Aberrant miR-378 expression promotes hepatic lipid accumulation via hijacking the bile acid-regulated autophagy

In this study, we investigated a cohort of 608 subjects, including 414 subjects without hepatic steatosis and 194 subjects with hepatic steatosis, and observed that subjects with hepatic steatosis exhibited higher levels of serum BAs (). Importantly, logistic regression analysis adjusted for age revealed that higher levels of circulating BAs were associated with a higher risk of hepatic stea­tosis (odds ratio [OR]: 1.154 per µmol/L increase; 95% confidence interval [CI]: 1.054–1.265;= 0.002;; model 1). More importantly, after further adjusting for triglycerides (TGs) and fasting blood glucose, higher levels of serum BAs were also associated with a higher risk of hepatic steatosis (OR: 1.147 per µmol/L increase; 95% CI: 1.043–1.262;= 0.005;; model 2). These results suggest that higher levels of serum BAs are independently associated with a higher risk of hepatic steatosis. Fig. 1a Fig. 1b Fig. 1b P P

In agreement with previous studies [], systemic venous and hepatic BA levels were elevated in mice with hepatic steatosis induced by HFD feeding (). Further analysis demonstrated that autophagy was impaired in the fatty livers of HFD-fed mice as evident from an accumulation of p62 protein (;). Moreover, by taking advantage of ­autophagic flux reporter (AFR) mice with hepatic expression of mCherry-EGFP-LC3 reporter gene we developed recently ­(), we observed a reduction of autophagic flux in the steatotic livers of HFD-fed mice, as evident from a decrease in the number of red LC3 puncta (). Furthermore, we also observed a positive correlation between circulating BA levels and hepatic p62 levels, as quantified by immunostaining, in HFD-fed mice (), further supporting the notion that elevated BA levels are associated with impaired autophagy in the steatotic liver. ²⁰ Fig. 1c Fig. 1d and e Supplementary Fig. S1a–c Supplementary Fig. S1d Fig. 1f Supplementary Fig. S1e

It has been shown that activation of FXR can suppress hepatic autophagy [,]. Given that the majority of BAs are FXR agonists and one of the physiological functions of BAs in the liver is to inhibit their own synthesis through FXR [], based on the above data, we speculated that increased circulating BAs might lead to hepatic FXR activation, thereby suppressing autophagy in the liver. Thus, we then tested whether enhancement of autophagy by using ursodeoxycholic acid (UDCA), an FXR antagonist, could improve liver steatosis in HFD-fed mice. We found that UDCA treatment not only increased hepatic autophagic induction, as evident from the increased number of LC3 puncta () and elevated LC3 protein levels () but also reduced the liver weight and hepatic TG content in HFD-fed mice (), suggesting that enhancement of autophagy by FXR inhibition may contribute to the improvement of liver steatosis. Notably, under our experimental conditions, UDCA treatment did not alter the AMP-activated protein kinase (AMPK) signaling (). These results suggest that its pro-autophagic effects are primarily mediated via FXR antagonism, although other mechanisms cannot be excluded. Thus, we speculated that elevated BAs might serve as a causative factor for hepatic steatosis by activating FXR to promote lipid accumulation through the suppression of autophagy. ^{18 19 14} Fig. 1g Supplementary Fig. S1f Fig. 1h and i Supplementary Fig. S1g

Pharmacological activation of FXR has been reported to inhibit hepatic autophagy by repressing the transcription of autophagy-related genes [,]. In agreement with the notion that FXR is a negative regulator of autophagy, inhibition of autophagic flux by the treatment of either the synthetic agonist GW4064 or the endogenous agonist chenodeoxycholic acid (CDCA) of FXR could be observed in hepatocytes upon autophagic induction (;), while attenuated accumulation of LC3 puncta was observed after bafilomycin A1 (Baf-A1) treatment (). Consistently, phosphorylation of Unc-51-like autophagy activating kinase 1 (ULK1) was increased following GW4064 treatment in primary hepatocytes (PHs) (), indicating upregulation of mTORC1 signaling. However, GW4064 treatment could not alter the mRNA levels of autophagy-related genes in PHs, suggesting the involvement of alternative mechanisms (). ^{18 19} Fig. 2a Supplementary Fig. S2a Supplementary Fig. S2b Fig. 2b Supplementary Fig. S2c

