Vitamin D receptor targets hepatocyte nuclear factor 4α and mediates protective effects of vitamin D in nonalcoholic fatty liver disease

Epidemiological studies have suggested a link between vitamin D deficiency and increased risk for nonalcoholic fatty liver disease (NAFLD); however, the underlying mechanisms have remained unclear. Here, using both clinical samples and experimental rodent models along with several biochemical approaches, we explored the specific effects and mechanisms of vitamin D deficiency in NAFLD pathology. Serum vitamin D levels were significantly lower in individuals with NAFLD and in high-fat diet (HFD)-fed mice than in healthy controls and chow-fed mice, respectively. Vitamin D supplementation ameliorated HFD-induced hepatic steatosis and insulin resistance in mice. Hepatic expression of vitamin D receptor (VDR) was up-regulated in three models of NAFLD, including HFD-fed mice, methionine/choline-deficient diet (MCD)-fed mice, and genetically obese (ob/ob) mice. Liver-specific VDR deletion significantly exacerbated HFD- or MCD-induced hepatic steatosis and insulin resistance and also diminished the protective effect of vitamin D supplementation on NAFLD. Mechanistic experiments revealed that VDR interacted with hepatocyte nuclear factor 4 α (HNF4α) and that overexpression of HNF4α improved HFD-induced NAFLD and metabolic abnormalities in liver-specific VDR-knockout mice. These results suggest that vitamin D ameliorates NAFLD and metabolic abnormalities by activating hepatic VDR, leading to its interaction with HNF4α. Our findings highlight a potential value of using vitamin D for preventing and managing NAFLD by targeting VDR.

Nonalcoholic fatty liver disease (NAFLD) 4 is characterized by overdeposition of lipids in hepatocytes and is acknowledged as the hepatic manifestation of metabolic syndrome (1,2). NAFLD progresses from simple steatosis to nonalcoholic steatohepatitis (NASH), cirrhosis, and hepatocellular carcinoma (3,4). Excessive lipid accumulation leads to inflammation and insulin resistance (IR), which in turn aggravates hepatic steatosis, thus creating a vicious cycle that promotes progression of NAFLD (5). NAFLD not only causes hepatic bullous steatosis and affects liver metabolic function, but also further develops into cirrhosis and participates in the pathogenesis of metabolic diseases such as diabetes, obesity, and dyslipidemia (6).
Recent studies suggest that vitamin D plays a role in the development of metabolic disorders, including diabetes, NAFLD, and metabolic syndrome (7,8). Vitamin D deficiency has been linked to the diagnosis of NASH and histological features consisting of ballooning, lobular inflammation, and fibrosis (9). Serum vitamin D levels have been suggested to represent a marker for predicting the severity of NAFLD (10). However, the mechanisms underlying the association between vitamin D deficiency and NAFLD remain unclear. The biological functions of vitamin D are primarily mediated via activation of the vitamin D receptor (VDR) (11), which is a member of the nuclear receptor (NR) family (12). A growing number of NRs and their cross-talk have been implicated in regulating the development of NAFLD (13), suggesting that vitamin D may modify NAFLD by targeting VDR and other NRs. Further elucidating the mechanisms underlying the association between vitamin D and NAFLD may provide a better understanding of the pathogenesis of NAFLD and guide novel treatments for its prevention/management.
In recent years, an increasing number of studies have investigated NR cross-talk. NRs are transcriptional factors that contain a DNA-binding domain at the N terminus and a ligand-binding domain at the C terminus (14). In addition to regulating gene expression in a DNA-binding manner, NRs can also affect physiological functions via cross-talk with one another (15,16). For example, a direct interaction between VDR and hepatocyte nuclear factor 4␣ (HNF4␣) in a VDR ligand-dependent manner has previously been demonstrated (17). HNF4␣ also belongs to the NR family; its structural features are similar to those of general NRs, but it has no specific ligands (18). A previous study has demonstrated that HNF4␣ expression is downregulated in NASH and diabetes, and HNF4␣ controls the expression of genes related to triglyceride (TG) transport, including microsomal TG transfer protein (MTTP) and apolipoprotein B (ApoB) (19). Loss of HNF4␣ leads to NAFLD development through impairment of lipid transport (20).
In this study, we discovered that vitamin D supplementation alleviated NAFLD by activating VDR, whereas hepatic-specific knockout of VDR abolished the ameliorative effects of vitamin D on NAFLD. Mechanistically, we found that VDR directly regulated HNF4␣ and thereby further ameliorated NAFLD.

