Targeting the hepatic circadian clock concomitant with tyrosine kinase inhibition reverses late-stage hepatocellular carcinoma

SUMMARY

Hepatocellular carcinoma (HCC) is a leading cause of cancer-related deaths. Most patients present at advanced stages, and the effectiveness of tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors is constrained by limited patient response. A subset of HCC shows elevated expression of the promoter 2 (“P2”)-driven hepatocyte nuclear factor 4 alpha (HNF4α) isoform, which directly transcriptionally represses the circadian brain and muscle ARNT-like protein 1 (BMAL1) transcription factor. This subtype of HCC is robustly inhibited by the plant-based flavonoid nobiletin (NOB), a circadian-fortifying compound. Using patient-matched human HCC and serum, we show that BMAL1-deficient HCC shows exaggerated carnitine palmitoyl transferase expression and related metabolite abundance and that P2-HNF4α regulates the carnitine palmitoyl transferase I gene. Finally, using different preclinical models, we show that the anti-tumor activity of TKIs, when they are co-administered with NOB, is maximal. Our results suggest that chronotherapy and combination therapy might provide improved clinical outcomes for individuals with BMAL1-deficient HCC.

In brief

Fekry et al. demonstrate that BMAL1-deficient hepatocellular carcinoma (HCC) is highly sensitive to the clock-activating flavonoid nobiletin. When combined with tyrosine kinase inhibitors, nobiletin most effectively halts tumor progression, underscoring the promise of circadian-based chronotherapy to enhance treatment efficacy in HCC.

Graphical Abstract

INTRODUCTION

Hepatocellular carcinoma (HCC) is the most common liver cancer, and it is associated with a high mortality rate, largely because it is usually detected at an advanced stage. Surgical resection is typical for early disease, but late-stage HCC requires systemic therapy. For years, the tyrosine kinase inhibitor (TKI) sorafenib (SOR) was first-line therapy, but other broadly acting kinase inhibitors have also been utilized, including lenvatinib (LEN) and regorafenib.^1,2 More recently, immune checkpoint inhibitors have shown promise, although the overall 5-year survival of HCC remains around 18%.^3,4 Thus, there is a pressing need to reduce the HCC incidence and provide more effective therapy at late stages of disease. The success of anti-viral treatments has dramatically reduced virus-mediated HCC; however, HCC is on the rise in several populations, driven by the increasing incidence of metabolic dysfunction-associated fatty liver disease (MAFLD), a byproduct of the obesity epidemic and a risk factor for HCC.^5–7 Although mechanisms by which MAFLD increases HCC risk are not fully understood, circadian disruption alone has been demonstrated to be sufficient to induce ectopic fat deposition and, ultimately, HCC in mice.^8,9

HCC is associated with several somatic mutations. However, each mutation has a rather low level of occurrence, making them less than ideal drug targets. Prior studies have revealed that human and mouse HCC is highly heterogeneous in the expression of hepatocyte nuclear factor four alpha (HNF4α), with high-HNF4α-expressing tumors showing loss of the circadian transcription factor brain and muscle ARNT-like protein 1 (BMAL1).¹⁰ HNF4α is a liver-enriched transcription factor that plays important roles in hepatocyte cell fate, glycolysis, gluconeogenesis, cholesterol metabolism, and lipoprotein secretion, among others.^11,12 We and others have demonstrated that the primary isoform of HNF4α driven by the “P1” promoter (P1-HNF4α) expressed in healthy adult liver functions acts as a tumor suppressor for numerous genes involved in cell cycle regulation and tumor progression, often in a circadian manner.^10,13–16 Overexpression of this isoform can attenuate liver fibrosis and cirrhosis in mouse models of liver injury.¹⁷ Global knockout of HNF4α in mice results in embryonic lethality, while liver-specific loss causes death by 6 weeks of age.¹⁵ Though some human HCCs have lost HNF4α expression altogether,¹⁸ a subset (approximately 50%) of human HCCs express high levels of an alternate isoform of HNF4α, typically abundantly expressed in fetal liver and driven by an alternate “P2” promoter (P2-HNF4α).¹⁹ HCCs expressing high levels of P2-HNF4α have greatly reduced BMAL1 expression, a result of direct transcriptional repression by the P2 isoform.¹⁶ Based on studies demonstrating the importance of the circadian clock in preventing early-stage liver disease and its ultimate disruption in the context of transformation to HCC,⁸ we hypothesized that targeting the clock in HCC in combination with existing pharmacological treatment might be advantageous in slowing tumor growth and development, particularly in high-HNF4α-expressing HCC.

The small plant-based flavonoid nobiletin (NOB) targets the circadian oscillator, bolstering rhythms and protecting against metabolic syndrome.²⁰ NOB, an activator of the retinoic acid receptor-related orphan receptor (ROR), activates BMAL1 expression in a ROR-dependent manner.^20,21 BMAL1 plays context-dependent roles in cancer, though its expression is generally reduced in many cancers.²² For example, it is known to be outcompeted at genomic target sites (CACGTG) by the oncogene MYC in neuroblastoma,²³ while its expression promotes leukemia stem cell self-renewal, driving acute myeloid leukemia progression.²⁴ NOB exhibits protective effects in a variety of disease states, encompassing metabolism-enhancing, anti-inflammatory, and anticancer properties.^20,25,26 NOB has anti-tumor effects against several cancer types, including breast, colon, liver, and gastric cancers, by targeting multiple tumor-inducing pathways.^26–31 Based on prior data showing that the overexpression or activation of BMAL1 in P2-HNF4α-positive and BMAL1-deficient HCC causes apoptosis and prevents tumor growth, we tested whether NOB could reduce the progression of late-stage HCC. Furthermore, we tested the impact of combination therapy, using TKIs currently employed in clinical practice for the treatment of advanced-stage HCC. Finally, using preclinical models and human HCC samples with patient-matched serum, we studied metabolites and genetic markers associated with HCC expressing low and high levels of BMAL1 and P2-HNF4α, respectively, in order to characterize a subset of HCC that is most likely to respond to blockade by NOB and combination therapies and to identify potential surrogate markers in the serum that may identify this HCC subset.

RESULTS

The STAM mouse is a preclinical model recapitulating P2-HNF4α-positive human HCC

We and others have previously shown that the P2-HNF4α isoform has pro-proliferative effects.^{10,16,32–34} This isoform is robustly induced in approximately half of human HCCs, as well as being upregulated in the liver following prolonged high-fat diet (HFD) feeding.¹⁶ Here, we implemented the STAM mouse model, a preclinical model mimicking the progression of liver disease observed in human HCC.^32,34 Mouse pups were treated on day 2 with streptozotocin (STZ) or vehicle (VEH), followed by a switch from vivarium chow to an HFD at 4 weeks of age (Figure 1A). P2-HNF4α is progressively expressed in both female and male mice as liver disease progresses, with maximal expression noted at the HCC stage (Figures 1B, 1C, and S1B). As expected, a rise in P2-HNF4a expression coincided with the loss of Bmal1 at 16 weeks (Figure 1C). Cultured HCC cells from the STAM model also revealed high P2-HNF4α expression, concomitant with low levels of BMAL1 (Figures 1D and S1C).¹⁶

Figure 1. Induction of the fetal isoform of HNF4α in HCC lends responsiveness to the circadian clock activator nobiletin.

(A) A schematic illustrating the STAM model and its disease stages (non-alcoholic steatohepatitis, NASH; hepatocellular carcinoma, HCC; streptozotocin, STZ; high-fat diet, HFD). The liver phenotype (middle) across stages of liver disease is shown. Hematoxylin and eosin (H&E) staining of each disease stage is shown (scale bar: 100 μm).

(B) Representative immunohistochemical image of P1- and P2-HNF4α expression in the liver of STAM mice at different liver disease stages.

(C) RT-PCR reveals P1- and P2-Hnf4a levels 16 weeks post-STZ injection. *p < 0.01 and **p < 0.001, one-way ANOVA test, Dunnett’s multiple comparisons test (n = 5–12).

(D) Staining of cultured primary HCC cells from STAM mice using antibodies against BMAL1 and P2-HNF4α, followed by DAPI staining (scale bar: 50 μm).

(E) RT-PCR demonstrates Bmal1 expression following treatment with 10 μM nobiletin (NOB) in AML12, Hepa1c1c7, and HepG2 cell lines and primary HCC cells.

(F) Western blot showing BMAL1 after NOB treatment.

(G) BMAL1 expression in cells transfected with scrambled siRNA (Scrm) or RORα siRNA (siRORα) and treated with vehicle (Veh) or NOB. Data are expressed as the fold change relative to Scrm-Veh; mean ± SEM (n = 3). Statistical analysis was performed using one-way ANOVA with post hoc test; **p < 0.01 and ****p < 0.0001.

(H) RT-PCR analysis reveals expression of P2-HNF4a in response to a 10 μM NOB treatment in HepG2 and primary HCC cells. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.01, **p < 0.001, and ***p < 0.0001 (n = 6).

(I) MTT assay reveals proliferation of AML12, Hepa-1c1c7, HepG2, and mouse primary HCC cells following NOB treatment. *p < 0.01 and **p < 0.001, one-way ANOVA test, Dunnett’s multiple comparisons test (n = 5–12).

(J) Migration assay reveals number of migrated cells following NOB treatment of Hepa1c1c7 and HepG2 cells. Two-tailed unpaired t test, **p < 0.005 (n = 6).

