Circadian Phase Determines Tissue-Specific Adaptations to Long-Term Exercise in Obese Mice

The global escalation of obesity continues to drive the burden of type 2 diabetes, cardiovascular disease, and metabolic dysfunction-associated steatotic liver disease (MASLD; formerly known as non-alcoholic fatty liver disease, NAFLD) [1,2,3,4]. Meta-analytic estimates place the prevalence of MASLD near 25% worldwide, with substantially higher rates among individuals with obesity or diabetes [4]. Population-based cohorts also reveal marked race- and sex-specific differences in obesity-related vascular risk [3]. While positive energy balance and sedentary behavior remain proximal drivers of weight gain [5,6], structured exercise is an effective, low-risk intervention that improves multiple domains of metabolic health and cardiorespiratory fitness [7,8]. Beyond energy balance per se, mounting evidence indicates that circadian disruption is an important, modifiable contributor to metabolic disease.

The circadian timing system comprises a central pacemaker in the suprachiasmatic nucleus and peripheral clocks across metabolic organs, collectively orchestrating behavior, hormone secretion, and substrate metabolism [9,10,11]. Disruption of this system increases vulnerability to metabolic diseases with inflammatory features, such as obesity and diabetes [12], while endocrine mediators including adiponectin exhibit circadian oscillations that are metabolically relevant [11]. Nutritional challenges, particularly a high-fat diet (HFD), dampen circadian amplitude and shift the phase of clock-controlled genes in peripheral tissues, leading to behavioral rhythm disturbances and impaired metabolic flexibility [13,14,15]. At the molecular level, HFD-induced reprogramming of the hepatic transcriptome further disrupts the normal diurnal balance between lipid synthesis and oxidation [16]. Given this tight interconnection between circadian regulation and metabolism, exercise has emerged as a potent nonphotic zeitgeber capable of modulating both central and peripheral clocks. Appropriately timed exercise can shift the phase of core clock gene expression and thereby contribute to the realignment of circadian metabolic rhythms [17]. Recent reviews further highlight that circadian-aligned behaviors—particularly the timing of exercise (chrono-exercise) and feeding (chrono-nutrition)—jointly regulate metabolic flexibility, insulin sensitivity, and mitochondrial function [18]. Taken together, these findings highlight exercise as a promising behavioral strategy to restore or re-entrain circadian metabolic rhythms.

Recent acute studies have revealed distinct time-of-day-dependent metabolic signatures following a single bout of exercise [19]. In adipose tissue, early active-phase exercise (e.g., ZT15, Zeitgeber time) preferentially induces genes for lipolysis and thermogenesis, whereas responses are attenuated in the early rest phase (ZT3) [20]. Notably, obesity diminishes this phase dependency, suggesting reduced metabolic plasticity under HFD conditions [21]. Collectively, acute studies indicate phase gating of exercise responses, but whether such gating persists after long-term training remains insufficiently defined.

Studies of long-term exercise interventions have likewise revealed pronounced time-of-day-dependent differences. Voluntary wheel running over several weeks was shown to alter clock gene expression, with phase advancement in the liver and adipose tissue and markedly increased expression levels in skeletal muscle [22]. Another study demonstrated that maximal running speed and endurance were enhanced in the late active phase, whereas no improvement was observed in the early active phase [23]. Forced treadmill training further supported these differences: endurance capacity was significantly improved when training occurred in the late active phase but was less effective in the early active phase, a discrepancy associated with liver glycogen content and feeding behavior [24]. In obesity models, comparison of early versus late active-phase exercise revealed that only late-phase training significantly ameliorated MASLD, whereas early-phase training was ineffective [25]. Collectively, these findings underscore the critical role of exercise timing in shaping long-term adaptations.

