b,c). The circadian phase of Per2 was sub- stantially delayed in Sirt7−/− MEFs. These data implicate that SIRT7 regulates the expression of endogenous core clock proteins.Next, we examined the mRNA levels of core clock genes inSirt7−/− livers. As shown, Sirt7 deficiency slightly down-regulatedthe mRNA levels of Bmal1, Rev-erbβ and Dbp, but not Per2 or Rev-erbα (Fig. 5e). Although the amplitude of the Cry1 daily oscil- lation was virtually unchanged, its mRNA level was reduced in Sirt7−/− livers specifically in the light phase. Of particular interest, the circadian phases of Bmal1, Cry1, Rev-erbβ and Dbp were sub- stantially advanced, whereas that of Per2 was delayed in Sirt7−/− livers compared with WT.

No circadian phase shift in Bmal1, Cry1, Dbp or Per2 was observed in the hypothalamus of Sirt7−/− mice (Extended Data Fig. 9). These data suggest that SIRT7 regulates the circadian phase of hepatic clocks.Acute AT changes the circadian phase of the hepatic clock. Our data so far suggest a BT–HSP70–SIRT7–CRY1 axis in the regula- tion of the hepatic clock. Acute AT challenge affected SIRT7 and CRY1 levels in mouse livers. We next examined whether various AT conditions modulate the hepatic clock. High-temperature treatment increased the protein level of HSP70 but decreased that of SIRT7; bycontrast, the cold temperature treatment elicited the opposite effect (Fig. 6a–c). Both high and low temperature dampened the oscillation of BT and HSP70 and SIRT7 protein levels. Interestingly, consistent with a low level of SIRT7, high-temperature treatment led to accumulation of CRY1 at ZT5 and ZT9, which is quite similar to that in Sirt7−/− liver. By contrast, CRY1 level was reduced at ZT17 at cold temperature compared with room temperature, probably because of a constitutively high SIRT7 level in the liver. Likely attributable to feedback repression of itself, Cry1 mRNA level was increased at cold temperature but reduced at high temperature (Fig. 6d). Whereas cold temperature increased Bmal1 gene expres- sion, both protein and mRNA levels of BMAL1 were reduced at high temperature. The protein level of PER2 was reduced at ZT21 at high temperature, but Per2 mRNA level was significantly reduced at ZT13 and ZT17. By contrast, cold temperature treatment led to increased expression of Per2 at ZT5 and ZT9 but hardly affected the protein. Again, similar to Sirt7 deficiency, both cold and high temperature attenuated the oscillation of Dbp and Rev-erbβ. The circadian phases of Bmal1, Cry1, Per2, Rev-erbβ and Dbp were all advanced upon both high and cold treatment. These results indicate that acute temperature challenge modulates the circadian phase of the liver clock.

We further examined the effect of long-term acclimation to AT challenges. The mice were acclimated at different ATs for 1 week. High-temperature treatment only slightly increased the BT at light phase and attenuated its oscillation (Extended Data Fig. 10a). Cold temperature treatment decreased the BT and also slightly attenuated the oscillation. The protein levels and circadian phases of HSP70 and SIRT7 were accordingly changed between high–cold and room temperature (Extended Data Fig. 10b,c), but less obvious compared with that in the acute treatment (see Fig. 6b,c). Notably, low AT increased protein level of SIRT7 but decreased that of HSP70 at ZT13 and ZT17 compared with room temperature. The expression levels and circadian phases of Bmal1, Per2 and Rev-erbβ mRNAs were merely changed at different conditions (Extended Data Fig. 10d). High temperature delayed the circadian phase of Rev- erbα and increased the expression level of Dbp. Although circadian phases of Cry1 mRNA were advanced at high and cold temperature compared with that at room temperature, the circadian phase of CRY1 protein was delayed (Extended Data Fig. 10b–d). We noticed that little difference in circadian locomotor activity was observed at different conditions (Extended Data Fig. 10e,f).SIRT7 counteracts with hepatic clock phase entrainment by RF. The circadian phase of hepatic clocks is coordinated by systemic cues from the central pacemaker and feeding activity. The SCN- orientated synchronizing signal counteracts feeding-induced phase entrainment of hepatic clocks.