NOB effectively targets P2-HNF4α-expressing HCC

Based on data showing the incompatibility of BMAL1 and P2-HNF4α expression in HCC, we examined the impact of the circadian-fortifying flavonoid NOB on normal (AML12) and cancerous liver cells, including primary mouse HCC. NOB induced BMAL1 mRNA expression, specifically in cancer cells initially exhibiting low BMAL1 expression, such as human HepG2 and primary mouse HCC (Figure 1E). Non-transformed AML12 cells also exhibited an increase in Bmal1 expression upon NOB treatment. However, this effect was not observed in cells that expressed Bmal1 robustly, such as in Hepa1c1c cells (Figure 1E). BMAL1 protein was also increased in HepG2 and AML12 lines treated with NOB (Figures 1F and S1D). To assess whether NOB-induced activation of BMAL1 is circadian-clock dependent, we performed RORα knockdown experiments (Figures 1G and S2A). NOB upregulation of BMAL1 was substantially reduced following RORα silencing by small interfering RNA (siRNA), indicating that BMAL1 induction by NOB is ROR dependent. Interestingly, NOB treatment suppressed P2-HNF4α expression in cancerous cells that initially lacked BMAL1 expression (Figure 1H). Testing the proliferation and migration of the corresponding cell lines, we observed that NOB inhibits cell proliferation and migration of BMAL1-deficient HCC while having no effect on HCC expressing high levels of BMAL1. NOB effects were negligible in normal liver cell lines^10,35,36 (Figures 1I and 1J). NOB reduced phosphorylated AKT(Ser473) in HepG2 cells, indicating suppression of the AKT signaling pathway (Figure S2B), which aligns with studies showing that NOB inhibits cancer progression by targeting AKT and downstream effectors.^37–39 We tested the effects of acute NOB treatment on cellular senescence using SA-β-galactosidase staining in HepG2 cells (Figure S2C). NOB treatment did not induce senescence, suggesting that its inhibitory effects are through alternative pathways, such as apoptosis in HCC.

NOB prevents HCC formation in the STAM model

Based on NOB’s inhibition of the proliferation and migration of BMAL1-deficient HCC, we tested its efficacy on tumor formation. 4-week-old mice were treated with VEH or 200 mg/kg NOB every other day for up to 16 weeks, with treatments administered at two different times, zeitgeber time (ZT)2 and ZT14 (Figure 2A). NOB significantly improved the phenotype of STAM mice, with their size and appearance closely resembling those of VEH-treated mice (Figure 2B) and all NOB-treated mice surviving the duration of the experiment (Figure 2C). Interestingly, NOB reduced rest-phase activity, measured via infrared-sensor-equipped cages (for actograms and period-grams, see Figure 2D). NOB significantly improved liver health, with treated livers showing a typical reddish brown color, and reduced both tumor incidence and tumors per liver (Figure 2E). Additionally, NOB lowered basal glucose levels significantly (Figure S1). Gene expression analysis from NOB-treated STAM and control mice revealed that while expression of P1-HNF4α was largely unchanged by treatment, NOB increased BMAL1 expression and reduced P2-HNF4α (Figure 2F). Furthermore, NOB reduced expression of the epithelial-mesenchymal transition (EMT)-related gene Snail (Snai1) while upregulating E-cadherin (Cdh1) and downregulating the expression of several proto-oncogenes, including the basic-helix-loop-helix (bHLH) transcription factor (Myc) proto-oncogene sarcoma Schmidt-Ruppin A-2 protein (Src), and genes associated with inflammation, including interleukin 6 (Il6). Additionally, it downregulated expression of proliferation-related genes, including platelet-derived growth factor receptors alpha and beta (Pdgfrα and Pdgfrβ, respectively). Expression of the HCC-related oncogene paternally expressed 10 (Peg10) and tumor suppressor phosphatase and tensin homolog (Pten) were upregulated by NOB treatment, while inhibitor of kappa B alpha (IκBα) expression was reduced, in line with previous findings that the anti-tumorigenic effects of NOB are partially nuclear factor-kappa B (NF-κB)-signaling dependent²⁶ (Figure 2F).

Figure 2. Nobiletin treatment blocks tumor formation in vivo and improves survival.

(A) Experimental timeline of nobiletin (NOB)- or vehicle (Veh)-treated STAM mice.

(B) Physical features of STAM mice treated with Veh or NOB.

(C) NOB effect on survival in STAM model. Kaplan-Meier analysis, p = 0.0036 (n = 7–12).

(D) Diurnal home cage activity, as measured by infrared sensors (top). Double-plotted actograms (middle) and periodograms (bottom) of Veh- or NOB-treated mice are shown.

(E) Livers of 16-week-old Veh- or STZ-treated STAM mice at zeitgeber time (ZT)2 or ZT4 (top) and percentage of tumor incidence and tumors per liver (bottom).

(F) RT-PCR reveals mRNA levels of genes involved in the circadian clock, EMT, proliferation, and HCC. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.03, **p < 0.005, ***p < 0.0005, and ****p < 0.0001 (n = 7–12).

Co-treatment with TKIs and NOB shows synergistic effect on proliferation and migration of HCC

The TKI SOR, though often utilized as a standard treatment for advanced HCC, can lead to chemoresistance.^40,41 Other TKIs are sometimes used in combination with SOR. Among them, LEN has shown efficacy in treating advanced HCC in cases of SOR resistance.^42,43 To determine whether the combination of TKIs and NOB could improve the HCC response, we treated HepG2, Hep3B, Huh7, and Snu499 cells with different concentrations of SOR and LEN in combination with 5 μm NOB, a concentration below the calculated IC50 for each cell line (Figure S2D), so that synergism or additive effects might be observed. MTT assays revealed a synergistic/additive effect when co-treating cells with NOB and low doses of both SOR and LEN, even in cell lines such as Hep3B and Huh7, which are reported to have SOR resistance.^44,45 Using the Bliss synergy model, we found that combining NOB with either SOR or LEN generally led to synergistic or additive effects, even at low doses. This suggests that NOB could enhance current treatment efficacy while possibly allowing for more tolerable doses of standard drugs (Figures S3A, S3B, and S4). Compared to individual treatments, we also observed a significantly enhanced anti-migratory effect in all tested cell lines when SOR or LEN was combined with NOB (Figures S3C and S3D). These findings underscore the potential for combination therapy by TKIs and NOB, even in HCC cells previously shown to harbor resistance to SOR.

Analysis of RNA-seq reveals distinct anti-proliferative and anti-tumorigenic gene signatures with NOB and TKI co-treatment

Though TKIs inhibit several kinases involved in tumor cell proliferation and angiogenesis, SOR inhibits rapidly accelerated fibrosarcoma (RAF), vascular endothelial growth factor receptor (VEGFR), and PDGFRs, among others.^41,46,47 SOR is still commonly used to treat inoperable HCC, despite its relatively modest effects.^48,49 To gain deeper insights into the biological mechanisms underlying the benefits of co-treatment on HCC, we subjected human HepG2 liver cancer cells to treatment with VEH, 5 μM of SOR, NOB, or a combination of SOR and NOB and performed RNA sequencing (RNA-seq). Using p < 0.01, we compared single vs. combination treatment. As revealed by heatmaps, numerous genes were differentially regulated across groups (Figure 3A). We identified 1,770 differentially expressed genes (DEGs) in response to SOR, with 998 downregulated and 772 upregulated in expression compared to the control (Figure 3B). Upon NOB treatment, 1,008 DEGs were downregulated and 604 upregulated. Compared to the control, combination therapy had the most substantial impact, with 4,376 DEGs (Figure 3B). Gene annotation and enrichment analysis revealed altered regulation of cancer-related pathways, including circadian entrainment, cell cycle regulation, transforming growth factor beta (TGF-β) signaling, VEGF signaling, WNT signaling, HCC pathways, signal transducers and activators of transcription (STAT)/IL-6 signaling, and NF-κB signaling (Figures 3C, 3D, S5A, and S5B). Venn diagrams reveal the number of significant DEGs across four key comparisons and the overlap between each set of genes (Figure S5C).

Figure 3. RNA-seq reveals distinct anti-proliferative and anti-tumorigenic gene signatures with NOB and TKI co-treatment.

(A) Heatmaps of RNA-seq data reveal gene expression of HepG2 cells following Veh (control [C]), NOB, SOR, or NOB + SOR treatment.

(B) Volcano plots reveal differentially expressed genes in HepG2 cells for each treatment group.

(C) Gene annotation reveals pathways downregulated after NOB + SOR treatment.

(D) Heatmap illustrating differential gene enrichment after NOB + SOR vs. control.

(E) RT-PCR analysis reveals gene expression changes associated with cancer-related processes, circadian regulation, cell proliferation, epithelial-mesenchymal transition (EMT), inflammation, and HCC signaling. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.03, **p < 0.005, ***p < 0.0005, and ****p < 0.0001 (n = 3).

Our data reveal significant synergistic regulation by NOB and SOR on HCC-related genes, including PEG10 and PTEN. PEG10, typically overexpressed in HCC, was effectively downregulated through the combination therapy. Interestingly, combination therapy improved PTEN expression, a tumor suppressor frequently inactivated in HCC.⁵⁰ As verified by quantitative RT-PCR (RT-qPCR), BMAL1 and P2-HNF4a displayed pronounced changes following combination therapy, with BMAL1 expression increasing and P2-Hnf4a decreasing (Figure 3E). Compared to the single treatment, genes related to tyrosine growth factor receptor (TGFR) signaling and cell proliferation, including TGFβ, PDGFRα, PDGFRβ, and Ki67, exhibited significant downregulation in the combination group, as did genes involved in IL6 and NF-κB signaling (Figure 3E). TP53 was significantly induced by SOR + NOB treatment, and genes involved in EMT, including beta-catenin (CTNNB1), CDH1, and cadherin 2 (CDH2), displayed pronounced changes, with positive regulators of EMT decreasing in expression and negative regulators of EMT increasing (Figure 3E). In summary, our data provide evidence that combined NOB and SOR treatment causes substantial amplification of anti-tumor responses compared to single-agent treatment.

Combined treatment with NOB and SOR inhibits HCC progression

We utilized the STAM model to evaluate the combined impact of SOR and NOB in vivo. In this model, HCC nodes are typically detectable at 11–12 weeks of age following the administration of STZ and HFD. To investigate the potential enhancement of SOR effects by NOB administration, we administered a low daily dose of SOR (10 mg/kg) alone, 200 mg/kg of NOB every other day, or a combination of the two for 4 weeks starting at 12 weeks of age (Figure 4A). Among male animals, the tumor number was significantly decreased following individual treatments with NOB and SOR, an effect more pronounced after combination therapy. In addition, the tumor incidence was decreased to less than 10% in males upon combination therapy and decreased to 0% in females (Figures 4B and 4C). Anti-glutamine synthetase (GS) and H&E staining confirmed the decrease in both tumor number and incidence provided by combination treatment (Figures 4D and S5D). Pathological assessment revealed reduced hepatocyte ballooning and microvesicular steatosis in the livers of STAM mice treated with SOR + NOB, as detailed in Table S1. RT-qPCR of livers from each treatment group revealed pronounced changes in Bmal1 and P2-Hnf4a, with combination therapy having the greatest effect on expression, with Bmal1 expression increasing and P2-Hnf4a decreasing (Figure 4E). Similarly, genes linked to TGF-β signaling and cell proliferation, such as Tgfβ, Pdgfrα, Pdgfrβ, and Ki67, exhibited marked downregulation after NOB + SOR treatment compared to control and single-agent treatment. IL6 signaling decreased significantly in combination treatment, and NF-κB signaling was downregulated, as demonstrated by IκBα enhancement. TP53 was induced after SOR + NOB, while genes related to EMT signaling, including Ctnnb1, Cdh1, and Cdh2, also changed, with EMT-promoting genes downregulated and EMT-inhibiting genes upregulated. Additionally, Peg10 and Pten were impacted by SOR + NOB treatment (Figure 4E). These findings suggest that the synergistic impact of NOB and SOR combination therapy holds significant promise as a therapeutic strategy for HCC, warranting further clinical investigation.