Vitamin D supplementation attenuates high-fat diet-induced hepatic steatosis
To explore the association between vitamin D with NAFLD, serum 25-hydroxyvitamin D (25-VitD) levels were evaluated in NAFLD patients and in high-fat diet (HFD)-fed mice. As shown in Fig. 1A, serum 25-VitD levels were decreased in NAFLD patients and in HFD-fed mice compared with those in healthy controls and standard chow diet (SCD)-fed mice, respectively.
To explore whether vitamin D supplementation is beneficial for NAFLD, 2000 IU of vitamin D per 4057 kcal were added to SCD-fed or HFD-fed mice for 8 weeks. As expected, HFD-fed mice had significantly higher body weights than those of SCDfed mice after 8 weeks of feeding (Fig. 1B). Vitamin D supplementation significantly elevated serum 25-VitD levels ( Fig. 1A) and reduced body weights in HFD-fed mice (Fig. 1B), although both groups of mice consumed comparable amounts of food during 8 weeks of feeding (Fig. S1A). Vitamin D supplementation also significantly reduced intrahepatic TG contents in HFD-fed mice (Fig. 1C). The ameliorated HFD-induced hepatic steatosis by vitamin D supplementation was further confirmed by hematoxylin and eosin (H&E) and Oil Red O staining (Fig.  1D).
Moreover, we found that vitamin D-treated HFD-fed mice showed lower hepatic protein expression of fatty acid synthesis-related genes (SREBP1c, FAS, ACC, and SCD1) but higher hepatic expression levels of fatty acid ␤-oxidationrelated proteins (CPT1␣ and PPAR␣) compared with these parameters in HFD-fed mice without vitamin D supplementation (Fig. 1E). Ketone bodies are the product of fatty acid ␤-oxidation; we found that vitamin D supplementation increased total ketone body levels (the sum of acetoacetate and ␤-hydroxybutyrate) in the serum of HFD-fed mice (Fig. 1F), indicating that fatty acid ␤-oxidation was activated via vitamin D treatment. The anti-steatotic effect of vitamin D was further confirmed by an in vitro experiment showing that vitamin D treatment significantly ameliorated palmitic acid (PA)-induced lipid accumulation in both QSG-7701 and L02 human hepatocyte cell lines (Fig. S2, A-C).

Vitamin D supplementation alleviates HFD-induced insulin resistance
We next evaluated the effect of vitamin D supplementation on insulin resistance, a common manifestation of NAFLD. We found that vitamin D supplementation significantly attenuated HFD-induced hyperglycemia and hyperinsulinemia in mice (Fig. 2, A-C). Vitamin D supplementation also significantly improved HFD-induced glucose and insulin intolerance, as determined by the glucose tolerance test (GTT) and insulin tolerance test (ITT) (Fig. 2, D and E). To examine the molecular mechanisms by which vitamin D improves insulin intolerance, we examined the activation of phosphatidylinositol 3-kinases (PI3K) and protein kinase B (AKT)-both of which are essential factors that mediate insulin signaling (21)-as well as the expression of forkhead box protein O1 (FOXO1), the expression of which is known to be attenuated by insulin through the activation of the PI3K/AKT-signaling pathway (22). As shown in Fig.  2F, vitamin D supplementation significantly up-regulated hepatic expression of PI3K and phosphorylation of AKT but down-regulated hepatic expression of FOXO1 in insulintreated HFD-fed mice, compared with these parameters in insulin-treated HFD-fed mice without vitamin D supplementation. Phosphoenolpyruvate carboxykinase and glucose-6-phosphatase, two key gluconeogenic enzyme genes under the transcriptional control of FOXO1 (23), were also down-regulated following vitamin D supplementation (Fig. 2F). Finally, in vitro results recapitulated our in vivo findings, such that incubation with vitamin D slightly enhanced PA-induced PI3K and AKT activation in QSG-7701 cells (Fig. S3A).

