Small genetic sequences control CLOCK protein production to influence daily rhythms and sleep

To elucidate the role of uORFs in circadian rhythm regulation, we first performed a systematic analysis of uORFs distribution in Drosophila circadian genes. We annotated 36,590 putative canonical uORFs (starting with AUG and ending with a stop codon UAA/UAG/UGA) based on the 5' UTR annotation of 13,390 protein-coding genes in D. melanogaster, employing a method previously described [24,39] (S1 Table; Methods).To examine whether uORFs are more prevalent in genes associated with circadian rhythms, we compiled a list of circadian rhythm-related genes following the methods described in previous studies [40–45] by using the gene ontology (GO) term GO:0007623, which represents explicitly “circadian rhythm”. This list comprises 152 protein-coding genes considered as circadian rhythm-related genes for the subsequent analysis. When considering all of these circadian rhythm-related genes, including those without uORFs (treated as having zero uORFs), the number of uORFs per gene was significantly higher compared to other genes [p = 2.64 × 10⁻²³, two-sided Wilcoxon rank-sum test (WRST); Fig 1A and S2 Table], demonstrating a strong enrichment of uORFs in genes involved in circadian rhythm. Since conserved uORFs are more likely to be functional [24,25,46], we then focused on uORFs highly conserved across Drosophila species. As the start codon (uATG) is the most pivotal and definitive feature of a canonical uORF [24,25], we investigated the conservation patterns of the uATGs based on multiple sequence alignments (MSA) of 23 Drosophila species. In total, we found 388 uATGs, within 250 genes, that are identical across the examined Drosophila species (S3 Table). These genes were significantly enriched for functions in neuronal axon genesis and transcriptional regulation, implicating the potential roles of uORFs in modulating these processes (S1A Fig). Notably, the circadian rhythm-related genes had a significantly higher proportion of highly conserved uATGs compared to other genes (29/1137 versus 359/35453; p = 2.3 × 10⁻⁵, Fisher’s exact test; Fig 1B). Among the 10 core clock genes (Table 1), 7 out of 82 uORFs had uATGs identical across Drosophila species, indicating higher uORF conservation compared to other circadian genes (7/82 versus 22/1055; p = 0.005, Fisher’s exact test; Fig 1B). These seven conserved uORFs were found in four genes, with Pdp1 containing four, and Clk, sgg, and cwo each containing one (Table 1).

In addition to comparing the proportion of conserved uORFs across gene categories, we also quantitatively assessed the conservation of uATGs by calculating their branch length score (BLS) across 23 Drosophila species as previously described [47]. A larger BLS value indicates a higher degree of conservation. uATGs in circadian rhythm-related genes exhibited significantly higher BLS values than the other genes (p = 8.58 × 10⁻¹⁶, WRST; S1B Fig), with core clock genes displaying even higher conservation (core versus non-core clock genes, p = 0.016; core clock genes versus other non-clock genes, p = 3.21 × 10⁻⁶; WRST; S1B Fig). As clock genes often have multiple isoforms with distinct 5' UTRs [48–50], we also restricted our analysis to the most abundant isoform per gene (estimated by kallisto [51]) and observed similar enrichment and conservation patterns (enrichment: p = 1.44 × 10⁻¹⁴, WRST, S2A Fig; conservation: core clock genes versus other non-clock genes, p = 0.04; non-core clock genes versus other non-clock genes, p = 2.96 × 10⁻¹²; WRST, S2B Fig). These findings suggest that circadian rhythm-related genes, particularly core clock genes, are more likely to be regulated by evolutionarily conserved uORFs.

Besides conservation, translation activity is also an indicator of the functional roles of uORFs. We analyzed previously published mRNA-Seq and Ribo-Seq data from D. melanogaster heads to quantify the translation efficiency of uORFs and CDSs [39]. Translational efficiency refers to the rate at which ribosomes translate mRNA into proteins, typically quantified by comparing the abundance of ribosome-protected mRNA fragments (RPF) to total mRNA levels (Methods). Among the uORFs expressed in male and female heads (mRNA RPKM > 0.1), about 80% (692/888 in males, 714/879 in females) showed evidence of translation (translational efficiency > 0.1, Table 1). Notably, 25 of the 29 highly conserved uORFs exhibited translation signals in both sexes. For core clock genes, about 78% of the expressed uORFs (52/67 in males and 49/63 in females) showed translation signals. These data suggest a potential regulatory role of uORFs in the translation of circadian rhythm-related genes.

To further assess the impact of uORFs on translation efficiency of the downstream CDS, we categorized the expressed circadian rhythm-related genes (149 in male and 147 female heads) into three groups based on uORF conservation and translation: Class I [genes with translated uORFs that are highly conserved uORFs (BLS ≥ 0.8)], Class II [genes with translated uORFs that are less conserved (BLS < 0.8)], and Class III (genes without translated uORFs). A statistically significant trend in the translational efficiency of CDS was observed across these classes: Class I < Class II < Class III (Fig 1C). These results suggest that the translation of uORFs, particularly the evolutionarily conserved ones, suppress the translation of downstream CDS. Notably, four core clock genes—Clk, cyc, per, and tim—showed reduced translation efficiency compared to the genomic average, with Clk, a Class I gene, exhibiting the lowest translation efficiency relative to other clock genes, pointing to a significant regulatory role of its uORFs on translation (Fig 1C).

