Circadian clock-independent ultradian rhythms in lipid metabolism in the Drosophila fat body

Abstract

The role of circadian clocks in regulating metabolic processes is well known; however, their impact on metabolic states across species and life stages remains largely unexplored. This study investigates the relationship between circadian rhythms and metabolic regulation in the Drosophila larval fat body, a metabolic hub analogous to the mammalian liver and adipose tissue. Surprisingly, the fat body of period null mutants, which lack a functional circadian clock in all tissues, exhibited 12-h rhythms in gene expression, particularly those involved in peroxisome function, lipid metabolism, and oxidative stress response. These transcriptomic rhythms were aligned with 12-h oscillations in peroxisome biogenesis and activity, reactive oxygen species levels, and lipid peroxidation. Furthermore, period mutants exhibited 12-h rhythms in body fat storage, ultimately leading to a net reduction in body fat levels. Collectively, our results identify clock-independent ultradian rhythms in lipid metabolism that are essential for larval survival and development.

Circadian rhythms are biological cycles with a period of approximately 24 h, which allow organisms to anticipate environmental changes and adjust their physiological functions accordingly. These self-sustained rhythms are generated by endogenous circadian clocks and synchronized (entrained) by timing cues such as light, temperature, and food availability (1, 2, 3). Numerous physiological processes and behaviors exhibit circadian rhythms, and their disruption has been linked to a wide range of diseases, including metabolic, cardiovascular, and neurodegenerative disorders. The rising prevalence of circadian disruption in modern society, due to factors like shift work, increased nocturnal light exposure, high-fat/sugar diets, and sedentary lifestyles, exacerbates these problems (4, 5, 6). Understanding the interaction between circadian rhythms and physiological processes is critical for insights into overall health and well-being.

Drosophila melanogaster is an excellent model for studying circadian regulation and its disruption due to its extensive genetic tools and partially conserved molecular clock mechanisms with vertebrates. Clocks consist of interlocked transcriptional/translational feedback loops across metazoans. In Drosophila, the core feedback loop comprises the CLOCK/CYCLE (CLK/CYC) complex, which activates the expression of period (per) and timeless (tim) genes. PER and TIM proteins form a complex that translocates to the nucleus and inhibits CLK/CYC activity, leading to the downregulation of their own expression. A stabilizing loop involving PDP-1 and VRILLE (VRI) further regulates Clk transcription (3, 7).

In adult flies, approximately 75 bilateral pairs of neurons in the brain, each containing a molecular clock, serve as the pacemaker neurons that control behavioral rhythms. Similar to mammals, the central clock also coordinates body-wide physiological processes by synchronizing peripheral clocks (8, 9). In larvae, nine pairs of pacemaker neurons play critical roles in controlling various developmental and behavioral timings (10). These include the coordination of developmental growth and transitions by the steroid hormone ecdysone and circadian phototaxis behavior (11). In larvae and pupae, a key peripheral clock resides in the prothoracic gland (PG), an endocrine organ connected to the brain. This arrangement resembles the hypothalamic-pituitary gland system in vertebrates, where both systems coordinate hormone production and release. PG synthesizes ecdysone hormone under the regulation of prothoracicotropic hormone (PTTH) (12, 13). Pacemaker neurons regulate PTTH neurons in the brain, leading to circadian secretion of the PTTH. This neuroendocrine coupling controls the timing of eclosion (14, 15, 16), exemplifying the precise regulation between central and peripheral clocks.

The PG also communicates with the fat body, the central metabolic organ analogous to the mammalian liver and adipose tissue. The fat body primarily consists of adipocytes and coordinates lipid and carbohydrate metabolism, protein synthesis, amino acid metabolism, and reactive oxygen species (ROS) detoxification (17, 18). In larvae, the fat body integrates environmental and nutritional signals to regulate growth and development through the insulin/insulin-like growth factor signaling (IIS) and the Target of Rapamycin (TOR) pathway (12, 19). Ecdysone hormone released from the PG regulates fat body remodeling by activating the ecdysone receptor, which leads to the inhibition of IIS and TOR pathways, attenuating lipid accumulation (20, 21). Conversely, fat storage levels in the fat body are sensed by the PG, which fine-tunes the timing of the ecdysone surge to initiate metamorphosis only when body fat levels reach a critical threshold (22).

Nutrient availability, sensed by the fat body, regulates the production of Drosophila insulin-like peptide six (DILP6), which signals to insulin-producing cells (IPCs) in the brain. IPCs release other DILPs, which activate IIS cascade involving PI3K, AKT, TOR, and S6K, while simultaneously inhibiting the forkhead transcription factor FOXO. These signaling mechanisms regulate the balance between lipid mobilization, biosynthesis, and degradation. This intricate body–brain crosstalk maintains metabolic balance and coordinates growth and energy storage in a manner similar to insulin signaling in vertebrates (23, 24, 25).

The adult fat body possesses a functional clock, which regulates food consumption and metabolism by interacting with pacemaker neurons (26, 27). However, aside from the central clock and PG, peripheral tissues, including the fat body, do not exhibit overt circadian rhythms until adulthood (13, 28). How the circadian system influences the dynamic metabolic state of the fat body during development, which crucially affects the physiological functions and survival of adult animals (29, 30), remains poorly understood.