VDR expression is up-regulated in steatotic livers
The biological functions of vitamin D are mediated via VDR binding/signaling, and hepatic VDR expression has been reported to be up-regulated in NAFLD mice and in NAFLD patients (24). Therefore, we investigated the involvement of VDR in hepatic steatosis in both diet-induced and genetically obese mouse models. As illustrated in Fig. 3A, HFD-fed mice showed significantly-higher hepatic nuclear expression of VDR than did SCD-fed mice. Similar up-regulation of VDR was observed in methionine/choline-deficient diet (MCD)-fed mice (Fig. 3B) and in genetically obese (ob/ob) mice (Fig. 3C). In vitro experiments showed that nuclear expression of VDR was significantly up-regulated in primary mouse hepatocytes or in the human hepatocyte QSG-7701 cell line after palmitic acid stimulation (Fig. 3, D and E).

Hepatic-specific VDR deletion exacerbates HFD-induced steatosis and insulin resistance
A previous study reported that global VDR knockout mice had decreased HFD-induced fatty livers compared with those in wildtype (WT) mice provided with a HFD (24). However, global VDR knockout may lead to systemic metabolic abnormalities, such as hypocalcemia, impaired bone formation, growth retardation, a lean phenotype, and resistance to dietinduced obesity, which may be attributed to reduced liver ste-Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD

Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD
atosis (25)(26)(27). To explore the specific role of hepatic VDR in the pathogenesis of NAFLD, we generated hepatic-specific VDR knockout mice (HKO) (Fig. S4, A-C). VDR flox/flox mice (Flox) served as the littermate WT control group. As illustrated in Fig. 4A, HKO mice showed higher body weight gain than did Flox mice, although both groups of mice consumed comparable amounts of food during 8 weeks of HFD feeding (Fig. S1B). HKO mice also had higher intrahepatic TG contents than those of Flox mice after HFD feeding (Fig. 4B), which was further confirmed by H&E and Oil Red O staining (Fig. 4C). Exacerbated steatosis in HKO mice versus Flox control mice was also observed in another model of NAFLD induced by feeding with an MCD diet (Fig. S5, A and B), indicating a protective role of VDR in NAFLD.
Moreover, compared with those of Flox mice, HKO mice had higher levels of hepatic protein expression of lipogenic genes (SREBP1c, FAS, ACC, and SCD1) but lower levels of hepatic fatty acid ␤-oxidation proteins (CPT1␣ and PPAR␣) after HFD feeding (Fig. 4D). HKO mice also had decreased ketone body levels (the sum of acetoacetate and ␤-hydroxybutyrate) in the serum compared with those in Flox mice after HFD feeding (Fig. 4E). HKO mice showed up-regulated hepatic SREBP1c and ACC but down-regulated hepatic CPT1␣ after MCD feeding compared with those of WT mice (Fig. S5C). Furthermore, knockdown of VDR via transfection with siRNA significantly aggravated PA-induced lipid accumulation in QSG-7701 and L02 cell lines (Fig. S6, A-C).
In addition to aggravated hepatic steatosis, HKO mice also showed higher fasting serum glucose, insulin, and homeostatic model assessment of insulin resistance (HOMA-IR) values compared with those of Flox mice after 8 weeks of HFD feeding (Fig. 4, F-H), indicating that hepatic-specific VDR deletion promoted insulin resistance. Furthermore, GTT and ITT analyses showed that HKO mice developed greater glucose and insulin intolerance than did Flox mice after 8 weeks of HFD feeding (Fig. 4, I and J). Insulin-induced PI3K/AKT activation was lower and the levels of insulin-induced gluconeogenesisassociated genes were higher in the livers of HFD-fed HKO mice, compared with those in Flox controls (Fig. 4K). Knockdown of VDR via siRNA also attenuated insulin-induced PI3K/ AKT activation in QSG-7701 cells (Fig. S6D).

