A Period2 Phosphoswitch Regulates and Temperature Compensates Circadian Period

Article

A Period2 Phosphoswitch Regulates and Temperature Compensates Circadian Period Graphical Abstract

Authors Min Zhou, Jae Kyoung Kim, Gracie Wee Ling Eng, Daniel B. Forger, David M. Virshup

Correspondence [email protected] (D.B.F.), [email protected] (D.M.V.)

In Brief It’s been long known that circadian clocks compensate for changes in temperature, but the molecular mechanism has been elusive. Zhou, Kim, and coworkers describe a dual-kinase, multisite phosphoswitch that regulates stability of the PERIOD protein and provides for environmental and metabolic control of the clock, including temperature compensation.

Highlights d

PER2 degradation follows three-stage kinetics, consistent with a phosphoswitch model

d

Inclusion of a phosphoswitch suggests a temperature compensation mechanism

d

This phosphoswitch integrates diverse environmental stimuli to tune PER2 stability

d

The phosphoswitch is a design feature of the clock

Zhou et al., 2015, Molecular Cell 60, 77–88 October 1, 2015 ª2015 Elsevier Inc. http://dx.doi.org/10.1016/j.molcel.2015.08.022

Molecular Cell

Article A Period2 Phosphoswitch Regulates and Temperature Compensates Circadian Period Min Zhou,1,7 Jae Kyoung Kim,2,3,4,7 Gracie Wee Ling Eng,1 Daniel B. Forger,4,5,* and David M. Virshup1,6,* 1Program

in Cancer and Stem Cell Biology, Duke-NUS Graduate Medical School, 8 College Road, Singapore 169857, Singapore of Mathematical Sciences, Korea Advanced Institute of Science and Technology, Daejeon 305-701, Korea 3Mathematical Biosciences Institute, The Ohio State University, Columbus, OH 43210, USA 4Department of Mathematics 5Department of Computational Medicine and Bioinformatics University of Michigan, Ann Arbor, MI 48109, USA 6Department of Pediatrics, Duke University Medical Center, Durham, NC 27710, USA 7Co-first author *Correspondence: [email protected] (D.B.F.), [email protected] (D.M.V.) http://dx.doi.org/10.1016/j.molcel.2015.08.022 2Department

SUMMARY

Period (PER) protein phosphorylation is a critical regulator of circadian period, yet an integrated understanding of the role and interaction between phosphorylation sites that can both increase and decrease PER2 stability remains elusive. Here, we propose a phosphoswitch model, where two competing phosphorylation sites determine whether PER2 has a fast or slow degradation rate. This mathematical model accurately reproduces the threestage degradation kinetics of endogenous PER2. We predict and demonstrate that the phosphoswitch is intrinsically temperature sensitive, slowing down PER2 degradation as a result of faster reactions at higher temperatures. The phosphoswitch provides a biochemical mechanism for circadian temperature compensation of circadian period. This phosphoswitch additionally explains the phenotype of Familial Advanced Sleep Phase (FASP) and CK1εtau genetic circadian rhythm disorders, metabolic control of PER2 stability, and how drugs that inhibit CK1 alter period. The phosphoswitch provides a general mechanism to integrate diverse stimuli to regulate circadian period. INTRODUCTION Circadian rhythms are widespread in nature and coordinate internal physiologic functions with external daily light-dark cycles (Pittendrigh, 1960). Key features of circadian rhythms include their persistence in the absence of external signals, their regulation by metabolic stimuli, and their ability to robustly compensate, and even over-compensate, for changes in external temperature (Bass and Takahashi, 2010; Buhr et al., 2010; Pittendrigh, 1954; Sweeney and Hastings, 1960). In vertebrates, multisite phosphorylation accompanied by b-TrCP-dependent

proteasomal degradation of PER2 regulates its accumulation in a daily cycle, and PER2 in turn facilitates repression of circadian transcription (Gallego and Virshup, 2007; Lee et al., 2001). PER2 accumulation and degradation are also influenced by metabolic and environmental stimuli that act via changes in phosphorylation (Badura et al., 2007; Gallego and Virshup, 2007; Kaasik et al., 2013; Kim et al., 2013; Mehra et al., 2009; Pilorz et al., 2014; Shanware et al., 2011; Vanselow and Kramer, 2007; Vanselow et al., 2006; Xu et al., 2007). However, the data on specific PER2 phosphorylations are complex, and no existing model fully explains the role of PER2 phosphorylation in clock timing. Two key phosphorylation events have been linked to PER2 degradation in mammals. Familial advanced sleep phase (FASP) is caused by a missense mutation S662G in human PER2 (S659 in mouse) that prevents priming phosphorylation by an as-of-yet unidentified priming kinase. Priming phosphorylation is required for CK1-dependent phosphorylation at a series of immediately downstream serine residues (Lowrey et al., 2000; Shanware et al., 2011; Toh et al., 2001; Vanselow et al., 2006; Xu et al., 2007). The mechanistic role of these FASP site phosphorylations is uncertain, as it was initially thought to destabilize PER2 by regulating nuclear localization (Vanselow et al., 2006; Xu et al., 2007), while more recent studies demonstrated that phosphorylation of this cluster actually stabilizes PER2 protein independent of location (Shanware et al., 2011). A second CK1-targeted PER2 phosphorylation region that appears to be priming-independent (S477–S479 in mPER2) is required for b-TrCP-dependent degradation of PER2 (Eide et al., 2005; Isojima et al., 2009). Recent studies in Drosophila and Neurospora suggest that similar widely separated phosphorylation sites on PER and FRQ interact in a temporal manner (Baker et al., 2009; Chiu et al., 2011; Garbe et al., 2013; Querfurth et al., 2011; Shanware et al., 2011), but how they combine to regulate circadian rhythms remains elusive. The interaction of the circadian clock with the environment has long been noted. Temperature fluctuations can entrain circadian pacemakers, but the core clock machinery is buffered against temperature-induced changes in clock period (Buhr et al., 2010; Pittendrigh, 1954). Higher temperatures generally increase biochemical reaction rates, but paradoxically, the circadian Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc. 77

Figure 1. PER2 Degrades in Three Distinct Stages Regulated by Clock Time and CK1

A 250 Cntrl Hr 19

Counts / Second

200

Hr 22 Hr 25

150

Hr 28 Hr 30

100 50 0 0

5

10

15

20

25

30

35

40

45

Time (hrs)

Plateau (hrs)

5 4 3 2 1

C Per2 Abundance, %

B

CHX

100

(A) PER2 shows three-stage degradation when CHX is added during the rising phase of PER2 oscillation. Per2Luc MEFs were synchronized with dexamethasone and the abundance of PER2::LUC was assessed by measuring the luminescence intensity in the Lumicycle. At the indicated time points following dexamethasone shock, CHX (40 mg/ml) was added. The second or plateau stage, defined as the time when the instantaneous half-life is greater than 5 hr (see also Figure S1A and Experimental Procedures for details), is indicated by the shaded area. (B) The duration of the plateau is dependent on the time of addition of CHX. Data from Figure 1A are represented as mean ± SD. (C) Inhibition of CK1 largely eliminated both the first and the second stage of degradation. Per2Luc MEFs were synchronized as above, and either CHX (40 mg/ml) or CHX (40 mg/ml) and PF670 (1 mM) together was added 22 hr after dexamethasone. Data were normalized to PER2::LUC abundance immediately prior to drug addition.

