Adiponectin modulates the diurnal hepatic transcriptome and energy metabolism in male mice

Abstract

Adiponectin is an adipocyte-derived hormone with insulin-sensitizing and lipid-lowering effects. Its expression and circulating levels show pronounced variation across the day, which opens the possibility that adiponectin influences metabolic programs in target tissues, such as the liver in a time-of-day dependent manner. To test this, we compared liver circadian transcriptome profiles (with sampling at 4-h intervals) between adiponectin-deficient (ADQ-KO) and wild-type (ADQ-WT) mice. Adiponectin loss led to tonic (i.e. time-independent) transcriptional changes in the liver with 1,393 differentially expressed genes (518 up- and 875 downregulated). These included upregulation of chromatin and RNA processing pathways and downregulation of immune and mitochondrial metabolic genes. At the same time, circadian analysis identified a marked reprogramming of transcriptome rhythms in ADQ-KO livers with changes in MESOR (n = 3,369 transcripts), amplitude (n = 386), and phase of gene expression (n = 603). Genes associated with mitochondrial respiration and fatty acid metabolism showed reduced rhythm amplitude and MESOR, whereas glycolytic genes exhibited increased MESOR. One of the identified adiponectin candidate targets and a regulator of hepatic metabolism, Hif1a, was further studied by functional assays in murine hepatocytes. Pharmacological adiponectin receptor activation promoted glycolysis and mitochondrial respiration under normoxia, but these effects were attenuated under hypoxia mimicry, consistent with HIF1a-dependent interference. These findings suggest adiponectin as a regulator of liver circadian metabolism, modulating both the timing and magnitude of energy-related gene expression programs, potentially in part through a HIF1a-mediated mechanism.

Introduction

A body-wide network of cellular circadian clocks – with a central pacemaker located in the hypothalamic suprachiasmatic nucleus (SCN) and subordinate oscillators in other brain and peripheral tissues (1) – coordinates the temporal adaptation of physiology and behavior across the 24-h day (2). Through a series of semi-redundant cues, such as behavioral signals (e.g. activity–rest cycle and food intake), neuronal signals, core body temperature rhythms, and endocrine rhythms, the SCN shares temporal information with peripheral-tissue clocks throughout the body, thus ensuring physiological coordination across tissues (3, 4).

Adiponectin is a peptide hormone synthesized by white adipose tissue that exerts a wide range of metabolic functions, such as glucose utilization and lipid breakdown (5). In mice, adiponectin transcription and secretion are circadian clock-regulated, peaking during the active phase (i.e. at night in mice) and aligning with higher metabolic demands around this time (6). Adiponectin signals via two membrane-bound receptors, AdipoR1 and AdipoR2, which are expressed in a tissue-specific manner (7). Its insulin-sensitizing effects depend on PPAR signaling (8). In the liver, adiponectin reduces glucose and lipid production (8, 9) and enhances fatty acid oxidation through changes in transcriptional regulation of associated metabolic pathways (10).

Both experimental and human studies have reported an inverse relationship between adiponectin serum levels and metabolic diseases (8, 11). In line with this, adiponectin has protective effects against metabolic liver diseases such as metabolic dysfunction-associated steatohepatitis (MASH) (9, 12, 13). Partially associated with such protective effects, the absence of adiponectin under MASH-promoting conditions leads to liver insulin resistance and mitochondrial dysfunction (e.g. altered morphology and reduced mitochondrial complex activity), representing an early susceptibility step in MASH development (14).

Only a handful of studies have investigated how adiponectin interacts with circadian rhythms. Under constant darkness conditions, adiponectin knockout (ADQ-KO) mice show elevated food intake. Moreover, they show lower hepatic triglyceride levels specifically in the late active phase, higher hepatic cholesterol in the late inactive phase, and tonically elevated serum glucose levels due to reduced insulin secretion. Consistent with this, insulin responses are notably impaired during the late active phase (15). In another study, increased food intake in ADQ-KO mice was associated with behavioral shifts and disrupted clock gene expression in the mediobasal hypothalamus (MBH), suggesting that adiponectin may act as a metabolic time signal to MBH clock and appetite rhythms. In line with this, rhythmic intracerebroventricular infusion of the adiponectin receptor agonist, AdipoRon, restored MBH clock gene expression and physiological feeding patterns in mice kept on a high-fat diet (HFD) (6).

