Synchronizing the Liver Clock: Time-Restricted Feeding Aligns Rhythmic Gene Expression in Key Metabolic Pathways

Eight- to ten-week-old male C57BL/6 mice were obtained from the Guangdong Medical Laboratory Animal Center and housed under specific-pathogen-free conditions. They were maintained on a 12 h light/12 h dark cycle (lights on at 7:00, ZT0), with free access to water and unrestricted movement. Sixty 8-week-old male C57BL/6 mice were randomly assigned to four groups: the control group (CON group), the control with Time-Restricted Feeding group (CON + TRF group), the high-fat diet group (HFD group), and the high-fat diet with time-restricted feeding group (HFD + TRF group) to minimize baseline bias. Mice in the CON and HFD groups had ad libitum access to the standard chow or high-fat diet, respectively, whereas CON + TRF and HFD + TRF mice were only provided access to their respective diets during the daily ZT12 (19:00) to ZT15 (22:00) window, with food removed at all other times. Following this feeding regimen for six months, daily food intake was recorded and body weight was measured every three days. After 22 weeks, mice at each time point were selected randomly from their respective groups and anesthetized with isoflurane at five distinct time points: ZT0 (07:00), ZT6 (13:00), ZT12 (19:00), ZT18 (01:00), and ZT24 (07:00); blood was then collected, followed by dissection during which liver tissues were harvested and either flash-frozen in liquid nitrogen, embedded in OCT (Optimal Cutting Temperature) compound for cryosectioning, or fixed in 10% neutral buffered formalin for preservation.

All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee of Shenzhen University Medical School (IACUC-202400026, 19 December 2023).

Mice were fed one of two diets:

Standard SPF Rodent Maintenance Diet: This diet (SPF-F02-001, SiPeiFu (Beijing) Biotechnology, Beijing, China) provided 3.01 kcal/g. The percentage of calories derived from macronutrients was as follows: fat (12.1%), protein (21.6%), and carbohydrates (64.6%);

High-Fat Diet (HFD): A 60% high-fat diet (D12492, Research Diets, New Brunswick, NJ, USA) was used, providing 5.24 kcal/g. This diet comprised 60% fat, 20% protein, and 20% carbohydrate-derived calories.

After 8 weeks of feeding, three randomly selected mice from each group were randomly selected for rectal temperature measurements at 3 h intervals from ZT0 to ZT48. For each measurement time point, rectal temperature was measured three consecutive times using a mouse rectal thermometer (Jinuotai, Beijing, China). The mean of these three measurements for each mouse was calculated and used as the final value.

After 12 weeks of feeding, three randomly selected mice from each group were housed in behavioral monitoring cages for 7 days to record their activity–rest rhythms. Locomotor activity over the 7-day period was analyzed using Ethovision XT video tracking software (Nordes, Wageningen, The Netherlands). Spontaneous motor activity was defined as the distance traveled per unit of time (3 min).

After 16 weeks of feeding, exercise capacity was assessed in three randomly selected mice from each group using a treadmill fatigue test. The treadmill fatigue test was performed according to an established protocol [34]. Briefly, mice underwent a 3-day progressive training regimen on a treadmill set at a 10° incline with a 0.5 mA electric shock grid at the base to encourage running. Following training, the formal test commenced at 12 m/min. The treadmill speed was increased by 1 m/min every 5 min. Fatigue was operationally defined as remaining in the "fatigue zone" (the treadmill belt section within approximately one body length of the shock grid) for 5 consecutive seconds. Upon meeting this criterion, the mouse was immediately removed, and its running time, distance, and final speed were recorded.

After 20 weeks of feeding, we subjected mice in each group to a 16 h fasting period. At ZT12, mice received an intraperitoneal injection of a 20% glucose solution (2 g/kg body weight). Following disinfection, blood glucose levels from the tail vein were measured at 0, 15, 30, 60, 90, and 120 min post-injection using an ACCU-CHEK Performa glucometer.