Despite growing evidence for the importance of exercise timing, several research gaps remain. First, most studies have focused on acute or short-term interventions [20,22], with limited verification of long-term adaptations. Although a two-week voluntary wheel-running study [22] showed that skeletal muscle rhythmicity and exercise performance peaked during the late active phase; however, voluntary activity differs fundamentally from forced treadmill exercise in both physiological load and experimental control. Furthermore, previous studies—whether voluntary or forced—have rarely examined exercise performed during the rest phase [23]. Thus, systematic comparisons of long-term exercise between the rest phase (ZT3) and the active phase (ZT15) in HFD-induced obese models remain scarce. Second, the synergistic mechanisms between the liver and adipose tissue that underlie homeostatic adaptations to exercise performed at different circadian phases remain poorly understood. Beyond single-tissue outcomes, the coordination between hepatic lipid handling and adipose mobilization represents a critical yet unresolved component of circadian exercise adaptation. Together, these gaps highlight the need to clarify how circadian timing governs tissue-specific metabolic adaptations to long-term exercise in obesity.

To address these questions, we employed an HFD-induced obese mouse model under controlled dietary and light–dark conditions, allow detailed molecular analysis of liver and adipose tissue responses under standardized experimental settings. We hypothesized that long-term exercise would induce tissue-specific adaptations governed by circadian phase, with active-phase training (ZT15) favoring lipid mobilization in adipose tissue and rest-phase training (ZT3) promoting a hepatic oxidative advantage. The results demonstrate that long-term exercise training produces distinct, tissue-specific adaptations governed by circadian phase, underscoring chrono-exercise as a critical design variable for metabolic interventions in obesity.

This controlled experimental study used seven-week-old healthy male C57BL/6J mice (n = 32), obtained from Takasugi Experimental Animals Supply Co., Ltd. (Kasukabe, Japan). All experimental procedures were approved by the Institutional Animal Care and Use Committee of Waseda University (Saitama, Japan; approval number: A23-126, approved on 11 April 2023) and conducted in accordance with the ARRIVE 2.0 guidelines.

The experiment comprised 8 weeks of HFD feeding followed by 8 weeks of treadmill training or sedentary control. Tissue collection was performed 48 h after the final exercise session following a 4 h fast, to minimize the acute effects of exercise. After euthanasia, tissues were rinsed with posphate-buffered saline (PBS) to remove surface blood, blotted dry, weighed, and immediately frozen in liquid nitrogen. All samples were stored at −80 °C until subsequent biochemical, histological, and molecular analyses. The prespecified primary endpoint was plasma triglycerides (TGs), and additional analyses of liver and adipose tissues were conducted to evaluate tissue-specific adaptations to exercise timing.

Mice were housed in plastic cages (2–3 per cage) under controlled temperature and maintained on a 12:12 h light–dark cycle (lights on at 08:00 and off at 20:00), with ZT 0 defined as lights on and ZT12 as lights off. Standard bedding was provided throughout the experiment and replaced weekly. To minimize baseline differences in body weight, mice were assigned to two groups using pairwise randomization based on body weight. After a 2-week acclimatization period, all mice were ranked by body weight, and the overall mean was calculated. Animals with similar body weights around the mean were sequentially paired, and one mouse from each pair was alternately assigned to the ZT3 or ZT15 group. All mice had ad libitum access to an HFD (Research Diets D12492, E.P. Trading Co., Ltd., Tokyo, Japan; protein 20 kcal%, carbohydrate 20 kcal%, and fat 60 kcal%), and water.

Two circadian phase groups were established (Figure 1):

ZT3 group (n = 16): Maintained under the standard light–dark cycle (ZT0 = 08:00).

ZT15 group (n = 16): On Day 3 of HFD feeding, mice were exposed to constant light for 24 h and then subjected to a 12 h reversal of the light–dark cycle (new ZT0 = 20:00) to align exercise timing with the mice’s natural active phase, while ensuring that training for the ZT3 and ZT15 groups was conducted at the same clock time and under identical environmental conditions. This procedure produced a stable 12 h phase delay while preserving a 24 h rhythm [26,27].

After 8 weeks of HFD feeding, each group was further subdivided into sedentary (ZT3-sed, n = 8, ZT15-sed, n = 8) and exercise (ZT3-exe, n = 8, ZT15-exe, n = 8) groups based on body weight (Figure 1). Sample size was determined based on previous similar studies and logistical feasibility [20,28]. Subgroup allocation was performed using the same pairwise randomization method described above to minimize baseline differences. During the entire intervention, body weight, food intake, and water intake were measured weekly, and feeding behavior was monitored 3–7 times per week at ZT0 (morning) and ZT12 (evening) to assess circadian rhythmicity. Food intake was recorded per cage (2–3 mice each) and averaged within each cage to obtain a single data point. Therefore, the statistical unit for feeding analyses was the cage (n = 6 cages per group during the HFD period and n = 3 cages per group during the exercise intervention).