Following these recent findings on adiponectin’s role in circadian regulation, this study aimed to characterize how adiponectin deficiency affects liver daily rhythms in mice. In the absence of adiponectin, we observed a substantial phase delay in the expression rhythm of genes associated with glucose and lipid metabolism. Furthermore, several genes associated with ATP and mitochondrial metabolism lost their 24-h expression rhythm in the absence of adiponectin. Bioinformatic analyses identified HIF1a as one candidate regulator of these dampened rhythms. Given HIF1a’s role in glycolytic and mitochondrial pathways, we hypothesized that HIF1a may modulate adiponectin signaling. Activating adiponectin signaling boosted glycolysis and maximal respiration in hepatocytes under normal oxygen conditions. However, its impact on glycolytic capacity was reduced under hypoxia-like conditions, suggesting that increased HIF1a limits adiponectin’s metabolic effects. Taken altogether, our findings provide a temporal perspective on how adiponectin regulates liver metabolism and highlight HIF1a as a possible mediator of these effects.

Materials and methods

Mouse model and experimental conditions

Three- to eight-month-old male ADQ-WT and ADQ-KO mice (B6; 129-Adipoq^tm1Chan/J; JAX strain 008195) on a C57BL/6J background were group-housed under a 12-h light:12-h darkness (LD, ∼300 lux in the light phase) cycle at 22 ± 2°C and a relative humidity of 60 ± 5% with access to standard chow and water ad libitum. Mice were culled by cervical dislocation at 4-h intervals, and tissues were kept on dry ice and stored at −80°C. Animal experiments were ethically in line with international guidelines for the ethical use of animals. Data are reported according to the ARRIVE guidelines. For each time point, a total of 3–4 mice were used.

Transcriptome analysis

Total RNA was extracted following the manufacturer’s instructions using TRIzol (Thermo Fisher, USA). Genome-wide expression analyses utilized Clariom S arrays (Thermo Fisher Scientific) with 100 ng RNA per sample, as per the manufacturer’s recommendations (WT Plus Kit, Thermo Fisher). Data analysis was conducted with Transcriptome Analyses Console (Thermo Fisher, version 4.0) with results expressed in log2 values. Microarray data from wild-type mice were obtained from a previous publication (16) (GSE199998). Gene expression data were processed by averaging repeated probes, and lowly expressed genes were filtered out using a log2 expression threshold of less than 3.8, corresponding to the 25th percentile. Only gene names associated with ENSEMBL codes were included in the analysis. After removing duplicates, a total of 14,456 genes were retained for further analysis.

Tonic DEG analysis

Principal component analysis (PCA) was performed using all samples and the factoextra package in R and the Hartigan–Wong, Lloyd, and Forgy–MacQueen algorithms (version 1.0.7). The analysis was conducted using Student’s t-test, with adjustments for multiple testing by applying a false discovery rate (FDR) correction (p_adj). Genes with a fold change of ± 1.5 (log2 FC = 0.58) and p_adj less than 0.01 were considered significantly differentially expressed.

Rhythm analysis

For rhythm analysis, CircaN (17), Metacycle (18), and JTK (19) methods were employed within the CircaN framework (17). In the knockout (KO) dataset, two samples (KO_ZT6d and KO_ZT18d) were recreated by averaging data from the same time point. This method was chosen as CircaN cannot run with an unbalanced number of samples. For MC and JTK, a P-value threshold of < 0.01 was applied. For CircaN, a combined P-value threshold of < 0.01 and correlation coefficient (r) > 0.4 was used as a cutoff. All rhythmic genes identified were merged and subsequently analyzed using CircaCompare (20). To directly compare the rhythm parameters MESOR and amplitude, CircaCompare fits were used irrespective of meeting rhythmicity thresholds. Phase comparisons were only performed when a gene was considered rhythmic in both genotypes (P < 0.05). Transcription factor target genes were sourced from the ChIPseq Atlas (21). Candidate HIF1a target genes were identified using ChIP-Atlas (https://chip-atlas.org/) by selecting genes with HIF1a binding sites located within ±10,000 bases of their transcription start sites in mice (Mus musculus, mm9 or mm10) and were further evaluated for rhythmicity using the CircaCompare dataset.

Enrichment analysis

Gene lists were analyzed using the ClusterProfiler package in the R environment, with the entire identified genes serving as the background dataset. For cases where identified biological processes exceeded 100 hits, the simplify function was applied with a cutoff value of 0.6 to streamline the results. In scenarios with fewer processes, only those involving more than three genes and exhibiting a P-value < 0.01 were considered for further analysis.