Given the critical role for SIRT7 in the SCN-driven synchronization of hepatic clocks and advanced circadian phase of core clock genes in Sirt7−/− livers, we reasoned that SIRT7-mediated synchronizing cues from the central pace- maker might counteract the feeding cues in the circadian phase entrainment of hepatic clocks. To test this hypothesis, we subjected Sirt7−/− and WT control mice to daytime RF (Fig. 7a). After 2 d of RF (RF2), the circadian phases of Bmal1, Cry1, Rev-erbα, Rev-erbβ and Dbp were slightly changed in WT mice. Intriguingly, the circadian phases adapted to RF more rapidly in Sirt7−/− mice. For instance, Bmal1, Cry1 and Rev-erbα levels showed a circadian phase advance of up to 8 h in Sirt7−/− mice, whereas such a phase shift was <4 h in mice fed ad libitum. Although two peaks of Per2 expression were observed at ZT4 and ZT16 in WT mice, a new phase of Per2 already appeared after RF2 in Sirt7−/− mice whereby its expression peaked at ZT4. At RF4, the circadian phase of all examined clock genes had completely reversed, and a difference was hardly observed between Sirt7−/− and WT mice. This finding suggests that SIRT7 counteracts the phase entrainment of hepatic clocks by RF.The reversal of phase of core clock genes in WT mice after RF4 (Fig. 7a) prompted us to investigate whether the BT–HSP70–SIRT7 axis is involved in resetting the phase of core clocks during RF. We found that BT oscillations were still maintained after RF2, but the level was significantly increased at ZT20 compared with feeding ad libitum (Fig. 7b). However, BT oscillations were completely abolished after 4 d, with a constitutive high BT. We measured HSP70, SIRT7 and CRY1 protein levels. Consistent with changes in BT, after RF2, the HSP70 and SIRT7 levels were altered at ZT20, and the cir- cadian phase of CRY1 was slightly delayed (Fig. 7c,d). After RF4, the HSP70 and SIRT7 protein levels lost rhythmicity and became constitutively high or low, respectively, across a circadian cycle. Meanwhile, CRY1 level was high after RF4 compared with feeding ad libitum, but the oscillation was attenuated. We further examined BT and HSP70 levels in Sirt7−/− mice during RF. As shown, Sirt7 deficiency had little effect on BT and HSP70 levels (Fig. 7f–i and Supplementary Fig. 3), suggesting a downstream role of SIRT7 in the BT–HSP70–SIRT7 axis. The difference in CRY1 level between Sirt7−/− and WT mice was prominent in the light period after 2 d, but this effect became negligible after RF4. These data imply that the BT–HSP70–SIRT7 axis counteracts the circadian phase reset- ting induced by RF.SIRT7 regulates rhythmic hepatic gluconeogenesis. Although the phase of hepatic clock genes was reversed after RF4, as compared with ad libitum (Ad), the rhythmicity of the BT–HSP70–SIRT7 axis was completely abolished at RF4. RF can down-regulate blood glucose44, and CRY1 regulates circadian gluconeogenesis45,46. The constant high level of CRY1 expression induced by RF and Sirt7 deficiency prompted us to investigate whether SIRT7 and CRY1 cooperate to regulate glucose homeostasis. To this end, we first examined blood glucose levels in mice fed ad libitum. The blood glucose level oscillated, with a trough at ZT1 and peak at ZT13 in WT mice (Fig. 8a). By contrast, attenuated blood glucose oscillations were observed in Sirt7−/− mice, and glucose levels were significantly reduced at ZT5 and ZT9. Notably, the overall blood glucose level was attenuated in Sirt7−/− mice. The blood glucose oscillation was also attenuated in WT mice at RF2, with a moderate glucose level maintained across the whole day. By contrast, the oscillation pattern of blood glucose levels was almost reversed in Sirt7−/− mice. Notably, the blood glucose level was low from ZT5 to ZT13 in Sirt7−/− mice compared with WT. The oscillation pattern in glucose levels was totally reversed in both genotypes at RF4, peaking