Hepatic VDR is required for the inhibitory effects of vitamin D on the development of NAFLD
To explore whether the benefit of vitamin D in inhibiting NAFLD is mediated by hepatic VDR, we supplemented VDR-HKO mice with 2000 IU of vitamin D per 4057 kcal during their HFD regimen. Vitamin D supplementation failed to reduce the HFD-induced body weight gain in VDR-HKO mice after 8 weeks of HFD feeding (Fig. 5A). VDR-HKO mice showed similar intrahepatic TG contents with or without vitamin D supplementation (Fig. 5B). H&E and Oil Red O staining revealed that hepatic-specific VDR knockout diminished vitamin D amelioration of hepatic lipid accumulation in HFD-fed mice (Fig. 5C). Moreover, VDR-HKO mice supplemented with vitamin D showed comparable hepatic levels of proteins mediating fatty acid synthesis (SREBP1c, FAS, ACC, and SCD1) and fatty acid ␤-oxidation (CPT1␣ and PPAR␣) compared with those of HFD-fed VDR-HKO mice without vitamin D supplementation (Fig. 5D). Moreover, vitamin D supplementation did not affect fasting serum glucose, insulin, or HOMA-IR values in HFD-fed HKO mice (Fig. 5, E-G). GTT and ITT analyses showed that HKO mice had similar HFD-induced glucose and insulin intolerance with or without vitamin D supplementation (Fig. 5, H and I). Vitamin D failed to affect insulin-induced PI3K/AKT activation or up-regulation of gluconeogenesis-related genes in the livers of HFD-fed HKO mice (Fig. 5J). These results suggest that hepatic-specific VDR deletion eliminates the beneficial effects of vitamin D in HFD-induced metabolic syndromes.

Activation of VDR up-regulates hepatic HNF4␣ expression
The downstream regulatory mechanisms of VDR in NAFLD remain unclear. VDR is an NR that interacts with HNF4␣, a liver-specific transcriptional factor that plays an important role in the pathogenesis of NAFLD (17,19,20); thus, we hypothesized that activation of VDR ameliorates NAFLD via the regulation of HNF4␣. As illustrated in Fig. 6, A and B, vitamin D supplementation significantly up-regulated hepatic HNF4␣ expression at the protein and mRNA levels in HFD-fed mice. Vitamin D supplementation also up-regulated hepatic MTTP and ApoB-which are two downstream molecules of HNF4␣ that export TGs from the liver into the circulation-in HFD-fed mice. In addition, vitamin D elevated serum very low-density lipoprotein (VLDL) levels in HFD-fed mice (Fig. 6C). In contrast, hepatic-specific deletion of VDR significantly down-regulated hepatic expressions of HNF4␣, MTTP, and ApoB, as well as decreased serum VLDL levels in HFD-fed mice (Fig. 6, A-C). We found similar results in in vitro studies (Fig. S7).
Immunofluorescent results showed that VDR and HNF4␣ co-localized in the nuclei of hepatocytes (Fig. 6D). Co-immunoprecipitation (Co-IP) experiments showed that HNF4␣ directly interacted with VDR in QSG-7701 cells overexpressing VDR and HNF4␣ (Fig. 6E). Because further investigation of the binding sites of these two proteins would be of great value for future functional studies, we next explored the domains required for VDR-HNF4␣ interaction. To this aim, a series of HNF4␣-deletion and VDR-deletion mutants were structured, and domain-mapping experiments were performed (Fig. 6, F and G). Co-IP experiments showed that amino acids (aa) 1-122 of VDR and aa 142-377 of HNF4␣ were required for the VDR-HNF4␣ interaction (Fig. 6, F and G). The 1-122-aa domain of VDR is the DNA-binding domain, and the 142-377-aa domain of HNF4␣ is the ligand-binding domain (28,29), suggesting that VDR may affect the ability of HNF4␣ to bind to its endogenous ligands (i.e. fatty acids) via its DNA-binding ability. In addition, the interaction between VDR and HNF4␣ in QSG-7701 cells was enhanced by vitamin D treatment (Fig. 6H). Furthermore, luciferase assays showed that vitamin D enhanced the transcription of HNF4␣ to induce the expression of MTTP and ApoB (Fig. 6I).