Phase set enrichment analysis

The PSEA method (22) was used to functionally classify the rhythmic transcriptome for each genotype. Mouse gene names were converted to their human equivalents using geneName (version 0.2.3). A minimum of 10 genes with a q-value of < 0.05 were selected against the background. Biological process IDs were derived from their names using version v2024.1 (https://www.gsea-msigdb.org/gsea/msigdb/human/collections.jsp#C5). Processes were filtered in REVIGO with a 0.7 reduction. The phase of biological processes was estimated by calculating the median peak expression of all genes enriched for that category. For each genotype, enriched processes were combined into a single list and manually assigned to higher hierarchical biological process classes. Then, the difference between the phases of similar processes was calculated and represented as a delta phase.

ChIP-X enrichment analysis – version 3 (ChEA3)

TF prediction was performed using the online application ChEA3 (https://maayanlab.cloud/chea3/) (23) on CRGs with reduced amplitude as input. Mean library ranks were used.

Western blotting

Frozen liver tissue samples were lysed in radioimmunoprecipitation assay (RIPA) buffer, followed by centrifugation to remove cell debris. The supernatant was used for western blotting. The supernatant samples were mixed at a 3:1 ratio with reducing gel loading buffer (ROTI Load 1, Carl Roth, Germany) and heated at 95°C for 5 min 20 μg protein samples were loaded on pre-cast gradient SDS page gels (SurePage 4–12%, GenScript, USA). Proteins were transferred to methanol-activated polyvinylidene difluoride (PVDF) membranes (Bio-Rad, Germany) and separated at 100 V for 1.5 h. Membranes were blocked for 1 h at RT with 5% w/v skimmed milk in PBS. Membranes were incubated with primary antibodies overnight at 4°C in 5% w/v skimmed milk and washed with PBS-T (0.1% v/v Tween 20 in PBS), followed by the secondary antibody for 1 h at RT. Primary antibodies against HIF1a protein (NB100-134; Novus Biologicals, USA) and beta-actin (#4970, Cell Signaling, The Netherlands) and a secondary antibody (polyclonal goat anti-rabbit immunoglobulins/HRP, Agilent Technologies, Germany) were used at 1:1,000 dilution for the primary and 1:4,000 dilution for the secondary antibody. After membranes were washed with PBS-T, bands were visualized by chemiluminescence (Western System, GE Healthcare Life Sciences, Germany). The signals were detected using the Azure imaging system (Biozym Scientific, Germany). The images were evaluated using ImageJ 1.52 k (24).

Seahorse experiment

AML-12 mouse hepatocytes, sourced from the ATCC Biobank (catalog no. CRL-2254) and transduced with Bmal1:Luc plasmid (25), were cultured in Gibco Dulbecco’s modified Eagle’s medium supplemented with nutrient mixture F12 (DMEM/F12, Thermo Fisher). The growth medium was enriched with 1% v/v penicillin–streptomycin, 1% v/v insulin–transferrin–selenium (ITS, Thermo Fisher), and 10% v/v non-heat-inactivated fetal bovine serum (FBS, Thermo Fisher). In addition, 10 nM dexamethasone (Sigma-Aldrich, Germany) was included to maintain the cells, which were incubated at 37°C in a humidified atmosphere with 5% CO₂.

Cellular metabolism, mitochondrial oxygen consumption rate (OCR), and extracellular acidification rate (ECAR) of AML-12 cells were determined on a Seahorse XF Pro 96 Analyzer (Agilent Technologies, Germany). Forty thousand cells were plated in each 96-well plate pre-coated with poly-D-lysine (Agilent Technologies, Germany). After 6 h of seeding, 20 μL CoCl₂ (diluted in cell culture medium, final concentration of 100 μM) or cell culture medium was added in the cell culture followed by a 12-h culture. Two microliters of AdipoRon (final concentration of 5 mM) or vehicle control (DMSO) were added to each well and incubated for 6 h. After cells were washed with Seahorse assay medium (XF DMEM, pH 7.4, supplemented with 10 mM glucose, 1 mM sodium pyruvate, and 2 mM L-glutamine), cells were incubated at 37°C in a non-CO₂ incubator for 1 h. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured at the basal condition, followed by sequential injections of oligomycin (final conc. 1.5 μM), carbonyl cyanide-p-trifluoro- methoxy-phenylhydrazone (FCCP; final conc. 2 μM), rotenone and oligomycin (final conc. 0.5 μM, each), and 2-deoxyglucose (2- DG; final conc. 100 mM) from ports A, B, C, and D, respectively.