This study investigates the interplay between circadian rhythms and metabolic processes in larval fat body, an organ essential for metabolic homeostasis. We demonstrate that wild-type larval fat body exhibit circadian rhythms in transcriptome and lipid metabolism, non-cell-autonomously controlled by circadian clocks. Unexpectedly, period null (per⁰) mutants, lacking a functional circadian clock, display 12-h period free-running rhythms in transcriptome, peroxisomal function, lipid peroxidation, and reactive oxygen species (ROS) levels. Peroxisome function cycles and ROS rhythms are inverse to each other, mitigating ROS elevation in the absence of clocks. These ultradian metabolic cycles in per⁰ fat bodies coincide with ultradian fluctuations and a net reduction in body fat levels. Our findings suggest the these newly identified peroxisomal ultradian rhythms play a critical role in regulating metabolic and energetic balance, potentially ensuring tissue and organism viability when molecular clocks are disrupted.

Results

Circadian clock-dependent and -independent transcriptomic rhythms in the larval fat body

To explore whether circadian clocks control fat body functions in larvae, we conducted circadian RNA-sequencing (RNA-seq) analysis on larval fat bodies of white-eyed control (w¹¹¹⁸) and per⁰ mutant animals in w background (31). The parent flies were entrained at 25 °C in a 12-h light/12-h dark (LD) cycle for 3 days. The embryos were transferred to constant darkness (DD) at 25 °C and maintained under this condition through molting. Starting from the second day in DD (DD2), the fat bodies of both genotypes were collected every 4 h over a 32-h period, encompassing the early to late third-instar larvae. Prepupae were excluded (Fig. 1A).

Figure 1.

Around-the-clock gene expression profiling of per⁰ and w¹¹¹⁸ larval fat bodies. A, schematic representation of the circadian fat body collection and the RNA-seq assay. B, heatmap depicting the genes expressed with temporal oscillations in DD in the fat bodies of per⁰ and w¹¹¹⁸ larvae, identified using the MetaCycle algorithm with a cut-off of p.adj  <  0.2. Each column corresponds to a single timepoint from a biological replicate. Each row represents a gene, and rows are ordered by the peak phases. Yellow and blue colors indicate high and low relative expression levels, respectively. C, percentage of genes rhythmic with the indicated period length (h) relative to the total number of cycling genes in w¹¹¹⁸ (purple) and per⁰ (Turquoise) larvae. D, circular plot representing the gene expression peaks in w¹¹¹⁸ (purple) and per⁰ (light blue) larvae. The number on each circle indicates the % of cycling genes peaking at a given CT for each genotype. E, Venn diagram depicting the number of genes expressed rhythmically during DD in the FB in w¹¹¹⁸ (purple) and per⁰ larvae (turquoise). F, examples of RNA-seq profiles of cycling genes in DD in per⁰ larvae (turquoise) that are non-cycling in w¹¹¹⁸ larvae (purple). X-axis indicates CT. Each small data point represents the normalized counts of the indicated gene in one replicate. Connecting lines and big points represent the means of the data points.

The RNA-seq analysis for cycling transcripts was performed using the Metacycle (32) algorithm, which incorporates multiple methods from ARSER, JTK CYCLE and Lomb-Scargle, and the average expression levels. This analysis revealed 150 rhythmically expressed genes in w¹¹¹⁸ with a period of 24 h. As expected, no canonical clock genes were expressed in the larval fat body, indicating that circadian rhythms in fat body transcriptome are non-cell-autonomously driven by circadian clocks located elsewhere. Surprisingly, we found 86 genes exhibiting a rhythmic expression pattern in the fat body of per⁰ larvae. Most cycling genes in per⁰ displayed a 12-h period, with a few exceptions showing a 16-h period (Fig. 1, B and C, and Fig. S1 and Data S1). The phase of transcript rhythms also differed between genotypes: w¹¹¹⁸ larvae had peak expression mostly at CT4 (Circadian Time 4, which refers to 4 h after subjective lights-on in DD) and 16, whereas per⁰ peaked at CT4 and 10 (Fig. 1D). Only six cycling genes were shared between w¹¹¹⁸ and per⁰ which displayed 24h and 12h rhythms, respectively (Fig. 1E). These results suggest the existence of a parallel regulatory system that drives ultradian rhythms, independent of the canonical molecular clock.

What is the biological relevance of these alternative rhythms, and what signals drive them? To shed light on these questions, we performed a gene ontology (GO) analysis of rhythmically expressed genes in per⁰ larvae. This analysis revealed that the genes involved in oxidative stress response, lipid metabolism, and transcriptional modulation were overrepresented in per⁰ cycling genes. These include the regulators of the c-Jun N-terminal kinase (JNK) signaling pathway, such as Mnn1, Atg9, and pont, and those in TOR signaling pathway, such as PRAS40, AMPKα, PDK, mio, Atg2, and REPTOR-BP (Table S1). Interestingly, we found a significant enrichment of cycling transcripts involved in the peroxisomal pathway, as evidenced by the rhythmic expression of the genes controlling peroxisome biogenesis and function: Pex5, Abcd3, Jasfrac2 (Fig. 1F), and Lsd-1 (Fig. S1). These genes were not rhythmically expressed in w¹¹¹⁸ larvae.

