Alternative splicing of Clock transcript mediates the response of circadian clocks to temperature changes

We first sought to determine whether Clk transcripts exhibit alternative splicing, as in the case of other key clock genes such as per and tim. When Drosophila Clk transcripts were first cloned in 1998 by three independent labs, two different cDNA products were identified (34–36). The longer transcript, which we termed Clk-long, encodes the canonical CLK protein. The other slightly shorter transcript, hereinafter termed Clk-cold, encodes a CLK protein with a four amino acid (aa) deletion at aa13-16 (Fig. 1A). The annotation of the genomic Clk sequence revealed a potential alternative 3’ splice site in exon 2 of Clk, which may have produced the two cDNA of different lengths. Indeed, both transcripts are expressed in fly heads, according to deep sequencing of circadian transcriptome in Hughes et al. (37). Cell-specific RNA-seq data from Wang et al. (38) also indicated that both Clk-long and Clk-cold are expressed in several circadian neuronal cell types, including small lateral ventral neurons (sLN_v), the key pacemaker neurons (SI Appendix, Fig. S1A↗). Given that the four aa deletion of CLK-cold is adjacent to the bHLH DNA-binding domain (aa17-62) (35, 36), we hypothesize that differential expression of these two transcripts could impact the function of the molecular clock.

To further confirm the expression of both Clk-long and Clk-cold, we generated Clk cDNA fragments that span exons 1 and 2 from RNA extracted from whole heads of w¹¹¹⁸ flies collected at a time-point when Clk mRNA expression is high (35) (SI Appendix, Fig. S1B↗). We noticed three potentially different cDNAs when we analyzed the amplified cDNAs on agarose gel (SI Appendix, Fig. S1C↗). Sanger sequencing revealed that the bottom band represents Clk-cold and the middle band represents Clk-long (SI Appendix, Fig. S1D↗). The top band represents a hybrid of Clk-cold and Clk-long, potentially an PCR artifact that exhibits slower mobility on the agarose gel. These data support published sequencing data (37, 38) reporting Clk-long and Clk-cold are expressed in fly heads.

To determine whether AS of Clk is temperature-sensitive, we evaluated daily relative abundance of Clk-long and Clk-cold from head extracts of w¹¹¹⁸ flies entrained under LD at 25 and 10 °C, respectively (Fig. 1B). 10 °C was chosen to better simulate a more naturalistic cold temperature, e.g., morning of a spring day, during which Drosophila flies are viable and active (39). Nested qPCR assays targeting the alternative 3′ splice site region allow quantitative analysis of both transcripts (SI Appendix, Fig. S1E↗). At 25 °C, Clk-long and Clk-cold were expressed at 1:2 ratio. However, the relative abundance of Clk-long decreases and Clk-cold becomes even more prevalent at 10 °C. AS of Clk did not exhibit daily oscillation at constant temperature, given the ratio does not oscillate under LD condition, as determined by rhythmicity analysis incorporating nonparametric methods (RAIN) (40) (25 °C: P = 0.81; 10 °C: P = 0.86, RAIN).

Even in endothermic organisms such as mice, body temperature rhythm was found to drive rhythmic AS of over 1,000 exons (12). We therefore hypothesize that environmental temperature cycles drive rhythmic AS of Clk in ectothermic Drosophila. We measured expression of Clk-long and Clk-cold under semi-natural conditions, where the incubators were set to mimic a typical day in May in Davis, CA (weatherspark.com↗), with diurnal temperatures ranging from 10 to 25 °C (Fig. 1C). We observed steady increase of relative Clk-long in response to increasing temperature in the first half of the day. As temperature decreases in the second half of the day, relative level of Clk-long decreases. Under environmental temperature cycles, AS of Clk is rhythmic (P = 0.034, RAIN). These data suggest that AS of Clk is sensitive to environmental temperature changes.