Figure 4. The combination of NOB and TKI inhibits HCC progression once already formed.

(A) Timeline of STAM mice treated with Veh (C) or combination therapy.

(B) Livers of STAM mice treated with vehicle (C), NOB, SOR, or combination NOB + SOR.

(C) Percentage of tumor incidence and tumors per liver after treatment.

(D) Hematoxylin and eosin (H&E) and glutamine synthetase (GS) immunohistochemistry of liver samples from STAM livers of each treatment group.

(E) RT-PCR analysis of gene expression for proliferation, circadian, inflammation, and epithelial-mesenchymal transition (EMT)-related genes in each treatment group. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.03, **p < 0.005, ***p < 0.0005, and ****p < 0.0001 (n = 8–10).

Unbiased metabolomics analysis reveals enhanced metabolic improvement in the STAM mice exposed to the co-treatment of SOR and NOB

The lack of reliable biomarkers and non-invasive diagnostic tools for HCC detection limits progress in clinical outcomes for HCC. In this study, we conducted targeted metabolomics on serum samples collected 16 weeks after STZ/VEH injection from control and STZ-treated mice, with (“C-HFD” and “HCC”) or without (“C-VEH” and “C-STZ”) HFD exposure. Our analysis revealed elevated abundance of specific metabolites within the “HCC” group only (mice exposed to a combination of HFD and STZ), in comparison to control mice (Table S2; Figures 5A–5C and S6A–S6C). Comparing control (C) and HCC groups, we noted significant changes in the abundance of cholesterol metabolites, essential amino acids, and acylcarnitines (Figures 5D–5F). In each case, metabolite levels were significantly higher in HCC compared to control with or without HFD or STZ treatment without HFD. To investigate the impact of NOB and SOR treatments on cholesterol, we measured total, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) cholesterol and triglycerides (TGs) in each experimental condition. When used separately, NOB and SOR each led to a reduction in total and LDL serum cholesterol in the STAM model, but combination treatment exhibited a synergistic effect on LDL and serum TG levels (Figure 5G).

Figure 5. Unbiased metabolomics of STAM mice reveals metabolites elevated in HCC.

(A) Heatmap of differentially expressed metabolites in serum of vehicle-treated mice (C-VEH), STZ-treated mice on chow (C-STZ), vehicle-treated mice on high-fat diet (C-HFD), and STZ-treated mice on HFD (HFD-STZ or HCC).

(B) Principal-component analysis (PCA) of biological replicates within treatment groups.

(C) Volcano plots reveal differences in serum metabolites between C-VEH and HCC mice.

(D–F) Metabolites with differential expression in the HCC group compared to control.

(G) Serum total, HDL, and LDL cholesterol and triglycerides in mice treated with C, nobiletin (NOB), or sorafenib (SOR) + NOB. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.01, **p < 0.001, ***p < 0.0001, and ****p < 0.0001 (n = 5).

Integrating transcriptomics and metabolomics reveals beneficial mechanisms of action of combination therapy

We performed signaling pathway network analysis using MetaboAnalyst (v.5.0), cross-referencing the gene list obtained from RNA-seq after NOB and SOR treatment with the list of metabolites overexpressed in the HCC group (Figure S6D). We then investigated gene expression for pathways highlighted in the analysis in STAM livers and in HepG2 cells in vitro. This analysis was carried out on cells treated with NOB, SOR, or a combination of NOB and SOR. Next, we verified the mRNA expression of particular genes correlated with the metabolites linked to HCC in the liver of the STAM model and in human HepG2 cells treated with NOB, SOR, or combination therapy. Pronounced gene expression changes were seen in single vs. combination therapy for several HCC survival genes, including tubulin beta 2B class Ilb (TUBB2B), cytochrome P450 family 27A1 (CYP27A1), steroid biosynthesis cytochrome P450 family 11A1 (CYP11A1), steroid sulfatase (STS), and regulation of TG and cholesterol synthesis angiopoietin-like 3 (ANGPTL3), with all genes showing maximal changes in expression after combination treatment, compared to control or single-agent administration (Figure 6A). Several studies have highlighted upregulation in ANGPTL3, an inhibitor of lipoprotein lipase (LPL), in several cancer types, including colorectal cancer, where it promotes liver metastases.⁵¹ One study revealed that NOB reduces ANGPTL3 levels in an LXR-dependent manner.⁵² Interestingly, co-administration of SOR and NOB further reduced ANGPTL3 mRNA expression in both the STAM model and human HepG2 cells, while SOR had negligible effects on its expression. TUBB genes belong to the tubulin beta class I and have been linked to the prognosis of various cancers. TUBB2B plays a role in HCC and is an independent prognostic factor for shorter overall survival in patients, promoting HCC progression in part by controlling cholesterol metabolism in an HNF4α/CYP27A1-dependent manner.⁵³ Low expression of CYP11A1, a gene involved in steroid biosynthesis, has been observed in HCC,⁵⁴ and its expression was reduced in the livers of STAM mice. However, treatment with NOB increased its expression, and combination therapy exaggerated this effect compared to single-agent treatment (Figure 6A). ANGPTL3 regulates circulating TGs and cholesterol levels by employing reversible inhibition of LPL and endothelial lipase enzymes.^55,56 Studies have demonstrated variations in ANGPTL3 expression in different cancer types. Like Cyp11a1, Angptl3 expression was greatly reduced by NOB + SOR in both STAM mice and in our human HepG2 cells treated with combination therapy.

Figure 6. Omics integration reveals potential mechanisms underlying the benefits of the NOB and SOR combination in HCC treatment.

(A–C) RT-PCR analysis showing gene expression associated with cholesterol metabolism (A), amino acids (B), and acylcarnitines (C) in STAM livers after sorafenib (SOR) and nobiletin (NOB) single or combination treatment, as well as in HepG2 cells.

(D) Chromatin immunoprecipitation (ChIP) experiments targeting P1-HNF4α and P2-HNF4α in HepG2 cells. ChIP-RT-PCR reveals direct binding of proteins to elements associated with acylcarnitine transport and lipid metabolism (CPT1A, CY27a1, and SLC7A6).

(E) Cpt1a mRNA expression in STAM liver at 4, 7, 10, and 16 weeks after vehicle or STZ treatment. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.01, **p < 0.001, ***p < 0.0001, and ****p < 0.0001 (n = 5–12).

Analyzing gene expression involved in amino acid metabolism, cathepsin K (CTSK), solute carrier family 7A6 (SLC7A6), solute carrier family 38A2 (SLC3842), and solute carrier family 6A14 (SLC6714) were also affected by combination vs. control or single-agent administration of NOB or SOR (Figure 6B). CTSK levels are elevated in individuals affected with breast, colon, prostate, and liver cancers.^57,58 HCC exhibits significant upregulation of CTSK expression. While its downregulation leads to the suppression of HCC cell proliferation, CTSK overexpression has the opposite effect, promoting HCC cell proliferation.⁵⁹ Here, SOR and NOB co-treatment led to a pronounced reduction in CTSK expression in livers with HCC compared to individual treatment (Figure 6B). SOR and NOB also significantly decreased SLC7A6 expression in STAM mice and human HepG2 cells. Upregulation of various amino acid transporters, including SLC7A6, results in a significant uptake of amino acids in Myc-dependent hepatocarcinogenesis,⁶⁰ and overexpression of SLC7A6 promotes HCC cell proliferation, migration, and glutamine uptake. Conversely, inhibition of SLC7A6 hinders HCC progression, in part by blocking amino acid uptake.⁶¹ Here, single NOB and SOR treatment led to a reduction in the expression of SLC7A6, which was more pronounced after combination therapy (Figure 6B).

HCC progression in STAM mice correlated with robust upregulation of carnitine palmitoyl transferase 1A (CPT1A), a gene associated with acylcarnitine metabolism, in HCC tumors and HepG2 cells. This upregulation was completely reversed by the combination therapy of NOB and SOR (Figure 6C). CPT1A provides an essential step in fatty acid oxidation (FAO). Past studies have revealed elevated expression of CPT1A in a range of malignancies.^62,63 While hepatic CPT1A deficiency can protect against HFD-induced liver damage and steatosis, cancer cells undergo metabolic reprogramming to support rapid growth and survival under unfavorable conditions.⁶⁴ Abnormal fat metabolism, mainly FAO mediated by CPT1A, plays a crucial role in tumor development and metastasis.⁶⁵ CPT1A inhibition can slow cancer spread and has emerged as a potential drug target in breast, prostate, glioblastoma, colon, gastric, multiple myeloma, and nasopharyngeal cancers, among others.^63,66–71 Our results reveal a substantial induction of CPT1A in the STAM livers, which was reduced by single treatment of NOB or SOR but more impressively by combination treatment (Figure 6C).

HNF4α exerts transcriptional regulation at the CPT1A locus.^72,73 To test whether HNF4α isoforms might be involved in treatment-dependent regulation of genes modulated by single or combination treatment, we designed primers surrounding DR1 sites in the promoter of several genes, including CPT1A, CYP27A1, and SLC7A6, previously identified as HNF4α targets^72,74 in HNF4α chromatin immunoprecipitation (ChIP)-seq experiments (published in ChIP-Atlas). ChIP-qPCR was performed in HepG2 cell lines using P1- or P2-HNF4α-specific antibodies to identify the direct binding of HNF4α to CPT1A, CYP27A1, and SLC7A6 genomic DR1 sites, emphasizing their regulatory relationship (Figures 6D and S7). In each case, P1- and P2-HNF4α isoforms bound to target D1 repeat sequences, though P2-HNF4α showed stronger binding at SLC7A6, in particular. In STAM livers, P2-HNF4α binding increased as liver disease progressed, suggesting a P2-HNF4α-dependent upregulation of CPT1A as the disease progresses (Figure 6E). The downregulation of P2-HNF4α by single and, more prominently, combination therapy (Figure 4B) suggests an HNF4α-dependent regulation of CPT1A, CYP27A1, and SLC7A6 expression by NOB and TKI inhibition in HCC.