Quality control and exclusion criteria: all histological and molecular analyses were conducted by investigators blinded to group allocation. A few animals were excluded from the final dataset due to unexpected death or severe fighting-related injuries. Predefined exclusion criteria included >15% body weight loss or visible injury, in accordance with institutional welfare policies. After exclusion, the final sample sizes ranged from n = 5 to 8 per group (Figure 1), ensuring data integrity and comparability.

Exercise training began in Week 9 of HFD feeding and continued for 8 weeks (5 days/week, 60 min/day). Before the formal intervention, mice underwent a two-day familiarization period (10 and 12 m/min, 10 min/day) to reduce stress and ensure adaptation to treadmill running.

Exercise groups (ZT3-exe, ZT15-exe) performed treadmill running at their assigned circadian phase:

ZT3 (early rest phase): 11:00 local time.

ZT15 (early active phase): 23:00 local time after phase reversal.

Training was performed on a motorized treadmill (KN-73, Natsume Seisakusho Co., Ltd., Tokyo, Japan) with a 0° incline, following a progressive moderate-to-vigorous aerobic protocol corresponding to approximately 60–70% of VO₂ max for obese C57BL/6J mice [29,30]. During exercise, if necessary, mice were gently encouraged to continue running by light tail tapping, and no electric shocks were applied. To minimize stress, the same operator manually adjusted the speed using a rotary dial at a steady pace, ensuring smooth acceleration while continuously monitoring the animals’ condition. The treadmill speed was set as follows: training started at 9 m/min in Week 9, increasing by 2 m/min every 10 min until the 20th minute, after which the speed was maintained for the remainder of the 60 min session. The initial speed increased by 1 m/min each subsequent week. By Week 13, the starting speed reached 13 m/min, rising to 16 m/min at 5 min and 17.5 m/min at 10 min, which was maintained thereafter. Training from Weeks 14–16 followed the same speed pattern. Sedentary controls (ZT3-sed and ZT15-sed) did not perform treadmill running but were placed around the treadmill at the same time to control for environmental factors.

At the end of the intervention, tissue sampling began 48 h after the final exercise session to avoid acute exercise effects. Prior to sampling, a 4 h fasting period was implemented to minimize the impact of food intake on metabolic parameters. Sampling and weighing were performed alternately between groups, with each tissue collected by the same investigator. Both the operator and the recorder of tissue weights were blinded to group allocation.

During the experiment, blood glucose levels were monitored at three time points: before HFD feeding (Week 0), before training (Week 8), and after training (Week 16). Blood glucose was measured using a glucometer (Free Style Precision Neo; Abbott Japan Co., Ltd., Tokyo, Japan) in combination with compatible electrodes (FS Precision Glucose Test Electrodes; Abbott Japan Co., Ltd., Tokyo, Japan). All measurements were conducted by the same operator, beginning at ZT2 for the ZT3 group and ZT14 for the ZT15 group, with alternate testing order to account for circadian influences. For each measurement, 1–2 μL of capillary whole blood was collected from the tail tip and applied to the electrode, and glucose concentration was automatically determined by the device via an electrochemical method.

Hepatic lipids were extracted using a modified Bligh and Dyer chloroform–methanol method. Briefly, approximately 100 mg of frozen liver tissue was placed into a pre-weighed 2 mL microtube, and the net weight was recorded. A 10-fold volume of distilled water was then added, and the tissue was homogenized. A 440 μL aliquot of the homogenate was transferred into a new 2 mL tube, to which 1100 μL of chloroform–methanol solution (1:2, v/v; Chloroform, Cat# C2432, Sigma-Aldrich Japan, Tokyo, Japan; Methanol, Cat# 19-2420-3, Sigma-Aldrich Japan, Tokyo, Japan) was added. After thorough vortexing and centrifugation, the lower organic phase (chloroform layer) was collected. The solvent was evaporated at 80 °C until only lipid residues remained. The lipid residue was dissolved in 100 μL isopropanol (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan). The investigator performing liver sampling and TG assays was blinded to group allocation throughout the experiment. TG levels in the liver tissue were quantified using a commercial kit (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) by colorimetric assay in a 96-well plate by a plate reader (SpectraMax iD5, Molecular Devices, San Jose, CA, USA). Assay reliability was verified by the standard calibration curve (R² > 0.99) included in each batch. Concentrations were calculated from the calibration curve and expressed as milligrams of TGs per gram of wet tissue weight [31,32,33].