Mouse whole blood was clotted at room temperature for 1 h, then centrifuged at 5000 rpm for 15 min at 4 °C to obtain serum. The serum supernatant was diluted fivefold with physiological saline for biochemical analysis of total protein (TP), albumin (ALB), total cholesterol (TC), total triglycerides (TG), aspartate transaminase (AST), alanine aminotransferase (ALT), Glucose (Glu), High Density Lipoprotein Cholesterol (HDL-c), Low-density Lipoprotein Cholesterol (LDL-c) using a blood biochemistry analyzer (Mindray BS-220, Shenzhen, China).

Frozen sections were prepared from the liver OCT embedded samples of three randomly selected mice per group, with a section thickness of 10 μm. Frozen sections of mouse liver tissue were circled along the periphery using a histological marker pen and equilibrated at room temperature for 5 min in a humidified chamber. Sections were then rinsed in 60% isopropanol for 20 s, stained with Oil Red O working solution for 10 min, and subsequently rinsed again in 60% isopropanol for 30 s. Following a distilled water wash, sections were counterstained with hematoxylin for 1 min, rinsed again in distilled water, and differentiated by a quick dip (1–2 s) in 1% acid-alcohol. Sections were then washed under slow running tap water for 30 min and finally mounted in glycerin gelatin for microscopic examination. Each mouse randomly selected one liver staining section for microscopic imaging. Three random fields (images) were captured from each section under the microscope for analysis. Images were captured at a magnification of 200× (with a 20× objective) for representative overviews and quantitative analysis. The quantitative analysis of the slices was accomplished using the ImageJ (1.54p) software.

Serum levels of insulin (JL11459, Jianglai, China), leptin (JL11317, Jianglai, China), and melatonin (JL10087, Jianglai, China) were quantified using commercially available ELISA kits according to the manufacturer's instructions. Mouse serum samples were diluted 5-fold prior to analysis. Leptin levels were determined using a sandwich ELISA format, while insulin and melatonin levels were measured using competitive ELISA formats.

At each of the five time points (ZT0, ZT6, ZT12, ZT18, and ZT24), liver tissues were harvested from three randomly selected mice in both the HFD and HFD + TRF groups for subsequent RNA sequencing. Liver RNA extraction, library preparation (using mRNA 221 enrichment), and paired-end sequencing on the BGIseq500 platform were performed as previously described [35]. Raw data were processed with SOAPnuke (v1.6.5) [36] to remove adapter sequences, reads containing >1% unknown bases, and low-quality reads (>40% 224 bases with Phred score ≤ 15).

All analyses were performed using GraphPad Prism 8.0.2 (GraphPad Software, San Diego, CA, USA) and R [37]. All data were shown as the mean ± standard deviation. Statistical differences among multiple groups were analyzed using one-way analysis of variance (ANOVA) or unpaired Student's t-test. p < 0.05 was considered to indicate statistical significance.

RNA-seq data analysis was performed as follows: clean reads were aligned to the reference genome and quantified using Rsubread [38] and featureCounts. Differential expression between HFD + TRF and HFD groups was identified with DESeq2 [39] (adjusted p ≤ 235 0.05). The most significantly upregulated and downregulated pathways refer to those 236 with the highest positive and negative Normalized Enrichment Score (NES), respectively, from the Gene Set Enrichment Analysis (GSEA) comparing the HFD + TRF group to the HFD group. The JTK_CYCLE algorithm [40] (period = 24 h, delta = 6 h, "independent" method) was used to identify the periodic characteristics of each group, as described previously [35]. Transcripts meeting the significance threshold (BH.Q < 0.05) were deemed rhythmic. Changes in rhythm parameters (mesor, amplitude, acrophase) between the HFD + TRF and HFD groups were evaluated with CircaCompare [41], where a significant difference (p < 0.05) in any parameter indicated rhythm alteration. Enrichment analyses were performed using GSEA (with NES ranking pathways) and GO term analysis via clusterProfiler [42].