Given that Clk is the only core clock gene harboring highly conserved and translated uORFs that appear to strongly suppress its CDS translation, we focused on Clk uORFs. The 5' UTR of D. melanogaster Clk mRNA contains five canonical uORFs, with the second (uORF2) and third (uORF3) being in-frame and overlapping (Fig 2A). Among these, the start codon of uORF2 (uATG2) is the most conserved, identical across all 23 Drosophila species (BLS = 1) (Figs 2A and S1C). As CYC is the binding partner of CLK, we also analyzed cyc gene for comparison. The 5' UTR of D. melanogaster cyc mRNA contains four canonical uORFs, but their start codons are generally less conserved than those of Clk, with none being identical across the 23 Drosophila species (S3A Fig). We observed abundant mRNA expression and ribosome occupancy for uORF2 and uORF3 of Clk in both male and female heads and bodies (Fig 2B). As for cyc, although mRNA expression was observed for all four uORFs, uORF2 likely had the most pronounced ribosome occupancy (S3B Fig).

To explore the regulatory impact of uORFs on CDS translation, we performed a dual-luciferase reporter assay in vitro using wild-type (WT) and mutated 5' UTRs of Clk and cyc mRNAs, where the start codons for uORF2 and uORF3 of Clk and uORF2, uORF3, and uORF4 of cyc were mutated from ATGs to ATTs (Fig 2C). In Drosophila S2 cells, in addition to luminescence intensity, we also measured the mRNA level for each sample to exclude the influence of mRNA expression or degradation (S4 Fig). We observed that the relative translation output, calculated by normalizing the luminescence intensity to the mRNA level, was much higher for constructs with uATG-mutated 5' UTRs compared to WT controls (Clk^WT versus Clk^mutant: p = 0.0022; cyc^WT versus cyc^mutant: p = 0.0014; WRST) (Fig 2C). These results confirmed the repressive function of these uORFs on the translation of CDSs in Clk and cyc mRNAs.

To investigate the function of Clk and cyc uORFs in vivo, we knocked out Clk uORF2 and uORF3 simultaneously in D. melanogaster by mutating their start codon ATGs to ATTs, and generated a homozygous knock-out line (Clk-uORF-KO) (Fig 2D). Similarly, we mutated the ATG start codons of uORF2–4 to ATTs in cyc and generated a homozygous knock-out line (cyc-uORF-KO) (Fig 2D). We carried out ribosome fractionation followed by quantitative PCR (qPCR) to assess the translational status of Clk and cyc mRNAs in flies maintained under 12-hr light/12-hr dark (12L12D) cycles. This technique separates mRNAs based on the number of bound ribosomes and enables the calculation of the polysome-to-monosome (P-to-M) ratio, which reflects translation efficiency (S5A Fig). A larger P-to-M ratio means more mRNAs are enriched in the polysome fractions and bound by more ribosomes, thus indicating higher translation efficiency [52–54]. We conducted experiments on heads from Clk-uORF-KO, cyc-uORF-KO, and WT flies at six Zeitgeber Time (ZT) points, starting from ZT0 (the onset of light period) to ZT20 at 4-hr intervals (S6 Fig). In male heads, the P-to-M ratio of Clk mRNA was significantly higher in Clk-uORF-KO flies at ZT0, ZT4, and ZT8, but showed no significant differences at ZT12 and ZT16, with a decrease at ZT20 (Fig 3A). The pattern in Clk-uORF-KO females mirrored that in males (Fig 3B). Conversely, the P-to-M ratio for cyc mRNA in cyc-uORF-KO flies was significantly higher only at ZT16 (S5B Fig). Overall, these patterns indicate Clk uORFs exert substantial repression on translation, especially during the light phase, while cyc uORFs exert modest effects, prompting a focus on Clk uORFs in subsequent analyses.

Next, we measured CLK protein abundance in heads at various Zeitgeber time points using immunoblotting (Fig 3C and 3D). CLK levels were significantly higher in male Clk-uORF-KO flies compared to WT during the morning hours (ZT0, ZT4, and ZT8; Fig 3E), with a similar pattern observed in females (Fig 3F). To determine whether this increase resulted from enhanced translation rather than reduced protein degradation, we treated fly heads with the proteasome inhibitor MG132. Notably, Clk-uORF-KO flies still exhibited elevated CLK protein levels (S7A–S7D Fig). Furthermore, no significant differences in CLK ubiquitination were detected between WT and Clk-uORF-KO flies (S7E and S7F Fig). These results support that uORFs repress CDS translation in Clk mRNAs rather than affecting protein degradation.

Given the well-established role of CLK phosphorylation rhythms in the cycling of CLK/CYC-driven transcription [55], we conducted additional experiments to profile CLK phosphorylation status. While we observed rhythmic CLK phosphorylation in both WT and Clk-uORF-KO flies, no significant differences were detected between the two genotypes (S8 Fig), suggesting Clk uORFs do not affect CLK phosphorylation status. We also assessed CLK abundance on the first day of constant darkness and found that CLK levels were significantly elevated during the earlier part of the subjective day (S9A–S9D Fig), mirroring the pattern under light/dark cycles, suggesting the translational repression mediated by Clk uORFs is independent of light. Furthermore, CLK level remained elevated in Clk-uORF-KO flies lacking per and tim (S9E–S9H Fig), indicating that the increased CLK level during the day in Clk-uORF-KO is not caused by the altered circadian phase. Together, these data demonstrate that Clk uORFs downregulated CLK protein translation during the daytime, with pronounced effects in the early hours.

Previous studies have demonstrated that the biogenesis of translation machinery is controlled by oscillating transcription and/or translation activities [7,38], with peak translation of several translational initiation factors occurring predominantly at midday in Drosophila [56]. Our previous proteomic data [57] further revealed that 24 proteins in the GO term “translation” (GO: 0006412) category exhibit rhythmic expression, most of which are ribosomal proteins and translational factors, with 19 peaking during the subjective day (S10 Fig). This rhythmicity of translation machinery components may provide instructive signals driving Clk translation, overriding uORF-mediated translational regulation at night. Thus, uORFs may play a “permissive” role in regulating Clk translation, influenced by “instructive” signals from the translational machinery.