Overexpression of HNF4␣ in the liver ameliorates steatosis and insulin resistance in HFD-fed VDR-HKO mice
To further explore whether HNF4␣ mediated the regulatory effects of VDR in NAFLD, a rescue experiment was performed by tail-vein injection of an HNF4␣-overexpressing AAV into Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD VDR-HKO mice. Injection of HNF4␣-overexpressing AAV significantly up-regulated hepatic expressions of HNF4␣ and its downstream molecules, MTTP and ApoB, in mice (Fig. 7A). Overexpression of HNF4␣ significantly restored fasting serum glucose, insulin, and HOMA-IR values (Fig. 7, B-D) and reduced intrahepatic TG contents in HFD-fed VDR-HKO mice (Fig. 7E). Serum VLDL levels were also restored via overexpression of HNF4␣ (Fig. 7F). The alleviation of hepatic steatosis in HNF4␣-overexpressed VDR-HKO mice was confirmed by H&E and Oil Red O staining (Fig. 7G). These results suggest that HNF4␣ mediates the regulatory effects of VDR on NAFLD.

Discussion
In this study, we investigated the effects and mechanisms of vitamin D on the pathogenesis of NAFLD. We found that vitamin D supplementation played a protective role in HFD-induced NAFLD via activation of VDR. Hepatic-specific knockout of VDR aggravated HFD-or MCD-induced liver steatosis. In addition, HFD or MCD treatment led to a compensatory increase in VDR expression, which interacted with HNF4␣ to protect the liver from damage caused by lipid deposition.
NAFLD is a widespread disease that affects 25.24% of the global population (30). The prevalence of NAFLD in China has currently reached 29.2% and has continued to increase year after year (31). It has been widely accepted that vitamin D deficiency promotes insulin resistance and NAFLD (9,32,33). Based on clinical findings revealing an association between vitamin D levels and NAFLD, vitamin D has been suggested as a therapeutic option for NASH (34). However, whether vitamin D supplementation improves NAFLD has remained controversial in clinical trials (35)(36)(37). Here, we explored the role of vitamin D in the pathogenesis of NAFLD, and we found that vitamin D supplementation decreased HFD-induced lipid accumulation and insulin resistance in mice, suggesting a potential therapeutic effect of vitamin D for the treatment of NAFLD.
Vitamin D exerts its regulatory role primarily through VDR binding and signaling (38). A previous study has suggested that VDR expression is negatively associated with the severity of liver histology in NASH patients (39). More importantly, expression of liver VDR is induced in mouse models of NAFLD and NAFLD patients (24). Collectively, these findings strongly indicate that VDR plays a critical role in NAFLD. However, a previous study suggested that global VDR knockout decreases HFD-associated liver steatosis in mice (24), which contradicts the results of this study. However, global VDR knockout can lead to systemic change because VDR is widely expressed in various tissues, and ϳ3% of the human genome is regulated by VDR (40); hence, HFD-induced hepatic phenotypes induced by global VDR knockout may be caused by these aforementioned systemic changes. Furthermore, it has been reported that VDR null mice exhibit a lean phenotype and growth retardation, and global VDR knockout mice are resistant to diet-induced obesity (25,27). Therefore, whole-body knockout of VDR may not represent the best model for exploring the effects of hepatic VDR on NAFLD. In this study, we used hepatic-specific VDR knockout mice-which showed no differences in growth or development characteristics compared with those of WT mice-to explore the hepatocyte-specific effects of VDR deletion on NAFLD. The use of hepatic-specific VDR knockout mice allowed us to decipher the specific roles of hepatic VDR in the pathogenesis of NAFLD. Our data showed that hepatic-specific VDR knockout mice exacerbated HFD-induced and MCD-induced liver steatosis, suggesting that VDR has an important role in NAFLD. We also found that hepatic-specific VDR deletion diminished the rescuing effect of vitamin D on hepatic steatosis and insulin resistance in mice, indicating that the regulatory role of vitamin D primarily depends on the activation of VDR.
It is well-established that HNF4␣ is a central transcription factor that regulates expression levels of genes involved in the progression of NAFLD in the liver (19,41,42). Previous studies have reported that loss of HNF4␣ leads to liver steatosis by reducing hepatic expression of MTTP and ApoB, both of which are essential to VLDL secretion (20,43,44). HNF4␣, which belongs to the NR family, has been demonstrated to interact with VDR (17). In this study, we demonstrated that hepatic VDR deletion in mice led to decreased protein and mRNA levels of HNF4␣, whereas vitamin D supplementation in WT mice restored these levels, as well as downstream levels of MTTP and ApoB (19). We also identified co-localization of VDR and HNF4␣ in the nucleus and confirmed their interaction as direct binding partners. Furthermore, it is noteworthy that we reported for the first time that the ligand-binding domain of HNF4␣ and the DNA-binding domain of VDR participated in this binding; the identification of this VDR-HNF4␣ interaction provides a foundation for future functional studies. We also demonstrated that vitamin D enhanced the binding of VDR and HNF4␣ and up-regulated the transcriptional activity of HNF4␣. Moreover, exacerbation of HFD-induced hepatic steatosis and IR in VDR-HKO mice was diminished by overexpression of HNF4␣ via AAV injections. Therefore, we speculate that the regulation of VDR on the progression of NAFLD and its associated complications depends on HNF4␣-mediated TG exporting. Hence, it may be worthwhile for future studies to investigate the correlation between the expression levels of VDR and HNF4␣ in the livers of NAFLD patients.

Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD
Based on our present findings, further studies are needed to clarify the mechanisms of VDR regulation in NAFLD. This study cannot fully exclude the role of VDR itself in the progression of NAFLD, because ligand-independent regulations of VDR have been found in some studies (45,46). Future studies testing hepatic VDR overexpression are needed to verify this distinction. Additionally, because we identified the interaction sites of VDR and HNF4␣, further investigation should be performed to elucidate the precise mechanisms by which VDR regulates the HNF4␣ activity via protein interactions. Our data indicate that the DNA-binding domain of VDR participates in the interaction of VDR and HNF4␣. As the most important part of VDR, the DNA-binding domain is sufficient for target gene discrimination and transcriptional regulation (47). Studies on the regulation of other transcriptional factors that modulate transcriptional activity of VDR have also been reported (48). It is possible that the interaction of VDR and HNF4␣ alleviates hepatic steatosis by affecting DNA-binding activity of VDR. For further confirmation, future studies should mutate the essential amino acids of the DNA-binding domain of VDR to decrease its ability to bind to VDR response elements of its target genes without affecting the interaction of VDR and HNF4␣. In this study, we found that the effect of vitamin D on FOXO1 was only evident in the modeling groups and not in the control groups. A possible explanation is that PA stimulation in the modeling groups may have perturbed cellular homeostasis, and vitamin D treatment may facilitate the homeostasis. In contrast, cellular homeostasis in the control groups was already achieved under

Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD
normal conditions, such that vitamin D supplementation may not have exhibited any further effect on such homeostasis. Future studies clarifying the different effects of vitamin D/VDR signaling in normal and NAFLD conditions may help to better elucidate the mechanisms and treatments for NAFLD. In this study, we also found an effect of VDR deletion in MCD models, which exhibit similar histological characteristics to those of human NASH (49). It has been widely accepted that vitamin D plays an important role in inflammatory processes (50). Consequently, further studies focusing on the role of vitamin D in inflammation are required. In this study, because vitamin D supplementation or VDR ablation had no effect on food intake, the observed decrease in body weight may be associated with increased energy expenditure. This hypothesis is partially supported by our present evidence that vitamin D supplementation and VDR ablation ameliorated hepatic steatosis and improved insulin resistance, the latter of which may consequently increase energy expenditure and inhibit body weight gain (51)(52)(53). Therefore, studies investigating the effect of vitamin D on energy expenditure are needed in the future. We found that vitamin D supplementation and VDR knockout did not change the serum nonesterified fatty acid (NEFA) concentration. The main source of serum NEFA is via lipolysis in adipocytes, and this process is regulated by several factors, such as the interplay of hormonal, neurological, and pharmacological stimuli (54). In addition, NEFA is metabolized by many metabolic pathways (55). Hence, the effects of vitamin D supplementation and VDR knockout on serum NEFA levels might be compensated by other factors that are related to NEFA metabolism; further studies are needed to clarify this issue.
In conclusion, we identified a novel mechanism by which vitamin D ameliorates NAFLD via the activation of hepatic VDR and its interaction with HNF4␣ (summarized in Fig. 8), providing a novel insight into improving treatment of NAFLD.

Patients
This study initially enrolled employees who were attending their annual health examinations (during the period between January 1, 2017, and June 31, 2019) at Zhejiang University Taizhou Hospital. A total of 2394 participants (1635 males and 759 females; between 16 and 86 years old) were included in the analysis. NAFLD was diagnosed by abdominal ultrasound by a trained ultrasonographist who was unaware of the clinical and laboratory data, using a GE Logic E9 machine (GE Healthcare). Hepatic steatosis was diagnosed according to conventional criteria (56). Venous blood samples were drawn after an overnight fast of at least 8 h. A Sciex API 3200 mass spectrometer was used to detect serum 25-VitD levels. The study protocol was approved by the Ethics Committee of Zhejiang University Tai-zhou Hospital (approval number K20160603) and abided by the principles of the Declaration of Helsinki.

Mice and treatments
All mouse procedures were approved by the Animal Care and Use Committee of the First Affiliated Hospital, College of Medicine, Zhejiang University. Male C57BL/6 mice (6 -8 weeks old) were housed on a 12-h light/dark cycle in a temperature-controlled environment (23 Ϯ 2°C) with ad libitum access to food and water. Mice were fed with a HFD (Research Diet, New Brunswick, NJ) for 8 weeks to establish an NAFLD model or an MCD (MPBio, Santa Ana, CA) for 5 weeks to establish a NASH model. For vitamin D administration, 2000 IU of vitamin D per 4057 kcal was added into the diet after weaning and maintaining until the end of the 8-week HFD. A vitamin D-supplemented diet and the corresponding controlled diet were purchased from Shuyishuer Bio (Changzhou, China). At the end of the mouse experiments, mice were fasted overnight for 16 h prior to being sacrificed. After an overnight fasting for 16 h, the mice were injected intraperitoneally with insulin (1 unit/kg) at 10 min prior to being sacrificed in order to investigate insulin signaling. VDR fl°x/ϩ mice from a C57BL/6J background were generated utilizing the CRISPR/CAS9 system by Biocytogen Biological Technology Co., Ltd. (Beijing, China). The third exon was flanked by loxP sites, and thus two single guide RNAs (sgRNA1 and sgRNA2) targeting VDR introns 2-3 and 3-4 were designed. The schematic diagram is shown in Fig.  S3A. The PCR primers used for identification are listed in Table  S1. Hepatocyte-specific VDR knockout mice (VDR-HKO) were generated by crossing VDR fl°x/ϩ mice with albumin-Cre mice (Biocytogen). The ob/ob mice were purchased from the Model Animal Research Center of Nanjing University and were fed normal chow. To overexpress HNF4␣ into the liver of

Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD
VDR-HKO mice, VDR-HKO mice were injected with AAVs through the tail vein at a dose of 4 ϫ 10 11 genome copies per mouse.

