Circadian clocks are present across almost all life forms and allow them to adapt their activities with daily changes in their surrounding environment. In eukaryotes, the molecular circadian pacemaker is a self-sustained transcriptional feedback loop that repeats itself every ∼24 h, thus producing molecular, physiological and behavioral rhythms that closely match the duration of the Earth’s revolution around its own axis (Dunlap, 1999). In Drosophila, the circadian clock comprises four core components (Hardin, 2011). The transcription factors CLOCK (CLK) and CYCLE (CYC) form a heterodimer that binds to E-box motifs upstream of the period (per) and timeless (tim) genes to promote their transcription. PER and TIM heterodimerize in the cytoplasm before entering the nucleus. Once inside the nucleus, the PER/TIM heterodimer binds to CLK/CYC and thus inhibits per and tim transcription. PER, TIM, and CLK undergo various post-translational modifications that affect the period of the clock, the most notable being the progressive phosphorylation driven by kinases such as DOUBLETIME (DBT) (Kloss et al., 1998; Price et al., 1998), SHAGGY (Martinek et al., 2001), Proline-directed kinases including p38 MAP Kinase (Ko et al., 2010; Dusik et al., 2014), CASEIN KINASE 1 alpha (CK1α) (Lam et al., 2018), CASEIN KINASE 2 (CK2) (Lin et al., 2002; Lin et al., 2005; Akten et al., 2003; Szabó et al., 2013), AMPK (Cho et al., 2019) and NEMO (NMO) (Chiu et al., 2011; Yu et al., 2011).

Circadian clocks exhibit three key properties (Pittendrigh, 1960). First, these clocks free-run with a period of ∼24 h in the absence of external temporal cues (referred to as Zeitgebers) such as the daily light and temperature cycles. Second, these Zeitgebers can reset the clocks, thus anchoring the clock’s phase to the day/night cycle. Third, the period of circadian clocks is temperature-compensated. As ambient temperature increases, most enzymatic reactions speed up. However, under constant conditions, circadian clocks exhibit ∼24 h periodicities over a wide range of physiological temperatures, even in ectotherms. This was first observed by C. Pittendrigh while studying eclosion rhythms of Drosophila pseudoobscura (Pittendrigh, 1954).

The mechanism of temperature compensation is one of the long lasting mysteries of circadian rhythms. Different models have been proposed. For example, the network model proposes that multiple temperature-sensitive reactions cancel each other with changes in temperature. Early studies on Dinoflagellates indeed suggested that two temperature-sensitive reactions with opposite effects would help achieving constant period (Hastings and Sweeney, 1957). At the biochemical level, phosphorylation mediated by CK1 and CK2 has received the most attention across species. Mutations in the β1 and α subunits of CK2 in Neurospora alter temperature compensation (Mehra et al., 2009). Temperature compensation in Arabidopsis is modulated by the inhibitory effect of CK2 phosphorylation on CIRCADIAN ASSOCIATED CLOCK A1 (CCA1) binding. CCA1 binding to its target promoters increases with temperature, which is kept in check by the opposing effect of CK2 phosphorylation (Portolé and Má, 2010). In mammals, phosphorylation of a PER2 peptide by CK1δ is temperature-insensitive, owing to opposite temperature sensitivity of substrate and product binding by this kinase (Isojima et al., 2009; Shinohara et al., 2017). Interestingly, the CK1ε ^Tau mutant shows an undercompensation phenotype (Tosini and Menaker, 1998; Zhou et al., 2015). Furthermore, CK1δ/ε activity on PER2 results in a temperature-dependent phosphoswitch involving two phosphorylation sites (Eide et al., 2005; Zhou et al., 2015; Masuda et al., 2020). One of them is Serine (S)-478, also called phosphodegron. S478 phosphorylation causes βTrCP (β-Transducing repeat Containing Protein)-dependent ubiquitination and degradation of mPER2 (Eide et al., 2005; Reischl et al., 2007; Masuda et al., 2020). The second is S659, the mouse homolog to the site of Familial Advanced Sleep Phase Syndrome (FASPS) mutations (Toh et al., 2001). S659 phosphorylation inhibits phosphorylation of the phosphodegron, and thus PER2 degradation (Shanware et al., 2011; Zhou et al., 2015). S478 is preferentially phosphorylated at colder temperatures while FASPS domain phosphorylation increases at warmer temperatures, consequently balancing out PER stability (Zhou et al., 2015; Narasimamurthy et al., 2018). Along with phosphorylation, other processes such as sumoylation in plants (Hansen et al., 2017), nuclear cytoplasmic ratio of PER (Giesecke et al., 2021) and TIM (Singh et al., 2019), PER intermolecular and PER-TIM interactions (Gekakis et al., 1995; Huang et al., 1993, Huang et al., 1995) have been proposed to modulate temperature compensation of circadian rhythms in flies.