Since CLK-cold is missing four amino acids (aa 13-16) adjacent to the N-terminal bHLH DNA binding domain (aa17-62), we hypothesize that CLK-cold displays altered CLK-DNA binding activity. To test this hypothesis, we performed CLK chromatin immunoprecipitation (CLK-ChIP) followed by qPCR using extracts from adult fly heads collected at 25 °C vs. 10 °C. In agreement with published data (41), we observed rhythmic CLK occupancy at the per CRS, a region within the promoter critical for generating rhythmic per expression (42) and at the vri E-box (CLK binding motif) at 25 °C (Fig. 2 A and B) (per CRS: P = 0.014; vri E-box: P = 0.002, RAIN). However, we observed dampening of daily rhythmicity of CLK occupancy at 10 °C (per CRS: P = 0.582; vri E-box: P = 0.142, RAIN), likely due to significantly higher CLK occupancy at both circadian promoters at ZT4. Our ChIP data suggest CLK-cold exhibits elevated DNA binding activity as compared to CLK-long, most notably at ZT4.

We next asked whether altered DNA binding activity regulates transcriptional activity of each CLK isoforms. We first assayed transcriptional activity of CLK in Drosophila S2 cells using a per-luc reporter assay (36). We compared per-luc reporter gene activity in S2 cells expressing CLK-long or CLK-cold (Fig. 2C). The 2.7-fold increased reporter gene activity in CLK-cold suggests a significantly elevated transcriptional activity of CLK-cold, as compared to CLK-long. Transcriptional activity of CLK-cold was also inferred by measuring expression of known CLK target genes in flies entrained in 12 h:12 h LD at 25 °C vs. 10 °C, respectively (Fig. 2D). At 10 °C where CLK-cold is elevated, mRNA levels of CLK targets, including per, vrille (vri), clockwork orange (cwo), goliath (dgol), are significantly higher than the levels observed at 25 °C at multiple time-points. Similarly, total tim and pdp1 mRNA levels are elevated at 10 °C, despite the fact that these two genes have temperature-sensitive AS (SI Appendix, Fig. S2↗). In contrast, mRNA levels of non-CLK targets including Clk and cryptochrome (cry) did not increase at 10 °C (Fig. 2E). This strongly indicates that the observed elevated expression at cold temperature is not a general phenomenon. Taken together, our results revealed that cold-induced AS of Clk results in a CLK protein that binds more readily to DNA at an early part of the day-night cycle, thereby increasing expression of CLK target genes.

We next sought to explain how the four-aa deletion in CLK-cold alters CLK transcriptional output. Since CLK phosphorylation status displays daily rhythmicity and is tightly correlated with its transcriptional activity (28, 30, 43), we investigated whether these four amino acids overlap with any kinase-targeted motif. Kinase prediction using group-based phosphorylation site predicting and scoring (GPS) 5.0 (44) showed that the serine 13 (S13) residue could be a potential substrate of kinases from several kinase families (SI Appendix, Table S1↗). Previous studies suggested that NEMO (45), SGG (46–48), CK2 (49), and CK1α (50) could be CLK kinases and CLK kinases could be recruited by PER to phosphorylate CLK. Among these kinases, GPS 5.0 kinase prediction algorithm identified CK1α to be the most likely candidate to phosphorylate CLK(S13).

To first determine whether phosphorylation of S13 can regulate CLK transcriptional activity, we generated Clk plasmids expressing nonphosphorylatable S13 to Alanine (A) or phosphomimetic S13 to Aspartic Acid (D) mutations, both in the context of the CLK-long isoform, and tested their transcriptional activity using per-luc luciferase assays in Drosophila S2 cells. Whereas CLK(S13A) was found to exhibit higher transcriptional activity when compared to CLK-long (WT), CLK(S13D) had significantly lower transcriptional activity (Fig. 3A). Our results suggest S13 could indeed be a potential phosphorylation site that regulates CLK transcriptional activity. We then performed a series of experiments to determine whether CLK(S13) is a bona fide CK1α-dependent phosphorylation site. We first determined whether CLK interacts with CK1α by performing coimmunoprecipitation (coIP) assays using protein extracts from Drosophila S2 cells coexpressing CLK-V5 and CK1α-cmyc (Fig. 3 B–D). We detected interactions between CLK and CK1α when using CLK-V5 as bait (Fig. 3 B and C). Reciprocal coIP using CK1α-cmyc as bait resulted in the same conclusion (Fig. 3 B and D). Control experiments were performed using extracts of S2 cells expressing either of the proteins alone to demonstrate minimal nonspecific binding (Fig. 3 B–D).