Interestingly, an approach using unbiased metabolomics (NODE: OEZ00021653) revealed that FXR activation in PHs by GW4064 ­treatment elevated the levels of acetyl-CoA, a vital metabolic cue that inhibits autophagy by activating mTORC1 signaling [,] ­(). This finding was subsequently validated by a targeted fluorescence-based assay (). More interestingly, administration of the ATP-citrate lyase (ACLY) inhibitor, hydroxycitric acid (HCA), could reduce acetyl-CoA levels and restore autophagic flux (), suggesting that the increased acetyl-CoA levels contribute to the anti-autophagic effect of FXR. Given that FXR is a central metabolic regulator, these results support the notion that acetyl-CoA serves as a downstream effector of FXR in regulating autophagy. Collectively, the above findings indicate that BAs may suppress autophagy via acetyl-CoA−mTOR signaling pathway. ^{21 22} Supplementary Fig. S2d Fig. 2c Fig. 2d and e

To investigate the mechanism underlying FXR-induced acetyl-CoA elevation, we examined the expression of pyruvate dehydrogenase kinase isoenzyme 4 (PDK4), which inhibits the conversion of pyruvate to acetyl-CoA through the pyruvate dehydrogenase complex. However, we found that GW4064 treatment enhancedmRNA expression rather than reducing it, failing to justify the observed accumulation of acetyl-CoA (). We next assessed the expression ofand acyl-CoA synthetase short-chain family member 2 (), two additional enzymes in acetyl-CoA synthesis, and observed that GW4064 ­treatment had no significant impact on their mRNA levels ­(). Consistent with the established role of ACOX1, a key enzyme in peroxisome-mediated acetyl-CoA production, in the regulation of autophagy [], a subsequent transcriptional profiling employing RNA sequencing (RNA-seq) (NODE: OEP00006649) and validation by qPCR in GW4064-treated PHs collectively revealed that the repression of hepatic autophagy by FXR activation was accompanied by an increase inexpression (). The regulation of ACOX1 by GW4064 treatment was confirmed at the protein level using western blotting (). Similar results could be obtained when CDCA, an endogenous FXR agonist, was employed instead of GW4064 (). Importantly, our chromatin immunoprecipitation (ChIP) analysis revealed that FXR could be recruited to the promoter region of(), indicating that FXR regulates ACOX1 expression at the transcriptional level. Together, these data suggest that BAs and FXR can inhibit autophagy via a previously undescribed mechanism involving ACOX1-mediated acetyl-CoA production. PDK4 ACLY ACSS2 ACOX1 ACOX1 Supplementary Fig. S2e Supplementary Fig. S2e ²¹ Fig. 2f Fig. 2g Supplementary Fig. S2f and g Fig. 2h

We also assessed the ACOX1 expression and hepatic acetyl-CoA content in HFD-fed mice with elevated BA levels. In line with the data from HFD-fed mice, increases in the mRNA levels of known FXR target genes, ACOX1 mRNA and protein levels, and acetyl-CoA levels were all observed in the livers of HFD-fed mice (;). Accordingly, the mRNA levels ofandwere not elevated in HFD-fed mice (). On the other hand, UDCA treatment reduced the mRNA levels of established FXR target genes, both the mRNA and protein levels of ACOX1, and the levels of p62 and p-ULK1 proteins, as well as the levels of acetyl-CoA in the livers of HFD-fed mice (;). These findings suggest that UDCA treatment might enhance mTORC1-controlled autophagy by targeting the FXR-ACOX1-mediated acetyl-CoA production, leading to a decrease in TG accumulation in these mice. Collectively, our above data support the conclusion that elevated circulating BAs promote hepatic steatosis by inhibiting autophagy. Fig. 2i and j Supplementary Fig. S2h Supplementary Fig. S2h Fig. 2k and l Supplementary Fig. S2i ACLY ACSS2

As prolonged fasting (also referred to as long-term fasting) is known to induce physiological liver steatosis, we also tested whether circulating BA levels were also elevated and involved in the development of hepatic lipid accumulation after prolonged fasting. Interestingly, we observed that prolonged fasting induced hepatic lipid accumulation in mice concomitant with elevated systemic and hepatic BA levels (). Moreover, hepatic acetyl-CoA content,mRNA levels, and mRNA levels of documented FXR target genes were all increased following prolonged fasting (;). Accordingly, the mRNA levels ofandwere not elevated after prolonged fasting (). Collectively, we propose that BAs function as negative regulators of hepatic autophagy through FXR/ACOX1/acetyl-CoA axis to exacerbate hepatic steatosis during prolonged fasting. Fig. 2m and n Fig. 2o and p Supplementary Fig. S2j Supplementary Fig. S2j ACOX1 ACLY ACSS2