To quantitatively assess the effects of Clk-uORF-KO on circadian rhythms, we utilized a mathematical model previously established for simulating the Drosophila molecular clock [58]. This model employs nonlinear ordinary differential equations to model the regulatory weights among mRNA and protein molecules within the circadian network, making it applicable for predicting the consequence of increasing Clk translation. By incrementing the regulatory weight from Clk mRNA to CLK protein from the baseline to higher values, we incorporated the anticipated enhancement in Clk translation due to the uORF KO into our model (S11A Fig).

The simulation indicated an immediate increase in CLK protein levels correlating with a higher regulatory weight from Clk mRNA to protein, mimicking the Clk-uORF-KO effect (S11B Fig). This, in turn, led to elevated levels of per and tim mRNA and protein (S11B and S11C Fig). In the simulations, we used the intervals between peaks of tim mRNA as a measure of circadian period length. The mean ± SEM of period length were 23.926 ± 0.003 hrs, 20.967 ± 0.078 hrs, 18.956 ± 0.061 hrs, and 17.494 ± 0.043 hrs for regulatory weights from Clk mRNA to CLK protein set at 0.026 (baseline), 0.027, 0.028, and 0.030, respectively (S11D and S11E Fig and S4 Table). These findings demonstrated a shortening of the circadian period as the regulatory weight increases, suggesting that enhanced Clk translation—resulting from the disruption of Clk uORFs—compresses the circadian period.

To corroborate the simulation outcomes, we measured the mRNA level of per and tim, observing significant elevation in both male and female Clk-uORF-KO fly heads (Fig 4A and 4B). As our simulations predicted, disrupting Clk uORFs reduced the period length of over one hour for the locomotor rhythm of Clk-uORF-KO flies. Specifically, the period length of WT males was 24.0 ± 0.25 hr, Clk-uORF-KO males 22.8 ± 0.21 hr, WT females 24.2 ± 0.14 hr, and Clk-uORF-KO females 22.7 ± 0.09 hr (Fig 4C and 4D and S5 Table). These empirical results, together with our simulation data, indicate that Clk uORFs modulated CLK translation, thereby impacting the pace of the molecular clock through the regulation of downstream genes such as per and tim.

Given that the sleep-wake cycle is a key manifestation of circadian rhythms and Clk mutants exhibit significantly reduced sleep [59], we examined the impact of knocking out Clk uORFs on sleep under 12L12D cycles. Deletion of Clk uORFs significantly lengthened sleep duration, with a more pronounced effect in females (9.5% increase; Figs 5A and S12A) compared to males (4.0% increase; S12B and S12C Fig). A closer examination demonstrated that in Clk-uORF-KO female flies, sleep duration was extended by 38.65% during the first 8 hrs of the light phase, indicating that Clk uORFs facilitate morning wakefulness (Fig 5B). This increase in sleep duration was attributed to longer average sleep bouts, suggesting enhanced sleep continuity upon Clk uORF deletion (Fig 5C). Furthermore, waking activity levels were higher in female Clk-uORF-KO flies, ruling out impaired motor function as the cause of the prolonged sleep (Fig 5C). In contrast, the sleep alterations in Clk-uORF-KO males were less pronounced than in Clk-uORF-KO females (S12D Fig). We further validated the sleep phenotype in Clk-uORF-KO females using a video monitoring system we previously established [60], confirming a comparable increase in sleep during the morning hours (S12E and S12F Fig). Due to the more pronounced sleep phenotype in females, all subsequent sleep-related studies were conducted in female flies.

Given that CLK suppresses the expression of the gene encoding tyrosine hydroxylase (ple), which in turn diminishes dopamine signaling and activity levels [61], we investigated whether dopamine signaling mediates the effects of Clk uORFs on sleep. Our qPCR assays revealed a notable decrease in ple mRNA in female Clk-uORF-KO heads compared to WT (Fig 5D), aligning with the hypothesis that elevated CLK protein levels—resulting from Clk uORF disruption—downregulates ple expression. To further explore this, we treated female flies with dopamine precursor L-DOPA, which facilitates dopamine synthesis [62]. Interestingly, this treatment reversed the prolonged sleep duration in female Clk-uORF-KO flies (Fig 5E and 5F). Thus, Clk uORFs appeared to promote morning alertness by modulating dopaminergic signaling.

Given the critical role of the circadian clock in seasonal adaptation [63], we hypothesized that uORFs might regulate Clk translation on a circannual basis. To test this hypothesis, we subjected Clk-uORF-KO and WT female flies to conditions mimicking seasonal changes. Because the most important seasonal signal for the circadian clock is the shortening/lengthening of photoperiod [63], we monitored locomotor rhythm and sleep under photoperiods of different lengths, ranging from 4 hrs of light and 20 hrs of dark (4L20D) to 20 hrs of light and 4 hrs of dark (20L4D). While Clk uORFs did not influence the timing of locomotor activity rhythm across photoperiods (S13 Fig), sleep duration exhibited a photoperiod-dependent pattern (Fig 6A). The difference in sleep duration between Clk-uORF-KO and WT females increased as photoperiods shortened (Fig 6B). Under 4L20D, Clk-uORF-KO females exhibited a 22.44% increase in sleep duration relative to WT (1,124 ± 18 min versus 918 ± 26 min), whereas this difference was reduced to 10.44% under 20L4D (1,195 ± 18 min versus 1,082 ± 21 min). Additionally, in WT females, sleep duration positively correlated with photoperiod length (Pearson’s r = 0.323, p = 6.5 × 10⁻⁸), but this correlation was weaker in Clk-uORF-KO females (Pearson’s r = 0.172, p = 0.006) (Fig 6C). Collectively, these findings suggest that Clk uORFs modulate sleep responses to seasonal changes, likely through translational regulation that fine-tunes sleep duration in response to photoperiod length.