at ZT1. At this stage, little difference was observed in the oscillation pattern and in the overall blood glucose levels between WT and Sirt7−/− mice, implicating an important role for SIRT7 in regulating RF-induced blood glucose reduction.A previous study indicates that loss of Sirt7 improved insu- lin resistance and suppressed blood glucose levels during high-fat diet (HFD) conditions47. We thus performed glucose tolerance test (GTT) and pyruvate tolerance test (PTT) in WT and Sirt7−/− mice with normal feeding conditions (Fig. 8b). The GTT data indi- cated better glucose tolerance in Sirt7−/− mice compared with WT, and the PTT suggested inhibition of gluconeogenesis in Sirt7−/− mice. Rhythmic hepatic gluconeogenesis is controlled by crypto- chromes45,46, partially by inhibiting the glucagon–CREB pathway. We examined phosphorylated CREB (p-CREB) during RF. The level of p-CREB peaked at ZT4 and ZT8 in the Ad scenario (Fig. 8c,d). At RF2 and RF4, the peak levels of p-CREB were observed at ZT0 and ZT20, suggesting that the glucagon–CREB pathway was rapidly reversed during RF. Meanwhile, the level of p-CREB was compa- rable between WT and Sirt7−/− mice. In addition, serum insulin and glucagon levels and food intake were not much changed in Sirt7−/− mice compared with WT (Fig. 8e–g), suggesting that reduced blood glucose levels in Sirt7−/− mice were most likely attributable to sup- pressed hepatic gluconeogenesis.atissues22,51–53. Although the molecular mechanisms by which feeding regulates local clock have been demonstrated2, how feeding defeats the SCN-derived signals during RF is still largely unknown. Here, we revealed SIRT7 as an early responsive element to light, transmit- ting timing information to the mouse liver. Like the SCN-lesioned animals, the kinetics of RF-mediated phase shifting is accelerated in Sirt7−/− livers. In mice fed ad libitum, Sirt7 depletion shifts the circa- dian phase of the liver, but not of the hypothalamus, in which Sirt7 is hardly detected. Previous studies have shown that Bmal1 and Per2 are differentially controlled by the SCN and food-derived resetting cues34,51, respectively. We found that SIRT7 in the liver is stabi- lized by light but is independent of BMAL1 and PER2, implicating SIRT7 as an early responsive element to light in the liver. SIRT7 was increased at ZT16 upon light exposure in fasted mice, supporting an essential role for the light–dark cycle rather than food availability in driving rhythmic SIRT7 expression.BT is a common resetting cue used by the SCN to entrain periph- eral clocks17–19,32. HSF1 is essential in resetting peripheral clocks via BT, and Hsps transcription is driven by HSF1 in a circadian man- ner17,32. At the molecular level, however, the connection between BT and local clock components is still largely unknown. Our results indicate that HSP70, a key molecular chaperone, interacts and pro- motes SIRT7 ubiquitination and proteasomal degradation. SIRT7 deacetylates CRY1 at K565/579 residues, two of ten cryptochrome lysines targeted for FBXL3-mediated ubiquitination and degrada- tion. A constitutively high level of CRY1 may suppress its own transcription in Sirt7−/− mice, specifically in the light phase of a circadian cycle. Similar to Sirt7-deficient mice, a delayed circadian phase of Per2 was observed in Fbxl3−/− livers54,55, supporting the notion that the disrupted circadian phase phenotype of Sirt7−/− mice is attributable, at least partially, to dysregulated CRY1. Thus, the BT–HSP70–SIRT7–CRY1 axis integrates systemic BT to the hepatic clock. Consistently, acute temperature challenge substantiallyinterferes with the oscillation and circadian phase of core clocks in liver. Cold temperature treatment enhanced oscillation of Bmal1, Cry1 and Per2. By contrast, the amplitude of Rev-erbα/β and Dbp was reduced, suggesting that clock genes are differentially regulated by