As both HFD feeding and prolonged fasting could increase hepatic lipid content, while concurrently upregulating hepatic miR-378 (;), a positive regulator of BA biosynthesis by targeting v-maf musculoaponeurotic fibrosarcoma oncogene homolog G (MAFG) [], which we identified previously [], we then tested whether miR-378 could affect hepatic lipid accumulation via BA-mediated autophagy suppression upon prolonged fasting. As expected, we found that overexpression of miR-378 in the livers of mice by adenoviral miR-378 (Ad-378) infection could elevate not only BA levels but also the expression of hepatic FXR target genes (;). In agreement with our notion that BAs act as negative regulator of hepatic autophagy, both immunofluorescence and immunohistochemistry analyses revealed a decrease in the number of LC3 puncta in the livers of mice infected with Ad-378 (). Consistently, decreased autophagic flux was observed in the livers of AFR mice infected with Ad-378 upon prolonged fasting, as evident from a decrease in the number of red LC3 puncta (). Fig. 3a Supplementary Fig. S3a ^{23–25 26} Fig. 3b and c Supplementary Fig. S3b and c Fig. 3d and e Fig. 3f

Accordingly, Ad-378 infection resulted in an increase in either the phosphorylation of ULK1 or the protein levels of p62 in the livers of mice under prolonged fasting conditions (). Meanwhile, increased ACOX1 mRNA and protein levels and acetyl-CoA levels were observed in the livers of these Ad-378-infected mice after prolonged fasting (). In line with these data, GW4064 treatment in PHs cultured in low-glucose medium supplemented with fatty acids, which mimics the nutritional status during prolonged fasting, led to increased expression of small ­heterodimer partner (), an FXR target gene, as well as elevated ACOX1 mRNA and protein levels (). As exacerbated hepatic steatosis was found in the livers of mice infected with Ad-378 (;), we speculated that hepatic miR-378 might act to promote lipid accumulation in the liver via a non-cell-autonomous mechanism involving BA/FXR/ACOX1/acetyl-CoA axis-mediated autophagy attenuation after prolonged fasting. Fig. 3g Fig. 3h and i Supplementary Fig. S3d and e Fig. 3j and k Supplementary Fig. S3f–h SHP

Surprisingly, overexpression of miR-378 increased rather than inhibited autophagic flux in cultured hepatocytes, while CDCA administration attenuated the pro-autophagic effect of miR-378 upon autophagic induction (). Based on thesedata and the abovedata, we speculated that hepatic miR-378 might regulate autophagy in the mouse liver through both cell-autonomous and non-cell-autonomous mechanisms. As expected, transfection of miR-378 mimics resulted in an increase in the number of LC3 puncta upon autophagic induction by Hank’s balanced salt solution (HBSS) administration in AML12 hepatocytes (). Moreover, increased accumulation of LC3 puncta was observed in AML12 cells transfected with miR-378 mimics after treatment of either Baf-A1 or chloroquine (CQ) (;). Similar results were obtained by examining LC3 protein levels and lipidation in HepG2 cells transfected with miR-378 mimics in the presence of Baf-A1 (). Consistently, attenuated LC3 lipidation was observed in HepG2 cells treated with miR-378 inhibitor in the presence of Baf-A1 ­(). Downregulation of p-ULK1 was observed after treatment of miR-378 mimics in PHs (). The effect of miR-378 on autophagic flux was further verified by using AML12 hepatocytes stably expressing mRFP-GFP-LC3 reporter gene or infected with adenoviral RFP-GFP-LC3 (Ad-GFP-RFP-LC3) or PHs isolated from AFR mice (;). Together, thesedata suggest that miR-378 is able to promote autophagic flux in hepatocytes in a cell-autonomous manner. Supplementary Fig. S4a Fig. 4a Fig. 4b Supplementary Fig. S4b Fig. 4c Supplementary Fig. S4c Supplementary Fig. S4d Fig. 4d Supplementary Fig. S4e and f in vitro in vivo in vitro