Overexpression of P2-HNF4α leads to an elevation in metabolite levels associated with the formation of HCC

To determine whether specific metabolites might be associated with the induction of P2-HNF4α in liver disease and HCC, we performed untargeted metabolomics analysis in hepatocytes isolated from wild-type (WT) mouse liver (which express tumor-suppressive P1-HNF4α) and used lentivirus to drive expression of the P2-HNF4α isoform, followed by untargeted metabolite analysis of lysates from control and P2-HNF4α-overexpressing cells. Employing MetaboAnalyst (v.5.0), we generated a list of approximately 120 matched compounds, representing our significantly predicted metabolites (Tables S3 and S4). Heatmaps and volcano plots reveal differentially expressed metabolites, and subsequent pathway analysis highlighted key pathways affected by P2-HNF4α overexpression, including amino acid metabolism, steroid hormone biosynthesis, and one-carbon metabolism (Figures 7A–7C). Overexpression of P2-HNF4α in primary hepatocytes harvested from WT mice led to an increase in a number of the metabolites also found in the serum of STAM mice with HCC. Particularly, there was an induction of cholesterol and cholesterol sulfate following P2-HNF4α overexpression (Figure 7D). Remarkably, certain acylcarnitines associated with CPT1A expression significantly increased after P2-HNF4α overexpression, and amino acids such as valine and leucine showed large increases, resembling patterns observed in the STAM model (Figure 7D).

Figure 7. P2-HNF4α expression in human HCC correlates highly with acylcarnitine metabolism changes in liver and human serum.

(A) Heatmap of differentially expressed predicted metabolites in mouse primary hepatocytes with control vs. P2-HNF4α overexpression.

(B and C) Volcano plot (B) and pathway analysis (C) of metabolites varying in abundance in control vs. P2-HNF4α+ primary hepatocytes. Node sizes reflect suggested pathway impact, determined through topological analysis. Node color indicates significance, determined by enrichment analysis.

(D) Metabolites with differing abundance in P2-HNF4α+ vs. control primary hepatocytes.

(E) Pearson correlation analysis revealed positive correlation (R = 0.836) between Cpt1a and P2-Hnf4a expression in STAM HCC tumors (left) and negative correlation (R = −0.6717) with Bmal1 (right). Two-tailed t test. p = 0.0001 and 0.0061, respectively.

(F) Human serum CPT1A quantified using ELISA in normal and HCC-positive individuals.

(G) Human HCC staining using antibody to P2-HNF4α.

(H) Pearson correlation between human serum CPT1A in patients with HCC and expression of P2-HNF4α in the matching liver specimens (R = 0.678; p = 0.0005).

(I) Timeline of serum shock synchronization (left) and treatments with vehicle (C), nobiletin (NOB), sorafenib (SOR), or NOB + SOR, as well as harvest timelines for in vitro analysis of HepG2 cells. RT-qPCR analysis (right) shows expression of BMAL1 and DBP after treatment.

(J) Scheme for serum shock and single or combination treatments, followed by harvest every 6 h. Proliferation of HepG2 cells measured by MTT assay following treatments and rhythmicity of proliferative response following serum shock and treatment (right table) are shown.

(K) Tumor bioluminescence in HCC xenograft model treated with NOB + SOR at two zeitgeber times (ZTs) 2 weeks after tumor detection. Excised tumors (bottom) from vehicle (VEH) and treatment groups at both ZTs are shown. Quantification of bioluminescence over time is shown on the right. Two-way ANOVA, Sidak’s multiple comparisons test, *p < 0.01, **p < 0.001, ***p < 0.0001, and ****p < 0.0001; (n = 6).

To determine whether there was a correlation between P2-HNF4α and Cpt1a expression, we analyzed P2-HNF4α and Cpt1a levels in HCC from STAM mice. A strong positive correlation (R = 0.836) emerged between P2-Hnf4a and Cpt1a gene expression in HCC from STAM mice (Figure 7E, left). We also noted a negative correlation (R = −0.6717) between Cpt1a gene expression and Bmal1 (Figure 7E, right), consistent with the inverse correlation between Bmal1 and P2-Hnf4a previously demonstrated in inducible BMAL1 hepatic knockout mice.⁷⁵

To test the relationship of P2-HNF4α, CPT1A expression, and acylcarnitine metabolism in individuals with HCC, we conducted CPT1A measurements on serum from humans with or without HCC using an ELISA. The ELISA revealed substantial upregulation of CPT1A in the serum from patients with HCC compared to the serum from healthy individuals (Figure 7F). To determine whether patients with high levels of CPT1A in the serum also showed higher levels of P2-HNF4α in their HCC, we analyzed P2-HNF4α expression in matched liver tissue from patients with HCC (Figures 7G and S8A). Remarkably, a strong positive correlation existed between the serum levels of CPT1A and the expression of P2-HNF4α in the corresponding HCC liver tissue (Figure 7H). As previously reported, P2-HNF4α expression exhibited heterogeneity across the HCC samples, with some HCC being HNF4α deficient but those having HNF4α expressing the P2-HNF4α isoform (Figure 7H). Notably, while P2-HNF4α was positively correlated with serum CPT1A in patients with HCC, we saw a negative correlation of BMAL1 expression with CPT1A expression, consistent with the transcriptional downregulation of BMAL1 by P2-HNF4α.¹⁶

To further test the relationship of P2-HNF4α and CPT1A in the context of NOB and TKI response, we tested whether single and combination therapy could reduce cell viability using siRNA to P2-Hnf4a, Cpt1a, or both. Single and combination therapy provided maximal inhibition of cell viability compared to the control when both P2-HNF4α and CPT1A were present, though some (albeit reduced) inhibition was present when both genes were inhibited using siRNAs (Figure S9A). Thus, we predict that these genes function in the same pathway. Finally, to test whether loss of P2-HNF4α and CPT1A altered mitochondrial efficiency in tumor cells, we used siRNAs to knock down each gene alone or together and measured mitochondrial efficiency using Seahorse assays (Figures S9B–S9D). These results revealed that both single and double knockdowns of P2-HNF4α and CPT1A were sufficient to not only reduce the basal respiration capacity of HepG2 cells but also to greatly diminish ATP production and proton leak (Figure S9B and S9C). Interestingly, dual knockdown of P2-HNF4α and CPT1A did not augment the mitochondrial defects in HepG2 cells, suggesting that these two genes work in concert and within the same pathway to alter mitochondrial FAO.

Timed administration of NOB and SOR suggests that chronotherapeutics for HCC treatment could be advantageous for tumor inhibition

The utilization of biological clocks to fine-tune the timing and efficacy of chemotherapy treatments represents a poorly explored area in the context of HCC. This approach, commonly referred to as “chronotherapy,” takes into account the body’s intrinsic time-keeping mechanisms to limit toxicity and improve efficacy. To test whether single or combination therapy resulted in a time-dependent effect on rhythmic gene expression, HepG2 cells were synchronized using 50% horse serum and subjected to VEH, NOB, SOR, and NOB + SOR. These treatments were administered 12 h following synchronization, and cells were collected at 6 h intervals. Subsequent RT-PCR revealed a clear circadian induction and the oscillation of circadian genes, including BMAL1 and DBP (Figure 7I), highlighting NOB’s potential for modulating circadian rhythms at the molecular level. The circadian oscillation of CPT1A has been reported, with peak levels occurring around ZT17.⁷⁶ Data from the CircadiOmics portal, comparing liver transcriptomics of animals on a normal chow diet to those on an HFD, revealed significant circadian oscillations of Cpt1a in livers of HFD-fed animals (Figure S6E). Such circadian reprogramming has been previously observed in response to nutritional challenge, with some genes robustly oscillating under HFD conditions, as opposed to normal chow.⁷⁷ In HepG2 cells, we observed circadian oscillation of CPT1A in synchronized HCC cells, which was reduced following treatment with NOB and SOR (Figure 7I).

To investigate the potential benefits of timed drug administration on treatment outcomes, we conducted a series of in vitro experiments to test the effect of single or combination treatment on the proliferation of cancer cells. HepG2 cells were synchronized using 50% horse serum, and MTT cell proliferation assays were carried out following VEH, NOB, SOR, or NOB + SOR treatment at 6 h intervals, commencing at the 12 h mark after drug administration and continuing for a duration of 30 h. Our findings revealed that while all treatments exhibited significantly reduced cell viability at all of the time points, the combination therapy had the greatest effect. The circadian oscillation in gene expression and cell proliferation was further validated through the JTK_CYCLE rhythmicity test (Figures 7J and S10).

Finally, to test whether the combination treatment was more effective at blocking HCC progression at specific time points, we utilized NSG immunodeficient mice. NSG mice were subcutaneously injected with HepG2 cells on both flanks. After 15 days of unimpeded tumor growth, we initiated NOB + SOR treatment at two antiphase ZTs, ZT8 (late resting phase/daytime) and ZT20 (late active phase/nighttime). This regimen involved the administration of a low daily dose of SOR (10 mg/kg) in combination with 200 mg/kg of NOB every other day continuously for 2 weeks. A significant therapeutic effect of combination therapy was observed, with particularly notable outcomes for the ZT20 time point, as depicted in Figures 7K and S8B. This experiment underscores the influence of timing on treatment efficacy and sheds light on the potential advantages of synchronizing drug administration with circadian rhythms in the quest for improved therapeutic outcomes.

DISCUSSION

Though MASLD is a risk factor for HCC, the feasibility of diagnosing liver disease and HCC at early stages remains problematic. Currently, minimal hope remains for accurately diagnosing liver disease outside of obtaining reliable biomarkers that can be acquired from non-invasive clinical sampling. For example, a gut-microbiome-based metagenomic signature has recently shown promise in diagnosing advanced fibrosis in patients with MASLD,^78,79 and some serum biomarkers have been shown to associate with specific hepatic tumor types.⁸⁰ Still, further stratification of tumor types, their potential for progression, and tailored treatment are needed. Furthermore, characterizing factors that play roles in the chronic process of developing MASLD and the progression to HCC will aid in identifying new strategies for their diagnosis and therapy.