BT. Moreover, the cold temperature treatment led to increased mRNA levels of Per2 at ZT4 and ZT8, but change at protein levels was merely observed. We speculate that in addition to transcrip- tional regulation, hepatic PER2 is predominantly controlled at posttranscriptional and posttranslational levels. High temperature led to dampened oscillation of SIRT7 with constitutively reduced protein levels across the circadian cycle. Similar to Sirt7-KO mice, the advanced circadian phases of Bmal1, Cry1, Dbp and Rev-erbβ were observed at high AT compared with room AT. Interestingly, long-term acclimation of AT challenge elicits a quite different effect on hepatic clock compared with that of acute challenge, which may be because of rescued oscillation of BT and HSP70 and SIRT7 protein levels. It is plausible to speculate that the central circadian system might engage metabolic reprogramming and/or tempera- ture compensation to adapt AT changes.Mammalian cryptochrome shares a conserved photolyase homology at the amino terminus, but a divergent carboxyl tail domain (CTD)56–58. Although it is generally accepted that CTD is crucial for cryptochrome functional diversity57,59, the molecu- lar mechanisms are not well understood. Our results demonstrate that CRY1, but not CRY2, is deacetylated by SIRT7, because the K565/579 residues are present only in CRY1. The data shed light on the functional importance of the CTD for CRY1 stability mediated by SIRT7 deacetylase. In addition, CRY1 protein stability is also reg- ulated by AMPK-mediated phosphorylation43. Our data implicate that SIRT7 and AMPK regulate CRY1 protein stability via indepen- dent pathways, involving ubiquitination at different residues.The SCN also indirectly regulates daily rhythms of BT via feed- ing and/or activity14. Under normal feeding conditions, elevated BT in the dark phase is attributable to increased activity and diet- induced thermogenesis60. We demonstrated that BT is not strongly changed in the dark phase in fasted mice (Extended Data Fig. 2). Interestingly, a previous study found that BT was not affected in the dark phase after food deprivation, likely because of maintained activity in the dark phase61. During prolonged feeding in the day- time, food-entrainable oscillator overcomes the effects of the SCN and acts as the master pacemaker14,60. Food-entrainable oscillator- controlled food-anticipatory activity leads to an increase of BT preceding the feeding time60. Meanwhile, feeding signals regulate diet-induced thermogenesis in the ventromedial hypothalamus and contribute to the constitutive high BT during daytime. It explains why the SCN-controlled BT oscillation was blocked at RF4, suggesting that BT-mediated synchronization of peripheral tis- sues is lost.Disrupted circadian rhythms lead to metabolic disturbances30,65. Long-term RF leads to metabolic disturbance with reduction of blood insulin and increase of glucagon and free fatty acid44,53, which can be rescued by glucose administration44,53. The circadian control of gluconeogenesis is complicated: multiple pathways are involved, including local clock components (cryptochrome), feeding signals (insulin and glucagon), the SCN-controlled corticosterone and autophagy45,46,66. Our data indicate that RF leads to constitutively high BT and HSP70, which disrupt SIRT7 oscillation and subse- quently lead to an increase in CRY1 level. CRY1 regulates hepatic gluconeogenesis by different pathways, that is, glucagon–CREB and glucocorticoid pathways. Cytoplasmic CRY1 inhibits the gluca- gon–CREB pathway during feeding time (ZT13)46, whereas nuclear CRY1 interacts with the glucocorticoid receptor (GR) and inhib- its gluconeogenesis45. Our results suggest that the glucagon–CREB pathway less likely contributes to attenuated gluconeogenesis by Sirt7 deficiency. Because SIRT7 is predominantly localized in the nucleus67,68, we