CHX+PF670

50

sites on PER2 regulate PER2 stability in opposing ways. We biochemically 0 5 10 15 20 -8 -7 -6 -5 -4 -3 -2 -1 confirmed this phosphoswitch between the FASP site and the b-TrCP binding Hrs before peak Time (hrs) sites on PER2. The phosphoswitch allows the tuning of the fraction of PER2 that is in clock keeps a 24-hr period regardless of ambient temperature, either a stable or unstable state. Tuning via the phosphoswitch a phenomenon known as temperature compensation. Circadian provides a robust mechanistic explanation for two important clocks actively compensate for the temperature-induced in- circadian clock mutation phenotypes that affect PER2 phoscreases in biochemical reaction rates. Almost 60 years ago, phorylation, CK1εtau and FASP. Furthermore, by tuning the fracHastings and Sweeney proposed that multiple reactions having tion of stable PER2, the phosphoswitch regulates circadian opposing contributions to period could induce temperature period in response to metabolic signals and temperature compensation of circadian clocks (Hastings and Sweeney, change. Importantly, the mathematical model predicted and 1957). However, the specific opposing molecular reactions that we experimentally confirmed that the phosphoswitch increases the stable PER2 fraction and slows down the clock at higher temresult in temperature compensation remain unknown. Reversible phosphorylations are a candidate mechanism to perature by exploiting a difference in temperature coefficients of produce temperature compensation. Indeed, recent studies the PER2 regulatory kinases. The PER2 phosphoswitch demonreveal that phosphorylation-regulated degradation of the key strates that temperature compensation is a design feature of the circadian regulator FRQ in Neurospora is involved in temperature core circadian machinery. compensation (Mehra et al., 2009). CK1 appears to be important in temperature compensation. Unlike most enzymes, CK1 is RESULTS relatively temperature insensitive (Isojima et al., 2009; Qin et al., 2015). Furthermore, the tau mutation in CK1ε, which dis- Three-Stage Degradation of PER2 rupts the ability of CK1 to phosphorylate primed substrates The kinetics of PER2::LUC degradation were monitored in a such as PER2, causes loss of temperature compensation (Low- LumiCycle utilizing mouse embryo fibroblasts (Per2luc MEFs) rey et al., 2000; Tosini and Menaker, 1998). These studies derived from homozygous Per2luc knockin mice (Yoo et al., suggest that CK1-mediated phosphorylation of PER2 plays an 2004). When protein translation was inhibited by addition of important role in temperature compensation, but the molecular cycloheximide (CHX) during the falling phase (CT 26–38 hr) of the circadian cycle, PER2 degradation was exponential, as exmechanism is unclear. Here, we investigated the stability of PER2 and made the un- pected (Figure 1A). Unexpectedly, when CHX was added during expected observation that endogenous PER2 shows three- the rising phase (CT 14–26 hr) of the cycle, PER2 degraded in stage degradation kinetics rather than simple exponential decay. three distinct stages (Figures 1A and S1A): an initial rapid decay Our revised mathematical model predicts that this three- (the first stage), a plateau-like slow decay (the second stage), stage degradation occurs when two competing phosphorylation and finally, a more rapid terminal decay (the third stage). This

0

0

78 Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc.

three-stage degradation was not simply due to the immortalization, as it occurred in primary Per2Luc MEFs as well (Figure S1B). Remarkably, the duration of the second stage of degradation was dependent on when in the circadian cycle the CHX was added (Figures 1A, 1B, and S1A). The dependence of the length of this plateau on circadian time (Figures 1B and S1A) suggested it might be a key feature of the clock mechanism. Interestingly, the addition of the CK1d/ε (CK1) inhibitor PF670462 (PF670) (Badura et al., 2007) together with CHX primarily eliminated the first and the second stage, so that the degradation appeared nearly exponential (Figure 1C). Thus, the plateau is dependent on CK1 kinase activity. PF670 was reported to increase period length through CK1 inhibition (Figure S1C) (Badura et al., 2007; Kim et al., 2013). However, CK1 inhibition had a minimal effect on the third stage of degradation, similar to the lack of effect seen if PF670 was added after the peak of accumulation (Figure S1D). The initial sharp increase in PER2 abundance in the first hour after CHX and PF670 addition is likely due to the asynchronous effects of PF670 and CHX (Figure 1C and see modeling below). LiCl, an inhibitor of GSK3, another kinase that phosphorylates PER2 (Iitaka et al., 2005), did not alter the three-stage degradation of PER2 (Figure S1E). Mathematical Modeling Predicts that the Phosphoswitch Generates the Three-Stage Degradation of PER2 Based on these multistage degradation kinetics, we revised and extended our mathematical circadian clock model (see Experimental Procedures for details) (Gallego et al., 2006a; Kim and Forger, 2012). Whereas existing models (Kim and Forger, 2012; Leloup and Goldbeter, 2011; Vanselow et al., 2006) were not able to explain the plateau (data not shown), the inclusion of a phosphoswitch that controls PER2 stability (Figure 2A) simulates both circadian rhythm of PER2 protein abundance (Figure S2A) and PER2 three-stage degradation (Figure 2B). The key feature of the revised model is that newly translated PER2 protein has two fates depending on which sites are initially phosphorylated (Figure 2A). If the initial phosphorylation occurs at the b-TrCP binding site, then PER2 is rapidly degraded (Eide et al., 2005). Conversely, phosphorylation at the FASP site by the combined activity of the priming kinase followed by CK1 inhibits the phosphorylation at the b-TrCP binding site, thereby stabilizing PER2 (Shanware et al., 2011; Vanselow et al., 2006; Xu et al., 2007). With the inclusion of the phosphoswitch, our model correctly predicted the effect of CHX at different clock times (Figures 1A and 2B), the effects of PF670 including the initial spike (Figures 1C and 2C), and the insensitivity of the third stage of degradation to CK1 inhibition (Figures S1D and S2B). Using the model, we explored the detailed mechanisms underlying the three-stage decay of PER2 (Figure 2D). The model predicts that the initial rapid decay is induced by the brisk degradation of PER2 phosphorylated on the b-TrCP binding site. The plateau is due to the primed phosphorylation on the FASP site, which stabilizes the PER2 protein. Finally, during the third phase (Figure 2D) and falling phase (Figure S2C), PER2 is already fully phosphorylated at the FASP cluster, and a monophasic rather than three-stage decay occurs. When CK1 is inhibited, most of the PER2 is monophosphorylated at the FASP site by the priming