Blood samples were collected from the posterior vena cava of mice under anesthesia, and plasma was separated by centrifugation. All sampling and measurements were performed by the same operator, with alternating group order to minimize potential effects of sampling time differences between groups on the results. Plasma levels of non-esterified fatty acids (NEFAs), glucose, and TGs were measured in duplicate using commercial biochemical assay kits (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) with a colorimetric method in 96-well plates and absorbance was determined using a microplate reader. Concentrations were calculated based on standard calibration curves.

Additional plasma biochemical parameters were analyzed by Kotobiken Medical Laboratories, Inc. (Tokyo, Japan), including liver function markers [aspartate aminotransferase (AST), alanine aminotransferase (ALT), and albumin]; lipid metabolism markers [total cholesterol (T-Cho), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C)]; myocardial and tissue damage markers [creatine kinase (CK) and lactate dehydrogenase (LDH)]; pancreatic function markers [amylase and lipase]; and renal function markers [blood urea nitrogen (BUN), creatinine (Cr), and uric acid (UA)]. These measurements provided a comprehensive evaluation of systemic metabolic status.

Frozen liver sections were air-dried at room temperature for 30 min, fixed in 10% neutral buffered formalin solution (Cat# 062-01661, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) for 10 min, and then briefly rinsed with 60% isopropanol. The sections were subsequently stained in working Oil Red O solution (Cat# 4049-1, Muto Pure Chemicals Co., Ltd., Tokyo, Japan) for 15 min, followed by quick washes with 60% isopropanol and distilled water. Counterstaining was performed with Mayer’s hematoxylin (Cat# 131-09665, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) for 1 min, rinsed with distilled water, and mounted with an aqueous mounting medium. Sections were photographed using an all-in-one fluorescence microscope (BZ-X800, KEYENCE, Osaka, Japan) equipped with a Plan Apo 20× objective lens (KEYENCE, Osaka, Japan).

For each mouse, 4–6 sections from different depths were analyzed. Five to six mice per group were selected for quantification to ensure representative morphology. Mice with extreme body weights or tissue abnormalities (e.g., excessive steatosis, damage) were excluded. A blinded investigator randomly selected ten non-overlapping fields per mouse, and ImageJ software (version 1.54f, National Institutes of Health, Bethesda, MD, USA) was used for quantification. The Oil Red O-positive area was expressed as a percentage of total tissue area (mean ± SEM, n = 5–6 per group; SEM = Standard error of the mean).

Total RNA was extracted using the NucleoSpin RNA Kit (Takara Bio Inc., Kusatsu, Japan) according to the manufacturer’s instructions, and RNA concentration and purity were determined using a spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). cDNA was synthesized using the PrimeScript^TM FAST RT Reagent Kit with gDNA Erase (Takara Bio Inc., Kusatsu, Japan). Quantitative PCR amplification was performed on a Thermal Cycler Dice^® Real Time System IV (Takara Bio Inc., Kusatsu, Japan) using TB Green^® Premix Ex Taq™ II FAST qPCR (Takara Bio Inc., Kusatsu, Japan), following the recommended protocol. To ensure accurate normalization, several candidate reference genes were initially evaluated for expression stability across groups. Among these, ribosomal protein S18 (Rps18) in liver and glyceraldehyde-3-phosphate dehydrogenase (Gapdh) in adipose tissue showed the lowest intra-group variability (coefficient of variation of Ct values < 2%) and were therefore selected as stable reference genes. Relative mRNA expression levels were calculated using the 2^–∆∆Ct method. Primer sequences for all target genes are listed in Table 1.