Raw data were analyzed on Agilent Seahorse Analytics (Agilent Technologies), and OCR, ECAR, and proton efflux rate (PER) values were obtained. Values of basal respiration, OxPhos-dependent ATP production, maximal respiration, spare capacity, proton leak, and non-mitochondrial respiration were calculated according to the manufacturer’s instructions. Basal glycolysis was determined by subtracting PER values at the fifth measurement of the basal measurement from the lowest PER values after the 2-DG injection, and maximal glycolysis was determined by subtracting the highest PER values after the FCCP injection from the lowest PER values after the 2-DG injection. All chemicals used in this assay, which are not specified above, were purchased from Sigma-Aldrich.

Data handling and statistical analysis of bioinformatic experiments

Statistical analyses were conducted using R 4.0.3 (R Foundation for Statistical Computing, Austria) using packages described above or in GraphPad Prism (version 10, GraphPad Software, USA).

Results

Absence of adiponectin suppresses liver immune and mitochondrial respiration pathways independent of time

To evaluate the role of adiponectin in hepatic transcriptional control, male APQ-KO mice and controls (ADQ-WT) were kept under a 12 h light:12 h darkness cycle (LD) with free access to food and water. Mice were sacrificed, and liver samples were collected at 4-h intervals across the day and processed for transcriptome analysis (Fig. 1A). PCA of expression data across samples revealed a clear separation of genotypes (Fig. 1B). In an initial assessment of tonic, i.e. time-independent, effects of adiponectin on liver transcription, all sampling time points were pooled, and differentially expressed genes (DEGs) were determined. This analysis yielded 518 up- and 875 downregulated genes (P_adj < 0.01 and log2 fold change > 0.58) (Fig. 1C). Enrichment analysis of these tonic DEGs suggested an upregulation in processes related to nucleic acid modification, such as chromatin regulation, mRNA processing, and splicing. Downregulated DEGs were enriched for genes associated with the immune system (e.g. antigen processing, tumor necrosis factor, or interleukin 6 signaling) and metabolic functions (e.g. mitochondrial respiration; lipid and carbohydrate metabolism) (Fig. 1D; Table S1 (see section on Supplementary materials given at the end of this article)). Representative 24-h expression profiles of selected genes from these processes are shown in Fig. 1E.

Figure 1.

Assessment of the tonic effects of adiponectin reveals a strong impact on the liver transcriptome. (A) The experimental design is illustrated. (B) A PCA plot of the microarray data is shown. (C) Differentially expressed gene (DEG) analysis groups all samples into two time-independent categories. (D) Enrichment analyses are conducted on the DEGs identified in (C). (E) Selected processes identified in (D) are represented by key genes. n = 3–4 per time point and group. Icon shown in (A) was obtained from NIH BIOART.

Lack of adiponectin phase delays liver transcriptome rhythms

To account for temporal (phasic) effects in our dataset, sampling time was considered to characterize diurnal rhythm effects using specialized algorithms (CircaN, JTK_cycle, and MetaCycle, P < 0.01). We identified 4,157 and 4,457 significantly rhythmic genes in ADQ-WT and ADQ-KO livers, respectively, with 41% of them (n = 2,493) being significantly rhythmic across genotypes (Fig. 2A). The acrophase distribution of rhythmic genes was overall comparable between the genotypes, but robustly rhythmic genes (i.e. significantly rhythmic in both genotypes) showed a modest phase delay (0.79 h, P < 0.0001) in ADQ-KO livers (Fig. 2B). Previous work reported that the loss of adiponectin was associated with dampened clock gene expression rhythms in the MBH (6). Interestingly, diurnal expression rhythms of core clock genes in the liver remained largely unaffected by the loss of adiponectin. However, we identified phase delays for Bmal1 (0.86 h), Clock (0.89 h), Per1 (2.18 h), and Per2 (1.07 h) in ADQ-KO livers (Fig. 2C). Overall, clock gene acrophases were significantly delayed (0.82 h; P = 0.003) (Fig. 2D), very similar to the observations from robustly rhythmic (non-clock) genes (Fig. 2B). To identify biological processes influenced by transcriptional phase changes, we conducted phase set enrichment analysis (PSEA). Interestingly, genes showing phase advances were enriched for transcripts associated with amino acid and DNA regulation, while genes showing phase delays in ADQ-KO mice were enriched for metabolic processes (e.g. lipid, carbohydrate, xenobiotic, and protein metabolism), circadian clocks, immune and stress responses (Fig. 2E; Table S2).

Figure 2.