kinase, but is not phosphorylated at the b-TrCP site (Figure 2E). The brief increase in luciferase activity seen in Figure 1C can be explained by assuming that PF670 acts faster than CHX, e.g., by greater cell permeability. This consistent with the very different calculated partition ratios of the two molecules (logP (calculated) PF670 = 6.42, CHX = 0.99), which reflects the hydrophobicity and lipid permeability of the molecules (Leo et al., 1971). To simplify the model simulations, we initially assumed that PER2 is phosphorylated at either S478 or S659 (Figure 2A). However, this assumption can be relaxed somewhat. Even if PER2 can be phosphorylated at both sites, the model behavior does not change substantially as long as (1) phosphorylation at pS478 indirectly inhibits the phosphorylation at S659 via rapid degradation (i.e., PER2 phosphorylated at both sites is unstable, similar to PER2 phosphorylated at S478) (Figure S2D) or (2) PER2 phosphorylated at S659 partially inhibits the phosphorylation at S478 (i.e., pS659 reduces the phosphorylation rate at S478) (Figure S2E). Because the simple model only included limited clock components (Figure 2A), we next tested if the phosphoswitch can induce the three-phase degradation in the presence of other post-translational modifications of PER2. We therefore included the phosphoswitch in the most detailed mathematical model available for mammalian circadian timekeeping, where PER2 binds with other clock components and is also phosphorylated by GSK3 (Figure S2F) (see Experimental Procedures for details). This detailed model with the phosphoswitch of PER2 also successfully simulates the three-stage decay (Figure S2G). Furthermore, we used this model to show that the predicted three-stage decay of PER2 is independent of GSK3 inhibition, matching our experimental data (Figures S1E and S2H). The Identification of the PER2 Phosphoswitch Our model requires that FASP phosphorylations stabilize PER2 by blocking the phosphorylation on b-TrCP binding sites. This is consistent with biochemical data in Drosophila (Chiu et al., 2011; Garbe et al., 2013). We tested this prediction in transient transfection experiments. Inspection of the major b-TrCP binding motif suggested that S478 on mouse PER2 was likely to be a key CK1 phosphorylation site that serves as a b-TrCP-binding phosphodegron for PER2. We confirmed that S478A mutant PER2 is markedly impaired in b-TrCP binding whereas wildtype PER2 co-immunoprecipitates b-TrCP (DF-box) (Figure 3A). Importantly, by developing a phosphoepitope-specific antibody, pS478, that recognizes phosphorylated S478 (Figures 3A, S3A, and S3B), we confirmed that mutation of PER2 FASP site S659 considerably increased CK1-dependent phosphorylation of PER2 S478 (Figures 3B and S3C). Using a luciferase fused PER2 construct, the half-lives of PER2 S478A and PER2 S659A were assessed in NIH 3T3 cells by transient transfection (Figure 3C). As expected, PER2 S478A had a significantly longer half-life than the wild-type protein, while the S659A mutation decreased the PER2 half-life. These data show that the PER2 phosphoswitch between the b-TrCP binding site and the FASP site regulates PER2 stability in an opposing manner. Conversely, mutation of the b-TrCP binding site (S478A) did not increase phosphorylation of PER2 at S659 (Figure S3D). However, even if phosphorylation of S478 did alter phosphorylation of S659, Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc. 79

Figure 2. A Mathematical Model Predicts the Phosphoswitch of PER2 as the Mechanism for Three-Stage Decay of PER2

A Per2

Per2

Per2

Per2

pS659 (FASP site)

B/C

Transcription

Translation

PER Per2

E-Box

Per2 RNA

Per2 pS478 ( TrCP site)

B/C BMAL1-CLOCK

Priming Phosphorylation on FASP site

Degraded Per2 protein

Phosphorylation on β-TrCP site CK1 phosphorylation on FASP site

B

C

PER2 Abundance (AU)

CHX

CHX+PF670

25

25

20

20

15

15

10

10

5

5

0 0

10

20

30

0 0

40

PER2 Abundance, %

E

CHX

100

50

50

0

5

10

Time (hrs)

Hr 24

10

20

30

40

CHX+PF670

100

0

Hr 16 Hr 18

Time (hrs)

Time (hrs)

D

Cntrl Hr 14

15

0 0

Total Unphosphorylated FASP (Prime) FASP (Complete) -TrCP Experimental Data

5 10 Time (hrs)

15

our mathematical model predicts that as long as the phosphorylation at S478 leads to rapid degradation, the three-stage degradation still occurs (Figure S2D). To biochemically address why phosphorylation of the FASP region protects the b-TrCP site from phosphorylation, we performed a limited proteolysis assay (Chiu et al., 2011). Wild-type and S478A mutant PER2 are more trypsin resistant, while mutation of S659 made PER2 more trypsin sensitive (Figure 3D). These results suggest that the polyphosphorylation of the S659 region alters the conformation of PER2 such that S478 is less accessible to CK1. This conformational flexibility is consistent with computational predictions that both S478 and the FASP domain are located on the borders of intrinsically disordered regions of PER2 (Figure 3E). Intrinsically disordered regions are likely to be involved in regulated signaling interactions, interac80 Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc.

(A) In the model, the translated PER2 protein has two fates, depending on the site of initial phosphorylation. A phosphorylation at one site (red dot) by CK1 induces b-TrCP binding and rapid degradation (Eide et al., 2005). A phosphorylation at another site, known as the FASP site, by the priming kinase (orange dot) induces sequential phosphorylation at neighboring downstream sites by CK1 (purple dots), which stabilizes PER2 protein (Shanware et al., 2011; Vanselow et al., 2006; Xu et al., 2007). Finally, fully phosphorylated PER2 protein becomes an active transcriptional repressor and degrades via a CK1- and b-TrCP-independent mechanism. See also Tables S1 and S2. (B) The model simulates the occurrence of threestage decay of PER2 exclusively during its rising phase. It also predicts the second-stage plateau increases when protein synthesis is inhibited earlier in the rising phase. To allow time for CHX to enter cells, the protein translation rate is assumed to be exponentially decreased with a half-life (0.13 hr) after CHX treatment. (C) When both translation and CK1 are inhibited during the rising phase, PER2 shows monophasic degradation after an initial spike. Here, to simulate the inhibition of phosphorylation by CK1, the binding rate of CK1 with PER2 is immediately reduced to 1% of the original binding rate. (D) The predicted fate of specific PER2 species. The model predicts that phosphorylation at b-TrCP binding site(s) induces the initial rapid degradation of PER2. The second, plateau stage is generated by the sequential phosphorylation around the FASP site (FASP [Prime] and FASP [Complete]), which stabilizes PER2. Finally, fully phosphorylated PER2 protein becomes unstable, which induces the third degradation stage. Time 0 is 22 hr following the prior PER2 peak, as in Figure 1C. (E) The model predicts that CK1 inhibition blocks the degradation of PER2 induced by b-TrCP site phosphorylation and promotes the accumulation of stable PER2 phosphorylated at the FASP site by the priming kinase. This causes the spikes immediately following CK1 inhibition. See also Figure S2.

tions that can be regulated by phosphorylation (Nishi et al., 2013; Xue et al., 2010). We found that three-stage decay was not observed when wildtype or mutant PER2::LUC proteins were transiently overexpressed (Figure 3C). Model simulations and overexpression studies confirmed that maintaining rhythmic PER2 expression at endogenous levels from circadian promoters is critical to observing the natural dynamics of PER2 degradation (Figures S3E–S3H). Therefore, we used pharmacological manipulation of the endogenous phosphoswitch (Figure 3F). Circadian rhythms are sensitive to metabolic status, in part via reversible O-GlcNAcylation of PER2 at the FASP phosphorylation site (Kaasik et al., 2013; Kim et al., 2012). We manipulated PER2 O-GlcNAcylation and hence FASP site phosphorylation by metabolic alteration. Alloxan (ALX), an O-linked N-acetylglucosamine