Statistical analyses were performed using GraphPad Prism 10.4.0 (GraphPad Software, La Jolla, CA, USA). Data normality was tested with the Shapiro–Wilk test, and variance homogeneity with the F test (two groups) or Brown–Forsythe test (fore groups). When assumptions were violated, data were log-transformed or analyzed using nonparametric tests. Two-group comparisons used unpaired Student’s t-tests (equal variances) or Welch’s t-tests (unequal variances). Four-group comparisons applied two-way analysis of covariance (ANOVA, phase × exercise) to test main and interaction effects. Significant interactions were followed by simple main effect analyses with Bonferroni post hoc corrections. To control false discovery rate, p values were adjusted using the Benjamini–Hochberg FDR method, yielding both raw p and adjusted q values. Effect sizes (partial η² for ANOVA; Hedges’ g for pairwise comparisons) and 95% confidence intervals (CIs) were reported. Significance was set at p < 0.05 (two-tailed). Data are presented as mean ± SEM, with 5–8 mice per group unless otherwise stated.

All analyses complied with ARRIVE 2.0 guidelines. In the main text, only statistically significant or biologically meaningful results are described, with nonsignificant findings summarized in the. Supplementary Materials

To assess the impact of exercise timing and circadian phase on hepatic steatosis and systemic metabolic health, we analyzed hepatic lipid accumulation, hepatic TG content, and plasma biochemical parameters. Statistical analysis was performed using two-way ANOVA (phase × exercise). Oil Red O staining of liver sections revealed a clear reduction in lipid droplet size in exercise-trained groups compared to sedentary controls (Figure 8a). Quantitative analysis confirmed a significant main effect of exercise on the percentage of lipid droplet area (F(1, 19) = 4.78, p = 0.0416, partial η² = 0.20, mean difference = 4.77, 95% CI [0.20, 9.33]; Figure 8b). A significant main effect of phase was also detected for hepatic TG content (F(1, 25) = 15.49, p = 0.0006, partial η² = 0.38, mean difference = −162.70, 95% CI [−259.00, −66.30]), with lower TG levels observed in the ZT3 groups compared with the ZT15 groups (Figure 8c). Analysis of plasma liver function markers revealed a significant main effect of phase on AST (p = 0.0295; Figure 8d, Table S3), while no significant differences were observed for ALT or albumin (Figure 8e,f). Additional plasma biochemical analyses of lipid metabolism markers (T-Cho, LDL-C, HDL-C), myocardial/tissue damage markers (CK, LDH), pancreatic function markers (amylase, lipase), and renal function markers (BUN, Cr, UA) revealed no significant differences among groups (Figure S2).

Overall, these findings indicate that exercise was associated with reduced hepatic lipid accumulation, whereas circadian phase corresponded to variations in hepatic TG content and plasma AST levels.

To comprehensively characterize the phase- and exercise-dependent regulation of hepatic gene expression, we analyzed a panel of genes involved in circadian rhythms, lipid metabolism, and inflammation. Gene expression was analyzed by two-way ANOVA (phase × exercise), and multiple testing was controlled using the Benjamini–Hochberg false discovery rate (FDR) correction. Statistical significance was defined at q < 0.05. Full p and q values for all genes are provided in Supplementary Table S5.

To elucidate the molecular adaptations in EPI in response to exercise timing and circadian phase, we analyzed the expression of genes involved in lipogenesis, lipolysis, fatty acid oxidation, and adipokine signaling. Differences in gene expression were evaluated using two-way ANOVA (phase × exercise). To account for multiple testing across genes, p values were further adjusted using the Benjamini–Hochberg false discovery rate (FDR) method, with significance defined as q < 0.05. The complete set of FDR-adjusted results is presented in Supplementary Table S5.

To further examine the relationships between plasma TG concentrations and the expression of lipid metabolism-related genes in liver and adipose tissue, we performed Pearson correlation and linear regression analyses. In EPI, both Pnpla2 (r = −0.49, R² = 0.24, slope = −9.27, 95% CI [−15.86, −2.68], F(1, 27) = 8.33, p = 0.0076) and Lipe (r = −0.42, R² = 0.18, slope = −10.39, 95% CI [−19.22, −1.57], F(1,27) = 5.84, p = 0.02) showed significant negative correlations with plasma TGs, with higher expression levels corresponding to lower plasma TG values. In the liver, Pparg also displayed a significant negative correlation with plasma TGs (r = −0.52, R² = 0.27, slope = −5.06, 95% CI [−8.35, −1.77], F(1, 27) = 9.97, p = 0.004), suggesting that lower hepatic Pparg expression was associated with lower plasma TG levels. Together, these results indicate a tissue-specific association between plasma lipid levels and the expression of lipid metabolism-related genes.