Peroxisomes are dynamic organelles enclosed by a single lipid bilayer membrane that play critical roles in lipid metabolism and the breakdown of ROS (33, 34). The peroxisomal membrane is composed of various peroxins (PEX proteins) and peroxisomal transporters (PST proteins). PEX proteins are essential for peroxisome assembly and maintenance, while PST proteins are vital for transporting substrates across the peroxisomal membrane, ensuring proper peroxisome function (34, 35). PEX5 acts as a receptor that recognizes PST1 in soluble peroxisomal proteins, facilitating their transport from the cytosol to the peroxisome matrix (36). ABCD3, also known as PMP70, is an ATP-binding cassette (ABC) transporter that plays a key role in transporting very long-chain fatty acids (VLCFAs) across the peroxisomal membrane for β-oxidation and detoxification (35, 37). Jasfrac2 encodes peroxiredoxin four (PRX4), a 2-Cys peroxiredoxin essential for reducing hydrogen peroxide and protecting cells from oxidative stress. PRX4 functions both as a cytoplasmic and secreted protein, modulating NF-κB and JNK pathways. It is primarily localized in the endoplasmic reticulum and extracellular space, where it provides antioxidant protection and supports cell survival (38, 39, 40).

Peroxisomes share several metabolic pathways with mitochondria and are pivotal for lipid metabolism, including the β-oxidation of VLCFAs and the detoxification of hydrogen peroxide via catalase, contributing to cellular homeostasis and metabolic balance (41, 42, 43). Notably, the redox state of peroxiredoxins has been shown to exhibit circadian cycles in red blood cells as well as in several prokaryotic and eukaryotic organisms. The peroxiredoxin redox cycle is thus believed to be a conserved non-transcriptional timekeeping machinery (44, 45, 46). Our finding that peroxisome-related genes, alongside those involved in oxidative stress and lipid metabolism, are enriched among the rhythmically expressed genes in per⁰ larvae suggests a potential link between the 12-h rhythm pathway and peroxisome function.

Peroxisome biogenesis and activity display ultradian rhythms in the absence of circadian clocks and light

To test this possibility, we sought to investigate the temporal changes in peroxisome abundance and activity over a 24-h day in the fat bodies of w¹¹¹⁸ and per⁰ larvae. PEX14 is a critical component of the peroxisomal membrane, required for importing matrix proteins into peroxisomes in collaboration with PEX5 and PEX7. PEX14 is present in every peroxisomal membrane and serves as a marker for identifying and quantifying peroxisomes (34, 47). ABCD3 aids in detoxifying cells by transporting potentially harmful VLCFA derivatives into the peroxisome for breakdown. This function is crucial for preventing lipid-related cellular damage and maintaining overall cellular health. ABCD3 levels are highly dependent on metabolic states; thus, ABCD3 immunolabeling has been used to detect peroxisomes engaged in lipid metabolism (48, 49). Therefore, we conducted an immunohistochemistry analysis of PEX14 and ABCD3 proteins in the fat bodies of w¹¹¹⁸ and per⁰ larvae every 6 h in DD.

In w¹¹¹⁸, PEX14 immunolabeling displayed 12-h rhythms, peaking at CT8 and 20 (Fig. 2, A and B), while ABCD3 levels exhibited circadian rhythms with a peak at CT20 (Fig. 2, A and C). These observations suggest that peroxisome abundance (detected by PEX14) exhibits intrinsic 12-h rhythmicity, while their activity (detected by ABCD3) follows circadian rhythms. In per⁰, both PEX14 and ABCD3 signals displayed 12-h rhythms with peaks at CT2 and CT14 (Fig. 2, D–F). These results taken together indicate that peroxisome biogenesis and turnover exhibit 12-h rhythms independent of circadian clocks. In contrast, the activity of peroxisomes is regulated by circadian clocks. A 6-h shift in the peak timings of the PEX14 rhythms between w¹¹¹⁸ and per⁰ suggests that circadian clocks also influence the phase of ultradian rhythms of peroxisome biogenesis.

Figure 2.

Peroxisome biogenesis and activity in the fat body display ultradian rhythms in per⁰ larvae. Staining of per⁰ larval fat bodies with anti-PEX14 and ABCD3 antibodies demonstrates clock-independent ultradian rhythms of peroxisomal biogenesis and activity. In w¹¹¹⁸ larvae, while the PEX14 levels exhibit 12-h rhythms, ABCD3 signal displayed circadian rhythms. A and D, Representative confocal images of the fat bodies of w¹¹¹⁸ (A) and per⁰ (D) larvae collected every 6 h in DD. Blue, DAPI, Green, PEX14. Magenta, ABCD3. Scale bar, 10 μm. B and C, quantification of the PEX14 (B) and ABCD3 (C) signals in the w¹¹¹⁸. E and F, Quantification of the PEX14 (E) and ABCD3 (F) signals in the per⁰. In all graphs, the center lines indicate the mean and the error bars represent the standard deviation (SD). Dots represent individual values. arb. units, arbitrary units. n = 30 to 35 per group. Statistical analysis in (B), (C), (E), and (F) was performed using the Kruskal–Wallis nonparametric test followed by Dunn's multiple comparisons test or the ordinary one-way ANOVA followed by Tukey's multiple comparisons test, depending on the distribution of the data. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. ns, not significant.