A

B

Figure 3. Molecular Characterization of the PER2 Phosphoswitch

(A) PER2 S478 phosphorylation regulates the binding of PER2 to b-TrCP. PER2, b-TrCP (DF-box), and CK1ε were co-expressed in NIH 3T3 cells. Twenty-four hours after transfection, Myc-PER2 was immunoprecipitated and binding partners assessed by SDS-PAGE and immunoblotting. Mutation of S478A in PER2 markedly decreased b-TrCP binding. WCL, whole-cell lysate. (B) PER2 FASP site S659A mutation promotes b-TrCP site phosphorylation. The indicated PER2 mutants were expressed in NIH 3T3 cells with or without CK1ε. b-TrCP binding site phosphorylation, detected with the pS478 antibody, was markedly increased in the presence of the FASP mutant (S659A). * Non-specific bands used as C D loading controls. (C) PER2 b-TrCP site mutant S478A increases while FASP site mutant S659A decreases PER2 stability. The indicated mutants were transiently expressed (10 ng of PER2 plasmids per 35 mm dish) in unsynchronized NIH 3T3 cells, and their abundance was assessed in a Lumicycle after the addition of CHX. Mean PER2 half-life is shown ± SD. (D) Limited proteolysis shows that PER2 S659A is more susceptible than WT PER2 to trypsin digest, indicating the S659A mutant causes a more open structure of PER2. The indicated PER2 wild-type and mutants were immunoprecipitated and digested with 1 mg/ml trypsin for 1 min on ice. E PER2 abundance was assessed by SDS-PAGE and immunoblotting. (E) PER2 is predicted to have an extended disordered domain between the b-TrCP site and the F FASP site, as well as after the FASP site. Mouse PER2 protein sequence was analyzed using PONDR-FIT from DisProt.org using default parameters (Sickmeier et al., 2007). A region is considered as intrinsic disordered when the Disorder Disposition score is greater than 0.5. PER2 domains and the location of S478 and S659 are indicated on the top of the graph. PAS, PAS domain 1 (dark blue) and 2 (purple); CLD, cytoplasmic localization domain (lime); CK1, CK1 binding domain (navy); NLS, nuclear localization signal (blue); CRY, Cry binding site (green). (F) Metabolic perturbation of the phosphoswitch affects the three-stage decay of PER2. ALX (10 mM) or PUG (200 mM) was added to PerLuc MEFs during the rising phase (5 hr before the peak), and degradation was assessed as above. These data were collected together with those shown in Figure 1C and the PF670+CHX curve is repeated here to aid in comparison. See also Figure S3.

(O-GlcNAc) transferase (OGT) inhibitor, and PUGNAc (PUG), an O-GlcNAc hydrolase (OGA) inhibitor, increases and decreases phosphorylation at FASP sites, respectively (Kaasik et al., 2013). When phosphorylation at the FASP site was increased by blocking OGT with ALX, PER2 degraded exponentially without the first and second stage (Figure 3F), similar to the effect of the CK1 inhibitor PF670. When the OGA was inhibited with PUG, which increases O-GlcNAcylation and reduces the phosphorylation at FASP site, the first stage of degradation was significantly accelerated (Figure 3F). Thus, metabolic control of FASP site phosphorylation also regulates first- and secondstage degradation, consistent with the phosphoswitch model.

Tau Mutation Phenotype Stems from Disruption of the Phosphoswitch The tau hamster, with a semidominant short circadian period caused by a missense mutation in CK1ε, has been extensively studied (Ralph and Menaker, 1988). The tau mutation decreases the activity of CK1ε on primed substrates (Lowrey et al., 2000). Through mathematical modeling and experimental validation, we previously showed that CK1εtau has increased activity on PER2 (Gallego et al., 2006a), yielding faster PER2 clearance, despite its decreased activity on primed substrates (Isojima et al., 2009; Lowrey et al., 2000). Indeed, we confirmed that in cells CK1εtau has decreased activity on the FASP site, as Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc. 81

A

B

C

D

Figure 4. CK1ε tau Mutation Destabilizes PER2 by Affecting the Phosphoswitch (A) CK1εtau is a loss of function on the FASP site. The indicated amount of CK1ε or CK1εtau expression plasmid was cotransfected with plasmid encoding wildtype or S659A myc-PER2 into NIH 3T3 cells. Phosphorylation at the FASP site was assessed in whole-cell lysates by immunoblotting with pFASP antibodies. Specific PER2 phosphorylation was calculated as the ratio of pFASP to total PER2 immunoreactivity. b-TrCP (DF-box) was co-expressed to block PER2 degradation. (B) CK1εtau is a gain of function on the b-TrCP S478 site, phenocopying the FASP mutation. Wild-type or tau mutant CK1ε (plasmid amount shown) was coexpressed with the indicated myc-PER2 variants as above. Phosphorylation on the b-TrCP binding site was assessed in whole-cell lysates by immunoblotting with pS478 antibodies. Mutation of the FASP site S659 minimizes the difference between CK1ε and CK1εtau activity on PER2. (C) tau MEFs show a larger first rapid degradation stage and a shorter second plateau stage than wild-type Per2Luc MEFs. The three-stage degradation of the Per2Luc and CK1εtau MEFs was measured as in Figure 1. The three-stage decay (CHX was added three [ 3] and five [ 5] hours before the second peak) shown is the mean of four duplicates. (D) The duration of plateau stage is shorter in Per2Luc/CK1εtau MEFs compare to Per2Luc MEFs. The plateau stage was quantified as in Figure 1B. Data are represented as mean ± SD. See also Figure S4.

measured by a phosphoepitope-specific antibody (gift from R.S. Tibbetts) (Figure 4A) (Shanware et al., 2011). The phosphoswitch model predicts that decreased phosphorylation of the FASP site allows increased phosphorylation of the b-TrCP site. Indeed, S478 phosphorylation increased significantly when CK1εtau was coexpressed with PER2 (Figure 4B, lanes 3–7), indicating that CK1εtau is a gain of function on S478. Importantly, as 82 Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc.

predicted by the phosphoswitch, the CK1εtau gain of function on the b-TrCP site was dependent on an intact FASP site, since when the FASP site was mutated (S659A) (Figure 4B, lanes 8–12), CK1ε and CK1εtau had similar activity on the b-TrCP site. Consistent with this, CK1εtau markedly shortened the halflife of wild-type PER2, but had only a small effect on the halflife of PER2 (S659A) (Figure S4).