Over the past several decades, circadian mutants have fuelled our understanding of the molecular clockwork. Therefore, to gain better understanding into the mechanism of temperature compensation in Drosophila we turned towards Drosophila mutants known to have a temperature compensation defect. Interestingly, of the numerous mutations impacting circadian rhythm period, only a subset of mutants show a defect in temperature compensation (Bao et al., 2001; Ewer et al., 1990; Hamblen et al., 1998; Konopka et al., 1994; Matsumoto et al., 1999; Peixoto et al., 1998; Rothenfluh et al., 2000; Sawyer et al., 1997). We were intrigued by the mutant per ^SLIH (Some Like It Hot, Figure 1A) due to its striking undercompensation phenotype: circadian period length decreases as temperature increases in this mutant instead of remaining approximately the same. (Hamblen et al., 1998). The per^SLIH mutation causes a Serine (S) to Tyrosine (Y) substitution at the 45th residue, which is a part of Drosophila PER’s phosphodegron (S44, 45, and 47) (Chiu et al., 2008). The PER phosphodegron is regulated in a similar manner as its mammalian homolog (Chiu et al., 2011): it is phosphorylated by the CK1δ/ε homolog DBT (Kloss et al., 1998; Price et al., 1998; Chiu et al., 2008), and its phosphorylation is inhibited by the per^short (per^s) phosphocluster, which is equivalent to the mammalian FASPS domain (Zhou et al., 2015). Phosphorylation events at the PER phosphodegron, S47 in particular, are critical for Supernumerary limbs (SLIMB)-mediated PER degradation and thus period length (Grima et al., 2002; Ko et al., 2002; Chiu et al., 2008). SLIMB is the fly homolog of βTrCP. Here, we show that, unexpectedly, the residues of the Drosophila phosphodegron are functionally heterogeneous and differentially modulate temperature compensation, even though they all lengthen circadian period. In addition, we found that the per^s phosphocluster, which negatively regulates S47 phosphorylation, also regulates temperature compensation. Furthermore, we show that substituting PER’s S47 residue to an Alanine (A) results in an overcompensated circadian period, i.e., circadian period lengthens when ambient temperature increases. This overcompensation phenotype correlates with increased hyper-phosphorylated PER levels in vivo at warm temperature and overcompensated PER degradation in a well-established cell culture model. These results converge to support an important role of the phosphodegron in temperature compensation in Drosophila.

Phosphorylation events play an important role in the temperature compensation of the circadian clock across phyla (Tosini and Menaker, 1998; Mehra et al., 2009; Portolé and Má, 2010; Zhou et al., 2015; Masuda et al., 2020). In Drosophila, PER phosphorylation has been studied in detail, and interestingly three per mutations affecting two PER domains undergoing DBT phosphorylation are undercompensated. per ^SLIH causes the most robust undercompensation phenotype (Hamblen et al., 1998), while per ^s and per ^T have milder effects (Ewer et al., 1990; Konopka et al., 1994; Rothenfluh et al., 2000; Bao et al., 2001). We therefore systematically tested phosphorylated residues in both the PER phosphodegron and the per ^s domain to understand their impact on temperature compensation. Overall, our results point at a central role for residue S47 through temperature-dependent modulation of PER degradation, and a differential, yet to be elucidated, role of S44 and S45. In addition we found that phosphorylated residues in the per ^s phosphorylation cluster also play a role in temperature compensation, as expected from their ability to inhibit S47 phosphorylation (Chiu et al., 2011).