reason that other pathways might be involved.The role of SIRT7 in hepatic glucose metabolism is seemingly controversial47,69. Loss of Sirt7 protects from HFD-induced hyper- glycaemia in mice. Consistently, our data revealed suppressed glu- coneogenesis and improved glucose tolerance in Sirt7−/− and LS7KO mice. Interestingly, better glucose and insulin tolerance have been reported in Sirt7−/− mice fed an HFD, which is most likely attrib- utable to an increased glucose disposal rate47. The mechanisms of HFD-induced hyperglycaemia and insulin resistance are compli- cated. SIRT7 also mediates adipogenesis and lipid metabolism, which may contribute to the pathological process47,70. Nevertheless, our results indicate that SIRT7 is required for the rhythmic hepatic gluconeogenesis. More importantly, SIRT7 couples BT and hepatic gluconeogenesis. Disruption of BT occur in diabetes mellitus and aging71,72. Altering BT may be critical to maintaining the robustness of peripheral clocks, providing a chronotherapeutic strategy to pre- vent metabolic diseases and aging. Collectively, we reveal a molecular network, that is, the BT– HSP70–SIRT7–CRY1 axis, wherein SIRT7 couples systemic BT cues to hepatic oscillators via HSP70 and ensures circadian phase coher- ence and glucose homeostasis in the liver. Sirt7-knockout (KO) alleles were created by CRISPR–Cas9-mediated genome editing in C57BL/6 mice via transgenic animal services from Cyagen. In brief, the Cas9 mRNA and Sirt7 guide RNA (gRNA) were generated by in vitro transcription and co-injected into fertilized eggs. Successful deletion was confirmed by PCR and DNA sequencing. Sirt7 heterozygous males were backcrossed with C57BL/6 females for at least five generations to preclude off-targeted mutations.The sequence of the gRNA was as follows, 5′-CTTGGCCGAGAGCGAGGATC-3′. Establishment of the Sirt7flox/flox mice was performed according to the method described in a previous report47 via transgenic animal services from Shanghai Model Organisms Center, Inc. In brief, LoxP sites were inserted into introns 5 and 9, respectively. The cytomegalovirus–Cre adenovirus was purchased from HanBio Technology. Male 8- to 10-week-old Sirt7+/+ and Sirt7flox/flox littermates were injected via the tail vein with 100 µl of adenoviral Cre (1–2 × 1010 plaque-forming units ml−1). The mice were maintained on a 12-h light–dark cycle (150 lx at the cage level from white light-emitting diode lamps) at 22 °C with access to a standard diet (70% kcal carbohydrates, 20% kcal protein and 10% kcal fat; Beijing Keao Xieli Feed Co., Ltd.) and water ad libitum. Male mice aged 8–12 weeks with littermate controls were used in all experiments. For temperature challenge, the mice were first placed in individual cages without nesting material and allowed to acclimate at 22 °C for at least 4 weeks in temperature-controlled incubators. The BT was measured using an infrared camera FLIRE6 (FLIR Systems). The mean temperature from all over the body was calculated. Rectal temperatures were measured using a Thermocouple Meter (Landwind Medical Industry Co., Ltd). For time RF, Sirt7−/− and WT control mice were fed ad libitum for 3 weeks and then subjected to RF in which food was available only during the daytime. Mice were sacrificed every 4 h after the inversion of the feeding regimen on either the second or fourth day. The kinetics of circadian phase shifting was determined by analyzing mRNA levels of clock genes by real- time PCR. All animals were housed and handled in accordance with protocols approved by the Committee on the Use of Live Animals in Teaching and Research. HEK293 cells were cultured in DMEM (Life Technologies), supplemented with 10% fetal bovine serum, 100 U ml−1 penicillin and 100 mg ml−1 streptomycin, and maintained in a humidified incubator at 37 °C with 5% CO2. Primary MEFs were prepared from E13.5 embryos of Sirt7+/− pregnant