Figure 5. PER2 Phosphoswitch Compensates Circadian Period in Response to Temperature Change

A

B

C

The phosphoswitch model predicts that CK1εtau, with decreased FASP site and hence increased b-TrCP site phosphorylation, should shorten PER2 half-life via markedly altered PER2 degradation kinetics. Indeed, three-stage degradation in MEFs from CK1εtau/tau/Per2luc mice had enhanced first-stage and markedly shortened second-stage degradation compared with Per2luc MEFs (Figures 4C and 4D). Enhanced b-TrCP site phosphorylation increased the first-stage degradation, and diminished FASP site phosphorylation prevented the stabilization of PER2, hence shortening the second stage. The Phosphoswitch Temperature Compensates Circadian Period Our mathematical model also predicts that the phosphoswitch plays a critical role in temperature compensation. CK1 is known to be relatively temperature insensitive. Building on this, we modeled several cases, including the assumption that the FASP priming kinase is more temperature sensitive than CK1 (Case 2 in Figure 5A). This would lead to increased FASP versus b-TrCP binding site phosphorylation as the temperature rises, and more PER2 would be directed to the slow second stage of degradation, effectively lengthening the period at higher temperatures and compensating the clock (Figure 5A). This assumption predicts that the length of the second-stage degradation, which relies on FASP site phosphorylation, will be more sensitive to temperature than first-stage degradation. Testing this prediction, we assessed period and PER2 degradation kinetics in Per2Luc MEFs incubated at different tem-

(A) Three scenarios for relative temperature sensitivity of the FASP site priming kinase and CK1. The mathematical model (Figure 2A) predicts that the observed temperature over-compensation occurs only when the priming kinase is more temperature sensitive than CK1 (Case 2: Q10 of priming kinase and CK1 are 4 and 1.33, respectively). (B) The plateau stage of three-stage decay is markedly shorter when Per2Luc MEFs were cultured at 30 C while the initial rapid degradation stages are similar. The plateau stage was quantified as in Figure 1B. (C) Temperature over-compensation mainly occurs during the rising phase. Per2Luc MEFs were cultured at either 30 C or 37 C. The duration of rising phase (bottom to peak) and falling phase (peak to bottom) were estimated with FFTnonlinear least-squares analysis (Izumo et al., 2006). The differences in these stages in 30 C and 37 C were quantified as shown on the right. The rising phase was 1 hr longer at 37 C. Data are represented as mean ± SD.

peratures. When the temperature was decreased to 30 C from 37 C, the second stage of PER2 degradation was markedly decreased, while the first stage remained intact (Figure 5B). As expected, Per2Luc MEFs show temperature over-compensation, i.e., the period lengthens at higher temperatures (37 C) (Figure 5C). Furthermore, we found that this difference in period length is mainly due to changes during the rising phase (Figure 5C), when the second degradation stage occurs (Figure 1A). The rising phase at 30 C is 1 hr shorter than at 37 C, while the falling phase is essentially unchanged. This indicates that the temperature compensation of the period mainly occurs when the FASP site is actively phosphorylated. These data support the prediction of the mathematical model: phosphorylation on the FASP site is more sensitive to temperature change than is phosphorylation on the b-TrCP binding site, and this is responsible for temperature compensation (Figure 5A). As CK1εtau has decreased activity on the FASP site (Figure 4), tau MEFs are predicted to have abnormal temperature compensation. Indeed, temperature over-compensation in CK1εtau/ Per2Luc MEFs was reversed compared with Per2Luc MEFs (Figure 6A), consistent with findings in the retina of tau mutant hamster (Tosini and Menaker, 1998). We also tested if temperature compensation was affected by the CK1 inhibitor PF670, which causes the loss of the second stage of the three-stage decay (Figure 1C). Similar to tau MEFs, over-compensation was eliminated in the presence of PF670 (Figure 6B). These results are consistent with the phosphoswitch model and indicate that regulation of the second stage of PER2 degradation via phosphorylation on the FASP site plays a critical role in temperature compensation. To further test this, we compared the half-life of transiently expressed PER2 at 30 C and 37 C. As temperature Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc. 83

A

B

Figure 6. Temperature Compensation Requires the Phosphoswitch (A) Circadian period is under-compensated in tau MEFs. Per2Luc and CK1εtau MEF cells were cultured in the Lumicycle as indicated to assess their temperature compensation. Period was calculated based on four replicates by Lumicycle data analysis software and represented as mean ± SD. (B) Temperature compensation does not occur when CK1 is inhibited. PF670 (0.1 mM) or vehicle was added at the middle of the third peak (red arrow) while Per2Luc MEF cells were cultured at 30 C and 37 C. (C) Half-life of PER2 becomes significantly shorter at low temperature (30 C) when CK1ε is coexpressed. However, this shortening of PER2 half-life does not occur when the phosphoswitch is disrupted with PF670 treatment, CK1εtau (Tau) coexpression, or mutation of the b-TrCP (S478A) or FASP (S659A) phosphorylation sites. Data are represented as mean ± SD.

S478A (Figure 6C) again suggests that CK1εtau retains the ability to phosphorylate additional, as of yet unidentified, degradation site(s) on PER2 in a temperature-insensitive manner. In summary, the phosphoswitch increases the fraction of stable PER2 protein by increased phosphorylation at the FASP site at higher temperature (Figure 7). This leads to longer second-stage degradation of PER2 during the accumulation phase, which lengthens the accumulation time and the period at higher temperature.

C

decreases to 30 C, the half-life of PER2 decreases significantly when CK1ε is coexpressed (Figure 6C), reflecting the shortening of the second stage of PER2 degradation (Figure 5B). However, when the phosphoswitch is disrupted by PF670, CK1εtau mutation, or mutation of PER2 at either S478 or S659, the half-life of PER2 no longer decreases substantially with decreasing temperature. The ability of CK1εtau to shorten the half-life of PER2 84 Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc.

DISCUSSION Our work demonstrates that the two-site phosphoswitch of PER2 regulates circadian timekeeping by controlling PER2 stability. The key insights came from observing the three-stage degradation of endogenous PER2::LUC and by sampling PER2 abundance frequently. The importance of studying PER2 at endogenous levels is consistent with previous studies showing the importance of the stoichiometry of diverse clock proteins (Kim and Forger, 2012; Lee et al., 2011a, 2011b). The disruption of three-stage degradation upon PER2 overexpression is consistent with the finding that overexpression of PER2, unlike other clock components (e.g., CRY or BMAL1), disrupts circadian rhythms (Chen et al., 2009). Our revised clock model incorporating the phosphoswitch provides a mechanistic explanation for a number of central phenomena in the clock including temperature compensation and metabolic regulation. The exquisite phosphorylation control

Figure 7. Model for Temperature Compensation with the Phosphoswitch As temperature decreases, the PER2 phosphorylated on the b-TrCP binding site dominates the PER2 phosphorylated on the FASP site (Figure 5B). This shortens the plateau stage decay of PER2 during rising phase, which accelerates the rising phase and shortens period in the low temperature (Figure 5C). Domain architectures were shown by colors. PAS, PAS domain 1 (dark blue) and 2 (purple); CLD, cytoplasmic localization domain (lime); CK1, CK1 binding domain (navy); NLS, nuclear localization signal (blue); CRY, Cry binding site (green).