Next, we determined whether CLK is phosphorylated by CK1α by assessing CLK mobility shift using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). We analyzed CLK in protein extracts of Drosophila S2 cells expressing either CLK alone or CLK coexpressed with CK1α. We observed slower-migrating CLK isoforms on a regular SDS-PAGE gel (SI Appendix, Fig. S3↗), likely representing phosphorylated CLK. Phos-Tag SDS-PAGE gel (53) was used to enhance phosphorylation-dependent mobility shift (Fig. 3 E and F). In addition, to test whether CK1α catalytic activity is responsible for the observed mobility shift, we coexpressed CLK with either CK1α(WT) or CK1α(K49R), a kinase-dead variant (50). We observed substantial slower-migrating CLK species in the presence of CK1α(WT). The amount of slower-migrating species was significantly reduced with CK1α(K49R) coexpression. These results indicate that CK1α kinase activity is required for CLK mobility shift.

To specifically determine whether CLK(S13) is phosphorylated by CK1α, we leveraged mass spectrometry (MS) to identify CK1α-dependent phosphorylation sites of CLK expressed in Drosophila S2 cells. This cell culture system has previously been used to map physiologically relevant phosphorylation sites on Drosophila PER (47, 54, 55), TIM (56) and CLK (51, 57). We coexpressed CLK tagged with FLAG epitope with either CK1α(WT) or kinase dead CK1α(K49R) in S2 cells and performed FLAG affinity purifications prior to MS analysis. We identified eight phosphorylation sites on CLK [SI Appendix, Table S2↗ and (58)]. Among them, we identified four sites that exhibited elevated phosphopeptide abundance when coexpressed with CK1α(WT) as compared to CK1α(K49R) [Fig. 3G and (58)]. These CK1α-dependent sites include S13, which is next to the bHLH DNA binding domain (35, 36); S258 and S311 next to PAS B protein binding domain (35, 36); S476 next to the nuclear localization signal (NLS) (31).

Finally, to further validate the phosphorylation of S13, we generated a S13 phosphospecific antibody (α-pS13) and assayed CLK(S13) phosphorylation in protein extracts of Drosophila S2 cells expressing Clk-V5 (WT) with or without ck1α (Fig. 3 H and I). Immunoblotting showed that CLK(pS13) significantly increased when ck1α was coexpressed (Fig. 3H, lanes 1 and 2). Importantly, there was little to no α-pS13 signal detected in extracts of S2 cells coexpressing Clk(S13A) and ck1α (Fig. 3 H, lane 3, Top), suggesting that α-pS13 antibody is phosphospecific. Taken together, our results strongly support that CLK(S13) is a CK1α-dependent phosphorylation site.