Circadian clock-independent ultradian rhythms of lipid peroxidation cycles and ROS levels

While medium- and long-chain fatty acids are mainly oxidized in mitochondria, VLCFAs are almost exclusively metabolized by β-oxidation in peroxisomes (41). Peroxisomal β-oxidation results in the production of hydrogen peroxide, which can be converted into more reactive radical species. At the same time, peroxisomes contain several enzymes that scavange ROS to protect cells from oxidative damage (42, 50, 51). Our finding that peroxisomal abundance and activity exhibit circadian clock-independent 12-h rhythms thus suggests that ROS production and its downstream effects may also follow these rhythms.

To test this hypothesis, we first measured the levels of lipid peroxidation using a fluorometry assay in the fat bodies of w¹¹¹⁸ and per⁰ larvae collected every 6 h in DD. In w¹¹¹⁸ larvae, lipid peroxidation levels displayed 24-h rhythms (Fig. 3A). In contrast, in per⁰, lipid peroxidation displayed 12-h rhythms with peaks at CT2 and 14 (Fig. 3B), mirroring the temporal patterns in peroxisome abundance and activity (Fig. 2, D–F).

Figure 3.

Clock-independent ultradian rhythms in lipid peroxidation and ROS accumulation. A and B, quantification of lipid peroxidation levels, normalized against the total protein content, in w¹¹¹⁸ (A) and per⁰ (B) larval fat bodies collected every 6h in DD. n = 48 to 72 larvae per group. C, representative confocal images of intracellular ROS levels in the fat bodies of w¹¹¹⁸ and per⁰ larvae, detected using the ROS indicator H2DCF. D and E, quantification of intracellular H2DCF levels in w¹¹¹⁸ (D) and per⁰ (E) larvae. n = 24 to 30 larvae per group. F, Comparison of the H2DCF levels over a 24-h day between genotypes. In all plots, the center lines indicate the mean and the error bars represent the SD. In (A and B, each dots represents the value per cuvette. Statistical analysis in (A), (B), (D), and (E), was performed using either the Kruskal–Wallis nonparametric test or the ordinary one-way ANOVA test, depending on the distribution of the data. In F, the Mann-Whitney test was used to compare values between genotypes. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. ns, not significant.

To measure intracellular ROS levels in the fat body, fat body samples from both w¹¹¹⁸ and per⁰ larvae were corrected every 6 h and stained with 2′,7′-Dichlorodihydrofluorescein diacetate (H2DCF), a cell-permeable, oxidation-sensitive fluorescent dye. In w¹¹¹⁸ larvae, ROS levels exhibited a 24-h rhythm, peaking at CT2 (Fig. 3, C and D). This temporal pattern reflects that ROS accumulation follows the peroxisomal activity (Fig. 2C) and leads to lipid peroxidation (Fig. 3A). In contrast, per⁰ larvae exhibited 12-h rhythms in ROS levels with peaks at CT8 and CT20 (Fig. 3, C and E), which are the inverse of the peroxisomal activity rhythms (Fig. 2F) and lead to the lipid peroxidation rhythms (Fig. 3B). Notably, the average ROS levels in the fat body over 24h were not significantly different between w¹¹¹⁸ and per⁰ larvae (Fig. 3F). These results indicate that the 12-h peroxisomal activity cycle maintains daily lipid homeostasis even in the absence of clocks, thereby preventing the elevation of overall ROS levels.

Clock-independent ultradian rhythms in body fat levels

Our findings of the clock-independent 12-h rhythms in peroxisome abundance and activity, lipid peroxidation, and ROS levels prompted us to explore the role of these ultradian rhythms at the organismal level. We thus performed a floating assay, which indirectly measures fat reserves based on buoyancy (52). This assay was performed every 6 h in LD and DD in w¹¹¹⁸ and per⁰ third instar larvae (Fig. 4A).

Figure 4.

Clock-independent ultradian lipid storage rhythms in Drosophila larvae.A, schematic representation of the buoyancy assay, an indirect measurement of fat reserves. Per⁰ and w¹¹¹⁸ third instar larvae were collected every 6h in LD and DD. B and C, Quantification of the percentage of floating w¹¹¹⁸ (B) and per⁰ (C) larvae in LD. D, comparison of the buoyancy between genotypes in LD, combined over a 24-h day. E and F, Quantification of the percentage of floating w¹¹¹⁸ (B) and per⁰ (C) larvae in DD. G, Comparison of the buoyancy between genotypes in DD, combined over a 24-h day. In all plots, the center lines indicate the mean and the error bars represent the SD. Each dot represents the value per cuvette containing eight larvae. A total of n = 48 to 72 larvae were used per group. H, developmental progression of individual w¹¹¹⁸ (n = 120) and per⁰ (n = 96) animals from the wandering third instar (L3) larval stage to eclosion under LD cycles. Time 0 represents the first observation of a wandering L3 larva. The number of animals at each developmental stage is shown. I and J, timing of pupation (I) and the eclosion (J) in individual w¹¹¹⁸ and per⁰ animals. Center lines indicate the mean, error bars represent the SD. Each dot corresponds to the time point at which an individual fly pupated or eclosed. For the statistical analysis in (B), (C), (E), and (F), depending on the distribution of the data, the Kruskal–Wallis nonparametric test or the ordinary one-way ANOVA test was used for comparing values between timepoints. In D, G, I and J, the Mann-Whitney test or t test was used to compare values between genotypes. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. ns, not significant.