miceas described previously73. In brief, after removing the head and internal organ, the embryos were trypsinized for 20 min and seeded in T-25 cell culture dishes.Primary MEFs were grown in DMEM supplemented with 15% fetal bovine serum, 10 mM HEPES (pH 7.0), 2 mM glutamine (Life Technologies), 8 mM nonessential amino acids (Life Technologies) and 1 mM sodium pyruvate (Life Technologies). The cells were transfected with the indicated constructs using Lipofectamine 3000 reagent (Invitrogen), according to the manufacturer’s protocols. Cycloheximide (CHX; 50 μg ml−1, Sigma-Aldrich), MG132 (20 μM, 6 h; Sigma-Aldrich) and NAM (10 mM, 6 h; Sigma-Aldrich) were added to the cultures as indicated.The FLAG-SIRT7, FLAG-CRY1, HA-SIRT7 and HA-CRY1 constructs were established as follows: the full-length coding sequences of human CRY1 and SIRT7 were amplified from the complementary DNA of HEK293 cells by real-time PCR; the PCR products were cloned between the XhoI and XbaIor BamHI and XhoI sites in the pcDNA3.1 vector. FLAG-SIRT7 H187Y was purchased from Addgene (53151). FLAG-His-HSP70 (CH833663), FLAG-His- CRY2 (CH872052) and FLAG-His-FBXL3 (CH804377) were purchased from ViGene Biosciences, Inc. Various KR mutations were introduced via PCR-based site-directed mutagenesis. The Myc-ubiquitination construct was provided byCRISPR–Cas9-mediated gene deletion. CRISPR–Cas9-mediated gene deletion was conducted as previously described74. In brief, guide RNA targeting SIRT7 was subcloned into the PX459 vector (48139; Addgene). The HEK293 cells were transfected with 1 μg PX459-gSirt7. After 48 h, clones were selected with 1 μg ml−1 puromycin (Invitrogen) and then expanded for further analysis. The mutations were confirmed by PCR, DNA sequencing and immunoblotting. To avoid nonspecific off-target effects, we used two independent gRNAs; the sequenceswere as follows: gSirt7-1-forward: 5′-CACCGCCGCCGCTTTGCGCTCGGAG-3′, gSirt7-1-reverse: 5′-AAACCTCCGAGCGCAAAGCGGCGGC-3′; and gSirt7-2-forward: 5′-CACCGTGTGTAGACGACCAAGTATT-3′, gSirt7-2-reverse: 5′-AAACAATACTTGGTCGTCTACACAC-3′.Antibodies, immunoprecipitation and western blotting. Anti-FLAG and anti-HA antibodies were purchased from Sigma-Aldrich. The antibodies used for immunoblotting included anti-SIRT1 (8469; Cell Signaling Technology, Inc.),anti-SIRT2 (ab67299; Abcam), anti-SIRT3 (5490; Cell Signaling Technology, Inc.), anti-SIRT4 (Sigma), anti-SIRT5 (ab108968; Abcam), anti-SIRT6 (NB100-2252; Novus), anti-SIRT7 (sc-135055; Santa Cruz Biotechnology, Inc.), anti-BMAL1 (ab93806; Abcam), anti-CRY1 (ab3518; Abcam), anti-PER2 (sc-25363; Santa Cruz Biotechnology, Inc.), anti-HSP70 (4872; Cell Signaling Technology, Inc.), anti-ubiquitin (3936; Cell Signaling Technology, Inc.), anti-α-tubulin (AT819-1; Beyotime Biotechnology, Inc.), anti-GST (2624; Cell Signaling Technology,Inc.), anti-His and pan-acetyl lysine antibody (PTM-101, PTM-105; PTM- Biolab), p-CREB Ser133 (9198; Cell Signaling Technology, Inc.) and p-AKT Ser473 (3787; Cell Signaling Technology, Inc.). The primary antibodies used for immunoprecipitation included anti-FLAG M2 Affinity Gel (Sigma-Aldrich), monoclonal anti-HA agarose (Sigma-Aldrich), anti-SIRT7 (sc-365344; Santa Cruz Biotechnology, Inc.) and anti-CRY1 (ab3518; Abcam). The secondary antibodies used for endogenous immunoprecipitation included rabbit anti-mouseimmunoglobulin G (IgG) (Light Chain Specific) and mouse anti-rabbit IgG (Light Chain Specific) (58802 and 93702; Cell Signaling Technology, Inc.).Whole-cell extracts were prepared in immunoprecipitation lysis buffer (20 mM Tris–HCl, pH 8.0, 250 mM NaCl, 0.2% Nonidet P40 (NP-40), 10% glycerol, 2 mM EDTA, 1 mM phenylmethyl sulfonyl fluoride and protease inhibitor cocktail). For immunoprecipitation, whole-cell lysates were mixed with 2 μg primary antibody as indicated or control IgG and precipitated using protein A/G agarose beads (Thermo