Indeed, the S47A substitution resulted in a striking overcompensation. This suggests that phosphorylation at this residue is important for the compensation. Actually, we found that PER levels were exaggeratedly temperature-dependent in vivo, with an increase in hyper-phosphorylated PER levels at high temperature. Moreover, PER degradation induced by DBT phosphorylation was slowed down at high temperature by the S47A substitution, as opposed to WT PER in a Drosophila cell culture assay. These results thus suggest that the S47 residue impacts the kinetics of PER degradation in a temperature-dependent manner, protecting degradation from being overcompensated. It should be noted that our molecular observations were made with whole-head extracts (and S2 cells), but rhythmic locomotor behavior is controlled by a small number of circadian pacemaker neurons, the small ventral lateral neurons (sLNvs), that contribute minimally to whole-head protein extracts (Zeng et al., 1994; Renn et al., 1999; Stoleru et al., 2005). Differences have been noted between the molecular clock of these neurons and other circadian neurons and tissues (Lin et al., 2002; Lim et al., 2011; Top et al., 2016; Reviewed in; Top and Young, 2018). We therefore cannot exclude that S47A impacts differently PER rhythms in sLNvs than in other tissues, but we note importantly that DBT-mediated PER phosphorylation and degradation is mechanistically conserved in clock neurons and peripheral tissues, and closely recapitulated in S2 cells (Kloss et al., 1998, 2001; Price et al., 1998; Ko et al., 2002; Chiu et al., 2008, 2011; Garbe et al., 2013).

Surprisingly, our results from cycloheximide chase assays suggest that S47A is less stable than WT at 18°C, while at 29°C we observed a mild increase in stability of S47A compared to WT. These results appear to differ from those reported in Chiu et al. (2008) that indicated a robust stabilization of S47A at 25°C. These discrepancies could be the result of slightly different growth conditions or other changes in the state of our S2 cell cultures, which could affect expression of endogenous proteins important for PER degradation. Nevertheless, since the increase in stability of the S47A protein at 29°C correlates well with the overcompensation of circadian period in S47A flies, our observations in S2 cells provide a coherent mechanistic explanation for S47A’s temperature-compensation phenotype. However, that circadian behavior period is lengthened at all temperatures in S47A mutant flies seems at odd with the kinetics of S47A degradation observed in cell culture, which is accelerated at 18°C or similar to wild-type PER at 29°C (Figures 5B,C). The cycloheximide chase assay we used models DBT-dependent PER degradation in the absence of any feedback such as PER repression of its own transcription. The S47A mutation could thus have a lengthening effect on period independent of its impact on temperature compensation, for example by modulating the ability of PER to function as a transcriptional repressor in vivo. A different mechanism for period determination and temperature compensation by the phosphodegron is in agreement with our observation that S to A substitutions at residues 44, 45 or 47 all cause period lengthening, while the valence of the temperature compensation phenotype is dependent on the specific residue being substituted.

Mutations in the per ^s phosphocluster provide further support to the key role played by phosphorylation at S47 in temperature compensation. Indeed, mutations in the per ^s phosphocluster, which inhibit S47 phosphorylation, have opposite effects on temperature compensation than S47A: flies carrying these mutations are undercompensated. Unexpectedly though, the phosphomimetic substitution S47D did not produce an undercompensation phenotype. Rather, an overcompensation phenotype was observed. This could be explained by the substitution to Aspartate not fully mimicking the structure of a phosphorylated Serine, or that constantly mimicking S47 phosphorylation impacts thermal compensation of the clock through a different mechanism than the S47A substitution. Notably it is not uncommon that phosphomimetic mutations do not produce opposite phenotype than phosphoinhibitory mutations, including in circadian proteins (Lin et al., 2005; Chiu et al., 2011; Top et al., 2018; Cai et al., 2021). Thus, S47D could lead to overcompensation via yet unknown, different mechanism. Given the opposite effect of S47A and S47D on SLIMB binding (Chiu et al., 2008) and consequently period length, PER-SLIMB interaction and thus PER degradation is unlikely to be the mechanism mediating S47D temperature overcompensation.