The floating assay revealed no significant temporal changes in buoyancy in w¹¹¹⁸ larvae in both LD and DD, indicating that they maintain consistent fat levels throughout the day (Fig. 4, B and E). In contrast, per⁰ larvae displayed a 12-h cycle in fat reserves in both LD and DD (Fig. 4, C and F), suggesting that these ultradian rhythms contribute to the regulation of body fat levels. Interestingly, fat levels averaged over 24h were significantly reduced in per⁰ larvae than in w¹¹¹⁸ in both LD and DD (Fig. 4, D and G). These results highlight that, although peroxisomal activity cycle persists in the absence of clocks, circadian clocks are nevertheless crucial for organisms' health and growth.

To further evaluate the effect of circadian clock disruption on growth and development, we tracked the developmental progression of individual w¹¹¹⁸ and per⁰ animals from the wandering third instar larval stage to eclosion (Fig. 4H). This assay revealed significant developmental delays in per⁰, with an average delay of approximately 12 h in pupation and 11 h in eclosion in per⁰ mutants compared to w¹¹¹⁸ controls (Fig. 4, I and J). Consistent with the established role of circadian clocks in entraining eclosion timing (53), eclosion of w¹¹¹⁸ flies were synchronized to lights-on transitions. In contrast, per⁰ flies displayed a broader distribution of eclosion (Fig. 4, H and J). These results underscore the crucial role of the circadian clocks in growth and defining developmental timing.

Additionally, to explore whether ultradian rhythms are also present in the fat bodies of adult flies, we analyzed previously published microarray data from adult Drosophila fat bodies for ultradian rhythmicity (54). We found that, in the flies with disrupted circadian clocks in the fat body (achieved by targeted expression of dominant-negative CLK), 258 genes exhibited 12-h periodicity, and none displayed 24-h cycling. In contrast, while 315 genes exhibited circa 24-h period, only 22 genes showed 12-h rhythms in wild-type flies. Furthermore, GO analysis of the 12-h cycling genes in dominant-negative CLK-expressing fat bodies revealed enrichment for peroxisomal genes, as well as genes involved in metabolism and detoxification, echoing the rhythmic biological pathways observed in per⁰ larval fat bodies (Supplementary Data S2). These results align with a metabolomics study using wild-type and per⁰ adult flies, which demonstrated the presence of certain metabolites exhibiting ultradian rhythmicity in per⁰ flies (55). Taken together, these findings support that ultradian rhythms are not restricted to the larval stage but are also present and functionally relevant in adult Drosophila.

Discussion

The interplay between circadian rhythms and metabolism is a conserved phenomenon across diverse species; however, specific metabolic variations regulated by clocks are still incompletely understood. Unexpectedly, we discovered clock-independent, 12-h period free-running rhythms in transcriptome and lipid metabolism in the fat body of Drosophila larvae, a dynamic metabolic organ crucial for coordinating metabolic demands with developmental timing. Ultradian peroxisomal activity cycles are coupled with redox cycles, thereby maintaining daily lipid homeostasis and keeping the ROS levels under control, even in the absence of circadian clocks.

Genes exhibiting ultradian rhythms in the fat bodies of per⁰ larvae include genes involved in peroxisome biogenesis and function (Fig. 1F and Table S1). A key function of peroxisomes is the β-oxidation of VLCFAs, a process that generates ROS as a byproduct. In wild-type larvae, anti-ABCD3 immunostaining levels—a proxy for peroxisome activity—exhibit circadian rhythms with a peak at CT20 (Fig. 2C). ROS levels peak at CT2 (Fig. 3D), suggesting a roughly 6-h delay between peak peroxisomal activity and ROS accumulation. This elevation of ROS, in turn, increases lipid peroxidation (56), leading to the peak of lipid peroxidation at CT8 in w¹¹¹⁸ (Fig. 3A). In contrast, in per⁰ larvae exhibit peroxisome activity peaks at CT2 and CT14 (Fig. 2F), followed by the elevation of ROS at CT8 and CT20 (Fig. 3E). Consequently, lipid peroxidation in per⁰ larvae peaks at CT2 and CT14 (Fig. 3B). Importantly, while peroxisomal activity increases ROS production, peroxisomes also play critical roles in scavenging ROS to protect cells from oxidative damage (42, 50, 51). Therefore, a 12-h peroxisomal activity cycle could explain how ROS levels are maintained in the absence of circadian clocks (Fig. 3F).

Clock-independent circadian rhythms have been observed in both procaryotes and eucaryotes, including human red blood cells (44, 45, 57). Additionally, circadian clock-independent ultradian rhythms have been found in mice (58). In mice lacking both Per1 and Per2, ultradian 16-h rhythms in gene expression were identified in the liver, encompassing peroxisomal-related genes and those involved in redox balance and metabolic homeostasis (58). Fly homologs of several of these genes, including abcd3, prx, peroxiredoxins, and Lsd-1, were among the rhythmically expressed genes in per⁰. In human red blood cells, non-transcriptional, clock-independent rhythms in peroxiredoxin redox states have been identified (45). Intriguingly, we found jafrac2, encoding peroxiredoxin 4, among the rhythmically expressed genes in per⁰ larvae. While our finding on peroxiredoxin rhythms pertain to gene expression levels rather than non-transcriptional rhythms, this similarity illustrates the conservation of circadian clock-independent rhythms across species. This underscores their potential importance in regulating metabolic and energetic balance, which may ensure the tissue and organism viability when molecular clocks are disrupted. It is noteworthy that ultradian rhythms in transcriptome and metabolome have been observed in mammalian liver also in the presence of circadian clocks (59, 60). This supports the evolutionary significance of ultradian metabolic rhythms, which can operate independent of and co-exist with circadian clocks.