Given the observed behavioral changes, we hypothesized that Clk uORFs mediate translational regulation in a photoperiod-dependent manner. To test this, we performed ribosome fractionation followed by qPCR to assess Clk CDS translation under 4L20D and 12L12D light-dark cycles. In WT female heads, Clk CDS translation efficiency (P-to-M ratios) was significantly lower under 12L12D compared to 4L20D at five of six time points (except ZT0) (Fig 7A); however, in Clk-uORF-KO, a significant reduction was observed only at ZT8 and ZT20, while ZT0 displayed an opposite trend compared to WT (Fig 7B). Consistently, overall CLK protein synthesis, measured as the area under the curve (AUC), was significantly reduced (approximately 30%) in WT under 12L12D compared to 4L20D (Fig 7C), whereas this reduction was minimal (approximately 8%) in Clk-uORF-KO flies (Fig 7D). These results demonstrate that the loss of Clk uORFs diminished translational regulation in response to changes in photoperiod length. Supporting this notion, a previous study identified 12 genes involved in translational regulation that exhibit photoperiod-dependent expression (16L8D versus 8L16D) [64]. We tested these genes and confirmed that 3 (RpS4, RpL34a, and RpL14) were differentially expressed under 4L20D relative to 12L12D (S14 Fig), implying that ribosomal proteins and other translational factors participate in adapting to photoperiod variation.

Despite higher Clk mRNA translation under 4L20D compared to 12L12D in WT flies, western blot analysis paradoxically revealed lower CLK protein levels under 4L20D—particularly from ZT4 to ZT12 (Fig 7E and 7F). In Clk-uORF-KO mutants, however, this reduction was only significant at ZT0 (Fig 7E and 7F). Moreover, the AUC analysis demonstrated that photoperiod changes significantly altered the total daily CLK level in WT but not in Clk-uORF-KO heads (S15 Fig). qPCR analysis showed that Clk mRNA levels were lower under 4L20D than 12L12D in both WT and Clk-uORF-KO (S16 Fig), which could contribute to the reduction of CLK protein level under 4L20D.

Together, these findings suggest that Drosophila integrate photoperiodic cues at multiple regulatory levels to facilitate seasonal adjustments in sleep/wakefulness via Clk, with Clk uORFs playing an essential role in this process.

CLK modulates the cyclical expression of many genes in Drosophila [64]. To investigate the impact of Clk uORFs on global gene expression rhythms, we performed mRNA sequencing on WT and Clk-uORF-KO fly heads collected at four-hr intervals under 12L12D cycles (three biological replicates per time point). Compared to WT females, Clk-uORF-KO females exhibited 625–1,422 differentially expressed genes (DEGs) per time point, while Clk-uORF-KO males showed 461–2,825 DEGs per time point. We observed a significant upregulation of tim and per expression in both male and female mutants (S17 Fig), consistent with the expression profiles detected by qPCR (Fig 4A and 4B).

Given the pronounced sleep alterations in Clk-uORF-KO mutants, we examined whether these DEGs were enriched for sleep-associated genes. We curated a list of sleep-related genes from the gene ontology term “sleep” (GO:0030431) and previous literature, including genes associated with sleep [60] and sleep latency [65], yielding 535 unique sleep-related genes (S6 Table). In females, we found a significantly higher proportion of sleep-related genes among DEGs compared to non-DEGs at 5 out of the 6 time points analyzed (p = 0.035, 0.45, 1.01 × 10⁻⁵, 4.30 × 10⁻⁴, 4.79 × 10⁻³, 7.39 × 10⁻⁴ for ZT0, 4, 8, 12, 16 and 20, respectively; Fisher’s exact test; Fig 8A), suggesting that Clk uORF deletion profoundly affects sleep-related gene expression. In contrast, this enrichment was weaker in males, reaching statistical significance only at ZT0 (p = 1.47 × 10⁻⁷) and ZT16 (p = 0.027; S18A Fig), consistent with the less pronounced sleep alterations in males. Moreover, gene ontology (GO) analysis revealed that DEGs in Clk-uORF-KO females were primarily enriched in metabolic pathways (Fig 8B). In males, DEGs were also enriched in metabolism pathways, though these enrichments were sparse across time points (S18B Fig). Together, these findings suggest that Clk uORFs broadly influence the expression of numerous genes involved in diverse biological processes beyond circadian rhythms and sleep.

To assess the impact of Clk uORFs on gene expression rhythms, we identified cycling genes with rhythmic expression using ARSER [66] and JTK_CYCLE [67] algorithms implemented in MetaCycle [68]. To ensure robust identification of rhythmic genes, we applied stringent false discovery rate (FDR) thresholds, requiring q values to be below 0.05 or 0.01 in both algorithms (S7 Table; Methods). Additionally, we incorporated an amplitude threshold (max/min fold-change ≥ 1.5) to account for both statistical and biological significance, consistent with previous studies [69]. Applying q-value thresholds of ≤0.05 or ≤0.01, we found that Clk-uORF-KO females consistently exhibited significantly more cycling genes than WT females (p = 8.70 × 10⁻¹¹ and 8.14 × 10⁻⁹, respectively; Fisher’s exact test; Fig 8C). In contrast, Clk-uORF-KO males exhibited a slight increase in cycling genes compared to WT males, but this was not statistically significant (S18C Fig). Only 40 to 60% of cycling genes overlapped between WT and Clk-uORF-KO flies, a trend consistent across different q-value cutoffs (Figs 8D and S18D). Using q ≤ 0.05, we identified 440 rhythmic genes in WT females and 648 in Clk-uORF-KO females, with 289 shared genes (Fig 8D). In males, 472 genes were rhythmic in WT and 512 in Clk-uORF-KO, with 230 shared genes (S18D Fig). The circadian rhythm-related genes we compiled earlier were significantly enriched among cycling genes in WT (p = 0.0008 for females, p = 0.002 for males; Fisher’s exact test; S19A Fig), while this enrichment was reduced in Clk-uORF-KO (p = 0.014 for females, p = 0.048 for males; S19A Fig), further supporting that the removal of Clk uORFs disrupted the rhythmic expression of many circadian rhythm-related genes. Moreover, knocking out Clk uORFs increased the relative oscillation amplitude (rAMP), with a more pronounced effect in females than in males (p = 4.74 × 10⁻¹⁵ in females versus p = 1.21 × 10⁻⁵ in males; Wilcoxon signed-rank test; S19B Fig).