We further examined the interactions of LDs with autophagosomes and lysosomes by employing 4,4-difluoro-4-bora-3a, 4a-diaza-s-indacene (BODIPY) to image intracellular LDs. Transfection of miR-378 mimics reduced the number of LDs in HepG2 cells either with or without palmitic acid (PA) treatment upon autophagic induction (). Moreover, increased colocalization of LDs with LC3 puncta was observed in PA-treated AML12 cells after Ad-378 infection upon starvation (). Meanwhile, increased colocalization of LDs with lysosome-associated membrane protein-2 (LAMP2) could be observed by using either BODIPY or Lipidtox, another lipophilic dye for imaging LDs, in AML12 cells overexpressing miR-378 (;). Collectively, the above results suggest that hepatic miR-378 can reduce the intracellular lipid storage by promoting lysosome-mediated autophagic degradation of LDs in a cell-autonomous manner. Fig. 4e Fig. 4f Fig. 4g Supplementary Fig. S4g

Subsequently, we investigated the role of miR-378 in regulating hepatic autophagy under short-term fasting, when BA action may not be considerable. After short-term fasting, overexpression of miR-378 in the livers of mice by Ad-378 infection decreased p62 protein accumulation and mTORC1 signaling, as evident from decreased phosphorylation levels of ULK1 and the ribosomal protein S6 kinase (S6K) (). Accordingly, the number of LC3 puncta was increased after hepatic miR-378 overexpression in these mice (). Similar results could be obtained from immunohistochemi­stry analysis of LC3 (). Moreover, Ad-378 infection increased the number of red LC3 puncta in the livers of AFR mice upon short-term fasting, suggesting an increased autophagic flux after miR-378 overexpression (). Our finding that the pro-autophagic effect of hepatic miR-378 overexpression was accompanied by decreased hepatic TG levels after short-term fasting ­() suggested that autophagy enhanced by miR-378 overexpression might contribute to the reduced hepatic lipid accumulation observed. Since miR-378 overexpression did not alter hepatic acetyl-CoA levels, ACOX1 expression, or the mRNA levels of FXR target genes, and circulating BA levels also remained unchanged in short-term fasted mice (), we speculated that the inhibitory effect of miR-378 on hepatic autophagy might be outweighed by its pro-autophagic effect, due to the absence of activation in the FXR/ACOX1/acetyl-CoA axis during short-term fasting. Fig. 5a Fig. 5b Fig. 5c Fig. 5d Supplementary Fig. S5a Supplementary Fig. S5b–e

Given that the pathways and BA composition regulated by hepatic miR-378 and its target genewere extremely similar () [], we then hypothesized that MAFG is a primary target gene mediating the physiological effects of miR-378 in the liver. To test whether MAFG is involved in the regulation of autophagy by miR-378, we explored the effect of MAFG on hepatic autophagy. Interestingly, overexpression ofby adenoviral(Ad-) resulted in a decrease in red fluorescent protein (RFP) signal in AML12 cells co-infected with Ad-RFP-GFP-LC3, suggesting that MAFG is a negative regulator of autophagic flux in hepatocytes (). On the other hand, inhibition of MAFG by adenoviralshRNA (Ad-sh) led to an increase in RFP signal in AML12 hepatocytes expressing RFP-GFP-LC3 reporter gene (). Accordingly, Ad-infection increased the phosphorylation of ULK1, while infection with Ad-shdecreased it in cultured hepatocytes (;). Thus, we speculated that MAFG suppresses autophagy by enhancing mTORC1 signaling in hepatocytes. As overexpression ofcould attenuate the effect of Ad-378 on either autophagic flux or p-ULK1 levels in cultured hepatocytes (), we hypothesized that MAFG is able to mediate the cell-autonomous pro-autophagic effect of miR-378 in hepatocytes. MAFG MAFG MAFG MAFG MAFG MAFG MAFG MAFG MAFG Supplementary Figs. S3b and S6a ²⁶ Fig. 6a Fig. 6b Fig. 6c and d Supplementary Fig. S6b Fig. 6e and f

As the downregulated MAFG protein expression was consistently accompanied with hepatic miR-378 overexpression either due to Ad-378 infection under both short-term and long-term fasting conditions or upon HFD feeding (), we speculated that miR-378-mediated repression of MAFG might be metabolic context independent. Since manipulation of either MAFG expression or CDCA treatment was sufficient to modulate autophagy in PHs from mice under various conditions, including short-term fasting, long-term fasting, and HFD feeding ­(), we collectively speculated that the role of MAFG or FXR in regulating autophagy might be cell-autonomous and independent of the metabolic context. Supplementary Fig. S6c Supplementary Fig. S6d and e