Peroxisome abundance (detected by PEX14) exhibits 12-h rhythmicity, while their activity (detected by ABCD3) follows circadian rhythms in wild-type larvae (Fig. 2, B and C). Therefore, circadian regulation appears to act downstream of peroxisomal function. The mechanism underlying the generation of 12-h peroxisome biogenesis rhythms remains to be elucidated. The fat body's metabolic state is constantly communicated with IPCs and the PG and regulated through molecular cascades involving PI3K, AKT, TOR, S6K, and FOXO. When lipid storage decreases, the fat body senses this change and transmits the information to the IPCs (23) and the PG (22). This inter-organ feedback loop ensures that developmental timing aligns with environmental conditions (61). In wild-type larvae, pacemaker neurons and clocks in the PG likely impose 24-h rhythms in gene expression and physiology of the fat bodies through rhythmic endocrine signaling. This non-cell-autonomous regulation by circadian clocks echoes similar findings (62, 63) and highlights the importance of the circadian system in coordinating rhythms organism-wide.

In the absence of clocks, peroxisomal β-oxidation follows peroxisomal abundance, leading to the 12-h anti-phasic rhythms in ROS levels. It is likely that cyclic peroxisomal β-oxidation and ROS production could communicate with the IPCs and the PG, potentially creating a feedback loop between the brain and the fat body that couples transcriptome rhythms and lipid metabolism (Fig. 5). GO analysis of genes rhythmic in per⁰ larvae identified regulators of the JNK signaling. The JNK pathway is activated by various intrinsic and extrinsic stresses and promotes metabolic homeostasis, tissue, and organism survival (64, 65). Therefore, it is plausible that, in per⁰ larvae, cyclic peroxisomal β-oxidation and periodic ROS accumulation activate the JNK signaling pathway, thereby driving rhythmic expression of stress-response genes.

Figure 5.

A model of ultradian lipid peroxidation cycle in larval fat body and its regulation by circadian clocks. Circadian clocks are present in the PG and pacemaker neurons in the brain. Pacemaker neurons regulate PTTH neurons (PTTHn) in the brain, leading to circadian secretion of the PTTH. PG synthesizes ecdysone hormone under the control of PTTH (12, 13). The ecdysone released from the PG is converted into its active form, 20-hydroxyecdysone (E20), in the fat body in a nutrient-dependent manner. E20 acts on the IPCs in the brain to regulate secretion of DILPs (73). DILPs released from the IPCs promote ecdysone synthesis in the PG. DILPs also trigger IIS and TOR signaling in the fat body, leading to the regulation of numerous genes. Ecdysone released from the PG also regulates transcriptome and lipid metabolism in the fat body (21, 74). In the fat body, peroxisome biogenesis exhibits 12-h rhythms. In the presence of circadian clocks, circadian rhythms override peroxisome activity and ultradian rhythms, resulting in circadian cycles of peroxisomal activity and ROS production (Left, w¹¹¹⁸). In the absence of circadian clocks, peroxisome activity follows its abundance, displaying 12-h rhythms. Lipid β-oxidation produces ROS, which are degraded by peroxisomes. This reciprocal regulation generates 12-h rhythms in lipid peroxidation and ROS production with inverse phases. Changes in lipid storage and ROS levels in the fat body are communicated back to the brain, creating a feedback loop that maintains metabolic balance and mitigates oxidative damages (Right, per⁰).

Additionally, genes regulated by and involved in the TOR signaling pathway are found among the per⁰ cycling genes, such as PRAS40, AMPKα, PDK, mio, Atg2, and REPTOR-BP (Data S1 and Fig. S1). REPTOR-BP is a co-factor for REPTOR, which is a downstream effector of TORC1 signaling. Under low nutrient conditions, TORC1 is inhibited, triggering the activation of a transcriptional program initiated by REPTOR and REPTOR-BP to promote catabolism (66). This could potentially underlie the 12-h oscillation of body fat levels observed in per⁰ larvae (Fig. 4, D and F). Collectively, our data suggest the possibility that 12-h rhythms in lipid metabolism and ROS cycling are interconnected with the activation of the JNK and TOR pathways.

It is noteworthy that clock-independent ultradian rhythms are also present in adult flies (Data S2) (54, 55), suggesting their importance in metabolic homeostasis throughout the life cycle of a fly. While circadian rhythms override 12-h rhythms in peroxisome activity and its downstream effects, our findings highlight that ultradian lipid metabolism cycles can operate independently of circadian clocks. Importantly, however, uncoupling these ultradian rhythms and circadian rhythms has overall negative consequences for organismal growth.

Experimental procedures

Fly strain of third-instar larvae experiments

D. melanogaster strains were maintained on standard cornmeal-agar food at 18  °C or 25  °C in a 12 h light-dark cycle and controlled humidity. The lines w¹¹¹⁸ and per⁰ were previously described (31). To obtain third-instar larvae, male and female flies of each genotype were transferred to tubes with food for 3 days and entrained in a 12h light-dark cycle. Then, they were flipped twice every day in 12 h:12 h LD cycles or in DD. After 3 to 4 days, the food of the flipped tube was scraped and rinsed with distilled water to expose larvae. A mixed population of male and female larvae was used in all experiments. The developmental stage was confirmed by the characteristic shape of the larval mouth hooks.