The absence of adiponectin phase delays core clock gene expression and several metabolic-related processes. (A) Venn diagram illustrating the number of rhythmic genes in each genotype. (B) The rose plot depicts the distribution of rhythmic genes based on their acrophase. The colors black, red, and purple represent ADQ-WT, ADQ-KO, and shared rhythmic genes between the genotypes, respectively. ****The overall phase difference of the robustly rhythmic genes calculated using a t-test against zero (P < 0.0001). (C) The diurnal profile of core clock genes is shown. The asterisk (*) indicates a tendency of phase effect (P < 0.1) detected by CircaCompare. (D) The overall phase difference of the core clock genes (including Per3) was calculated using a t-test against zero (P < 0.01). (E) The plot displays the phase difference between the groups of the identified biological process using the PSEA method. n = 3–4 per time point and group.

Lack of adiponectin dampens liver metabolic gene rhythms

To evaluate phasic effects of adiponectin deficiency on liver diurnal transcription, we used CircaCompare, which determines changes in Midline Estimating Statistic of Rhythm (MESOR), phase, and amplitude of temporal expression profiles. MESOR represents the mean expression level around which a rhythm oscillates, amplitude reflects the strength of oscillation, and phase represents the time of the highest expression within a cycle. Alterations in MESOR (n = 3,369), phase (n = 603), and amplitude (n = 386) were seen in the liver transcriptome of ADQ-KO mice (Fig. 3A; Table S2). Enrichment analysis of these genes yielded processes related to chromatin, histone, and mRNA regulation in genes with elevated MESOR in ADQ-KO livers. Conversely, genes with decreased MESOR were enriched for oxidative phosphorylation, mitochondrial organization, and fatty acid metabolism. Circadian-regulated genes (CRGs) with reduced amplitude were associated with mitochondrial morphogenesis as well as ATP and carbohydrate metabolism. Enrichment analysis of CRGs with increased amplitude revealed associations with lipid metabolism (isoprenoid and terpenoid metabolism) and steroid hormone regulation. Phase-advanced CRGs were enriched for DNA damage, cell cycle, and steroid metabolism-associated processes. In line with PSEA results (Fig. 2E), phase-delayed genes were associated with energy metabolic pathways (e.g. carbohydrate and triglyceride metabolism, gluconeogenesis, cholesterol, fatty acid, and glycerophospholipid metabolism), mitochondrial function (e.g. membrane organization, fusion, and complex activity), and the circadian rhythm (Fig. 3B; Table S2).

Figure 3.

Assessment of the phasic effects in the absence of adiponectin reveals a rewiring of the liver transcriptome with increased glycolytic signaling, decreased oxidative phosphorylation and enhanced Hif1a expression. (A) UpSet plots depicts the circadian-regulated genes (CRCs) identified by CircaCompare. (B) Enrichment processes from the different CRC classes identified in (A). (C) Prediction of transcription factors (TFs) using the CRCs with a reduced amplitude in the CheA3 algorithm. (D) Selected genes associated with oxidative phosphorylation (two first rows) and glycolysis (last row) are depicted. n = 3–4 per time point and group.

We further concentrated on CRGs with a reduction in amplitude and/or MESOR to explore the consequences of dampened rhythms in the liver correlating with alterations in energy metabolism transcripts (Fig. 3B; Table S2). We predicted possible transcription factors (TFs) associated with these CRGs using the CheA3 algorithm. Focusing on TFs that similarly showed reduced rhythm amplitudes, several TFs (e.g. Thap11, Hif1a, Atf2, Rara, Tcf12, and Ncoa1) were identified (Fig. 3C and D), most of which also showed increased MESOR. In ChEA3, transcription factors are ranked by mean library rank, where lower numerical scores reflect stronger enrichment of target genes, indicating higher predicted relevance. Among these was Hif1a, considering its function in energy metabolism regulation (26) and MASH development (27). In response to the observed elevated Hif1a expression in ADQ-KO livers, we anticipated a preference for glycolysis over oxidative phosphorylation in the absence of adiponectin (28). We, thus, examined transcripts linked to oxidative phosphorylation and glycolysis within our dataset, finding decreased MESOR for several genes related to oxidative phosphorylation and mitochondrial complex activity, while glycolysis-related genes exhibited increased MESOR in the livers of ADQ-KO mice (Fig. 3E).