Sixty 8-week-old male C57BL/6 mice were randomly assigned into four groups: ad libitum standard chow-fed (CON), time-restricted feeding with standard chow (CON + TRF), ad libitum high-fat diet-fed (HFD), and time-restricted feeding with high-fat diet (HFD + TRF). CON and HFD groups had unrestricted access to food, while CON + TRF and HFD + TRF groups were fed exclusively during a 3 h window from ZT12 (19:00) to ZT15 (22:00) (Figure 1A). All groups were maintained on their respective diets for 22 weeks. HFD mice exhibited continuous weight gain, showing a significant difference from CON mice starting at week 4 (p < 0.05), whereas CON + TRF and HFD + TRF mice maintained body weights at ~90% of CON mice (Figure 1B). This indicates that a 3 h feeding window effectively reduces body weight in high-fat diet-fed mice to levels comparable to standard chow-fed mice, independent of dietary fat content. Daily food intake analysis showed that CON + TRF mice consumed ~16% less food than CON mice, while HFD + TRF mice consumed ~31% less than HFD mice (Figure 1C). Caloric intake analysis revealed that CON + TRF mice consumed ~83% of the calories of CON mice, and HFD mice consumed ~147% of CON calories. Notably, the HFD + TRF group exhibited caloric intake comparable to the CON group (Figure 1E,F). These findings demonstrate that a 3 h feeding window significantly reduces body weight in both standard and high-fat diet-fed mice. Additionally, the similar caloric intake of HFD + TRF and CON mice suggests that time-restricted feeding enhances energy expenditure in high-fat diet-fed mice. Energy expenditure primarily includes basal metabolic maintenance, physical activity, growth and repair, and the thermic effect of food. We monitored the locomotor activity of mice over a one-week period. As nocturnal animals, mice exhibit higher activity during the dark phase and rest during the light phase. Compared to CON mice, HFD-fed mice showed disrupted diurnal rhythms, with a loss of night-time activity and day-time rest patterns, along with a significant reduction in total daily activity (Figure 1G,H), particularly in night-time activity (Figure 1H). A 3 h TRF regimen restored the disrupted circadian rhythm induced by HFD and significantly increased night-time activity, especially during ZT12-ZT21 (Figure 1I). In summary, 3 h TRF effectively prevented circadian rhythm disruption caused by prolonged HFD feeding and significantly enhanced daily activity, particularly during the night. During the treadmill fatigue test, it was found that a high-fat diet severely impaired the exercise capacity. Compared with the control group mice, the time required for the mice in the high-fat diet group to reach the limit exercise state was reduced by approximately 80% (high-fat diet group: 10 min; control group: 52 min; p < 0.001). It is notable that the 3 h intermittent fasting protocol significantly improved this defect, restoring the running endurance to a level similar to that of the control group mice (high-fat diet + intermittent fasting group: 42 min; compared with the high-fat diet group, p < 0.001; compared with the control group, p > 0.05). Similarly, through intermittent fasting intervention, both the total running distance and the maximum speed also returned to normal (Figure 1J). To further explore the energy regulation mechanism, we monitored the 48 h body temperature rhythm of each group of mice. The mice in the control group (CON group) exhibited a body temperature rhythm with lower levels during the day and higher levels at night. While the mice fed with a high-fat diet (HFD) maintained their body temperature at a relatively high level of 38 °C, with a reduced amplitude of daily temperature fluctuations. Notably, the 3 h intermittent fasting (TRF) protocol effectively restored the body temperature rhythm and average body temperature of the HFD-fed mice to levels close to those of normal mice (Figure 1K).