So far, our results suggest that S13 phosphorylation reduces CLK transcriptional activity. We also show that temperature-sensitive AS at cold temperature led to increased production of CLK-cold that lacks this inhibitory S13 phosphorylation, thus promoting CLK target mRNA expression in the cold. To characterize the function of CLK(S13) phosphorylation in vivo, we generated transgenic fly lines expressing nonphosphorylatable CLK(S13A) or phosphomimetic CLK(S13D) variants. These mutated Clk genes are expressed under endogenous Clk promotor (51). p{Clk(X)-V5} (X represents WT, S13A or S13D variants) transgenic fly lines were crossed into Clk^out background (51) to remove endogenous Clk expression. Next, we monitored daily locomotor activity rhythms of Clk transgenic flies, given it is a robust behavioral output of the circadian clock (59) (Fig. 4A and SI Appendix, Table S3↗). Flies were entrained for 4 d in 12 h:12 h light:dark (LD) cycles followed by release in 7 d in constant darkness (DD) at 25 °C to assess free-running rhythms. As expected, Clk^out null mutant exhibited arrhythmic locomotor activity in DD, similar to published results (51). Clk^out flies expressing two copies of Clk(WT) transgene displayed robust daily activity rhythms with a ~24-h period, indicating effective rescue of the arrhythmic Clk^out mutation (Fig. 4A and SI Appendix, Table S3↗). Even one copy of Clk(WT) transgene was sufficient for the rescue of arrhythmicity of the Clk^out null mutant, although two copies of Clk(WT) transgenes resulted in stronger behavioral rhythm. As compared to Clk(WT), Clk(S13D) flies with two copies of the transgene showed significantly dampened rhythm and period-lengthening of 1.2 h (P < 0.001, One-Way ANOVA), while Clk(S13D) flies with one copy of the transgene were mostly arrhythmic. Clk(S13A) flies with two or one copy of the transgene also displayed period-lengthening by 0.9 to 1.1 h and reduced rhythmicity as compared to Clk(WT) flies (two copies: P < 0.001; one copy: P < 0.001, One-Way ANOVA). Finally, both Clk(S13D) and Clk(S13A) can rescue arrhythmic Clk^out under LD, indicating light entrainment is not affected upon genetic manipulation of CLK(S13) phosphorylation. Taken together, our data suggest that CLK(S13) phosphorylation is required for robust circadian timekeeping.

To determine whether CLK(S13) phosphorylation-mediated downregulation in transcriptional activity observed in Drosophila S2 cells translates to whole animals, we quantified the mRNA of known CLK targets including per, tim, and vri in Clk(S13D) mutants at 25 °C. We observed significant reduction in levels and cycling amplitude of all mRNAs measured in Clk(S13D) mutants as compared to Clk(WT) flies, as determined by CircaCompare (60) (Fig. 4 B–D and SI Appendix, Table S4↗). The dampened mRNA oscillation in Clk(S13D) flies is consistent with dampened behavioral rhythmicity (Fig. 4A). It is important to highlight that the reduced expression of CLK targets in Clk(S13D) mutants is opposite of the higher CLK target gene expression when wild type flies were maintained at 10 °C rather than at 25 °C (compare Figs. 2D and 4 B–D). This supports that the S13 phosphorylation is a key regulatory event that fails to occur when flies express CLK-cold at 10 °C.

Since previous studies showed that phosphorylation-dependent reduction in CLK stability (49) is also a plausible mechanism to reduce CLK transcriptional activity, we wanted to rule out the possibility that CK1α can modify CLK stability (SI Appendix, Fig. S4 A and B↗). We measured CLK degradation independent of TTFLs by performing cycloheximide (CHX) chase assays in Drosophila S2 cells expressing CLK in the presence or absence of CK1α. We observed similar rates of CLK protein degradation in the two conditions, suggesting that CK1α does not regulate CLK stability. In agreement with S2 cell results, we did not observe significant difference in CLK abundance between Clk(WT) and Clk(S13D) flies maintained in LD cycles (SI Appendix, Fig. S4 C and D↗).