Activation of adiponectin receptors increases glycolysis in murine hepatocytes specifically under normoxic conditions

Exploring HIF1a’s role as a candidate mediator of adiponectin’s effects in the liver, we analyzed the regulation of established HIF1a targets (from a public database (21)) in our dataset. Most HIF1a target genes exhibited MESOR changes in ADQ-KO mice (Table S2). Genes with increased MESOR were enriched in glycolytic and pyruvate processes (e.g. Pgk1, Tpi1, Ddit4, and Ldha), while genes related to glucose metabolism (e.g. Ppp1r3c, Pgam1, and Gpi1) showed a MESOR decrease (Fig. 4A). These findings led us to investigate HIF1a protein levels at two time points, i.e. in the late inactive (ZT10) and late active phase (ZT22) (Fig. 4B). However, between these two time points, we observed no differences in HIF1a levels in ADQ-WT livers. This suggests either a lack of rhythmicity at the protein level or a rhythm with phasing that was not captured by the coarse two-time point analysis. Given the limited temporal resolution of this assessment, additional sampling across the 24-h cycle would be required to more definitively evaluate potential circadian variation in hepatic HIF1a protein abundance. In contrast, HIF1a protein level was significantly higher at ZT 10 in ADQ-KO mice compared to ADQ-WT, with less of a difference at ZT 22 (Fig. 4C, genotype effect P = 0.0002).

Figure 4.

Activation of pharmacological adiponectin receptors enhances glycolysis and mitochondrial respiration under normal oxygen conditions, but such effects are reduced under hypoxia-like conditions. (A) Enrichment analyses of known HIF1a target genes identified in the diurnal analyses of ADQ-KO mice. (B) Representative western blot of HIF1a levels in the livers of ADQ-WT and ADQ-KO mice. (C) Quantification of HIF1a protein levels. n = 3–4 per time point and group. (D, E, F, G, H, I, J, K) Mitochondrial respiration assay of AdipoRon-treated murine hepatocytes under normoxic or hypoxia mimicry (treatment with CoCl₂) conditions. Data from at least two independent experiments are shown. In (D) and (E), means are shown as standard deviation (SD).

To investigate the functional impact of adiponectin on the regulation of liver oxidative phosphorylation and glycolysis, as indicated by the diurnal transcriptome data, we used murine hepatocytes. We performed a Seahorse metabolic assay to evaluate oxygen consumption rate (OCR) and proton efflux rate (PER) in both normoxia and under hypoxia mimicry, with the latter a condition known to stabilize HIF1a levels (29, 30) (Fig. 4D and E). We employed AdipoRon, a pharmacological agonist that binds to both adiponectin receptors (AdipoR1 and AdipoR2) with comparable affinity (31) to mimic the effects of adiponectin. A 6-h incubation with AdipoRon did not alter basal but enhanced maximal respiration in hepatocytes (Fig. 4F and G). In addition, both basal and maximal glycolysis were increased (Fig. 4I and J). To explore the functional interaction between HIF1a and adiponectin in cellular metabolism, we treated hepatocytes with AdipoRon under hypoxic mimicry conditions by using CoCl₂ for over 18 h, which is known to increase HIF1a levels (29, 30). This approach was considered more appropriate than knock-down/overexpression since HIF1a protein is rapidly degraded under normoxic conditions, which would render genetic strategies rather ineffective.

In CoCl₂-treated cells, basal respiration was not affected, but the maximal respiration was significantly decreased compared to that under normoxic conditions, which was independent of AdipoRon treatment (Fig. 4F and G). At the same time, basal glycolysis was significantly increased compared to that in controls kept under hypoxia mimicry conditions and further increased by AdipoRon (Fig. 4I). However, no additional increase in maximal glycolysis in response to AdipoRon under the hypoxic mimicry condition was observed (Fig. 4J). As expected, hypoxic mimicry conditions led to a higher basal and maximal glycolysis in comparison with oxidative phosphorylation (Fig. 4H, I, J, K).

In summary, in hepatocytes, AdipoRon enhances basal glycolysis and maximal respiration under normoxia. Under hypoxic mimicry conditions, AdipoRon’s action is reduced without further effect on maximal glycolysis after CoCl₂ treatment, suggesting a functional interaction between HIF1a and adiponectin in the regulation of hepatic energy metabolism.