In our subsequent experiments, we monitored the metabolic rhythms of mice. Long-term high-fat diet (HFD) significantly increased the liver weights of mice and induced hepatic steatosis, a result that was confirmed by Oil Red O staining. In contrast, 3 h time-restricted feeding (TRF) significantly reversed both the HFD-induced increase in liver weight and lipid accumulation in the liver (Figure 2A and Figure S1A,B). Additionally, 3 h time-restricted feeding (TRF) reduced the serum levels of total cholesterol (TC), triglycerides (TG), and low-density lipoprotein cholesterol (LDL-c) in mice on a high-fat diet (HFD) to normal levels (Figure 2B). It also markedly lowered the aspartate aminotransferase (AST) levels, a key marker of liver function, back to baseline. Interestingly, regardless of the diet type (standard or high-fat), 3 h TRF significantly decreased the serum glucose content in mice (Figure 2B).

Next, we assessed glucose tolerance. Continuous 24 h blood glucose monitoring revealed that the high-fat diet (HFD) group exhibited persistently elevated random blood glucose levels. In contrast, 3 h time-restricted feeding (TRF) effectively reduced random blood glucose levels in both mice that were fed a normal diet and those on an HFD (Figure 2C). When compared to the normal diet group, HFD-fed mice showed significantly higher blood glucose levels and a delayed clearance rate following glucose injection at Zeitgeber Time 0 (ZT0), ZT12, and ZT24. This resulted in a markedly increased area under the curve (AUC) for glucose tolerance tests (Figure 2D,E), indicating that long-term consumption of an HFD significantly impaired glucose tolerance. Interestingly, HFD-fed mice subjected to 3 h TRF displayed normalized glucose tolerance at all three tested time points (ZT0, ZT12, and ZT24) (Figure 2D,E). This normalization suggests that 3 h TRF can counteract the detrimental effects of an HFD on glucose tolerance. Therefore, our findings indicate that 3 h TRF not only improves glucose tolerance but also helps maintain normal blood glucose levels in mice fed an HFD, highlighting its potential as a dietary intervention for managing glucose homeostasis.

We examined the secretion of metabolism-related hormones. In the CON group, melatonin secretion was low during the light phase and elevated during the dark phase, following a typical circadian rhythm. However, HFD-fed mice lost this rhythmicity, and 3 h TRF did not significantly restore the melatonin rhythm in HFD mice (Figure 2F). In comparison to the normal diet group, the HFD group showed significantly reduced leptin levels at ZT18, while leptin levels were elevated at ZT0/24. Time-restricted feeding improved the leptin secretion rhythm. Regarding insulin secretion, TRF significantly corrected the circadian rhythm of insulin, although secretion levels at all time points remained lower than those in the normal diet group.

To further investigate the effect of 3 h TRF on hepatic gene expression in HFD mice, we performed mRNA transcriptome sequencing on the livers of both HFD and HFD-TRF groups at ZT0, ZT6, ZT12, ZT18, and ZT24. Principal component analysis (PCA) of the three samples from each group at each time point revealed clear separation between the groups (Figure 3A,E,I,M,Q). Differentially expressed genes (DEGs) were identified using DESeq2 (p-adj < 0.05). At ZT0, TRF treatment significantly upregulated 400 genes and downregulated 355 genes (Figure 3B); at ZT6, 13 genes were upregulated and 41 downregulated (Figure 3F); at ZT12, 4 genes were upregulated and 5 downregulated (Figure 3J); at ZT18, 39 genes were upregulated and 19 downregulated (Figure 3N); and at ZT24, 56 genes were upregulated and 122 downregulated (Figure 3R). Gene Set Enrichment Analysis (GSEA) was used to detect significant pathway changes at each time point. The analysis showed that at ZT0, complement and coagulation cascades were significantly enriched and upregulated in the HFD-TRF group (Figure 3C), while fatty acid metabolism pathways were significantly downregulated (Figure 3D). At ZT6, PI3K-Akt signaling pathways were significantly enriched and upregulated in the HFD-TRF group (Figure 3G), while oxidative phosphorylation pathways were significantly downregulated (Figure 3H). At ZT12, complement and coagulation cascades were again enriched and upregulated in the HFD-TRF group (Figure 3K), while herpes simplex virus infection pathways were significantly downregulated (Figure 3L). At ZT18, IL-17 signaling pathways were significantly enriched and upregulated in the HFD-TRF group (Figure 3O), while valine, leucine, and isoleucine degradation pathways were significantly downregulated (Figure 3P). At ZT24, NOD-like receptor signaling pathways were significantly enriched and upregulated in the HFD-TRF group (Figure 3S), while tryptophan metabolism pathways were significantly downregulated (Figure 3T).