suggests that regulated dephosphorylation of PER2 by intracellular phosphatases such as PP1 and PP2A (Gallego et al., 2006b; Lee et al., 2011b; Shanware et al., 2011) can also regulate the length of the plateau. The model explains how CK1εtau, a mutant extensively studied in circadian biology, can have decreased activity on one site and therefore increased activity on other sites. Finally, the model demonstrates that the first identified human inherited circadian rhythm mutation, the FASP mutation, causes a decrease in PER2 stability through the same mechanism as CK1εtau. While the model requires the addition of an unidentified priming kinase, this is fully consistent with the known substrate recognition properties of CK1 and the identification of a priming kinase in Drosophila (Chiu et al., 2011). Furthermore, as seen with Drosophila PER, the phosphorylation at the FASP region by the priming kinase followed by CK1 alters the protease sensitivity of PER2. This is likely due to conformational changes that

block the phosphorylation at S478 of PER2. This is consistent with the predicted intrinsically disordered region between and after the two elements of the phosphoswitch that may become more structured after post-translational modification. Temperature compensation of the clock was first noted in poikilotherms, animals whose internal temperature can vary quite widely (Pittendrigh, 1954). Remarkably, temperature compensation is retained in large homeotherms such as mammals, where the temperature of the suprachiasmatic nucleus is constant (Barrett and Takahashi, 1995; Buhr et al., 2010; Tosini and Menaker, 1998; Zatz et al., 1994). Our results indicate that the phosphoswitch, which provides a robust mechanism for temperature compensation, is a design feature of the clock that evolutionarily precedes the switch from poikilo- to homeothermy. Hence, mammalian temperature compensation due to the predicted differences in the temperature coefficients of CK1 and the priming kinase is built into the core clock mechanism, regardless of any current evolutionary advantage. It would be interesting in future work to identify the phosphoswitch and investigate its role in the circadian clock of other organisms, such as flies, fungi, and bacteria. Our study provides an explanation for a long-studied naturally occurring mutation in CK1ε, the CK1ε tau mutation. How a lossof-function point mutation in CK1ε can produce an short period, while drugs that inhibit CK1d and CK1ε product long periods, has not been well understood. The tau mutation blocks the ability of CK1ε to recognize primed substrates (Gallego et al., 2006a; Isojima et al., 2009; Lowrey et al., 2000). Our data confirm that CK1εtau is indeed a loss of function on the primed FASP site, but it retains significant activity on the non-priming-dependent b-TrCP binding site, leading to an overall protein clearance gain in function. Hence, the CK1εtau mutation alters the balance of the phosphoswitch and drives PER2 into the unstable fraction. Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc. 85

In contrast, CK1d/ε inhibitors block phosphorylation of both the FASP and b-TrCP sites, slowing the degradation of PER2 (Badura et al., 2007; Meng et al., 2010). This current work provides both simple and detailed mathematical models, which, by properly accounting for this phosphoswitch, can be useful in future circadian research as prior modeling studies have been (Gallego et al., 2006a; Hirota et al., 2012; Kim et al., 2012, 2013; Leloup and Goldbeter, 2011; Vanselow et al., 2006). Together with previous modeling studies that consider sequential phosphorylation of PER rather than a phosphoswitch, our work shows how the kinetics of multisite PER2 phosphorylation regulate PER2 stability and hence circadian period (Leloup and Goldbeter, 2011; Vanselow et al., 2006). One feature of the phosphoswitch model is that the mutations, drugs, and environmental stimuli that most affect the circadian period bias PER2 toward one of the two states. This phosphoswitch appears to be a fundamental design principle of the clock, as it shows properties that have been observed in Drosophila and Neurospora circadian rhythms (Chiu et al., 2011; Garbe et al., 2013; Querfurth et al., 2011). Drosophila PER can also act as an interval timer, which is similar to the plateau (the second stage) we observed (Meyer et al., 2006). Furthermore, in a switch reminiscent of the phosphoswitch, temperature compensation and metabolic control of circadian rhythms may also be regulated by alternate splicing of PER and FRQ (Akman et al., 2008; Diernfellner et al., 2005; Majercak et al., 1999; McGlincy et al., 2012). Considering that PER2 stability has a more remarkable effect on circadian period than does overall protein levels (Baggs et al., 2009; Dibner et al., 2009; Forger, 2011; Kim and Forger, 2012), tuning the fraction of stable and unstable PER2 via the phosphoswitch provides an effective mechanism for temperature compensation, pharmacologic manipulation, and metabolic control of circadian rhythms. EXPERIMENTAL PROCEDURES Circadian and Three-Phase Decay Measurement CK1εtau Per2Luc mice in a C57BL/6 background were obtained from Andrew Loudon (Meng et al., 2008). CK1ε wild-type Per2luc mice were obtained by breeding the CK1εtau Per2Luc mice with C57BL/6 inbred mice (Yoo et al., 2004). MEFs were derived and immortalized as described (Proffitt and Virshup, 2012). MEFs were cultured in 35-mm Petri dishes until 80% confluent and then synchronized by dexamethasone (Balsalobre et al., 2000) for 2 hr in 37 C. Cells were washed with PBS and cultured in phenol red-free medium. Culture dishes were then sealed by coverslips using vacuum grease and transferred into the Lumicycle (Actimetrics) for luminescence recording. At the indicated time points, luminescence recording was paused for drug or vehicle addition. All the drugs were added in small volumes into the culture dishes directly without changing the medium. Luminescence data were recorded and exported by Lumicycle data collection software (Ver. 1). Period was assessed by the Lumicycle data analysis software (Actimetrics). The duration of plateau phases were estimated by calculating the instantaneous half-life of PER2 (Figure S1A). All mouse procedures were approved by the SingHealth Institutional Animal Care and Use Committee (IACUC). Data Analysis To track the change of instantaneous half-life, the degradation curves were divided into segments of 1 hr with a sliding window, which was moved at increments of 0.01 hr. Each segment of degradation curve was fitted to the expo-

86 Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc.

nential decay curve to estimate the instantaneous half-life. The plateau phase was defined as the region where the instantaneous half-life was longer than 5 hr (Figures S1A, 1B, and 4D). For the overall half-life of PER2, rather than fitting to an exponential curve, half of the time to reach 25% of initial amount was used because the degradation curves were not exponential (Figures 3C and 5F). The phase of PER2 rhythms was estimated by using FFT-nonlinear least-squares analysis (Izumo et al., 2006). The phases at bottom and peak were used to estimate the duration of rising phase and falling phase (Figure 5C). Mean and error for all experiments in the Lumicycle were calculated based on at least four duplicates. The Simple Mathematical Model of Mammalian Circadian Clocks The simple mathematical model of mammalian circadian clocks (Figure 2A) is described with ordinary differential equations, which can be found in Supplemental Equations. The description of variables and parameters of the simple mathematical model can be found in Tables S1 and S2. The parameters of the model were estimated by simulated annealing method (a stochastic global parameter-searching algorithm) (Gonza´lez et al., 2007), so that the models can generate rhythms of 24-hr and three-stage decay during the accumulation phase (Figures 2 and S2). SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, Supplemental Equations, four figures, and four tables and can be found with this article online at http://dx.doi.org/10.1016/j.molcel.2015.08.022. AUTHOR CONTRIBUTIONS M.Z. and J.K.K. contributed equally to this work. M.Z. designed, performed, and analyzed most of the experiments with input from J.K.K., G.W.L.E., D.B.F., and D.M.V. J.K.K. developed, simulated and analyzed the mathematical models and developed quantification tools for some experimental data (half-life, plateau phase, and phase duration) with input from M.Z., D.M.V., and D.B.F. G.W.L.E. generated mutant PER2 constructs and supervised mouse breeding. M.Z., J.K.K., D.B.F., and D.M.V. wrote the manuscript. D.M.V. and D.B.F. sponsored the project. ACKNOWLEDGMENTS We thank Randal S. Tibbetts for providing the pFASP antibodies and Andrew Loudon for CK1εtau mice and the Per2::luc mice originally developed by Joseph Takahashi. This work was funded by grants IRG10nov023 to D.M.V. from National Medical Research Council, Singapore; a program grant from the Human Frontiers of Science Foundation RPG 24/2012 to D.B.F.; and DMS-0931642 from National Science Foundation to Mathematical Biosciences Institute and J.K.K. Received: September 19, 2014 Revised: June 1, 2015 Accepted: August 25, 2015 Published: October 1, 2015 REFERENCES Akman, O.E., Locke, J.C.W., Tang, S., Carre´, I., Millar, A.J., and Rand, D.A. (2008). Isoform switching facilitates period control in the Neurospora crassa circadian clock. Mol. Syst. Biol. 4, 164. Badura, L., Swanson, T., Adamowicz, W., Adams, J., Cianfrogna, J., Fisher, K., Holland, J., Kleiman, R., Nelson, F., Reynolds, L., et al. (2007). An inhibitor of casein kinase I epsilon induces phase delays in circadian rhythms under free-running and entrained conditions. J. Pharmacol. Exp. Ther. 322, 730–738. Baggs, J.E., Price, T.S., DiTacchio, L., Panda, S., Fitzgerald, G.A., and Hogenesch, J.B. (2009). Network features of the mammalian circadian clock. PLoS Biol. 7, e52.