Fat body collection and RNA-seq analysis

Fat bodies of w¹¹¹⁸ and per⁰ larvae, ranging from early to late third-instar, were dissected at nine timepoints, every 4h over a 32-h period in DD in two biological replicates. At each time point, the full fat body of five larvae was dissected and collected into 350 μl Buffer RLT Plus (QIAGEN RNeasy Plus Micro Kit, Cat. # 74034). Then, the tissue was dissociated using an electric Pellet Pestle Cordless Motor, five. Total RNA extraction was performed according to the manufacturer's protocol (QIAGEN RNeasy Plus Micro Kit, Cat. # 74034) and eluted in 14 μl of nuclease-free water. All samples were flash-frozen in liquid nitrogen and stored at −80  °C until further processing. All steps required for the RNA integrity check, library preparation, and sequencing were performed by experts at the iGE3 Genomics Platform at the University of Geneva. Total RNA integrity and quantitation were performed using the Bioanalyzer High Sensitivity RNA 6000 Pico kit (Agilent, Cat. # 5067-1513). cDNA library preparations were conducted using the SMART-Seq v4 Ultra Low Input kit from TaKaRa (Cat. # 634894) according to the manufacturer's instructions. Libraries were sequenced on the HiSeq2500 Illumina machines using single-end 100-bp reads to generate more than 10 million reads per library. The iGE3 Genomics Platform performed quality control, trimming of sequencing adaptors, and filtering of low-quality reads.

For RNA-seq data analysis, raw reads were aligned with the Dm6 Drosophila genome using STAR software (v2.7.9a) (67). The raw counts of reads associated with each of the exons were determined using featureCounts (68) function of the Rsubread package (v1.22.2). All libraries generated on average ∼18 million single-end 100 bp reads, from which approximately 80% (∼14 million reads) were mapped to the genome. Genes presenting no expression in at least one of the conditions in one of the replicates, as well as genes whose expression was below a count threshold were excluded from further analysis. This threshold was set up based on the distribution of the gene counts, established with edgeR function. After that, the DESeq2 package (69) (v.1.30.1) was used to normalize the gene expression levels. The normalized counts were used for downstream analysis.

Immunohistochemistry and microscopy

All the imaging experiments were performed at least in duplicate. Fat bodies from early to late third-instar w¹¹¹⁸ and per⁰ larvae were dissected at four timepoints (CT2, 8, 14 and 20). They were then fixed in 4% paraformaldehyde (PanReac AppliChem, Cat. #A3813) + 0.3% Triton X-100 for 1 h on ice and washed quickly twice and then twice for 20 min in PBST-0.3 (PBS, 0.3% Triton X-100). Subsequently, the fat bodies were blocked in blocking solution (55% normal goat serum, PBST-0.3) (CTS Cell Signaling Technology Inc. Cat. #5425S) for 1 h at room temperature, and incubated with the primary antibodies for 48 h at 4  °C. Guinea pig anti-Pex14 (1:1000) (70) and rabbit anti-Abcd3 primary antibody (1:500) (49) (gifts from Dr Andrew Simmonds) were used as primary antibodies, and Alexa 647 goat anti-rabbit IgG (1:500, Thermo Fisher, Cat. # A21245) and Alexa 568 goat anti-guinea pig IgG (1:500, Thermo Fisher, Cat. #A11075) were used as secondary antibodies. Samples were then mounted on slides with Vectashield with DAPI (Vector Laboratories, Cat. #H-2000). Images were acquired using a Nikon AX confocal microscope.

Floating assay

Late third-instar larvae were isolated at four timepoints in LD and DD (ZT/CT2, ZT/CT8, ZT/CT14, and ZT/CT20), excluding larvae that had begun to pupariate (detected by everted spiracles). For each replicate, a total of 24 larvae were transferred to three separate 1 ml clear plastic cuvettes (8 larvae per cuvette) containing 1 ml of 20% sucrose (w/v) solution (Sigma-Aldrich, Cat #84097) dissolved in PBS(52). Images were captured of each cuvette. Using ImageJ, images of each replicate were analyzed by drawing a line from the base of the cuvette to the center of the liquid meniscus. The positions of larvae relative to the cuvette base were then measured, and their vertical position was calculated as a fraction relative to the distance to the meniscus for each cuvette. If this fraction is > 0.5, then the larvae are considered floating; if it is ≤ 0.5, it is considered not floating. The percentage of floating larvae per cuvette was combined for three biological replicates of each genotype and analyzed using GraphPad Prism (v.8.1).

Developmental timing assay

Embryos from per⁰ and w¹¹¹⁸ control flies were collected over a 1.5-h period and transferred to vials containing standard food (25–30 embryos per vial). Vials were kept in a controlled environment at 25 °C in a 12 h/12h LD cycle. For continuous video monitoring, vials were illuminated from the backside using an infrared light panel (Standard Light Back-200 x 200-IR, Basler AG), and development was recorded using an infrared camera (ace Classic acA1920-25um, Basler AG) equipped with an IR filter (MidOpt BP850-25.4, Midopt Midwest Optical Systems) and a Fujinon HF6XA-5M lens (Fujifilm Corp) controlled by a Raspberry Pi computer. The appearance of wandering L3 larvae and the time points of pupation and eclosion for individual animals were manually scored at hourly intervals. Experiments were performed in duplicate, with three vials observed in each replicate.