Cross-sex comparisons revealed that only approximately 50% of cycling genes were shared between WT male and female heads (S20A Fig), consistent with prior findings in Drosophila [70], mice [71], and humans [72,73]. This baseline sex-specificity extended to Clk-uORF-KO mutants, where only approximately 50% of cycling genes overlapped between male and female heads (S20B Fig), and a few genes that lost or gained rhythmicity in mutants were shared across sexes (S20C and S20D Fig). Nevertheless, GO analysis of these genes that lost or gained expression rhythmicity showed some similarities between males and females, with enrichment in metabolism pathways, though specific terms varied (S20E Fig). Overall, these findings indicate that Clk uORF deletion disrupted rhythmic expression in a large set of genes while inducing rhythmicity in hundreds of others, in a sex-dependent manner.

Transcriptome analysis revealed that numerous genes with altered expression and/or rhythmicity were enriched in several essential physiological pathways, primarily converging on metabolic processes. Since metabolism is fundamental to all physiological functions, we hypothesized that its disruption could ultimately manifest in Drosophila adaptation and fitness. Indeed, the Clk-uORF-KO flies showed significantly faster mortality rates under starvation (a metabolic stress) compared to WT flies for both sexes (p = 8.8 × 10⁻¹³ for females and p = 2.4 × 10⁻⁹ for males; log-rank test; S21 Fig). In addition, reproduction is intimately related with metabolic state [74,75]. We found that mated Clk-uORF-KO females laid significantly fewer eggs than their WT counterparts during an eight-day monitoring duration, with the total egg production of mutants markedly less (p < 0.01; S22A Fig). This pattern persisted in Clk-uORF-KO virgin females, which also laid fewer eggs than WT in both segmented (2-days) and pooled (10-days) measurements (S22B Fig). Accordingly, Clk-uORF-KO flies produced significantly fewer offspring than WT counterparts at 25°C (p = 1.5 × 10⁻⁵; WRST; S22C Fig). Even at 29°C, when both mutants and WT flies exhibited reduced offspring production compared to 25°C, the offspring counts for Clk-uORF-KO are still significantly fewer than that of WT (p = 0.00029, WRST; S22C Fig). Together, these results indicate that Clk uORFs profoundly affect fecundity and starvation resistance, with implications for fitness beyond their established roles in circadian rhythms and sleep regulation.

All flies were reared on standard cornmeal-yeast-sucrose medium and kept in light/dark (LD) cycles at 25°C except for the specific experimental design denoted in the text. Food was accessible ad libitum. Flies aged 3 to 5-days were used for experiments. {nos-Cas9}attP40 fly stocks used for embryo injection were from the Tsinghua Fly Center. The tool fly stocks (y sc v and Dr, e/TM3, Sb) used for mutant screening were obtained from the Tsinghua Fly Center. Clk-uORF-KO and cyc-uORF-KO mutants were generated and backcrossed to the Canton-S strain for nine generations to minimize the influence of genetic background. Throughout this study, homozygous Clk-uORF-KO or cyc-uORF-KO mutants were used, and the Canton-S strain served as the WT control flies unless otherwise stated. In addition, we backcrossed the Clk-uORF-KO mutant flies with the isogenic w¹¹¹⁸ strain (BDSC:5905) for seven generations, and compared the resulting homozygous Clk-uORF-KO flies with the isogenic w¹¹¹⁸ strain. This comparison ensured that the observed phenotypic effects in the mutants were not due to differences in genetic background. Clk^Jrk (BDSC:80927) were obtained from Bloomington Drosophila Stock Center (BDSC). Isogenic Clk-uORF-KO flies (w¹¹¹⁸ strain) were crossed with isogenic per⁰ (BDSC:80917) and tim¹(BDSC:80922) to generated homozygous triple knock-out flies (per⁰;tim¹;Clk-uORF-KO). Drosophila S2 cells used in this study were purchased from Life Technologies Corp (http://www.lifetechnologies.com↗).

We downloaded the 100-way vertebrate genome alignments (maf) based on human (hg38) and 27-way insect alignments based on D. melanogaster (dm6), and the corresponding phylogenetic trees from UCSC Genome Browser (genome.ucsc.edu↗). We used the Galaxy platform [89] to parse the multiple sequence alignments of 5' UTRs in vertebrates or in Drosophila.

Leveraging the genome annotation of D. melanogaster (FlyBase r6.04, https://flybase.org/↗), the 5' UTR sequence of each annotated transcript of D. melanogaster and its corresponding sequences in other 22 Drosophila were extracted from the genome alignments. The start codons of putative uORFs were identified by scanning all the AUG triplets (uAUGs) within the 5' UTRs of D. melanogaster. uATG triplets that overlapped with any annotated CDS region were excluded, resulting in a total of 78,458 uORFs. After removing overlaps and merging uORFs with identical uATG genomic coordinates, 36,590 non-redundant uORFs remained. Similarly, the start codon of each uORF (uATG) was identified in humans and other vertebrate species based on the genome alignments.