Production of AAVs
For viral production, 293T cells were maintained in 150-mm plates. For each transfection, 15 g of AAV plasmid carrying the construct of interest, 15 g of AAV helper plasmid, and 15 g of pAAVDJ serotype-packing plasmid were added to 4.5 ml of serum-free DMEM. Then, 90 l of Turbofect (Invitrogen) was added to the mixture and incubated at room temperature for 15-20 min. After incubation, the mixture was added to media and was mixed gently yet thoroughly. Cells were harvested between 48 and 72 h post-transfection. Subsequently, the recombinant AAV was purified via a ViraBind TM AAV purification kit (Cell Biolabs, San Diego, CA) according to the manufacturer's protocol. Purified AAV was quantified via the QuickTiter TM AAV quantitation kit (Cell Biolabs).

Measurement of biochemical indices and metabolic analysis
Blood glucose was measured with a glucometer (Onetouch, LifeScan Inc., Milpitas, CA). Serum insulin levels were measured via ELISAs (Millipore, Billerica, MA). Hepatic TG contents were assayed by corresponding kits (Applygen, Beijing, China). Serum ketone bodies were assayed by corresponding kits (Sigma). Serum VLDL levels were measured by ELISA kits (Sab Biotech, College Park, MD), which detected repeated fragments of VLDL. All procedures were performed according to the instructions provided by the manufacturers.

GTT and ITT
For GTT, mice were fasted for 16 h followed by intraperitoneal injection of 1 g/kg glucose (Sigma). For ITT, mice were fasted for 6 h followed by intraperitoneal injection of 0.75 units/kg of regular human insulin (Wanbang, Xuzhou, China). Blood glucose was determined at 0, 15, 30, 60, 90, and 120 min after the injections.

Histological analyses
Livers were sectioned after being fixed in 10% neutral formalin and were then embedded in paraffin, after which H&E staining was performed. For Oil Red O staining, livers were freshly embedded in optimal-cutting-temperature compound at Ϫ20°C and were then sectioned. Staining was performed as described previously (57).

Immunofluorescence
Cells were fixed in 4% paraformaldehyde for 20 min and washed three times in PBS, followed by permeabilization in 0.2% Triton X-100 (Sangon, Shanghai, China) for 10 min. After being washed three times in PBS, the cells were blocked in goat serum (ZSGB Bio, Beijing, China) for 30 min and incubated with rabbit VDR antibody and mouse HNF4␣ antibody at 4°C overnight, followed by incubation in fluorescein isothiocyanate (FITC)-conjugated secondary antibody (1:300) and DAPI staining. Details of antibodies were as follows: HNF4␣ (ab41898) and VDR (ab3508).

Quantitative real-time PCR
RNAs were isolated by the TRIzol method and were reversetranscribed into cDNAs. Quantitative real-time PCR was conducted using SYBR Green (Takara Shuzo, Otsu, Japan) with a Bio-Rad CFX384 system. The primers for quantitative realtime PCR were listed in Table S2. The relative expression levels of target RNAs were normalized using Gapdh as an internal control.

Vitamin D-VDR interacts with HNF4␣ to ameliorate NAFLD
Luciferase assay L02 cells were transfected with firefly luciferase reporter plasmids (PGL3-basic vector) driven by the HNF4␣ promoter (HanyinBio, Shanghai, China). Renilla luciferase-expressing plasmids were co-transfected and used for normalization. Both firefly and Renilla luciferase activities were measured by using the Dual-Luciferase assay kit (Promega, Madison, WI) for 48 h after plasmid transfection. Firefly luciferase units were normalized against Renilla luciferase controls. Results are presented as firefly/Renilla luciferase activities.

Statistical analysis
Values are reported as the mean Ϯ S.D. Statistical differences were determined via unpaired, two-tailed Student's t tests or one-way analyses of variance (ANOVAs) with Tukey's correction when appropriate. All tests were performed in Prism 8 (GraphPad). p Ͻ 0.05 was considered statistically significant.