We next sought to explore potential clock output that mediates impaired behavioral rhythmicity in Clk(S13D) mutant. Pigment-dispersing factor (PDF), a key neuropeptide in clock neuronal circuits, has been shown as an important molecular clock output to regulate rhythmic locomotor activity in flies (61). PDF level is modulated by a number of CLK target genes, including vri (62), hormone receptor-like 38 (HR38), (63) and stripe (63). Because we observed significant alteration of CLK target genes in Clk(S13D) flies (Fig. 4 B–D), we hypothesized that altered diurnal changes of PDF level at the dorsal projection of sLN_v neurons may cause dampened behavioral rhythms in Clk(S13D) mutant. We monitored PDF level using whole-mount immunocytochemistry, and observed that PDF exhibits diurnal changes in abundance in Clk(WT) flies, which is consistent with previous studies (62, 64) (Fig. 4 E and F). Diurnal changes in PDF abundance are abolished in Clk(S13D) mutant, despite normal behavioral rhythmicity in LD cycles. This is in agreement with previous findings indicating rhythmic PDF level is not required for maintenance of behavioral rhythmicity in LD (61, 62). Our results do not rule out that PDF rhythms are phase-shifted such that the difference observed in Clk(WT) flies at ZT3 and ZT15 was not observed in Clk(S13D) flies. This further supports the crucial role of S13 phosphorylation in maintaining diurnal changes of PDF level and therefore robust locomotor activity rhythms that do not occur when flies express CLK-cold at 10 °C.

In addition to Clk(S13D) flies, we also assayed CLK target gene expression in Clk(S13A) mutants. In contrary to what we expect based on our observation of elevated transcriptional activity of CLK(S13A) variant determined by per-luc reporter assay in Drosophila S2 cells (Fig. 3A), we observed that the expression of CLK target mRNAs in Clk(S13A) flies was also reduced when compared to Clk(WT) (Fig. 4 G–I and SI Appendix, Table S4↗). This discrepancy between the extent to which the Clk(S13A) mutation impacts CLK target gene expression in tissue culture vs. in whole animals is further explored in the following sections.

Temperature can entrain circadian clocks (65, 66). Given the temperature-sensitive nature of Clk splicing (Fig. 1 B and C), we hypothesize AS of Clk can mediate temperature entrainment via regulating the availability of S13 residue for phospho-regulation. Therefore, Clk(S13A) and Clk(S13D) flies are expected to exhibit compromised temperature entrainment. To test this, we monitored daily locomotor activity rhythms of Clk transgenic flies for 4 d in 10 h:14 h 25 °C/17 °C temperature cycles in DD (67) followed by release in 7 d in 17 °C constant temperature to assess free-running rhythms (Fig. 5A and SI Appendix, Fig. S5A↗). We observed anticipation of cold phase in Clk(S13D) and Clk(S13A) flies (SI Appendix, Fig. S5A↗), which is a clear indication of thermic entrainment. Surprisingly, free-running rhythms are compromised in Clk(S13D) and Clk(S13A) after 4 d of thermic entrainment. To rule out the possibility that reduced rhythmicity at 17 °C is due to incomplete entrainment, we employed photic entrainment using LD cycles under several temperatures lower than 25 °C and monitored locomotor activities in DD (Figs. 4A and 5B and SI Appendix, Fig. S5B↗). Clk(WT) flies were rhythmic under 18 °C with similar period length as compared to 25 °C due to temperature compensation, while Clk(S13D) and Clk(S13A) both displayed reduced rhythmicity similar to that entrained by temperature steps (Fig. 5A). Under 14 °C, over 50% Clk(WT) individuals remained rhythmic, as seen in previous findings (68). However, Clk(S13D) and Clk(S13A) became completely arrhythmic. Moreover, we observed gradual phase-advanced evening activity peak for Clk(WT) under LD as we lowered temperature from 18 to 10 °C, a behavioral response for seasonal adaptation (13, 69). Clk(S13D) and Clk(S13A) did not show an obvious evening peak under LD at 10 to 18 °C. Together, these data suggest dynamic phosphorylation of CLK(S13) is required for the adaptation of behavioral rhythms to cold temperature, but not for temperature entrainment.