We and others have demonstrated that HNF4α plays key roles in liver diseases.^12,81–86 Importantly, we found that different HNF4α isoforms exert divergent functions on the pathological changes during the development of MASLD and HCC.^10,16,34 Though not all human HCCs are HNF4α positive, collectively, our studies have demonstrated that approximately half of all human HCCs robustly express the predominantly fetal isoform of the protein P2-HNF4α. Though P2-HNF4α is largely shut down in the adult liver, it can be induced in the context of alcoholic fatty liver disease and HFD feeding and under prolonged fasting.^12,16,81 Ultimately, the circadian signature for P2-HNF4α-positive tumors is the loss of BMAL1 expression to a large extent, as this set of HNF4α isoforms directly binds to and transcriptionally inhibits expression at the BMAL1 locus.¹⁶ This study is the first to show that specific metabolic pathways are elevated in P2-HNF4α-positive and BMAL1-deficient HCC and that this subtype responds uniquely to the circadian-fortifying flavonoid NOB, with HNF4α-deficient HCC showing minimal response.

In an effort to identify the metabolic features of P2-HNF4α-positive and BMAL1-deficient HCC, we used the STAM model as well as P2-HNF4α overexpression in primary hepatocytes to identify possible metabolite signatures of this HCC subtype. Interestingly, overexpression of P2-HNF4α alone was sufficient to robustly increase the abundance of several metabolites also found in our STAM model of HCC, including specific amino acids, acylcarnitines, and cholesterol metabolites. Elevated cholesterol levels have been previously implicated in the progression of MASLD to HCC, and there is growing evidence that the cholesterol pathway plays a critical role in the development of HCC.⁸⁷ Though not all studies show a correlation between serum cholesterol levels and the incidence and mortality of HCC,⁸⁸ cholesterol plays a vital role in both normal bodily functions and tumor survival.⁸⁹ While research on the role of cholesterol in tumors is advancing, the application of cholesterol-lowering drugs in HCC treatment remains limited.⁹⁰ Some studies have found that statins, commonly used to lower cholesterol, can effectively reduce the risk of HCC.^91,92 Interestingly, statins also have anti-proliferative effects on HepG2 cells and have been shown to reduce HCC growth in rats.⁹² Our study reveals that overexpression of P2-HNF4α alone is sufficient to greatly increase specific cholesterol metabolites, mirroring the elevated levels we found in the serum of STAM mice with HCC.

Though NOB has been previously shown to have circadian-clock-dependent benefits in the context of metabolic disease,^{20,21,25,28,52} we find a remarkable augmentation of its effects on hepatic metabolism with tyrosine kinase inhibition. Though NOB can be detected in serum 24 h after oral administration in rodents, its half-life is relatively short in the plasma (1.8 h).⁹³ The peak of NOB in plasma occurs at 1 h following a single 50 mg/kg oral dose, resulting in an approximately 2 μg/mL concentration. Though we expect that the circulating levels here were higher, based on doses used for in vivo studies, its short half-life suggests why it might be more effective when administered at one ZT over another. Indeed, both in vitro (Figure 7J) and in vivo (Figure 7K), NOB co-treatment appears to have a time-of-day effect on cancer growth. This may be due to the robust circadian expression of its primary target, RORα.⁹⁴ Combined with SOR, this approach has synergistic effects on the inhibition of MASLD and HCC as displayed by different mouse models. NOB treatment reduced rest-phase activity, and its effect on locomotion patterns could certainly contribute to the improved hepatic phenotype, as circadian disruption is correlated with ectopic lipid deposition and HCC.⁸ However, one of the metabolic pathways rescued by the combination treatment was acylcarnitine metabolism. Acylcarnitines play critical roles in hepatic lipid metabolism, serving as carriers of long-chain fatty acids into mitochondria for energy production. High acylcarnitine levels are correlated with poor prognosis of HCC.^95,96 Consistent with these studies, we find a sizable increase in medium- and long-chain acylcarnitine levels in STAM mice with HCC, and surprisingly, overexpression of P2-HNF4α in primary hepatocytes from WT mice is sufficient to significantly raise acylcarnitine levels.

Though previously linked to carcinogenesis, our study reveals a pronounced regulation of CPT1A by P2-HNF4α. CPT1A upregulation has been seen in other tumors, including triple-negative breast cancer (TNBC), where high levels are associated with poor survival, and its inhibition triggers apoptotic cell death.^63,97 Furthermore, inhibitors targeting mitochondrial FAO or CPT1 and CPT2 suppression result in decreased autophosphorylation of SRC, decreasing tumorigenic potential.^98–100 Our data reveal that combination treatment decreases hepatic CPT1A expression in both STAM mice and HCC, though single treatment also has a lowering effect. Based on published ChIP-seq data, we predict that the reduction of Cpt1a is primarily due to reduced P2-HNF4α expression, as HNF4α isoforms bind directly to the Cpt1a locus, with differential binding sites in cancer vs. healthy liver cells¹⁴ (Figure S7). BMAL1, on the other hand, appears to bind weakly at the Cpt1a locus.¹⁰¹ Though there may be synergistic effects of combination treatment on CPT1A expression, that the knockdown of P2-HNF4α, CPT1A, or both together alters mitochondrial energetics in HCC cells similarly suggests that these proteins work collinearly to regulate mitochondrial FAO and energetics. Intriguingly, our research suggests that serum CPT1A may serve as a potential indicator of P2-HNF4α expression in the liver, as both mouse and human serum analysis revealed a strong correlation with P2-HNF4α-positive HCC in the liver. This correlation holds promise for guiding treatment decisions in individuals with HCC.

Amino acids play diverse roles in tumor biology and its microenvironment, acting as biosynthetic precursors, energy sources, and regulators of redox balance. Amino acid derivatives also contribute to epigenetic regulation and immune responses linked to tumorigenesis and metastasis. Identifying potential metabolic vulnerabilities, such as transporters and transaminases, offers promising targets for therapeutic intervention,¹⁰² as essential amino acids can be used as diagnostic biomarkers for HCC through metabolic analysis.¹⁰³ Though our findings are consistent with this body of research, the extent to which combination therapy can lower the expression of amino acid transporters, such as Slc38a2, for example, has not previously been demonstrated. Thus, combination therapy shows great promise in reprogramming the metabolic landscape in the context of HCC.

The potential benefits of chronotherapy are 2-fold: it is an opportunity not only for improved treatment effectiveness but also to minimize adverse side effects commonly associated with chemotherapy. Pioneered by some in the cancer field,^104,105 this strategy holds great potential to transform cancer therapy approaches. Taking into account the internal biological rhythms of patients, but also understanding tumor rhythms and those of their microenvironment to optimize therapeutic impact of drugs, could have enormous benefits, especially in the challenging landscape of HCC treatment.^106,107 Some studies have revealed time-of-day differences in proliferation and DNA damage in HCC,¹⁰⁸ while others have revealed diurnal differences in the toxic effects of HCC irradiation treatment.^109,110 Our study suggests that HCC be stratified to tailor treatment choices and that responses to NOB and combination NOB and TKI therapy may be optimized for patients depending on the time of day of the administration. Though TKIs generally have a half-life of between 15 and 20 h, our data suggest that in combination with NOB, increased tumor blockade may be achieved by time-of-day-dependent administration. Not only did cell proliferation defects show a time-of-day effect, but suppression of HCC progression was maximal when combination therapy was applied at ZT20 in the xenograft model. Though testing in humans would be required to validate that chronotherapy may hold promise for advanced HCC, these data suggest that there may be significantly improved treatment methods that take into account the circadian clock of HCC and its host.

Limitations of the study

Although TKIs are maximally beneficial when co-administered with NOB, we did not test the potential benefits of NOB in combination with immune checkpoint therapy, which is most often used clinically in combination with TKI inhibition due to the high toxicity of TKIs. However, the fact that a lower dose of TKI was needed when concomitantly administered with NOB compared to single administration suggests that NOB could mitigate the toxicity associated with TKIs in patients with HCC. Another unknown is whether acylcarnitine and CPT1A activity levels reach a specific threshold in BMAL1-deficient HCC that is not reached at earlier stages of liver disease, such as MAFLD. Though our metabolomics datasets on the STAM mouse model suggest that acylcarnitine production is highest under conditions of HCC, this must be determined in patients with HCC.

RESOURCE AVAILABILITY

Lead contact

Further information, requests for resources, etc., should be directed to Kristin Eckel-Mahan (kristin.l.mahan@uth.tmc.edu). She is responsible for coordinating and fulfilling all requests.