Each uAUG in D. melanogaster was determined the presence or absence in the orthologous sites in the other species based on multiple genome alignments. For each uAUG, we calculated the sum of the branch lengths of the subtree composed of the species in which the uAUG was present in the orthologous sites (B_sum) and then calculated the BLS value by dividing B_sum by the total branch lengths for the phylogenetic tree of the 27-insect species. Similarly, the BLS was calculated for the start codon of each human uORF across 100 vertebrate species.

The Ribo-seq data and matched RNA-seq data of D. melanogaster were obtained and processed from the previous study [39]. In brief, after removing 3' adaptors and quality controls, the NGS reads were mapped to the reference genome of D. melanogaster (FlyBase, r6.04) using STAR [90]. In each sample, we assigned a mapped RPF (27–34 nt in length) to its P-site and calculated the reads per kilobase per million (RPKM) values for a feature (CDS or uORF) with the mRNA or RPF data as: RPKM = 10⁹ × n/(L × N), where n is the number of reads that align to the feature, L is the length (nucleotides) of that feature, and library size N is the total number of reads uniquely mapped to the transcriptome. For uORFs that overlapped with CDSs, only the non-overlapping regions of the uORFs were used to calculate RPKM. The translation efficiency for a feature (CDS or uORF) was calculated as the ratio of RPF RPKM over mRNA RPKM [91].

We searched for possible sgRNA target sites near the start codon of the uORFs of Clk and cyc genes using the Benchling website (https://www.benchling.com/crispr/↗) to design optimal single guide RNA (sgRNA) sequences with high specificity and low off-target effects. Two pairs of sgRNAs were used for each gene. We then synthesized single-stranded complementary DNAs (ssDNAs) and annealed them to obtain double-stranded DNA (dsDNA), which served as the template for sgRNA expression. The dsDNA was then ligated into the BbsI-digested pU6B vector. To achieve single base editing (ATG to ATT) by CRISPR/Cas9-mediated Homology-Directed Repair (HDR), a plasmid DNA donor with about 2 kilobases flanking the targeted site was constructed following a previous protocol [92]. The donors and pU6B-sgRNA plasmid were co-injected into the embryos of transgenic Cas9 flies collected within one hour of laying at the Tsinghua Fly Center as described in Ni and colleagues [93]. The injected embryos were kept at 25°C and 60% humidity until adulthood (G0). The G0 adult flies hatched from the injected embryos were individually crossed with another strain (y sc v) to increase the number of offspring. Then, the F1 progeny were crossed with flies carrying an appropriate balancer (Dr, e/TM3, Sb). After F2 spawning, the F1 individuals were screened for mutations of interest by genotyping. The F2 progeny whose parents showed positive genotyping results were then screened for the y⁻ allele to separate the chromosome carrying nos-Cas9. The screened F2 males were crossed with flies containing the same balancer as mentioned above. After F2 genotyping, the progeny (F3) of positive F2 individuals carrying the same mutation were crossed individually to generate homozygous mutants in the F4 generation. The original homozygous mutants were backcrossed with Canton S flies for 9 generations or with w¹¹¹⁸ for 7 generations, respectively. The template sequences of the sgRNA uATG-KO and primers used for genotyping are listed in S8 Table.

The mutant fly was homogenized with DNA sequencing extract buffer [10 mM Tris-HCl at pH 8.0, 1 mM EDTA at pH 8.0, 25 mM NaCl, and 0.02% proteinase K (CW2584, CWBIO)] and incubated at 37°C for 30 min. PCR was performed with Taq Plus MasterMix (CW2849, CWbiotech). The PCR reaction was performed as follows: 94°C for 2 min followed by 94°C for 10 s, 57°C for 15 s, and 72°C for 1 min 20 s for 35 cycles. The primers used are listed in. PCR products were sequenced using commercial Sanger sequencing (AuGCT Co.). S8 Table

S2 cells were cultured in Schneider’s Drosophila Medium (Sigma) containing 10% heat-inactivated fetal bovine serum (FBS), 100 U/ml penicillin, and 100 μg/ml streptomycin (Thermo Fisher) at 25°C for 24 hr to reach 2–4 × 10⁶ cells/ml before further treatments. Plasmid transfection was conducted with Lipofectamine 3000 (L3000001, Thermo Fisher) according to the supplier’s protocol.

The wild-type (WT) 5' UTR of Clk and cyc was cloned from cDNA by PCR, and uATG mutations were introduced into 5' UTR by the amplification primers. The PCR primers used for this purpose are listed in S8 Table. The WT and mutated 5' UTR sequences were ligated into a linearized reporter plasmid, a modified psiCHECK-2 vector (Promega) with ORF codons of Rluc and Fluc optimized for D. melanogaster. The entire sequences of all the plasmids were validated by Sanger sequencing. The Renilla luciferase activity associated with WT or uORF-mutated 5' UTRs was measured according to the manual of the Dual-Luciferase Reporter Assay System (Promega) 32 hr after transfection and was normalized to the activity of firefly luciferase. The matched RT-qPCR for each dual-luciferase assay was performed with PowerUp SYBR Green Master Mix (Thermo Fisher) after reverse-transcribing RNA into cDNA using PrimeScript II first Strand cDNA Synthesis Kit (Takara). The qPCR primers used for this purpose are listed in S8 Table.

Flies were collected within 3–7 days of eclosion and entrained in 12-hr L: 12-hr D (12L12D) at 25°C for three days. After that, flies were collected and frozen at −80°C. Frozen flies were vortexed for 10 s to separate the head from the body on dry ice.