Taken together, our results support the hypothesis that miR-378 in the liver exerts dichotomous effects on hepatic autophagy, with its net outcome being ultimately determined by the metabolic context (). Under short-term fasting conditions, where BA signaling is minimal and the FXR/ACOX1/acetyl-CoA axis remains inactive, the pro-autophagic effect of miR-378, which stems from its suppression of MAFG, dominates. Conversely, after prolonged fasting or HFD feeding, the downregulation of MAFG by miR-378 leads to substantial BA accumulation. This, in turn, activates the FXR/ACOX1/acetyl-CoA axis, which exerts a potent inhibitory effect on autophagy that outweighs the direct pro-autophagic effect of MAFG suppression (). Thus, our study unveils previously unrecognized physiological and pathological functions of BAs in regulating hepatic autophagy and establishes both hepatic miR-378 and circulating BAs as bona fide regulators of hepatic autophagy and lipid catabolism during fasting. Supplementary Fig. S6f Supplementary Fig. S6g

Finally, as autophagic zonation has been reported very recently [], we then tested whether miR-378 exhibits a zonated expression pattern, thereby being critically involved in the spatiotemporal regulation of hepatic autophagy under different nutritional status. We isolated the pericentral hepatocytes around central vein (CV) (CV hepatocytes) and periportal hepatocytes around PV (PV hepatocytes), respectively, by using flow cytometry, and confirmed the expression of marker genes (). Interestingly, after short-term (ST) fasting, miR-378 was expressed mainly in pericentral hepatocytes, accompanied by a zonated autophagy with a similar bias, as reflected by a pericentral enrichment of red LC3 puncta (ST-CV versus ST-PV) (). These results suggest that the zonal pattern of hepatic autophagy after short-term fasting might be attributed, at least in part, to the zonated expression of miR-378 and its cell-autonomous pro-autophagic action. ²⁷ Supplementary Fig. S7a–c and e Fig. 7a and b

Autophagy was further enhanced in pericentral hepatocytes after long-term (LT) fasting, accompanied by further elevation of miR-378 levels (ST-CV versus LT-CV) (). As the levels of pericentral ACOX1 were increased after long-term fasting (ST-CV versus LT-CV) (), likely due to the elevation of BA levels (), we speculated that the greatly enhanced pericentral autophagy after long-term fasting might be principally attributed to the cell-autonomous pro-autophagic effect of miR-378 rather than the increase in anti-autophagic ACOX1 expression, although the action of circulating BAs became important after long-term fasting (). Fig. 7a and b Fig. 7c Fig. 2n Fig. 2n

Notably, the degree of autophagy enhancement around PV (ST-PV versus LT-PV) was not as significant as that around CV (ST-CV versus LT-CV) (). As only the mRNA levels of periportalshowed an increase after long-term fasting, not the levels of periportal miR-378 (), and the uptake of portal venous BAs is predominantly in the PV zones [], we speculated that the hindered increase in periportal autophagy might be mainly ascribed to the anti-autophagic effect of BAs, mediated through FXR-induced upre­gulation of ACOX1 after long-term fasting when the action of circulating BAs became considerable (). Nevertheless, we could not overlook the impact from non-increasable expression of periportal miR-378 (). Similarly, taking into account the importance of BAs returning to the liver, we also speculated that the hindered increase in periportal autophagy might contribute to the lipid accumulation in the periportal hepatocytes, as evident from stronger staining of LDs, after long-term fasting (). Fig. 7b Fig. 7a and c ²⁸ Fig. 2n Fig. 7a Fig. 7d ACOX1