Materials availability

No new unique reagents were generated in this study. All materials used are commercially available and listed in the key resources table.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
P1/P2-HNF4α	R&D Systems	Cat# PP-H1415-00; RRID:AB_2263954
P1-HNF4α	R&D Systems	Cat# PP-K9218-00; RRID:AB_2116902
P2-HNF4α	R&D Systems	Cat# PP-H6939-00; RRID:AB_2295209
BMAL1	Abcam	Cat# ab93806;RRID:AB_10675117
BMAL1	Novus Biologicals	Cat# NB100-129; RRID:AB_100025053
P84	GeneTex	Cat# GTX70220; RRID:AB_372637
ACTIN	Sigma aldrich	Cat# A5441; RRID:AB_476744
P-AKT(Ser473)	Cell Signaling Technology	Cat# 3787S (736E11); RRID:AB_915783
AKT	Cell Signaling Technology	Cat# 4691S (C67E7); RRID:AB_2921252
Anti-Mouse-HRP	BioRad	Cat# 1706516; RRID_2921252
Anti-Rabbit-HRP	BioRad	Cat# 1791019;RRID:AB_11125143
Biological samples
Mouse Serum	This paper	NA
Mouse Liver	This paper	NA
Biobank Human serum	This paper	NA
Biobank Human paraffinized sections	This paper	NA
Chemicals, peptides, and recombinant proteins
Nobiletin (NOB)	Selleckchem.com	#S2333
Sorafenib (SOR)	Sigma aldrich	N/A
Lenvatinib (LEN)	Sigma aldrich	SML3017
Streptozotocin (STZ)	Sigma aldrich	#S0130-1G
β-Galactosidase Staining Kit	Cell Signaling Technology	# 9860
SsoAdvanced Universal SYBR Green	BIORAD	Cat#4385618
iScript cDNA reverse transcription kit	BIORAD	Cat#1708891
Lipofectamine RNAi MAX	THERMOFISHER	Cat#13778075
penicillin/streptomycin Solusion	GIBCO	Cat#1514022
EMEM	ATCC	Cat#302003
DMEM	GIBCO	Cat#12491-015
MEM-alpha	GIBCO	Cat#12561-056
DMEM/F12	GIBCO	Cat#12634-010
Opti-MEM	GIBCO	Cat#31985-070
FBS	GENCLONE	Cat#25-514
Horse Serum	GIBCO	Cat#26050088
TRIzol reagent	LIFE TECHNOLOGIES	Cat#15596018
Protease inhibitors	ROCHE	Cat#11697498001
NAM (nicotinamide)	SIGMA ALDRICH ALDRICH	Cat#N0636
Trichostatin A	SIGMA ALDRICH ALDRICH	Cat# T8552
Dynabeads Protein G	INVITROGEN	Cat#10003D
DSG (disuccinimidyl glutarate)	THERMOFISHER	Cat#2053
CellTiter 96(MTT)	PROMEGA	Cat#G4000
Clarity Chemiluminescence Reagent	BIORAD	Cat#170-5061
Deposited data
RNA seq	Gene Expression Omnibus	GSE245588
Targeted Metabolomics	NMDR	https://doi.org/10.21228/M80R9W
Untargeted Metabolomics	NMDR	https://doi.org/10.21228/M83G11
Experimental models: Cell lines
Hep 3B	ATCC	Cat#HB8064
Hepa1c1c7	ATCC	Cat#CRL2026
Hep G2	ATCC	Cat#HB8065
Huh7	ATCC	Cat#HB8561
SNU449	ATCC	Cat#CRL2234
AML12	ATCC	Cat#CRL2254
Experimental models: Organisms/strains
NSG	Jackson lab	Cat#005557
C57BL/6J	Jackson lab	Cat#000664
Oligonucleotides
human KI67 _F- GAAAGAGTGGCAACCTGCCTTC human KI67_ R-GCACCAAGTTTTACTACATCTGCC	This paper	N/A
human ANGPTL3_F- CCTGAAACTCCAGAACACCCAG human ANGPTL3_R- TTCCACGGTCTGGAGAAGGTCT	This paper	N/A
human STS_F- GCTGGCAAAAGTCAACACGGAG human STS_R- GTCCGATGTGAAGTAGATGAGGG	This paper	N/A
human PRKCH_F- CATTTCGGCTCTCATGTTCCTCChuman PRKCH_R- GCACATTCCGAAGTCTGCCAGT	This paper	N/A
human CYP11A1_F- TGGCATCCTCTACAGACTCCTG human CYP11A1_R- CTTCAGGTTGCGTGCCATCTCA	This paper	N/A
human CYP27A1_F- GTGCTGCCTTTCTGGAAGCGAT human CYP27A1_R- TAGCCAGACACCTGGATGCCAT	This paper	N/A
human TUBB2B_F- GCACGATGGATTCGGTTAGGTC human TUBB2B_R- TCGGCTCCCTCTGTGTAGTGG	This paper	N/A
human PTEN_F- TGAGTTCCCTCAGCCGTTACCT human PTEN_R- GAGGTTTCCTCTGGTCCTGGTA	This paper	N/A
human PEG10_F- ACCACCAGGTAGATCCAACCGA human PEG10_R- TGTCAGCGTAGTGACCTCCTGT	This paper	N/A
human E-CADHER-IN _F- GTCAGTTCAGACTCCAGCCC human E-CADHER-IN _R- AAATTCACTCTGCCCAGGACG	This paper	N/A
human H-CADHER-IN _F- CCTCCAGAGTTTACTGCCATGAC human H-CADHER-IN _R- GTAGGATCTCCGCCACTGATTC	This paper	N/A
human SNAIL1_F- ACCACTATGCCGCGCTCTT human SNAIL1_R- GGTCGTAGGGCTGCTGGAA	This paper	N/A
human_ B-CATENIN _F- AAGACATCACTGAGCCTG CCAT human_B-CATENIN _R-CGATTTGCGGG CAAAGG GCAA	This paper	N/A
human TGF-B_F- TACCTGAACCCGTGTTGCTCTC human TGF-B_R- GTTGCTGAGGTATCGCCAGGAA	This paper	N/A
human IL6_F- AGACAGCCACTCACCTCTTCAG human IL6_R- TTCTGCCAGTGCCTCTTTGCTG	This paper	N/A
human CTSK_F- GAGGCTTCTCTTGGTGTCCATAC human CTSK_R- TTACTGCGGGAATGAGACAGGG	This paper	N/A
human SLC7A6_F- CAGGACACGTTCACTTACGCCA human SLC7A6_R- CTCAAAGGCGTCCTGAAAGTGC	This paper	N/A
human SLC38A2_F- AGCCAACAGCTCTTGTACCTGC human SLC38A2_R- GGAAGAACAGCAGGATGACAGAC	This paper	N/A
human SLC6A14_F- TGACCACTTCTGTGCTGGATGG human SLC6A14_R- AGCAAGCTCTCCACCATAGCCA	This paper	N/A
human SRC_F- CTGCTTTGGCGAGGTGTGGATG human SRC_R- CCACAGCATACAACTGCACCAG	This paper	N/A
human_IKBA_F- TCCACTCCATCCTGAAGGCTAC human_IKBA_R- CAAGGACACCAAAAGCTCCACG	This paper	N/A
human MYC_F- ACCACCAGCAGCGACTCT human MYC_R- GCTGTGAGGAGGTTTGCTGT	This paper	N/A
human P1/P2 HNF4A_F- ACCAAGAGGTCCATGGTGTTT human P1/P2 HNF4A_R- GTGCCGAGGGACGATGTAG	This paper	N/A
human P1HNF4A _F- CATGGATATGGCCGACTACAG human P1HNF4A _R- GCCCGAATGTCGCCATTGATCCCAGAGA	This paper	N/A
human P2HNF4A _F- CATGGTCAGTGTGAACGCGCCCCTCGG human P2HNF4A_R- GCCCGAATGTCGCCATTGATCCCAGAGA	This paper	N/A
human PDGF-R-A_F- GGTCTTGGAAGTGAGCAGT human PDGF-R-A_R- ACATCTGGGTCTGGCACATA	This paper	N/A
human PDGF-R-B_F- GTGCTCACCATCATCTCCCT human PDGF-R-B_R- ACTCAATCACCTTCCATCGG	This paper	N/A
human BMAL1 _F- GCAGTGCCACTGACTACCAAGA human BMAL1 _R- TCCTGGACATTGCATTGCAT	This paper	N/A
human CPT1A_F- GATCCTGGACAATACCTCGGAG human CPT1A_R- CTCCACAGCATCAAGAGACTGC	This paper	N/A
human P53 _F- CCTCAGCATCTTATCCGAGTGG human P53 _R- TGGATGGTGGTACAGTCAGAGC	This paper	N/A
human 18S _F- AAGTCCCTGCCCTTTGTACACA human 18S _R- GATCCGAGGGCCTCACTAAAC	This paper	N/A
Mouse Ki67_F- GCCTTGTGAAGCAGAAGAGAAAAC	This paper	N/A
Mouse Ki67_R- CTGGACCTCAAACACCTAACATCAG	This paper	N/A
Mouse Angptl3-F-_F- ACGAAAAGGGCTTTGGGAGGCT	This paper	N/A
Mouse Angptl3-F-_R- CGTAGTGCTTGCTGTCTTTCCAG	This paper	N/A
Mouse Sts_F- CCAACTGCTTCCTCATGGACGA	This paper	N/A
Mouse Sts_R- GTTCCTCCGCAGGAAGAGCGC	This paper	N/A
Mouse PR-kch_F- CTCTTCCAGCAGATTCGGCATC	This paper	N/A
Mouse PR-kch_R- AGCACCTTCACGGCGTACAGTT	This paper	N/A
Mouse Cyp11a1_F- TGCTCAACCTGCCTCCAGACTT	This paper	N/A
Mouse Cyp11a1_R- ACTGGCTGAAGTCTCGCTTCTG	This paper	N/A
Mouse Cyp27a1_F- TCAGGAGACCATCGGCACCTTT	This paper	N/A
Mouse Cyp27a1_R- CCAGTCACTTCCTTGTGCAAGG	This paper	N/A
Mouse Tubb2b_F- CCAGGCACAATGGACTCAGTCA	This paper	N/A
Mouse Tubb2b_R- TGCCCAGTTATTTCCTGCACCAC	This paper	N/A
Mouse Pten_F- AATTCCCAGTCAGAGGCGCTATGT	This paper	N/A
Mouse Pten_R- GATTGCAAGTTCCGCCACTGAA CA	This paper	N/A
Mouse Peg10_F- CCCCGGGCGCTGGTGTTG	This paper	N/A
Mouse Peg10_R- AGCGGGGCCGGGGAGTTTC	This paper	N/A
Mouse E-CadheR-in _F- CAAAGTGACGCTGAAGTCCA	This paper	N/A
Mouse E-CadheR-in _R- TGATGACACGGCATGAGAAT	This paper	N/A
Mouse H-CadheR-in _F- CCTCCAGAGTTTACTGCCATGAC	This paper	N/A
Mouse H-CadheR-in _R- CCACCACTGATTCTGTATGCCG	This paper	N/A
Mouse Snail1_F- CACCCTCATCTGGGACTCTC	This paper	N/A
Mouse Snail1_R- CTTCACATCCGAGTGGGTTT	This paper	N/A
Mouse B-Catenin _F- ACAAACTGTTTTGAAAATCCA	This paper	N/A
Mouse B-Catenin _R- CGAGTCATTGCATACTGTCC	This paper	N/A
Mouse TgFb_F- AGCCCGAAGCGGACTACTAT	This paper	N/A
Mouse TgFb_R- TTCCACATGTTGCTCCACAC	This paper	N/A
Mouse Il6_F- TACCACTTCACAAGTCGGAGGC	This paper	N/A
Mouse Il6_R- CTGCAAGTGCATCATCGTTGTTC	This paper	N/A
Mouse Ctsk_F- AGCAGAACGGAGGCATTGACTC	This paper	N/A
Mouse Ctsk_R- CCCTCTGCATTTAGCTGCCTTTG	This paper	N/A
Mouse Slc7a6_F- TCTACCTTCGCTGGAAAGAGCC	This paper	N/A
Mouse Slc7a6_R- GCCACCAGAAACAAGGAGCAGA	This paper	N/A
Mouse Slc38a2_F- GCGTTGGCATTCAATAGCACCG	This paper	N/A
Mouse Slc38a2_R- TCGTAGATGGGAAGAACAGCGG	This paper	N/A
Mouse Slc6a14_F- GCTTGCTGGTTTGTCATCACTCC	This paper	N/A
Mouse Slc6a14_R- TACACCAGCCAAGAGCAACTCC	This paper	N/A
Mouse Src_F- GTTGCTTCGGAGAGGTGTGGAT	This paper	N/A
Mouse Src_R- CACCAGTTTCTCGTGCCTCAGT	This paper	N/A
Mouse Iκbα_F- GCCAGGAATTGCTGAGGCACTT	This paper	N/A
Mouse Iκbα_R- GTCTGCGTCAAGACTGCTACAC	This paper	N/A
Mouse Myc_F- AACACAAACTTGAACAGCTAC	This paper	N/A
Mouse Myc_R- ATTTGAGGCAGTTTACATTATGG	This paper	N/A
Mouse P1/P2 HnF-4A_F- ACCAAGAGGTCCATGGTGTTT	This paper	N/A
Mouse P1/P2 HnF-4A _R- GTGCCGAGGGACGATGTAG	This paper	N/A
Mouse P1hnF-4A_F- CATGGATATGGCCGACTACAG	This paper	N/A
Mouse P1hnF-4A _R- GCCCGAATGTCGCCATTGATCCCAGAGA	This paper	N/A
Mouse P2hnF-4A _F- CATGGTCAGTGTGAACGCGCCCCTCGG	This paper	N/A
Mouse P2hnF-4A _R- GCCCGAATGTCGCCATTGATCCCAGAGA	This paper	N/A
Mouse Pdgfra _F- CAAACCCTGAGACCACAATG	This paper	N/A
Mouse Pdgfra _R- TCCCCCAACAGTAACCCAAG	This paper	N/A
Mouse Pdgfrb _F- TGCCTCAGCCAAATGTCACC	This paper	N/A
Mouse Pdgfrb _R- TGCTCACCACCTCGTATTCC	This paper	N/A
Mouse Bmal1 _F- GCAGTGCCACTGACTACCAAGA	This paper	N/A
Mouse Bmal1 _R- TCCTGGACATTGCATTGCAT	This paper	N/A
Mouse Cpt1a_F- GGCATAAACGCAGAGCATTCCTG	This paper	N/A
Mouse Cpt1a_R- CAGTGTCCATCCTCTGAGTAGC	This paper	N/A
Mouse P53 _F- TGAAACGCCGACCTATCCTTA	This paper	N/A
Mouse P53_R- GGCACAAACACGAACCTCAAA	This paper	N/A
Mouse 18s _F- CGCCGCTAGAGGTGAAATTC	This paper	N/A
Mouse 18s _R- CGAACCTCCGACTTTCGTTCT	This paper	N/A