The heads obtained from WT and mutant flies were collected and homogenized with lysis buffer, the lysates were clarified by centrifugation at 4°C and 20,000 g for 10 min, and the supernatants were applied to sucrose gradient ultracentrifugation (see S1 Text for details). The RNA in the monosome and polysome fractions was extracted separately using TRIzol reagent (Life Technologies, Inc.) and chloroform (Beijing Chemical Works) following the manufacturer’s instructions and were reverse transcribed into cDNA using the PrimeScript II first Strand cDNA Synthesis Kit (Takara). RT-qPCR analysis of Clk and cyc cDNA was performed using PowerUp SYBR Green Master Mix (Thermo Fisher) following the manufacturer’s instructions. The primer sequences employed for RT-qPCR are listed in S8 Table. For each sample, the ratio of Clk and cyc mRNA abundance in the polysome fraction to that in the monosome fraction was calculated as the P-to-M ratio. Six biological replicates were performed for each sample.

Proteins were extracted from fly heads and cultured cells using SDS lysis buffer [50 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1% Triton X-100, 0.5% SDS, 1 mM EDTA (pH 8.0), 2 mM DTT, 1× proteinase and phosphatase inhibitors (4693116001 and 04906845001, Roche)]. After homogenization, protein lysates were centrifuged at 12,000 g for 15 min at 4°C, then mixed with the 4× loading buffer [0.2 M Tris-HCl (pH 6.8), 8% SDS, 6 mM bromophenol blue, 4.3 M glycerol and 0.4 M DTT] and incubated at 95°C for 5 min. Equal amounts of protein were loaded into each well on 6% SDS–PAGE gels and then transferred to nitrocellulose membranes for 2 hr at 90 V. Membranes were incubated with primary antibody at 4°C overnight, followed by secondary antibody at room temperature for 1 hr. The primary antibodies used were as follows: guinea pig anti-CLK (1:1000, gift from Dr. Joanna Chiu) and rabbit anti-ACTIN 57B (1:5000, AC026, ABclonal, CN). Donkey secondary antibodies (1:10000 dilution) were conjugated either with IRDye 680 or IRDye 800 (926–68072 and 926–32213, LI-COR Biosciences, US). Bands were visualized and quantified with an Odyssey Infrared Imaging System (LI-COR Biosciences).

For immunoprecipitation of fly heads, 150 mg fly heads were homogenized with 1 ml PBS buffer and then centrifuged at 5,000 g for 10 min at 4°C. Pellets were suspended with 500 μl IP buffer (50 mM Tris-HCl at pH 7.4, 150 mM NaCl, 1% NP-40, 0.1% SDS, 2 mM EDTA at pH 8.0, 0.15 mM spermine, 0.5 mM spermidine, 330 nM TSA, 10 mM Nicotinamide, 1× proteinase and phosphatase inhibitors). Homogenates were then sonicated for 18 × 5 s at 10% power and centrifuged at 17,000 g for 5 min at 4°C. Twenty microliters of 450 μl supernatant was used as input, and the remaining was divided into two parts. Five micrograms of CLK antibody or guinea pig normal IgG were added to the two parts. The samples were incubated at 4°C overnight, and then 50 μl of Protein A magnetic beads (Bio-Rad) were added, followed by 4 hr incubation at 4°C. Beads were washed with IP buffer for 3 × 10 min, and protein precipitates were eluted by SDS–PAGE loading buffer for Western blotting.

We adopted a previously published mathematical model of the Drosophila circadian clock [58]. This system relies on a nonlinear, autonomous, first-order system of ordinary differential equations to simulate the regulatory weights among mRNAs or protein molecules in the molecular clock network. Details of the simulation can be found in the (see S1 Text for details).

RNA was extracted from frozen fly heads maintained at −80°C. Fifty fly heads per sample were homogenized in Total RNA Isolation (TRIzol) Reagent (15596026, Invitrogen) by using a homogenizer. After mixing with trichloromethane, homogenates were centrifuged at 12,000 g, and the suspension was precipitated with 75% ethanol. After air drying, total RNA was reverse-transcribed into cDNA, and genomic DNA was removed with One-step cDNA Synthesis SuperMix (AT311, Transgen). Quantitative real-time PCR was performed with One-Step RT-PCR SuperMix (AT141, Transgen). The PCR reaction was performed as follows: 45°C for 5 min; 94°C for 2 min; 94°C for 5 s, 60°C for 15 s, 72°C for 20 s for 40 cycles (Applied Biosystems). The ΔΔCT method was used for quantification. Beta-Actin was used as an internal control. The primers used are listed in S8 Table.

Locomotor activity levels of adult flies were monitored by Drosophila activity monitoring system (DAM system, TriKinetics) for 7 days of LD cycle at 25°C followed by 7 days of DD at 25°C. Flies were loaded into activity tubes containing 5% sucrose and 2% agar. The protocol of the DAM system was previously described [94]. For DD rhythmicity, chi-squared periodogram analyses were performed by Clocklab (Actimetrics). Rhythmic flies were defined as those in which the chi-squared power was ≥10 above the significance line. Period calculations considered only rhythmic flies. Dead flies were defined by 0 activity on DD7 and removed from the analysis.

Drosophila activity monitor system (Trikinetics) was used to record fly sleep. Flies were loaded into activity tubes containing 5% sucrose and 2% agar and entrained under LD at 25°C for 3 days, and then their activities in the next 3 days under LD conditions were recorded. Sleep is defined as 5 min consecutive inactivity. Sleep was analyzed with SleepMat following previously published protocol [95].

Video capture and object tracking analysis of fruit flies were performed as previously described [60]. In brief, 3–5 days-old female flies were entrained for 1.5–2 days in 12-well plates lined with fly food containing 2% agar, 5% sucrose, and 0.1% propanoic acid, and sleep was monitored for 24.5 hr. Daily sleep was quantified by the video-to-location (VTL) method.