Lipid peroxidation measurement

Peroxidized lipids were quantified using the lipid peroxidation kit (Sigma-Aldrich, Cat. # MAK085) following the manufacturer's instructions. For each replicate, four pools of six third-instar larvae were lysed in 300 μl malondialdehyde (MDA) lysis buffer and homogenized using an electric Pellet Pestle Cordless Motor (KIMBLE, Cat. #749540) at four timepoints in DD conditions (CT2, 8, 14, and 20). Samples were centrifuged at 13,000 g for 10 min to remove insoluble material. 200 uL of undiluted sample was used for the fluorometric measurement. Total protein concentration was quantified using the Pierce BCA protein assay kit (Thermo Scientific, Cat. #23227) on 25 μl of undiluted lysate, following the manufacturer's instructions. Colorimetric and fluorometric measurements were performed at 532 nm using a Pekin Elmer Victor X5 plate reader. All measurements were performed in triplicates, and lipid peroxidation levels were normalized against the total protein content of the sample.

ROS detection

The level of ROS production in the w¹¹¹⁸ and per⁰ third-instar larval fat body was measured at four timepoints (CT2, 8, 14 and 20) in DD using a cell-permeable dye, 2′7′-dichlorofluorescin diacetate (H2DCF) (Sigma-Aldrich, D6883), following the protocol described in (71). The fluorescence intensity for the fat body was measured using Image J/Fiji software (72) (v2.1.0/1.53c) after generating a Z-SUM projection from10 Z-stacks.

Rhythmic gene expression analysis

Expression profiles were analyzed for rhythmicity in R (r-project.org) using MetaCycle (32) (v1.2.0) an N-version programming (NVP) method to explore periodic data. This method implements ARSER (ARS), JTK_CYCLE (JTK), and Lomb-Scargle (LS) algorithms which uses a nonparametric test with the Fisher's correction. Each algorithm independently estimates period, phase, and p-values for rhythmicity. MetaCycle's meta2d function then combines these outputs. The period is determined using the arithmetic mean of individual algorithm estimates. Phase is calculated as the circular mean, accounting for the cyclical nature of time. p-values are merged using Fisher's method. Amplitude is calculated using least squares fitting to a general sinusoidal model, based on the final integrated period and phase.

RNA expression profiles and heatmaps were produced using ggplot2 (v3.1.0) in R. In the heatmaps, genes were sorted based on the phase, and the normalized counts were represented. The significance threshold was set to an adjusted p-value (p.adj) < 0.2. Gene Ontology (GO) term enrichment analysis of cycling genes was performed with the topGO (R package version 2.59.0) package (https://bioconductor.org/packages/topGO) using the runTest function with the “classic” algorithm and the Fisher statistics. The significance threshold was set to a p-value  <  0.005.

Analysis of a previously published microarray dataset generated from the fat body of adult flies (54) were performed as follows. Since the data were collected using microarrays rather than RNA-seq and comprised a single biological replicate per time point, we followed the original study's normalization strategy using the MAS5 method. Subsequently, we applied harmonic regression based on autoregressive spectral estimation (ARSE), implemented in the MetaCycle R package, for rhythm detection. This approach is suitable for detecting rhythmic transcripts in single-replicate time-series data. Periodicity was tested in two distinct ranges: 20 to 24 h (circadian) and 12 to 16 h (ultradian), using an adjusted p-value threshold of < 0.2, consistent with our own transcriptomic analysis. Gene ontology (GO) term enrichment analysis was conducted using the same method described above.

Image analysis

Confocal Z-stacks were analyzed using the Image J/Fiji software (1) (v2.1.0/1.53c). A sum-Z projection was generated from five Z-stacks, maintaining this number of slices across all samples to ensure comparability. To quantify fluorescence intensity, the PEX14 signal was used to create a mask and define the region of interest (ROI) in the FB section. This ROI was then used to extract fluorescence intensity values for both PEX14 and ABCD3.

Statistics

Statistical analysis and data visualization were performed using GraphPad Prism (v.8.1). Normally distributed data were compared using parametric tests, and non-normally distributed data were analyzed using nonparametric tests. Statistical comparison of fluorescence intensities, fluorometric assay, and buoyancy assay data was performed using the ordinary one-way ANOVA with Turkey's multiple comparison test when data were normally distributed. For non-normally distributed data, Kruskal–Wallis one-way ANOVA with Dunn's multiple comparison test was employed. To compare overall fluorescence intensity, fluorometric assay, and buoyancy assay data between the per⁰ and control genotypes, a t test was used for normally distributed data, and a Mann-Whitney test was used for non-normally distributed data. Significant values in all figures are: ∗p < 0.05, ∗∗p < 0.01. ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. “ns” indicates not significant. All the experiments were repeated at least twice.

Data availability

RNA-seq data has been deposited at the NCBI Sequence Read Archive (SRA) database (BioProject ID: PRJNA1188771) and at the Gene Expression Omnibus (GEO database (accession number: GSE282559).

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Qi-Qun Tang