Baker, C.L., Kettenbach, A.N., Loros, J.J., Gerber, S.A., and Dunlap, J.C. (2009). Quantitative proteomics reveals a dynamic interactome and phasespecific phosphorylation in the Neurospora circadian clock. Mol. Cell 34, 354–363. Balsalobre, A., Marcacci, L., and Schibler, U. (2000). Multiple signaling pathways elicit circadian gene expression in cultured Rat-1 fibroblasts. Curr. Biol. 10, 1291–1294. Barrett, R.K., and Takahashi, J.S. (1995). Temperature compensation and temperature entrainment of the chick pineal cell circadian clock. J. Neurosci. 15, 5681–5692. Bass, J., and Takahashi, J.S. (2010). Circadian integration of metabolism and energetics. Science 330, 1349–1354. Buhr, E.D., Yoo, S.-H., and Takahashi, J.S. (2010). Temperature as a universal resetting cue for mammalian circadian oscillators. Science 330, 379–385. Chen, R., Schirmer, A., Lee, Y., Lee, H., Kumar, V., Yoo, S.-H., Takahashi, J.S., and Lee, C. (2009). Rhythmic PER abundance defines a critical nodal point for negative feedback within the circadian clock mechanism. Mol. Cell 36, 417–430. Chiu, J.C., Ko, H.W., and Edery, I. (2011). NEMO/NLK phosphorylates PERIOD to initiate a time-delay phosphorylation circuit that sets circadian clock speed. Cell 145, 357–370. Dibner, C., Sage, D., Unser, M., Bauer, C., d’Eysmond, T., Naef, F., and Schibler, U. (2009). Circadian gene expression is resilient to large fluctuations in overall transcription rates. EMBO J. 28, 123–134. Diernfellner, A.C.R., Schafmeier, T., Merrow, M.W., and Brunner, M. (2005). Molecular mechanism of temperature sensing by the circadian clock of Neurospora crassa. Genes Dev. 19, 1968–1973. Eide, E.J., Woolf, M.F., Kang, H., Woolf, P., Hurst, W., Camacho, F., Vielhaber, E.L., Giovanni, A., and Virshup, D.M. (2005). Control of mammalian circadian rhythm by CKIepsilon-regulated proteasome-mediated PER2 degradation. Mol. Cell. Biol. 25, 2795–2807. Forger, D.B. (2011). Signal processing in cellular clocks. Proc. Natl. Acad. Sci. USA 108, 4281–4285. Gallego, M., and Virshup, D.M. (2007). Post-translational modifications regulate the ticking of the circadian clock. Nat. Rev. Mol. Cell Biol. 8, 139–148. Gallego, M., Eide, E.J., Woolf, M.F., Virshup, D.M., and Forger, D.B. (2006a). An opposite role for tau in circadian rhythms revealed by mathematical modeling. Proc. Natl. Acad. Sci. USA 103, 10618–10623. Gallego, M., Kang, H., and Virshup, D.M. (2006b). Protein phosphatase 1 regulates the stability of the circadian protein PER2. Biochem. J. 399, 169–175. Garbe, D.S., Fang, Y., Zheng, X., Sowcik, M., Anjum, R., Gygi, S.P., and Sehgal, A. (2013). Cooperative interaction between phosphorylation sites on PERIOD maintains circadian period in Drosophila. PLoS Genet. 9, e1003749. Gonza´lez, J.R., Armengol, L., Sole´, X., Guino´, E., Mercader, J.M., Estivill, X., and Moreno, V. (2007). SNPassoc: an R package to perform whole genome association studies. Bioinformatics 23, 644–645. Hastings, J.W., and Sweeney, B.M. (1957). On the mechanism of temperature independence in a biological clock. Proc. Natl. Acad. Sci. USA 43, 804–811. Hirota, T., Lee, J.W., St John, P.C., Sawa, M., Iwaisako, K., Noguchi, T., Pongsawakul, P.Y., Sonntag, T., Welsh, D.K., Brenner, D.A., et al. (2012). Identification of small molecule activators of cryptochrome. Science 337, 1094–1097. Iitaka, C., Miyazaki, K., Akaike, T., and Ishida, N. (2005). A role for glycogen synthase kinase-3beta in the mammalian circadian clock. J. Biol. Chem. 280, 29397–29402. Isojima, Y., Nakajima, M., Ukai, H., Fujishima, H., Yamada, R.G., Masumoto, K.-H., Kiuchi, R., Ishida, M., Ukai-Tadenuma, M., Minami, Y., et al. (2009). CKIepsilon/delta-dependent phosphorylation is a temperature-insensitive, period-determining process in the mammalian circadian clock. Proc. Natl. Acad. Sci. USA 106, 15744–15749.