Given reduced behavioral rhythmicity of Clk(S13) mutants at cold temperature, we hypothesize that diurnal changes of PDF level, which is important for robust behavioral rhythm, are abolished in Clk(S13) mutants. We observed lower PDF level at the dorsal projection of sLN_v neurons at 10 °C as compared to 25 °C as well as diurnal changes (ZT3 vs. ZT15) of PDF level in Clk(WT) flies even under 10 °C (Fig. 5C), consistent with previous findings (69). However, changes in diurnal PDF level at 10 °C were abolished in Clk(S13D) flies, with no significant reduction at 10 °C as compared to 25 °C. In Clk(S13A) flies, diurnal changes in PDF level at 10 °C were also abolished, with a significant decrease at ZT3 (25 °C vs. 10 °C) but an increase at ZT15 (25 °C vs. 10 °C), suggesting compromised response of PDF level to cold temperature. Our results align with the model that S13 phospho-regulation mediates a large part of the effect of temperature-sensitive AS of Clk. Since S13 phosphorylation only occurs in CLK-long, Clk(S13D) flies are not expected to exhibit reduced PDF levels when the temperature is lowered (Fig. 5C). Conversely, Clk(S13A) flies resembling flies expressing CLK-cold should exhibit lower levels of PDF as compared to Clk(WT) independent of the temperature (SI Appendix, Fig. S6↗). In addition, we noticed Clk(S13A) flies displayed diurnal changes in PDF level at 25 °C (Fig. 5C), which helps to explain better rhythmicity of Clk(S13A) flies at 25 °C as compared with Clk(S13D) flies (Fig. 4A). Taken together, our data support temperature-dependent availability of S13 residue for phospho-regulation can act through PDF to regulate behavioral rhythms at cold temperature.

Now that we showed that CLK(S13) phosphorylation is important for maintenance of circadian rhythms in a physiological range of temperature, we further investigated the molecular basis of CLK(S13) phosphorylation in regulating circadian rhythms by characterizing the impact of S13 phosphorylation in modulating CLK-DNA interactions in vitro and in flies. We overlaid an AlphaFold (70) model of Drosophila CLK-bHLH (aa1-71 of CLK-long including the bHLH domain) to the crystal structure of human CLOCK-BMAL1-DNA (71) and found plausible contacts between S13 and the negatively charged DNA backbone (Fig. 6A). This hints at phosphorylation being a direct modulator of CLK-DNA binding by suppressing CLK-DNA interactions via charge–charge repulsion. To test this hypothesis, we expressed and purified aa1-71 fragment of CLK-bHLH with and without the S13D mutation from Escherichia coli (SI Appendix, Fig. S7 A and B↗) and measured their binding affinity to a 21-bp per promoter as bait (72) using biolayer interferometry (BLI) (73) (Fig. 6B). The EC₅₀ between WT CLK-bHLH and DNA binding was estimated to be 0.27 μM (95% CI = [0.19, 0.37], Fig. 6C). Importantly, introduction of the phosphomimetic S13D mutation abolishes the binding of CLK-bHLH to DNA (Fig. 6C), demonstrating the potent inhibition of S13 phosphorylation on the ability of CLK to bind to DNA and act as a transcription factor. To demonstrate the specificity of the CLK-per promoter interaction, we performed a control experiment where the DNA is composed of a scrambled DNA sequence. We observed a reduction in CLK binding affinity on scrambled DNA compared to the 21-bp per promoter sequence, suggesting that the CLK-DNA binding in our BLI assay is selective for CLK target gene promotors (SI Appendix, Fig. S7C↗).

We next performed CLK-ChIP followed by qPCR using extracts from adult fly heads to further determine CLK-DNA binding in fly tissues collected at 25 °C (Fig. 6 D and E). We observed significantly lower CLK occupancy in Clk(S13D) mutants as compared to Clk(WT) flies at ZT19 at the per CRS (Fig. 6D). CLK occupancy at the tim E-box is also significantly lower in Clk(S13D) mutants at all time-points tested (Fig. 6E). Together with data from in vitro experiments (Fig. 6 A–C), our results revealed that CLK occupancy at CLK target gene promoters decreases upon CLK(S13) phosphorylation, which explains lower expression of CLK target genes in Clk(S13D) flies.