Discussion

Our findings show that adiponectin deficiency disrupts liver transcription across the day in a tonic and phasic manner. This distinction highlights an additional layer of adiponectin regulation that has been overlooked. Tonic, time-independent effects reflect constitutive transcriptional changes likely resulting from the chronic absence of adiponectin receptor, whereas phasic, time-dependent effects represent alterations in the temporal organization of the hepatic transcriptome, mediated by circadian systemic cues such as body temperature, food intake, and/or hormonal signaling. Specifically, downregulation of immune function and mitochondrial respiration genes, along with the upregulation of chromatin remodeling and mRNA, qualify as tonic effects. At the phasic level, a modest phase delay in clock gene expression and in rhythmic genes shared across genotypes, reduced amplitude and MESOR of oxidative phosphorylation genes, and increased MESOR of glycolytic genes, together suggesting a shift in hepatic energy metabolism, were identified. In this regard, HIF1a emerged as a potential mediator of these phasic metabolic effects, supported by functional assays showing that adiponectin receptor activation promotes both glycolysis and mitochondrial respiration under normoxic conditions.

Adiponectin’s effects on the liver are catabolic, mainly reducing glucose and lipid biosynthesis (15, 32, 33) and increasing fatty acid oxidation (8, 9, 15). In ADQ-KO mice, reduced expression of Hnf4a as well as decreased binding of HNF4a to its target genes (e.g. Chrebp and Cyp2e1) was shown (10). In contrast, under our experimental conditions, we observed increased MESOR and gain of rhythmicity in Hnf4a mRNA expression in the livers of ADQ-KO mice (Table S2). Our tonic DEG analysis also identified a downregulation of genes involved in glucose and lipid biosynthesis, but upregulated DEGs were not enriched for lipid oxidation. Enrichment of (phasic) CRGs yielded a more complex picture. For instance, glucose-associated genes were enriched in CRGs with loss of amplitude. Conversely, processes associated with fatty acid oxidation and oxidative phosphorylation were enriched in genes showing a reduction in MESOR. Several pathways involved in glucose (e.g. gluconeogenesis) and lipid metabolism (e.g. catabolism, biosynthesis, and oxidation) were enriched in phase-delayed CRGs. Such phase effects on metabolic pathways of glucose and lipid metabolism were confirmed by PSEA. These phase effects must be considered in a context where ADQ-KO mice exhibit increased food intake, including during their resting phase (6). Therefore, one could suggest that part of the phase effects observed in the liver could be attributed to changes in daily feeding profiles.

Our findings highlight the benefits of performing circadian/diurnal profiling to account for time-dependent effects. Previous studies have shown that incorporating time into transcriptome analyses uncovers additional processes in MASH, which helps to explain the lack of concordance observed in MASH studies (34). It is important to note that the mice in our study were still young and fed a standard chow, with no overt (liver) disease symptoms. However, detailed circadian analysis may detect early-stage processes that might predict an increased disease risk at a rather early stage. In a similar approach with thyroid hormone receptor beta (THRB) knockout mice, previous work identified several molecular changes that mark these mice as more susceptible to metabolic disturbances (35). Previous studies have shown that adiponectin may affect liver circadian clock function. In a model of metabolic syndrome (KK/Ta mice), core clock gene expression profiles in both liver and skeletal muscle display significant phase advances. Transgenic expression of human adiponectin in these mice was able to restore these rhythms, and the restorative effects were more prominent in the liver compared to skeletal muscle (36). In another study, the absence of adiponectin in mice kept under DD conditions had a minor impact on the liver’s circadian clock machinery (15), which aligns with our observations. The effects of adiponectin deficiency are also tissue-dependent, as clock gene expression in the MBH, but not in the SCN, is disrupted in ADQ-KO mice kept in DD. Interestingly, clock disruption in the MBH can be rescued by rhythmic AdipoRon infusion (6). Aged ADQ-KO mice show increased amyloid-beta deposition in the hippocampus and reduced BMAL1 expression. Studies in neuronal lines indicate that adiponectin reverses the effects of exposing cells to amyloid-β protein (37).

Given its metabolic effects, adiponectin has a strong potential to be used as a modulator of fatty liver diseases. For instance, adiponectin levels are inversely correlated with parameters such as insulin resistance and body fat (38) and liver fibrosis (39). Experimental data further support a beneficial role of adiponectin in MASH (9, 12, 13, 40), including stimulation of mitochondrial activity and decreasing hepatic lipid accumulation (14). Corroborating these findings, both tonic (DEGs) and phasic targets (CRGs) identified in our study point to a downregulation of mitochondria-associated genes. We also identified a loss of amplitude in genes involved in mitochondrial organization. Following the marked phase-delay shift in energy metabolism, several processes associated with mitochondria (e.g. organization, fusion, and respiratory complex) were also enriched in phase-delayed genes.