PCA of liver RNA-seq data from the HFD and HFD + TRF groups revealed a clear separation between the two groups (Figure 4A), indicating a profound and systemic reprogramming of the hepatic transcriptome by the 3 h TRF intervention. We then compared the rhythmic transcriptomes of liver tissues from HFD and HFD-TRF mice using JTK-cycle rhythmicity analysis, with genes having an adjusted p-value (p-adj) < 0.05 considered rhythmic. For shared rhythmic genes between the two groups, we used the CircaCompare method to compare median, amplitude, and phase differences. Genes with at least one p-value < 0.05 for median, amplitude, or phase were classified as having rhythmic changes, while genes with p-values ≥ 0.05 in all three parameters were considered unchanged. Analysis revealed that 3 h TRF dramatically increased the number of rhythmic genes in HFD-fed mice. The HFD group had 184 unique rhythmic genes (Group I), while the HFD-TRF group had 2688 unique rhythmic genes (Group II), representing a ~14.6-fold induction and underscoring the potent circadian-reinforcing effect of extreme TRF. There were 148 shared unchanged rhythmic genes (Group III) and 52 shared changed rhythmic genes (Group IV) (Figure 4A,B,D,F,H). Notably, genes in Group I were significantly enriched in pathways related to ribosome biogenesis and basic cellular "housekeeping" functions (e.g., mitochondrial and ER protein complexes) (Figure 4C), suggesting that under metabolic stress, a minimal set of core cellular processes retains circadian regulation. Group III genes showed similar rhythmic oscillations in both groups (Figure 4D), and were enriched in pathways related to ER-to-Golgi vesicle-mediated transport, Golgi vesicle transport, and ER chaperone complexes (Figure 4E). In contrast, the vast majority of genes that gained robust rhythmicity under TRF (Group II) were enriched in central metabolic pathways such as autophagy, lipid metabolism regulation, and mitochondrial gene expression (Figure 4G). This stark dichotomy strongly indicates that TRF specifically restores circadian timing to metabolic pathways critical for energy homeostasis, which are otherwise dysregulated by HFD. The small subset of genes whose rhythm parameters were altered (Group IV) were linked to ATP biosynthesis and ER stress response (Figure 4I), hinting at TRF's role in modulating energetic and proteostatic rhythms. We aimed to investigate how time restricted feeding induces rhythmic gene expression under metabolic stress conditions. Using the heatmap clustering with the Ward. D method applied to gene expression patterns, we identified that 2688 genes in Group II could be further subdivided into two distinct subgroups, designated II-1 and II-2 (Figure 4F). Subgroup II-1 (893 genes) transitioned from complete arrhythmia under HFD to acquiring de novo circadian rhythmicity after TRF (Figure 4J,K). These genes were selectively enriched in autophagy and fatty acid metabolism (Figure 4L), implying that TRF can act as a strong zeitgeber to "re-clock" and synchronize these key catabolic pathways. Conversely, subgroup II-2 (1795 genes) showed preexisting, low-amplitude oscillations under HFD that were significantly amplified by TRF (Figure 4M,N). These were linked to protein complex assembly and mitochondrial translation (Figure 4O), suggesting that TRF enhances the amplitude of nascent or dampened metabolic rhythms, thereby reinforcing temporal compartmentalization. Collectively, these data reveal that TRF employs at least two distinct transcriptional strategies—de novo induction and amplitude amplification—to achieve comprehensive intra-pathway synchronization, which likely underlies its efficacy in restoring metabolic efficiency.