In the case of Clk(S13A) mutant, CLK-ChIP qPCR showed significantly higher CLK occupancy at the per CRS and tim E-box at ZT3 (Fig. 6 F and G). Increased CLK-DNA binding in early day in Clk(S13A) mutant (ZT3, Fig. 6 F and G) resembles our observation of higher CLK-DNA binding in WT flies at 10 °C (ZT4, Fig. 2 A and B). These results support our model that AS allows CLK-cold to escape inhibitory phosphorylation at S13 to promote CLK-DNA binding. Surprisingly, CLK occupancy at ZT11 (per CRS and tim E-box) and ZT15 (tim E-box) in Clk(S13A) mutant is significantly lower as compared to that in Clk(WT). How might increased CLK-DNA binding in early morning contribute to a reduction of CLK-DNA binding in the beginning of the night? Upon CLK removal from DNA by CWO (74) and nuclear export (31), a fraction of CLK undergoes further phosphorylation followed by degradation (31) (Fig. 6H). Whereas another fraction of CLK is dephosphorylated by CKA-PP2A complex to replenish the pool of hypophosphorylated CLK for the next round of CLK-activated transcription (30). We hypothesize in Clk(S13A) mutants, CLK is still removed off DNA by CWO in late night. Without S13 phosphorylation, CLK would therefore exhibit premature DNA binding activity in early morning. These transcriptionally inactive CLK proteins cannot be exported out of the nucleus to be dephosphorylated at other residues and are eventually degraded in the nucleus. Without replenishing transcriptionally active CLK via nuclear export and dephosphorylation, CLK occupancy at circadian promoters in the beginning of the night is reduced (Fig. 6 F and G; ~ZT11-ZT15). To test this hypothesis, we analyzed daily CLK phosphorylation in Clk(S13A) (Fig. 6 I and J). As we expected, CLK phosphorylation level in Clk(S13A) was significantly reduced in early day (ZT3) as compared to Clk(WT). This is likely due to increased DNA-binding activity and the inability of CLK to be exported to the cytoplasm for further phosphorylation prior to degradation. Notably, reduced CLK occupancy at circadian promoters when CLK target transcription peaks (ZT11-15) (Fig. 6 F and G) is consistent with lower CLK target mRNA levels in Clk(S13A) mutants (Fig. 4 G–I). We cannot however rule out the possibility that additional feedback mechanisms absent in tissue culture system could have contributed to the discrepancy between reduced CLK target gene expression in Clk(S13A) flies (Fig. 4 G–I) and increased transcriptional activity of CLK(S13A) in S2 cells (Fig. 3A). Nonetheless, the timing of S13 phosphorylation primarily in the morning (Fig. 7 F and G), as shown by detection of S13 phosphorylation in flies using phosphospecific antibody also supports our model.

Finally, we sought to determine the molecular requirements for CK1α to phosphorylate CLK(S13). It has been proposed that PER-TIM repressor complexes recruit yet uncharacterized kinases for timely CLK phosphorylation to enhance repression (32, 75). Since CK1α has been shown to interact with PER in both the cytoplasm and the nucleus (50), we hypothesize that PER promotes CK1α-dependent phosphorylation of CLK (Fig. 7A). We first performed per-luc reporter assay to measure CLK transcriptional activity in S2 cells expressing CLK and PER in the absence or presence of CK1α (Fig. 7B). As expected, PER expression down-regulates CLK transcriptional activity. CK1α coexpression further reduced CLK transcriptional activity, indicating an enhanced PER repression via PER-dependent CK1α phosphorylation of CLK. We observed no significant difference between baseline luciferase activity and cells expressing CLK and PER in conjunction with CK1α, suggesting that PER and CK1α together essentially abolished transcriptional activity of CLK.