Data and code availability

• RNA-seq data available through NCBI GEO under accession number GEO: GSE245588.

• Targeted metabolomics datasets are publicly accessible via the Metabolomics Workbench under https://doi.org/10.21228/M80R9W (study ID: PR002527).

• Untargeted metabolomics have been deposited in the Metabolomics Workbench under https://doi.org/10.21228/M83G11 (study ID: ST004042).

• No custom code was generated for this study.

• Other information required to reanalyze data can be obtained from the lead contact.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

The utilization of animals in this study was granted approval by the UT Health Institutional Animal Care and Use Committee. The mice were collectively housed in a pathogen-free environment and provided unrestricted access to chow (specifically, PicoLab Rodent Diet 5053). These animals were subjected to 12 h of light and 12 h of dark cycle for entrainment.

STAM mouse model and drug treatment

The NASH-HCC model, also known as the “STAM” mouse model, was established through a single subcutaneous injection of 200 μg of streptozotocin (STZ) from Sigma, administered two days after birth in C57BL/6J mice. When these mice reached the age of 4 weeks, they were transitioned to a high-fat diet (HFD) from Research Diet (Catalog #D12492). Both NOB and SOR treatment were administered via gavage, with a dosage of 200 mg/kg of NOB or Veh control (DMSO) given every other day. A dose of 10 mg/kg of body weight was used for sorafenib. The dosing Veh for sorafenib consisted of 12.5% Cremophor, 12.5% DMSO, and 75% sterile saline. SOR was initially dissolved in a mixture comprising 50% Cremophor and 50% DMSO. Sterile saline was added to achieve the final volume immediately before application. In the control group, which received only the Veh, animals were administered a similar fluid mixture without the inclusion of the drug. Cremophor EL was procured from Sigma (Sigma-Aldrich).

Xenograft model

HepG2 cells were genetically modified through lentivirus-mediated gene transfer to stably express the luciferase gene. Subsequently, 2 million of these cells were implanted subcutaneously into experimental animal subjects’ right and left flanks. The animals underwent periodic monitoring of luciferase expression, with the initial assessment conducted 6 h post-implantation, followed by weekly checks for four consecutive weeks. To enable bioluminescent tracking, a 100 μL solution of 50 mg/mL D-luciferin was intraperitoneally injected into the animals 15 min prior to each luciferase measurement. Tumor growth was tracked by quantifying luciferase activity using the IVIS Spectrum instrument from PerkinElmer, USA. After the completion of the four-week monitoring period, the resulting tumors were surgically excised for further analysis.

Circadian activity

Mice were housed in cages equipped with infrared sensors (STARR Life Sciences, PA) to monitor total activity. In experiments conducted under a 12 h light/12 h dark cycle, cages were outfitted with infrared sensors to assess the relative level of total home cage activity during the rest period. Data were acquired using VitalView software (STARR Life Sciences), and subsequent locomotion data analysis was carried out using ClockLab software (ActiMetrics, IL).

Human samples

Human samples were acquired in accordance with IRB protocol HSC-MS-12–0652, approved by the Institutional Review Board at the University of Texas Health Sciences Center. All patients provided informed consent, and were above 18 years of age. Patients with liver diseases were enrolled after diagnosis as candidates for liver or kidney transplants or other hepatobiliary surgeries. Tissue samples were obtained from patients without end-stage organ disease or those not on immunosuppressive medications, including those who required immunosuppression but didn’t need transplant therapy.

METHOD DETAILS

Cell culture

Cell lines were acquired from the American Tissue Culture Collection (ATCC) located in Manassas, VA. All cell lines were authenticated by the vendor and confirmed to be free of mycoplasma contamination prior to experimentation. Hepa1c1c7, HepG2, and Snu449 cells were cultured in Eagle’s minimal essential Medium, while AML12 cells were grown in a 1:1 combination of Dulbecco’s Modified Eagle’s Medium and Ham’s F12 medium. All cell culture media were supplemented with 10% fetal bovine serum, 100 units/ml of penicillin G sodium, and 100 μg/mL of streptomycin. High-grade NOB was procured from Selleck Chemicals, and Sorafenib sulfate was obtained from Sigma-Aldrich. Both compounds were prepared as stock solutions by dissolving them in dimethyl sulfoxide (DMSO).

MTT assay

Cell viability was evaluated using the MTT proliferation assay, which utilized the CellTiter 96 Non-Radioactive Cell Proliferation Assay Kit (Promega). The assay was conducted in accordance with the manufacturer’s provided instructions. The absorbance of the samples was measured at 570 nm using a Cytation 5 reader.

Cell transfection

HepG2 cells (5 × 10⁵) were seeded in 12-well plates and cultured until ~80% confluency. Cells were then transfected with 1μM siRNA (either targeting RORα¹¹¹ or scrambled control) using Lipofectamine RNAiMAX transfection reagent (Invitrogen, Cat# 13778–150), following the manufacturer’s protocol.

Transwell migration assays

Cells were initially seeded into the upper chamber of a Transwell insert, using serum-free medium, at a density of 50,000 cells per well. For conducting migration assays, a medium containing 20% FBS was placed in the lower chamber to serve as a chemoattractant. Subsequently, the cells were incubated for a duration of 48 h in a CO2 incubator. Following the incubation period, non-migrating cells located in the upper chamber were carefully removed by scraping. The migrated cells, which had moved to the lower chamber in response to the chemoattractant, were then fixed with methanol for a period of 10 min. After fixation, these cells were stained with a 0.1% crystal violet solution. To quantify the migrated cells, images were captured using a Cytation 5 imaging system, and for each sample, four randomly selected fields were chosen for analysis and quantification.

Senescence-associated β-galactosidase (SA-β-gal) activity assay

HepG2 cells were seeded at a density of 5 × 10⁴ cells per well in 12-well culture plates and treated with either vehicle (DMSO) or Nobiletin (10 μM) for 48 h. Senescence-associated β-gal activity was detected using the Senescence β-Galactosidase Staining Kit (Cell Signaling Technology, Cat# 9860) following the manufacturer’s protocol at pH 6.0. Brightfield images were captured using a Nikon microscope at 10× and 20× magnification. All experiments were performed in triplicate. The percentage of positive cells was calculated by counting the number of SA-beta-gal (positive cells/total number of cells) X 100.

Histology, immunohistochemistry, and immunofluorescence staining

Mouse liver samples collected at specific time points were fixed in 10% neutral buffered formalin for 48 h and subsequently embedded in paraffin. Tissue sections were processed by undergoing two washes with xylene for 10 min each, followed by rehydration through a series of ethanol gradients, including two rounds of 100% ethanol for 5 min each, 90% ethanol for 5 min, 80% ethanol for 5 min, and finally 70% ethanol for 5 min. These rehydrated five-micrometer-thick sections were then mounted onto glass slides. Subsequently, the sections were stained using two different methods: hematoxylin and eosin (H&E) staining, Glutamine Synthetase (GS), or P2-HNFα antibody. Sections were counterstained with hematoxylin, mounted on slides, covered with coverslips, and imaged using a Cytation 5 imaging system. Scoring details can be found in Table S1, Histopathological Scoring of Mouse Liver.