To further investigate the underlying mechanisms, we performed an in-depth analysis of the key pathways enriched among TRF-induced rhythmic genes. Analysis of 45 genes significantly enriched in the autophagy pathway revealed that, compared to the arrhythmic expression patterns observed in the HFD group, most genes in the TRF group exhibited a peak at ZT6 and a trough at ZT18 (Figure 5A–C). Similarly, analysis of 35 genes enriched in the fatty acid metabolic process pathway showed that, unlike the arrhythmic expression patterns observed in the HFD group, most genes in the TRF group peaked at ZT12 and reached a trough at ZT18 (Figure 5D–F). Analysis of genes enriched in the proteasome-mediated ubiquitin-dependent protein catabolic process and regulation of protein containing complex assembly pathways revealed that, under long term 3 h of TRF treatment, expression amplitudes were increased and phase synchrony was enhanced, with peaks around ZT3 and troughs near ZT15 (Figure 5G–L). These findings suggest that time-restricted feeding promotes circadian synchronization of genes within key metabolic and proteostatic pathways. The increased amplitude and precise temporal alignment likely enable cells to perform specific functions more efficiently within defined time windows, minimizing unnecessary energy expenditure and enhancing overall metabolic economy under conditions of stress.

To further investigate these enriched pathways, we generated expression profiles of pathway-specific genes using the CircaCompare algorithm to assess circadian parameters such as phase, amplitude, and MESOR. Consistent with Figure 5A–C, genes in the autophagy pathway predominantly peaked at ZT6 and displayed pronounced rhythmic oscillations, including Anxa7, Atg5, Foxo3, and Ulk1 (Figure 6A). However, a subset of autophagy-related genes, such as Nampt, showed peak expression at ZT12, indicating some degree of phase heterogeneity within the pathway. In the fatty acid metabolism pathway, most genes reached peak expression at ZT12 with evident rhythmic oscillations, including Acsl4, Cyp7a1, and Decr2 (Figure 6B). A few genes, such as Elovl5, showed troughs at ZT12, whereas others, including Gstm7 and Slc27a5, showed peak expression near ZT6 (Figure 6B). Genes involved in the proteasome-mediated ubiquitin-dependent protein catabolic process generally reached their expression peak at ZT3 with significant rhythmic oscillations; examples include Araf, Btrc, and Derl1 (Figure 6C). Similarly, genes in the regulation of protein-containing complex assembly pathway also peaked at ZT3 and exhibited robust rhythmic oscillations, such as Add1, Camsap3, and Capn1 (Figure 6D). Given these observations, we speculate that the 3 h TRF regimen induces a core clock component to participate in regulating the coordinated oscillation of genes within the significantly enriched pathways, thereby enabling the liver to perform metabolic functions in an efficient, coordinated manner. Consequently, we analyzed the diurnal expression of core clock genes. Among them, genes such as Arntl, Clock, and Rorc showed no significant expression differences, whereas Rora, Nr1d2, and Nr1d1 exhibited more pronounced alterations (Figure 6E), suggesting that these core circadian genes may play important roles in the TRF-mediated improvement of liver function.

In summary, long-term 3 h TRF intervention in high-fat-fed mice not only markedly increased the number of rhythmically expressed hepatic genes but also, for the first time, induced a higher-order reorganization of the circadian transcriptome characterized by intra-pathway synchronization. This synchronization led to coherent peak expression of functionally related genes within defined temporal windows: autophagy-related genes predominantly at ZT6, fatty acid metabolic genes at ZT12, and proteostasis-related genes around ZT3. Notably, this temporal restructuring was accompanied by altered expression of specific core clock genes (e.g., Rora, Nr1d1), suggesting a potential molecular link between the feeding regimen and the observed transcriptional reprogramming. Collectively, these findings demonstrate that stringent TRF promotes a more robust and temporally precise circadian regulation of hepatic metabolism, which likely underlies the restoration of metabolic homeostasis disrupted by a high-fat diet.