CK1α regulates PER repression activity by favoring PER nuclear entry and promoting PER-DBT interaction and phosphorylation-dependent degradation (50). To remove the confounding effect of CK1α on PER nuclear localization and degradation in our luciferase assay, we therefore expressed nuclear localization sequence (NLS)-tagged PER (PER-NLS) and DBT(K/R), a kinase-dead DBT variant (76) to specifically examine the role of CK1α in modulating CLK activity (Fig. 7C). We observed significant reduction in CLK-dependent per-luc activity upon CK1α coexpression with PER-NLS and DBT(K/R). This suggests that in addition to the regulation of CK1α on PER nuclear entry and stability, the enhanced PER repression of CLK seen in Fig. 7B is also mediated through CLK phosphorylation by CK1α.

To directly determine whether PER-DBT scaffold promotes CLK(S13) phosphorylation by CK1α, we assayed CLK(S13) phosphorylation in protein extracts of Drosophila S2 coexpressing FLAG-Clk and ck1α in combination with either per(WT) or per(Δ), a per mutant encoding a variant lacking PER-DBT binding domain (77) (Fig. 7 D and E). Western blotting with S13 phosphospecific antibody showed that CLK(pS13) was significantly reduced when per(Δ) is coexpressed, as compared to per(WT) (Fig. 7D, lane 1-2). As expected, little to no CLK(pS13) signal was detected in Clk(S13A) (Fig. 7D, lane 3), showing the specificity of α-pS13.

In flies, PER interacts with CLK upon nuclear entry (33). PER-DBT can promote CLK(S13) phosphorylation for CLK removal from DNA in the late night. Alternatively, CLK(S13) phosphorylation can occur in the early morning when CLK is removed from DNA, which prevents premature DNA binding of CLK. To determine the timing of CLK(S13) phosphorylation, we detected CLK(S13) phosphorylation level by Western blots using α-pS13 in head extracts from flies collected over a daily cycle (Fig. 7 F and G). We observed that S13 phosphorylation occurs in the daytime (ZT4, ZT8, and ZT12), supporting the function of CLK(S13) phosphorylation as a prevention of premature CLK binding to DNA. Together, our results support that PER-DBT serves as a scaffold to enhance CK1α-dependent CLK(S13) phosphorylation to sustain off-DNA state of CLK.

Drosophila construct design and transformation of Clk transgene was performed as described by Mahesh et al. (51). Details are provided in SI Appendix↗.

Luciferase reporter assay was performed as described by Darlington et al. (36). Details are provided in SI Appendix↗.

ChIP and quantitative PCR were performed as described by Kwok et al. (75). Details are provided in SI Appendix↗.

Methods for coIP and western blotting and information about antibodies were as described by Lam et al. (50). Additional details are provided in SI Appendix↗.

Phos-Tag SDS-PAGE was performed as described by Lam et al. (50). Details are provided in SI Appendix↗.

Expression of CLK in Drosophila S2 cells, purification of CLK proteins, and subsequent mass spectrometry analysis were as described by Chiu et al. (54). Additional details are included in SI Appendix↗.

CLK(S13) phosphospecific antibodies was generated as described by Chiu et al. (54). Details are included in SI Appendix↗.

Fly activity monitoring was performed as described by Hidalgo et al. (69). Details are included in SI Appendix↗.

Immunofluorescence was performed as described by Cai et al. (75). Details are provided in SI Appendix↗.

Cycloheximide chase assay was performed in Drosophila S2 cells as described by Lam et al. (50). Details are provided in SI Appendix↗.

Biolayer interferometry to assay CLK-DNA binding was performed as described by Abdiche et al. (73). Additional details are included in SI Appendix↗.

Statistical analyses were performed as described by Cai et al. (75). Details are included in SI Appendix↗.