An interaction between HIF1a and circadian rhythms has previously been suggested. For instance, in patients with obstructive sleep apnea (OSA), levels of clock proteins (e.g. CLOCK and PER) and HIF1a, measured in serum, were higher. Of note, these effects on HIF1a in humans were phasic, with differences observed primarily in the morning. Correlational analyses further indicate a possible interaction between clock genes/proteins and HIF1a (41). Using muscle-specific HIF1a knockout mice, the authors identified that HIF1a plays a crucial role in mediating time-of-day-specific metabolic adaptations in skeletal muscle, particularly by supporting glycolytic metabolism during the early rest phase (ZT3). In control mice, HIF1a drives stronger transcriptomic and glycolytic responses to exercise at ZT3 compared to the active phase (ZT15). These effects are abolished in HIF1a-KO mice, which instead show a metabolic shift toward oxidative pathways and altered glucose partitioning, while maintaining intact rhythmic clock gene expression (42). A recent study demonstrated that HIF1a accumulation in the liver is temporally regulated and dependent on BMAL1. Combined HIF1a and BMAL1 knockout under hypoxic stress leads to increased mortality (43).

Our transcriptomic and cell-based data suggest an interaction between adiponectin signaling, the circadian clock, and HIF1a action in hepatocytes. Importantly, the transcriptomic analyses are associative, identifying co-regulated gene signatures and increased expression of HIF1a target genes in ADQ-KO livers, but they do not establish directionality or causality. The cell-based experiments provide functional evidence that adiponectin receptor activation influences glycolytic and mitochondrial metabolism and that these effects are modulated under HIF1a-stabilizing conditions, suggesting a functional interaction rather than demonstrating a direct mechanism. This metabolic shift could serve as a compensatory response to reduced mitochondrial gene expression, consistent with a HIF1a-driven transcriptional program. The AMPK pathway is activated by adiponectin signaling, leading to the inhibition of ATP-consuming pathways, including fatty acid synthesis, cholesterol synthesis, and gluconeogenesis. Additionally, AMPK promotes glucose uptake and oxidation to boost cellular energy production (44). Our in vitro results align with this: the increase in glycolysis upon AdipoRon treatment may be due to enhanced glucose uptake. Although the primary metabolic effect of AdipoRon was an increase in glycolytic activity, it is noteworthy that the treatment also significantly elevated proton leakage under normoxic conditions. Ucp2 expression is significantly reduced in the liver of ADQ-KO mice (14), suggesting that adiponectin signaling may promote UCP2 expression. This could represent a potential mechanism underlying the increased proton leak observed in our study.

Taken together, our findings provide new insights into how adiponectin regulates the liver’s circadian rhythm. Adiponectin’s effects on liver transcription are both tonic and phasic, especially with regard to energy metabolism-associated genes. Corroborating previous studies, we show that adiponectin impacts mitochondrial function, suggesting a shift toward glycolysis in the absence of this hormone. HIF1a signaling emerged as one potential interactor between adiponectin signaling and mitochondrial gene signature alterations. At the same time, our transcription factor prediction identified additional regulators such as Thap11, Atf2, Rara, and Ncoa1, whose relevance warrants further investigation. Furthermore, the effects of adiponectin may not be limited to hepatocytes. It should also be noted that our in vitro assays of AdipoRon effects were not performed at different time points, which limits conclusions about potential time-of-day dependence. Together with the fact that our analyses were primarily male-focused, this highlights the need for future studies to address both circadian timing and possible sex-dependent differences in adiponectin action. Another potential limitation is the age range of the animals used, which may introduce age-related metabolic variability despite all mice having no overt liver pathology.

Declaration of interest

Henrik Oster is a Senior Editor of Endocrine Connections. Henrik Oster was not involved with the peer review of this manuscript, on which he is listed as an author. The remaining authors declare no competing interests.

Funding

This work was supported by the German Research Foundation (DFG) through grants HO 353-11/1, INST 392/167–1 FUGGGRK-1957, CRC/TR 296 LOCOTACT (ID 424957847, TP13), and TRR 418: Foundations of Circadian Medicine (ID 541063275, B03 and C05). LVM de Assis was supported by the Knut and Alice Wallenberg Foundation as a Wallenberg Molecular Medicine Fellow and by the German Research Foundation (DFG) under grant TRR 418 (ID 541063275, B04).

Ethical statement

Animal experiments were ethically approved by the Animal Health and Care Committee of the Government of Schleswig-Holstein and in line with international guidelines for the ethical use of animals.

Data availability

Microarray data were deposited in Gene Expression Omnibus under code GSE305385.

Artificial intelligence statement

The authors used Grammarly to improve readability and language and then reviewed and edited the content, taking full responsibility for the publication content.