Interestingly, although HFD feeding resulted in an increase in miR-378 expression in pericentral hepatocytes (), the increased miR-378 expression did not result in a marked increase in pericentral autophagy in these mice as it did in mice fed a chow diet after long-term fasting (LT-CV versus HFD-CV) (;), indicating a miR-378 resistance status in the pericentral hepatocytes of HFD-fed mice. Since HFD feeding also resulted in an increase in pericentral ACOX1 expression (), we speculated that after HFD feeding when BA levels were elevated and their action became considerable (), the increased expression of ACOX1 might serve as a mechanism for the development of miR-378 resistance, impairing the ability of miR-378 in pericentral hepatocytes to promote autophagy. Notably, we observed an elevation of both miR-378 andmRNA levels in the periportal hepatocytes of HFD-fed mice (). The levels of portal venous BAs were higher than those of systemic venous BAs in HFD-fed mice (). As LC3 puncta were almost undetectable in the periportal hepatocytes of HFD-fed mice (), we speculated that the absence of autophagy might be primarily attributed to the anti-autophagic effect of BAs, mediated through FXR-induced upregulation of ACOX1 upon HFD feeding when the action of BAs became considerable (). Based on our findings here, we also speculated that the impaired autophagy in both pericentral and periportal hepatocytes contributes to the abnormal lipid accumulation in the liver after HFD feeding (), regardless of the underlying differences in mechanisms. Fig. 7e Fig. 7b and f Supplementary Fig. S7d Fig. 7e Fig. 1c Fig. 7g Fig. 7h Fig. 7f Fig. 1c Supplementary Fig. S7f ACOX1

Although this study reveals a pivotal role of the BA/FXR/ACOX1/acetyl-CoA axis in suppressing hepatic autophagy, several limitations remain. First, the role of alternative acetyl-CoA-generating pathways, such as those mediated by ACLY and ACSS2, was not fully explored. Therefore, the precise quantitative contribution of ACOX1-derived acetyl-CoA relative to other sources remains unclear. Future studies employing isotope-tracing and detailed metabolic flux analysiswill be essential to comprehensively understand the regulation of autophagy and lipid metabolism under different nutritional and pathological conditions. Second, the intriguing zonation of miR-378 and autophagy, while suggestive of a spatially resolved mechanism, remains correlative. The relationship between pericentral miR-378 expression and zonated autophagy requires direct experimental validation. Future studies employing zonation-specific gene manipulation will be able to dissect the cell-autonomous and non-cell-autonomous mechanisms of miR-378 in distinct hepatic zones. in vivo

All animal experiments were performed in accordance with institutional guidelines for the care and use of animals. Animal protocols were approved by the Ethics Committee of Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences Animal Care Committee (SINH-2022-YH-1). Male C57BL/6 mice and transgenic mice on the C57BL/6 background aged approximately 2–3 months were used in this study. Mice were housed under a 12 h light/12 h dark cycle in a specific pathogen-free animal facility. AFR mice (mCherry-EGFP-LC3; Shanghai Model Organisms Center, Inc.) were crossed with Alb-Cre mice. Adenoviruses were delivered via tail-vein injection to overexpress hepatic miR-378. Each mouse received Ad-378 (5 × 10plaque-forming units [PFU]) or Ad-Ctrl (5 × 10PFU) for 2 weeks. For short-term fasting experiments, mice were kept in new cages without food for 4–6 h before sacrifice. For long-term fasting experiments, mice were kept in new cages without food for 24–36 h before sacrifice. All mice had unlimited access to water. To induce hepatic steatosis, mice aged 6–8 weeks were fed on an HFD containing 60 kcal% fat (D12492, Research Diets) for 12 weeks. For UDCA treatment, following 12 weeks of HFD feeding, mice were maintained on an HFD supplemented with or without 0.5% (w/w) UDCA for an additional 2 months. 8 8

Data for 608 human subjects with and without hepatic steatosis were obtained from Zhongshan Hospital, Fudan University. The clinical characteristics are shown in. The diagnosis of hepatic steatosis was based on abdominal ultrasound examination. The study was approved by the Ethics Committee of Zhongshan Hospital, Fudan University. Informed consent was obtained from all participants. Supplementary Table S1