Izumo, M., Sato, T.R., Straume, M., and Johnson, C.H. (2006). Quantitative analyses of circadian gene expression in mammalian cell cultures. PLoS Comput. Biol. 2, e136. Kaasik, K., Kivima¨e, S., Allen, J.J., Chalkley, R.J., Huang, Y., Baer, K., Kissel, ek, L.J., and Fu, Y.H. (2013). Glucose H., Burlingame, A.L., Shokat, K.M., Pta´c sensor O-GlcNAcylation coordinates with phosphorylation to regulate circadian clock. Cell Metab. 17, 291–302. Kim, J.K., and Forger, D.B. (2012). A mechanism for robust circadian timekeeping via stoichiometric balance. Mol. Syst. Biol. 8, 630. Kim, E.Y., Jeong, E.H., Park, S., Jeong, H.-J., Edery, I., and Cho, J.W. (2012). A role for O-GlcNAcylation in setting circadian clock speed. Genes Dev. 26, 490–502. Kim, J.K., Forger, D.B., Marconi, M., Wood, D., Doran, A., Wager, T., Chang, C., and Walton, K.M. (2013). Modeling and validating chronic pharmacological manipulation of circadian rhythms. CPT Pharmacometrics Syst. Pharmacol. 2, e57. Lee, C., Etchegaray, J.P., Cagampang, F.R., Loudon, A.S., and Reppert, S.M. (2001). Posttranslational mechanisms regulate the mammalian circadian clock. Cell 107, 855–867. Lee, H.M., Chen, R., Kim, H., Etchegaray, J.P., Weaver, D.R., and Lee, C. (2011a). The period of the circadian oscillator is primarily determined by the balance between casein kinase 1 and protein phosphatase 1. Proc. Natl. Acad. Sci. USA 108, 16451–16456. Lee, Y., Chen, R., Lee, H.-M., and Lee, C. (2011b). Stoichiometric relationship among clock proteins determines robustness of circadian rhythms. J. Biol. Chem. 286, 7033–7042. Leloup, J.-C., and Goldbeter, A. (2011). Modelling the dual role of Per phosphorylation and its effect on the period and phase of the mammalian circadian clock. IET Syst. Biol. 5, 44–49. Leo, A., Hansch, C., and Elkins, D. (1971). Partition coefficients and their uses. Chem. Rev. 71, 526–616. Lowrey, P.L., Shimomura, K., Antoch, M.P., Yamazaki, S., Zemenides, P.D., Ralph, M.R., Menaker, M., and Takahashi, J.S. (2000). Positional syntenic cloning and functional characterization of the mammalian circadian mutation tau. Science 288, 483–492. Majercak, J., Sidote, D., Hardin, P.E., and Edery, I. (1999). How a circadian clock adapts to seasonal decreases in temperature and day length. Neuron 24, 219–230. McGlincy, N.J., Valomon, A., Chesham, J.E., Maywood, E.S., Hastings, M.H., and Ule, J. (2012). Regulation of alternative splicing by the circadian clock and food related cues. Genome Biol. 13, R54. Mehra, A., Shi, M., Baker, C.L., Colot, H.V., Loros, J.J., and Dunlap, J.C. (2009). A role for casein kinase 2 in the mechanism underlying circadian temperature compensation. Cell 137, 749–760. Meng, Q.-J., Logunova, L., Maywood, E.S., Gallego, M., Lebiecki, J., Brown, T.M., Sla´dek, M., Semikhodskii, A.S., Glossop, N.R.J., Piggins, H.D., et al. (2008). Setting clock speed in mammals: the CK1 epsilon tau mutation in mice accelerates circadian pacemakers by selectively destabilizing PERIOD proteins. Neuron 58, 78–88. Meng, Q.-J., Maywood, E.S., Bechtold, D.A., Lu, W.-Q., Li, J., Gibbs, J.E., Dupre´, S.M., Chesham, J.E., Rajamohan, F., Knafels, J., et al. (2010). Entrainment of disrupted circadian behavior through inhibition of casein kinase 1 (CK1) enzymes. Proc. Natl. Acad. Sci. USA 107, 15240–15245. Meyer, P., Saez, L., and Young, M.W. (2006). PER-TIM interactions in living Drosophila cells: an interval timer for the circadian clock. Science 311, 226–229. Nishi, H., Fong, J.H., Chang, C., Teichmann, S.A., and Panchenko, A.R. (2013). Regulation of protein-protein binding by coupling between phosphorylation and intrinsic disorder: analysis of human protein complexes. Mol. Biosyst. 9, 1620–1626. Pilorz, V., Cunningham, P.S., Jackson, A., West, A.C., Wager, T.T., Loudon, A.S.I., and Bechtold, D.A. (2014). A novel mechanism controlling resetting speed of the circadian clock to environmental stimuli. Curr. Biol. 24, 766–773.

Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc. 87

Pittendrigh, C.S. (1954). On temperature independence in the clock system controlling emergence time in Drosophila. Proc. Natl. Acad. Sci. USA 40, 1018–1029.

Toh, K.L., Jones, C.R., He, Y., Eide, E.J., Hinz, W.A., Virshup, D.M., Pta´cek, L.J., and Fu, Y.H. (2001). An hPer2 phosphorylation site mutation in familial advanced sleep phase syndrome. Science 291, 1040–1043.

Pittendrigh, C.S. (1960). Circadian rhythms and the circadian organization of living systems. Cold Spring Harb. Symp. Quant. Biol. 25, 159–184.

Tosini, G., and Menaker, M. (1998). The tau mutation affects temperature compensation of hamster retinal circadian oscillators. Neuroreport 9, 1001– 1005.

Proffitt, K.D., and Virshup, D.M. (2012). Precise regulation of porcupine activity is required for physiological Wnt signaling. J. Biol. Chem. 287, 34167–34178. Qin, X., Mori, T., Zhang, Y., and Johnson, C.H. (2015). PER2 Differentially Regulates Clock Phosphorylation versus Transcription by Reciprocal Switching of CK1ε Activity. J. Biol. Rhythms 30, 206–216. Querfurth, C., Diernfellner, A.C.R., Gin, E., Malzahn, E., Ho¨fer, T., and Brunner, M. (2011). Circadian conformational change of the Neurospora clock protein FREQUENCY triggered by clustered hyperphosphorylation of a basic domain. Mol. Cell 43, 713–722. Ralph, M.R., and Menaker, M. (1988). A mutation of the circadian system in golden hamsters. Science 241, 1225–1227. Shanware, N.P., Hutchinson, J.A., Kim, S.-H., Zhan, L., Bowler, M.J., and Tibbetts, R.S. (2011). Casein kinase 1-dependent phosphorylation of familial advanced sleep phase syndrome-associated residues controls PERIOD 2 stability. J. Biol. Chem. 286, 12766–12774. Sickmeier, M., Hamilton, J.A., LeGall, T., Vacic, V., Cortese, M.S., Tantos, A., Szabo, B., Tompa, P., Chen, J., Uversky, V.N., et al. (2007). DisProt: the Database of Disordered Proteins. Nucleic Acids Res. 35, D786–D793. Sweeney, B.M., and Hastings, J.W. (1960). Effects of temperature upon diurnal rhythms. Cold Spring Harb. Symp. Quant. Biol. 25, 87–104.

88 Molecular Cell 60, 77–88, October 1, 2015 ª2015 Elsevier Inc.

Vanselow, K., and Kramer, A. (2007). Role of phosphorylation in the mammalian circadian clock. Cold Spring Harb. Symp. Quant. Biol. 72, 167–176. Vanselow, K., Vanselow, J.T., Westermark, P.O., Reischl, S., Maier, B., Korte, T., Herrmann, A., Herzel, H., Schlosser, A., and Kramer, A. (2006). Differential effects of PER2 phosphorylation: molecular basis for the human familial advanced sleep phase syndrome (FASPS). Genes Dev. 20, 2660–2672. Xu, Y., Toh, K.L., Jones, C.R., Shin, J.-Y., Fu, Y.H., and Pta´cek, L.J. (2007). Modeling of a human circadian mutation yields insights into clock regulation by PER2. Cell 128, 59–70. Xue, B., Dunbrack, R.L., Williams, R.W., Dunker, A.K., and Uversky, V.N. (2010). PONDR-FIT: a meta-predictor of intrinsically disordered amino acids. Biochim. Biophys. Acta 1804, 996–1010. Yoo, S.-H., Yamazaki, S., Lowrey, P.L., Shimomura, K., Ko, C.H., Buhr, E.D., Siepka, S.M., Hong, H.K., Oh, W.J., Yoo, O.J., et al. (2004). PERIOD2:LUCIFERASE real-time reporting of circadian dynamics reveals persistent circadian oscillations in mouse peripheral tissues. Proc. Natl. Acad. Sci. USA 101, 5339–5346. Zatz, M., Lange, G.D., and Rollag, M.D. (1994). What does changing the temperature do to the melatonin rhythm in cultured chick pineal cells? Am. J. Physiol. 266, R50–R58.