Cells were initially seeded on glass coverslips within a six-well plate, and they were subsequently fixed with 3.7% methanol-free formaldehyde for a duration of 10 min. Following fixation, the cells underwent permeabilization, which was achieved by treating them with 0.1% Triton X-100 for 5 min. To minimize nonspecific interactions, the cells were then subjected to blocking using a solution of 3% BSA dissolved in 1× PBS (pH 7.4) for a period of 1 h. After this blocking step, the cells were incubated overnight at 4°C with specific primary antibodies. Subsequently, fluorescent secondary antibodies were applied, and the samples were mounted on slides for imaging.

Serum triglyceride and cholesterol measurements

Serum triglyceride (TG) levels were analyzed using the Cayman kit (Catalog # 10010303), following the manufacturer’s instructions. In the assay, ten microliters of serum samples or standards were added to the reaction mix. Additionally, total serum cholesterol, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) levels were quantified using the BioVision kit (Catalog #k613–100). The serum samples were prepared according to standard laboratory protocols. For serum cholesterol and lipoprotein assessment, the procedure outlined in the kit’s instructions was meticulously followed to ensure precise and reliable results. This comprehensive approach allowed for the accurate evaluation of both serum triglycerides and cholesterol levels, along with the specific characterization of LDL and HDL fractions in the samples.

Serum CPT1A measurements

The measurement of human serum CPT1A (Carnitine Palmitoyl transferase 1A) was conducted using a human Carnitine Palmitoyl transferase 1A ELISA kit obtained from MyBiosource (Cat#MBS452146). The assay was carried out following the manufacturer’s provided instructions. To determine the concentrations of CPT1A in the serum samples, OD was measured at a wavelength of 450 nm. These OD values were then utilized to determine the concentrations of CPT1A by referencing them against a set of reference standards provided in the kit.

RNA extraction and RT-PCR

Total RNA extraction was carried out using TRIzol reagent following the manufacturer’s guidelines (Thermo Fisher Scientific). Subsequently, 1 μg of total RNA was employed for cDNA synthesis with the assistance of the iScript cDNA Synthesis Kit (Bio-Rad). The Advanced Universal SYBR Green Supermix from Bio-Rad was utilized with a Bio-Rad C1000 instrument to perform RT-PCR amplification. The PCR protocol entailed initial denaturation at 95°C for 30 s, followed by a series of cycles: 95°C for 10 s, 62°C for 30 s, and then 39 cycles at 65°C for 31 s and 65°C for 5 s. Expression levels of ribosomal subunit 18S were employed as a reference for gene expression normalization. The fold change in mRNA expression for each gene was determined using the 2−ΔΔC method. The expression levels were normalized to the control condition at ZT2 in experiments involving multiple time points. Human and mouse RT-PCR Primers are listed in key resources table.

Transcriptomic profiling (RNA-seq) and analysis

RNA extraction from the cells was performed using TRIzol reagent according to the manufacturer’s instructions. Subsequently, the extracted RNA samples were sent to Novogene for library preparation and sequencing. All raw data and processed read count have been submitted to Gene Expression Omnibus (GSE245588).

ChIP

Chromatin immunoprecipitation (ChIP) on HepG2 cells was performed through the following steps: Initially, cells cultured in a 15 cm dish were washed with PBS containing protease inhibitors (1% NP-40, 1 mM PMSF, 1 mM NaF, 400 mM NAM, and 3.3 mM TSA). Tissue samples were then minced in PBS with these inhibitors and subjected to cross-linking using DSG (final concentration of 2 mM) in PBS containing MgCl (final concentration of 10 mM) for 45 minutes at 25°C. Additional cross-linking was performed using formaldehyde (final concentration of 1%) at 25°C for 15 minutes. Subsequently, the samples were centrifuged at low speed and resuspended in PBS with protease inhibitors for homogenization. After homogenization, the samples were centrifuged at low speed once more and resuspended in 2 mL of ChIP sonication buffer (comprising 1% Triton X-100, 50 mM Tris [pH 8.1], 5 mM EDTA, 0.1% deoxycholate, and 150 mM NaCl) with protease inhibitors. Following sonication, the samples were centrifuged at 16,000×g using a microcentrifuge, and the resulting supernatants were diluted at a 1:4 ratio in ChIP dilution buffer (containing 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris [pH 8.1], and 167 mM NaCl). Immunoprecipitation was carried out overnight using P1 or P2 HNF4α antibodies or control mouse IgG. Protein: DNA complexes were captured using blocked protein G agarose beads, and following elution, cross-linking was reversed by heating at 65°C overnight. Protein digestion was performed using 0.17 μg/μL proteinase K (New England Biolabs), and DNA extraction was carried out with phenol-chloroform, followed by ethanol precipitation and dissolution in 100 μL Tris-EDTA buffer (comprising 10 mm Tris-Cl [pH 8.0] and 1 mm EDTA). The ChIP samples were further diluted in 120 μL ddH2O, and 6 mL of ChIP samples were subjected to RT-PCR analysis using gene-specific primers. ChIP Primers listet at key resources table.

Immunoblotting

Cultured cells were subjected to sonication using a Qsonica instrument located in Newton, USA, for a duration of 10 s at an amplitude setting of 30%, all while immersed in RIPA buffer. Following sonication, the samples underwent centrifugation at 10,000 × g for 10 min, and the protein concentration in the resulting supernatant was quantified using the bicinchoninic acid (BCA) assay method. The cell lysates were subsequently analyzed through electrophoresis on an 8% SDS polyacrylamide gel, and the separated proteins were later transferred onto nitrocellulose membranes, and targeted proteins were detected with specific antibodies. Antibodies are listed in key resources table. All uncropped blots are provided in Figure S11.

Targeted metabolomics

STAM model serum metabolomics was carried out by Metabolon. Serum samples were collected and prepared following the recommended procedures provided by the company. Metabolon performed data analysis, which encompassed metabolite identification, quantification, and statistical analysis to evaluate variations among sample groups. The study results were presented in terms of targeted metabolite concentrations and their relevance within the study’s context. Additional analysis was performed using MetaboAnalyst (version 5.0), a specialized platform designed specifically for the analysis of metabolomics data. Processed data from metabolomics listed in Table S2:Targeted metabolomics data. The data for this project are publicly available at Metabolomics Workbench under project https://doi.org/10.21228/M80R9W (Study ID ID: PR002527).

Untargeted metabolomics

Untargeted metabolomics analysis on mouse hepatocytes was performed at the University of Texas Health Science Center Clinical and Translational Proteomics Service Center. The samples were analyzed using an Agilent 6538 UHD Accurate-Mass Quadrupole Time-of-Flight (Q-TOF) mass spectrometer coupled with an Agilent 1200 series HPLC. The samples were diluted in a solution of 25% acetonitrile, 75% water and 0.1% formic acid before injection for LC/MS analysis. The LC separation was carried out using an Agilent ZORBAX Extend-C18 Rapid Resolution HT column (2.1 × 50 mm, 1.8 μm). The mobile phases were: Buffer A - 2% acetonitrile and 98% water with 0.1% formic acid, and Buffer B - 80% acetonitrile and 20% water with 0.1% formic acid. A gradient of 25%–90% Buffer B was applied over 20 min at a flow rate of 0.2 mL/min. The Q-TOF was operated in ESI positive mode using the following settings: capillary voltage at 3.5 kV, drying gas flow rate at 5 L/min, nebulizer at 5psi, and source temperature at 300°C. The MS1 spectra were acquired in the mass range of 100–2000 m/z.

The raw data were processed using Agilent MassHunter BioConfirm software (Version B.04.00), which uses the Maximum Entropy deconvolution algorithm for the accurate molecular mass calculation.

Significant metabolites identified in this analysis were used for predicted pathway analysis and differential expression analysis of metabolites. Subsequent analysis was conducted using MetaboAnalyst (version 5.0), a specialized platform designed for metabolomics data analysis. Metabolites and matched compound data from untargeted metabolomics metabolites are listed in Tables S3 and S4:Untargeted metabolomics data. This dataset is publicly available at the Metabolomics Workbench under https://doi.org/10.21228/M83G11 (Study ID: ST004042).

Seahorse analysis

HepG2 cells were transfected with siRNA targeting CPT1A, P2-HNF4A, or both, along with a scrambled control, using Lipofectamine RNAiMAX (Invitrogen, Cat# 13778150) following the manufacturer’s guidelines. Mitochondrial function assessments were conducted using the Agilent Seahorse XFe24 Analyzer. Twenty-four hours post-transfection, approximately 10,000 cells were seeded and cultured in XFe24 cell culture plates. The cell medium was substituted during the experiment with pre-warmed assay medium (Seahorse base medium supplemented with 1 mM pyruvate, 10 mM glucose, and 2 mM glutamine, pH = 7.4). The plate was then incubated at 37°C in a non-CO2 incubator for 1 h. Following incubation, the plate was moved to the Seahorse XFe24 Analyzer to monitor the cells’ Oxygen Consumption Rate (OCR) at various time intervals. Different compounds, including 5 μM oligomycin, 1 μM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP), and a mixture of 5 μM rotenone and 15 μM antimycin A, were introduced into the assay medium to evaluate cellular mitochondrial function. Key parameters of mitochondrial respiration, such as basal respiration, proton leak, and maximal respiration, were computed using the Seahorse Wave software. The OCR readings were normalized to the cell numbers in each well.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical method

Results were reported as mean values with standard error of the mean (SEM). For experiments involving two or more variables, statistical analysis was carried out using one-way or two-way analysis of variance (ANOVA) in conjunction with multiple comparisons test using Prism 9 software. The Kaplan-Meier, non-parametric statistical technique was employed to calculate the comparison of the survival based on lifetime data from STAM animals. Rhythmic gene expression was analyzed using the JTK_CYCLE algorithm implemented in the MetaCycle package in R, with significance defined as Benjamini–Hochberg–adjusted q < 0.05.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116313.

Highlights.

• Nobiletin prevents HCC by targeting the circadian clock

• Nobiletin and tyrosine kinase inhibition synergistically inhibit tumor progression

• P2-HNF4α regulates CPT1A, linking clock disruption to lipid metabolism

• HCC treatment can be optimized using chronotherapy