AML12 cells were cultured in DMEM/F12 supplemented with 10% fetal bovine serum (FBS), 1% insulin-transferrin-selenium (ITS), 1% penicillin/streptomycin and 40 ng/mL dexamethasone. Primary mouse hepatocytes and HepG2 cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. Adenoviruses for the overexpression of miR-378 (Ad-378), MAFG (Ad-MAFG), and knockdown of(Ad-sh) were generated as described previously []. Cells were infected with the adenoviruses at a multiplicity of infection (MOI) of 10 for 36 h. GMR-miRmicroRNA double-stranded mimics for miR-378 (ACU GGA CUU GGA GUC AGA AGG), an inhibitor for miR-378 (CCU UCU GAC UCC AAG UCC AGU), and scrambled controls were purchased from GenePharma. To investigate the role of miR-378, cells were transfected with the indicated RNA oligonucleotides using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s instructions. AML12 cells were infected with Ad-RFP-GFP-LC3 ­(Hanbio Co. Ltd) for 36 h. AML12 cells stably expressing mRFP-GFP-LC3 were established via lentiviral infection. For HBSS starvation, cells were incubated in HBSS media for 2 h before analysis unless otherwise stated. For lipophagy studies, cells were incubated with 100 μmol/L PA for 24 h, followed by 2-h starvation in HBSS before fixation. To block autophagic flux, cells were treated with either 200 μmol/L CQ or 200 nmol/L Baf-A1 for 2 h before collection. For CDCA treatment, cells were treated with 25 μmol/L CDCA or vehicle for 24 h. For GW4064 treatment, cells were exposed to 2 μmol/L GW4064 or vehicle in complete medium for 24 h unless otherwise indicated. For (-)-hydroxycitric acid (HCA) treatment, cells were treated with 200 μmol/L HCA for 2 h. To mimic the nutritional status during prolonged fasting, PHs were incubated in low-glucose (1 g/L) DMEM supplemented with 50 µmol/L oleic acid (OA), 50 μmol/L PA, 10% FBS, and 1% penicillin/streptomycin for 24 h. MAFG MAFG in vitro ²⁶ TM

Liver tissues were fixed with 4% paraformaldehyde and embedded in paraffin or snap-frozen in optimal cutting temperature (OCT) compound. For immunofluorescence staining, tissue sections were fixed in 4% paraformaldehyde and deparaffinized with xylene and ethanol, followed by heat-activated antigen retrieval. Slides were blocked with 5% goat serum for 1 h at room temperature and incubated overnight at 4 °C with the following primary antibodies: anti-LC3B (2775S, CST, 1:100), glutamine synthetase (GS, 11037-2-AP, Proteintech, 1:200), p62 (66184-1-Ig, Proteintech, 1:200), and E-cadherin (3195S, CST, 1:200). The next day, slides were washed in PBS and incubated with appropriate fluorescent secondary antibodies (1:1000) for 1 h at room temperature. Finally, nuclei were stained with DAPI. For LD staining, frozen liver sections were incubated with LipidTOX (, Invitrogen). For cell imaging, hepatocytes were fixed in 4% paraformaldehyde, blocked, permeabilized, and incubated with the primary antibodies including anti-LC3B (2775S, CST, 1:100) and anti-LAMP2 (34141, CST, 1:100), followed by incubation with secondary antibodies. For LD staining in cells, cells were incubated with LipidTOX or BODIPY (493/503) (D3861, ­Invitrogen). DAPI was used to stain nuclei. For histological staining, paraffin-embedded sections were stained with hematoxylin and eosin (H&E). For immunohistochemistry, paraffin sections were stained with the primary antibody anti-LC3B (2775S, CST, 1:100). H34476

Liver TG content was measured according to the manufacturer’s instructions (290-63701, Wako). Briefly, approximately 50 mg of liver tissue was homogenized in 300 μL of isopropyl alcohol and centrifuged. After centrifugation, liver TGs were assessed using the LabAssay Triglyceride measurement kit. Hepatic acetyl-CoA levels in the liver were measured using an acetyl-CoA assay kit (MAK039, Sigma) according to the manufacturer’s instructions. Total BA levels in the serum and the liver were analyzed using the Total Bile Acid Assay Kit (E003-2-1, Jian-Cheng). The individual BA species in the serum were quantified using an ultra-high performance liquid chromatography (UHPLC) system coupled with an Orbitrap Exploris 120 mass spectrometer (Thermo Fisher Scientific) (). Supplementary Table S2

Total RNA was extracted from tissues or cells using TRIzol reagent (15596018, Invitrogen) according to the manufacturer’s instructions. Reverse transcription was subsequently performed with the RT Reagent Kit (RR037B, Takara). For real-time quantitative PCR (RT-qPCR) analysis, reactions were conducted using SYBR Green PCR Master Mix (11203ES08, YEASEN) on a QuantStudio 6 Flex Real-Time PCR System (ThermoFisher). Relative gene expression levels were calculated using the comparative ΔΔCt method. Cycle threshold (Ct) values were normalized to 18S rRNA as the internal reference. Primer sequences are provided in. RNA-seq libraries were constructed using the Illumina TruSeq Stranded mRNA Library Preparation Kit (Illumina). Samples were pooled for deep sequencing on NovaSeq X Plus platform. Differential gene expression analysis was performed using the DEGSeq. Data analysis was carried out via the Majorbio Cloud Platform (majorbio.com). The data have been deposited in the NODE database (OEP00006649). Supplementary Table S3