Fisher). To evaluate the acetylation of CRY1, we prepared whole-cell extracts in immunoprecipitation lysis buffer supplemented with 10 mM sodium butyrateand 10 mM NAM. Equal quantities of cell lysates or immunoprecipitated protein samples were separated by SDS–PAGE and transferred onto a polyvinylidene difluoride membrane (Millipore). The primary antibodies were added and incubated at 4 °C overnight. After incubation with secondary antibodies conjugated to horseradish peroxidase (Jackson Laboratories), the signal was visualized using an enhanced ECL chemiluminescence detection system (Sage Creation Science).RNA extraction and PCR with reverse transcription analysis. Total RNAs were extracted using TRIzol reagent (Invitrogen). Complementary DNA was synthesized from 2 μg RNA using PrimeScript RT Master kit (Takara, Japan) according tothe manufacturer’s protocol. Quantitative PCR with reverse transcription was performed on a Bio-Rad CFX Connect Real-time PCR system with SYBR Ex Taq Premixes (Takara). The relative quantification delta-delta threshold cycle method was used for analysis. Mouse 36b4 and human ACTIN were used as internal controls to normalize all data. The primers are listed in Supplementary Table 1.GST pull-down assay. GST pull-down assay was performed using recombinant GST-SIRT7 and His-CRY1 purified from BL21 E. coli. GST or GST-CRY1 protein(2 μg) was immobilized on glutathione–Sepharose 4B and incubated with His-CRY1 in GST binding buffer (20 mM Tris–HCl, pH 7.4, 0.1 mM EDTA, and 150 mM NaCl, 0.2% NP-40, protease inhibitors cocktail). The beads were washed three times with GST wash buffer (20 mM Tris–HCl, pH 7.4, 0.1 mM EDTA, and 250 mM NaCl, and 0.2% NP-40). Bead-bound CRY1 was analyzed by SDS–PAGE and western blotting. For purification of recombinant His-CRY1 from BL21 E. coli, 1 mM flavin adenine dinucleotide was added to the lysis buffer (50 mM PBS, pH 7.4, 0.5 M NaCl, 1 mM phenylmethyl sulfonyl fluoride, and protease inhibitors cocktail).In vitro deacetylation assay. FLAG-CRY1 was overexpressed in HEK293 cells and immunoprecipitated on the Anti-FLAG M2 Affinity Gel (Sigma-Aldrich). For the deacetylation assay, purified FLAG-CRY1 was incubated with 1 μg GST-SIRT7 in deacetylation buffer (50 mM Tris–HCl, pH 8.0, 4 mM MgCl2, 0.2 mM DTT, 1 mM NAD+ and protease inhibitors cocktail) for 30 min with constant agitation. The acetylation level of CRY1 was monitored by western blotting using anti-acetyl lysine antibodies.The gel lanes were cut from the gel and subjected to in-gel digestion with trypsin. The digested peptides were resuspended and analyzed by liquid chromatography mass spectrometry with a QTRAP 6500 mass spectrometer(Applied Biosystems). Data analysis was performed using the Mascot search engine against International Protein Index-HUMAN and National Center for Biotechnology Information databases for protein identification. Metabolic analysis. For glucose measurement, the mice were fasted overnight. For GTT, glucose (1 g kg−1) was intraperitoneally injected. For PTT, pyruvate (2 g kg−1) was intraperitoneally injected. Glucose levels were measured using a glucometer (OneTouch Ultra Vue; Johnson) at indicated time points. The serum insulin levels were measured using a Mouse Ultra Sensitive Insulin Jumbo ELISA kit (ALPCO). Glucagon levels in serum were measured using a Glucagon Quantizing ELISA kit (R&D Systems).Data are presented as the means ± s.e.m. Statistical analyses were performed JG98 using GraphPad Prism (version 6.0; GraphPad Software, Inc.). Statistical tests were two-tailed Student’s t-test (for comparison between two groups) or two-way ANOVA followed by Bonferroni’s multiple comparisons test (for comparison among three groups or in two groups at multiple time points).*P < 0.05 was considered to indicate a statistically significant difference. Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.