Another unexpected observation is that Serine to Alanine substitution at residues 44 and 45 have opposite effects on temperature compensation than S47A, even though all these substitutions lengthen circadian period. Residues of the phosphodegron are thus functionally heterogeneous when it comes to temperature compensation. Indeed, unlike S47A, S45A did not cause obvious change in temperature-dependent PER stability in vivo or in cell culture. It should be noted that unlike S47, S44 or S45 phosphorylation has not been unambiguously demonstrated (Chiu et al., 2008). It remains possible that these residues are not phosphorylated and play a structural role in the phosphodegron, and this could explain functional diversity in temperature compensation between S44/45 and S47. It is also possible that S44/45’s mechanism of action on period length and/or temperature compensation is independent of S47. Clearly, further studies are needed to elucidate S44 and S45’s exact function in thermal compensation. Of note, the S45F substitution, which introduces a more bulky amino acid than S45A, caused a more severe phenotype that better mimicked the per ^SLIH phenotype resulting from a S45Y substitution [Figure 1B; Table 1 (Hamblen et al., 1998)]. This could indicate an important structural role for this amino acid, perhaps controlling specific protein-protein interactions not implicated in PER degradation. We note that a putative Nuclear Localization Signal (NLS) is located close to the phosphodegron (Figure 1A, red mark) and perhaps a bulky amino acid at S45 can interfere in a temperature dependent manner with nuclear translocation. However, at least in S2 cells, this NLS has little impact on nuclear localization (Chang and Reppert, 2003).

Interestingly, our work, combined with previous work in mammals, converge to support the model that CKIδ/ε-mediated phosphorylation of PER proteins and their proteasomal degradation are critical for temperature compensation of the circadian clock. Surprisingly, however, there appear to be important mechanistic differences. Indeed, in Drosophila, mutation of the key Serine residue (S47) in the phosphodegron increases temperature compensation, while in mammals mutation of the homologous S478 residue has the opposite effect [Figures 1B, 2B; Table 1 (Zhou et al., 2015; Masuda et al., 2020)]. Intriguingly, in mammalian cell culture assays, phosphodegron mutation (S478A in mammals) causes degradation kinetics to become essentially insensitive to temperature (Zhou et al., 2015), while S47A degradation in flies is mildly overcompensated. However, in mammals, wild-type mPER2 degradation kinetics accelerates when temperature drops, while in flies we observed the opposite effect, perhaps because the phosphodegrons, though conserved, are located in different location in the PER and mPER2 proteins. This explains why mutations that both renders PER degradation less sensitive to temperature have opposite effect on temperature compensation.

Finally, it is important to point out that so far, any mutation or manipulation of phosphorylation of key circadian protein in any system only partially disrupts temperature compensation, or even increases it. It is therefore clear that multiple mechanisms are implicated, and recent results by Giesecke and collaborators (Giesecke et al., 2021) in flies for example point to an important role for nuclear export. It is therefore clear that temperature compensation is proving to be a complex, systemic process, as first proposed by Hastings and Sweeny in their seminal work on Dinoflagellates (Hastings and Sweeney, 1957).

Whether difference in period at 29 and 18°C were statistically significant was determined by using a two-way ANOVA followed by Sidak’s multiple comparison tests. Since in Figure 1 control flies showed a significant period shortening at warm temperatures compared to cold, we did factor this drift prior to run statistical test by subtracting the average period lengthening observed in S45SI, and S45SII (1.03 h) from the period observed at 18°C for each genotypes.

For graphical representation in Figures 1, 2, mean difference between 29 and 18°C were plotted on the Y axis and Standard Error (SE) of difference were used for error bars. Mean difference and SE of difference were obtained from the Two-way Anova statistics details is Prism. Western blots were analyzed using Two-way ANOVA with Tukey’s multiple comparison test. Rhythms in PER phosphorylation and abundance in Figure 3 were analyzed by JTK-CYCLE as in (Hughes et al., 2010).