This study systematically investigated how circadian phase influences long-term exercise adaptations in HFD-induced obese mice. Plasma TG was defined as the primary endpoint, representing the most sensitive systemic marker of metabolic adaptation. Three main observations emerged: First, plasma TG showed a significant circadian phase × exercise interaction, being lowest after training during the active phase (ZT15). Second, hepatic TG content was primarily determined by circadian phase, being lower in ZT3 groups, while exercise mainly reduced hepatic lipid droplet area. Third, active-phase exercise enhanced adipose expression of lipolysis- and oxidation-related genes, whereas rest-phase training coincided with a hepatic oxidative advantage. Overall, these findings indicate that exercise effects are both time- and tissue-dependent: active-phase exercise promotes lipid mobilization, while rest-phase conditions facilitate hepatic lipid clearance. Notably, several gene-level signals became weaker after multiple-comparison correction; therefore, these results should be interpreted as exploratory trends rather than definitive conclusions.

Building on these primary findings, the metabolic effects of long-term exercise interventions are both time-dependent and tissue-specific, showing consistent patterns at the behavioral and systemic levels. Plasma TG displayed a clear phase × exercise interaction, being lowest in the active-phase exercise group (ZT15-exe), suggesting that the time dependency observed in acute exercise is preserved at the steady-state endpoint after long-term training [20]. Hepatic TG content was mainly governed by phase, with lower levels in ZT3 than in ZT15, while exercise primarily reduced lipid droplet area. Although the pattern of hepatic TG changes observed in this study differs slightly from that reported previously, possibly due to differences in training time points, both studies consistently suggest a phase-dependent influence of exercise timing on long-term metabolic adaptation [25]. In addition, while exercise timing affected body weight and feeding rhythms, these effects were insufficient to explain the observed metabolic outcomes. In the ZT15-exe group, the reduced amplitude of feeding rhythm may indicate either altered energy allocation or a partial attenuation of rhythmicity; however, further studies are required to clarify this relationship.

The results demonstrate that hepatic metabolic state is primarily governed by circadian phase and secondarily modulated by exercise. Hepatic TG content was significantly higher at ZT15 than at ZT3, indicating that lipid accumulation tends to occur during the active phase. Histologically, Oil Red O staining showed that hepatic lipid droplet area was markedly reduced in exercised mice, suggesting that long-term training facilitates hepatic lipid clearance and turnover. Meanwhile, Cpt1a expression was consistently higher in ZT3 than in ZT15 groups and in sedentary than in exercised mice, indicating strong circadian rhythmicity in hepatic fatty acid oxidation, with exercise potentially enhancing lipid breakdown on this background. Core clock genes and lipogenic regulators (Bmal1, Cry1, Per2, Nr1d1, Srebf1, Acaca, and Fasn) displayed clear phase-dependent expression, further confirming that hepatic lipogenesis networks are rhythmically regulated by intrinsic clock systems. This is consistent with previous circadian transcriptome atlases, which revealed time-of-day-dependent expression of lipid metabolism genes in the liver, highlighting that the endogenous clock coordinates lipid synthesis and oxidation to maintain metabolic homeostasis [34]. Moreover, Pparg expression was negatively correlated with plasma TGs, suggesting a potential transcriptional mechanism whereby exercise and circadian phase jointly influence hepatic lipid droplet homeostasis, although this mechanism requires further validation.

Overall, hepatic changes were mainly phase-driven, with exercise acting to optimize lipid handling and relieve metabolic stress rather than directly reducing total hepatic TGs. This finding is generally consistent with previous reports showing that late active-phase exercise more effectively attenuates hepatic steatosis than early active-phase exercise [25], although the specific response patterns observed in this study were not entirely identical.