For western blot analysis, proteins were extracted from cells or tissues using radio-immuno precipitation assay (RIPA) lysis buffer supplemented with protease and phosphatase inhibitors, and quantified using the Bradford assay (23236, ThermoFisher). Lysates were separated by SDS-PAGE and transferred onto polyvinylidene fluoride (PVDF) membranes. Membranes were blocked with 5% BSA and incubated overnight at 4 °C with primary antibodies, including: p62 (610833, BD Biosciences), p-S6K(T389) (9234, CST), S6K (34475, CST), β-Actin (A5316, Sigma), LC3B (2775S, CST), Hsp90 (4874S, CST), p-ULK1(S757) (6888S, CST), ULK1 (8054S, CST), MAFG (NBP2-15019, Novus Biologicals), ACOX1 (10957-1-AP, Proteintech), and GAPDH (10494-1-AP, Proteintech). Chemiluminescent signals were detected using BeyoECL Plus (P0018S, Beyotime).

ChIP assays were conducted using the EZ Magna ChIP G kit (17-409, Merck/Millipore) following the manufacturer’s recommended protocol. Primary mouse hepatocytes were incubated with 2 μmol/L GW4064 or vehicle for 24 h. Chromatin lysates were immunoprecipitated with an anti-FXR antibody (sc-13063, Santa Cruz) or anti-mouse IgG (12-371B, Millipore) as a negative control. The PCR products were analyzed by electrophoresis on a 2% agarose gel. Primer sequences for ChIP assay are provided in. Supplementary Table S3

The isolation of PHs was performed as previously described [], with minor modifications. In brief, after mice were anesthetized, the livers were perfused with perfusion buffer (PBS with 1 mmol/L EGTA), followed by digestion buffer (0.05% collagenase type I). Isolated hepatocytes were filtered through a 70-µm cell strainer. The resulting cell suspension was centrifuged at 50for 2 min at 4 °C. To remove dead hepatocytes, the pellet was resuspended with 12.5 mL DMEM and added to 11.25 mL Percoll mixed with 1.25 mL 10× HBSS. Cells were then centrifuged at 1000for 10 min. The centrifuged cells were collected and counted. For FACS analysis, hepatocytes were centrifuged at 50for 2 min at 4 °C, resuspended in FACS buffer (2 mmol/L EDTA and 0.5% BSA in 1× PBS), and filtered through a 70-µm cell strainer. Dead cells were stained with Live/Dead Violet Viability Kit (Invitrogen) and excluded during analysis. An anti-mouse CD16/32 antibody (Invitrogen) was used to block non-specific binding. To sort pericentral and periportal hepatocytes, cells were stained with APC anti-mouse CD73 (1:100, BioLegend) and PE anti-mouse/human CD324 E-cadherin (1:100, BioLegend). The protocol was developed as described before []. Flow ­cytometry data were acquired using a CytoFLEX S flow cytometer ­(Beckman). FACS was performed on MoFlo Astrios EQ sorter. ^{33 34} g g g

Targeted metabolomics analysis was performed using a UHPLC system coupled to a triple quadrupole mass spectrometer, with metabolites detected in both positive and negative electrospray ionization modes. Widely targeted metabolites were quantified in multiple-reaction monitoring (MRM) mode, with all transitions from large-scale metabolites (QMT1000 Kit, Shanghai BioProfile) detected under optimized declustering potential and collision energy. The original MRM data were extracted using MultiQuant 3.0.2 software, and the peak area of each metabolite was obtained for quantification across all samples. The data have been deposited in the NODE database (OEZ00021653).

Randomization and blinding strategy were used whenever possible. All experiments were performed at least twice, and representative data are shown. Data were expressed as means ± SEM. GraphPad Prism and two-tailed unpaired Student’stest were applied. Logistic regression models were used to estimate the values of OR with 95% CI. Zen 3.8 (Blue edition) software was used for acquisition of images from confocal microscope. FlowJo was used for FACS analysis. A-value of less than 0.05 was considered significant. t- P