Adult female flies were entrained for 3 days at 25°C and anesthetized with CO₂. Brains were dissected in PBS buffer containing 3.7% formaldehyde. After fixation at room temperature for 30 min, the brains were rinsed for 10 min twice in PBS and incubated in PBS with 1% Triton for 30 min at room temperature. The brains were then incubated in 5% donkey serum diluted in PBT (PBS with 0.5% Triton) for 30 min at room temperature, followed by incubation in 1:50 mouse anti-PDF (DHSB, DF C7) antisera in PBT containing 5% donkey serum at 4°C. After PBT rinses for six times, the brains were incubated with 1:500 donkey anti-mouse AlexaFluor 647 (Thermo Fisher) for PDF immunostaining in PBT overnight at 4°C. After final rinses in PBT, brains were mounted in 80% glycerol diluted in PBS. PDF-labeled specimens were photographed with × 20 lens by Olympus FV3000 laser scanning confocal microscope (Olympus). The microscope, laser, and filter settings for a given experiment were held constant.

The complexity of neuronal axon terminal projections was analyzed using the Sholl Analysis plugin in Image J. The center was defined as the inflection point of axonal curvature, with the maximum radius set to the farthest distal point of the axon. Concentric circles were drawn at 5 μm increments, and the number of intersections between the axon and the first 10 concentric circles was quantified. The average number of intersections across experimental groups was calculated to evaluate axonal branching complexity.

We collected heads from Clk-uORF-KO mutants, and WT flies raised under LD cycles at 25°C at ZT 0, 4, 8, 12, 16, and 20. Library construction and sequencing with PE150 were conducted by Annoroad on the Illumina Nova6000 platform. Three biological replicates were sequenced for each time point. The clean data were mapped to the reference genome of D. melanogaster (FlyBase, r6.04) using STAR [90]. Reads mapped to the exons of each gene were tabulated with htseq-count [96]. Then the fragments per kilobase per million (FPKM) value was calculated for each gene as: FPKM = 10⁹ × n/(L × N), where n is the raw reads count that aligns to the gene, L is the merged length (nucleotides) of all exons of the gene, and library size N is the total number of reads uniquely mapped to the transcriptome. The genes with median FPKM ≥ 1 across six-time points were considered expressed genes and were further analyzed for rhythmic expression.

For a sample, we used DESeq2 [97] to determine the size factors of the library and normalized the mRNA read counts by dividing the raw counts with the corresponding size factors. Differential gene expression analysis was performed using DESeq2. We assumed genes with p. adjust < 0.05 as differentially expressed genes between WT and mutants. The normalized read counts were used to determine the circadian cycling using MetaCycle [68]. We used the two algorithms implemented in Metacycle, ARSER (ARS) [66] and JTK_CYCLE (JTK) [67], to identify cyclic genes. Significant cyclic genes were initially determined under two stringent levels of false discovery rate (FDR, q value) thresholds: q ≤ 0.05 and 0.01 in both JTK and ARSER, to evaluate the potential bias due to the selection of different q-value cutoffs. For the simplicity of the following analysis, a gene was determined to be significantly cycling if it had q ≤ 0.05 in both JTK and ARSER and a max/min fold-change ≥1.5, a criterion by a previous publication for rhythmicity detection [69]. We used the relative amplitude (rAMP) calculated by the MetaCycle to compare the amplitude of overlapped cyclic genes between WT and mutant.

Gene ontology (GO) enrichment analyses of the DEGs and genes that lost/gained rhythmic expression were performed by the Metascape platform (www.metascape.org↗) with a setting of the threshold at p. adjust < 0.05 [98].

For pharmacological experiments, 5 mg/ml L-DOPA (BBI Life Sciences) was mixed in the fly food (5% sucrose and 2% agar), and the flies were fed with or without L-DOPA food during sleep monitoring under LD at 25°C for 6 days.

Newly hatched virgins were picked out and allowed to mature for two days in separate vials. They were then mated by placing one virgin female with three male flies for two days. Subsequently, 10 female parents from each group were transferred to a fresh dish containing a grape juice-based medium overlaid with yeast. This procedure was replicated in five separate dishes for each strain.

After one day of egg laying, the number of eggs deposited was meticulously documented. Following this assessment, the female parents were relocated to a new dish. This egg quantification process was conducted over eight days. For the virgin female, egg counts were performed every two days, extending over 10 days.

Newly hatched virgins were picked out and allowed to mature for two days in individual vials. They were then mated by placing one virgin female with three male flies for two days. After that, each female parent was transferred to a new vial, and the number of offspring produced in 10 days was measured at 25 and 29°C. All assays were performed with 20 females per genotype.

We selected 3–5-day-old adult males and females and placed them in a starvation medium (1.5% agar), with 10 flies per vial and 10 vials for both males and females from each strain. We observed every 6 hrs or 12 hrs to count the number of deaths under starvation conditions until all flies starved to death. The survival curves were plotted using the ggsurvplot package in R.

Quantifications in all data graphs represent the mean of at least four biological replicates, and error bars represent the standard error of the mean (SEM). Two-tailed Mann–Whitney U test, Wilcoxon rank-sum test, Wilcoxon signed-rank test, log-rank test, and two-tailed Student t test was used to calculate statistical significance. All statistical analysis was carried out using Statskingdom or R language. The number of biological replicates (n) and significant p values were noted in each statistical analysis in corresponding figure legends.

The raw fastq files for the RNA-seq libraries were deposited to NCBI Sequence Read Archive (SRA) with BioProject accession PRJNA1026336. The raw data and custom code used to perform the analyses and generate the figures are available at via Zenodo at the following DOIs: https://doi.org/10.5281/zenodo.15208358↗.