The results indicate that adipose tissue metabolism exhibits robust circadian characteristics and is further modulated by exercise timing. In EPI, core clock genes and Srebf1 showed significant phase-dependent expression, reflecting rhythmic regulation of lipogenic pathways, while Fasn was elevated in exercised mice, suggesting that long-term training might contribute to increased de novo lipogenesis in adipose tissue. Such an adaptive remodeling likely reflects the tissue’s metabolic buffering role in systemic lipid homeostasis, facilitating lipid uptake and re-esterification to stabilize plasma TG levels [35,36]. This interpretation is consistent with the notion that healthy adipose tissue supports systemic metabolic health through dynamic lipid turnover [35]. Pnpla2 expression was inversely associated with plasma TGs, implying that mild enhancement of adipose lipolytic activity may contribute to systemic lipid improvement.

In summary, long-term exercise may induce rhythmic adaptation in adipose tissue, rebalancing between “mobilization–oxidation” and “synthesis–storage” phases, thereby improving lipid handling and transport efficiency. This trend is consistent with Pendergrast et al. [20], who reported that active-phase exercise enhances fat mobilization and mitochondrial function, further supporting the tissue specificity of exercise timing effects.

Integrating these findings, we propose a “tissue × time” cooperative model: the metabolic benefits of exercise are jointly determined by tissue type and circadian phase, with the significant phase × exercise interaction in plasma TG concentration reflecting this systemic coordination. During the rest phase (ZT3), the liver exhibits stronger fatty acid oxidation, and exercise may act synergistically with this oxidative background to reduce lipid droplet load and optimize lipid handling. In contrast, during the active phase (ZT15), adipose tissue appears more responsive at the transcriptional level. Exercise increased Fasn expression, while Pnpla2 and Lipe were inversely associated with plasma TG levels, suggesting rhythmic regulation of lipid uptake and re-esterification, which may underlie the observed TG reduction. Taken together, TG reduction following active-phase exercise may correspond to adipose transcriptional remodeling, whereas rest-phase exercise appears to reflect improved hepatic lipid processing.

Our study provides preclinical evidence for the concept of “chrono-exercise,” indicating that the metabolic outcomes of exercise depend on the timing of training. Human studies likewise support this circadian dependence. Morning exercise enhances parasympathetic activity, whereas evening sessions increase sympathetic tone [37], sustain higher lipid utilization during recovery [38], and improve insulin sensitivity and lipid profiles more effectively than morning exercise [39]. In addition, exercise has been shown to modulate the circadian expression of core clock genes such as Cry1, Per2, and Bmal1 in human skeletal muscle [40]. Collectively, these findings suggest that the optimal exercise time may vary depending on individual circadian characteristics and metabolic goals.

At the clinical level, circulating adipokines such as adiponectin and leptin are measurable indicators of metabolic health. Although this study examined adipokines only at the transcriptional level and found no significant group differences, clinical evidence indicates that decreased leptin and increased resistin levels are associated with malnutrition and dyslipidemia [41]. These observations emphasize the potential value of circadian-aligned exercise programs in preventing metabolic disorders. Overall, the proposed “tissue × time” model provides a framework for integrating circadian rhythms with exercise physiology, but further mechanistic and clinical validation across sexes and populations is required to establish its broader applicability.

Several limitations should be acknowledged. First, sampling was performed 48 h after the final exercise session following a 4 h fast, which minimized acute effects but limited the ability to infer dynamic lipid flux. Therefore, the mechanistic interpretations should be considered hypothesis-generating and require further validation. Second, protein-level or functional analyses (e.g., ATGL, HSL, CPT1A) were not conducted, reducing the certainty of pathway-related conclusions. Third, circadian phase was mainly inferred from feeding behavior, while other physiological rhythms were not monitored, limiting system-level interpretation. Fourth, the light–dark manipulation used for ZT15 mice may have introduced phase-specific confounding despite re-entrainment before the experiment. Fifth, forced treadmill exercise may have induced stress responses interacting with circadian regulation. Finally, this study was limited to male C57BL/6J mice with modest sample sizes, and the findings should therefore be regarded as preclinical and hypothesis-generating, warranting further confirmation across sexes, strains, and human populations.

The original contributions presented in this study are included in the article and. Further inquiries can be directed to the corresponding author. Supplementary Materials