FGF21 is a peptide hormone, secreted predominantly from the liver, and its expression is increased in the fasting state (Badman et al., 2007; Inagaki et al., 2007). FGF21 has been shown to regulate the metabolism of lipids and glucose, and FGF21 agonism holds significant potential for treatment of NASH (Ritchie et al., 2020).

To identify a novel secreted protein regulating systemic glucose and lipid metabolism, we have previously performed microarray analysis of livers of mice with hepatic steatosis (Shen et al., 2014). Preliminary data suggest that mRNA levels of FGF9, another member of FGF family, are also increased in the liver of ob/ob mice. Our real-time PCR and Western blotting data confirmed the microarray analysis (Figures 1A,B). We obtained similar results in DIO mice (Figures 1C,D). Interestingly, similar to FGF21, fasting-induced hepatic FGF9 expression, while refeeding, reversed this effect (Figures 1E,F).

These results indicate that FGF9 expression is regulated by different nutritional statuses, and FGF9 might have an important effect on hepatic glucose and lipid metabolism.

Given that FGF9 expression is induced in the liver of DIO mice, we explored the effects of long-term hepatic FGF9 deficiency on systemic metabolism. We first generated an adeno-associated virus expressing FGF9-specific short hairpin RNA (AAV-shFGF9) and infused AAV-shFGF9 into mice via the tail vein. Feeding a chow diet or high-fat diet after eight weeks, the mice were sacrificed for further studies. As a result, the knockdown of FGF9 did not affect hepatic lipid metabolism or systemic glucose metabolism in mice fed with a chow diet (Supplementary Figures S1B–E). However, an increased liver-to-body weight ratio was observed in mice injected with AAV-shFGF9 (Figure 2A). Moreover, AAV-shFGF9 treatment further aggravated glucose intolerance and insulin resistance induced by HFD (Figures 2B,C), and impaired hepatic insulin signaling (Figure 2D). Furthermore, H&E and Oil Red O staining revealed lipid accumulation in the liver (Figure 2E). The biochemical analysis also confirmed the significant increase in hepatic TG levels in AAV-shFGF9-injected mice (Figure 2F).

We also explored the molecular mechanism for the aggravated fatty liver phenotype in AAV-shFGF9-injected mice. We found that FGF9 knockdown in the liver increased the expression of genes involved in lipid synthesis (ChREBP, Fasn, and PPARγ) and fatty acid transport (Fabp1, Fabp4, and CD36). Meanwhile, AAV-shFGF9 treatment also decreased the expression of genes involved in lipid oxidation (Cpt1a, Cyp4a10, and Cyp4a14), as shown in Figures 2G,H. The results suggest that FGF9 plays an important role in the hepatic lipid metabolism.

We next studied whether the effect of FGF9 deficiency on hepatic lipid metabolism is cell autonomous. We generated an adenovirus expressing FGF9-specific shRNA (Ad-shFGF9) and controlled adenovirus Ad-shCon. Primary hepatocytes treated with OA&PA, which stimulates cellular lipid synthesis, were infected with Ad-shFGF9 to observe the short-term effects of FGF9 deficiency on cellular lipid metabolism. As a result, the knockdown of FGF9 expression in primary hepatocytes increased lipid accumulation, as shown in Figures 3A,B.

Consistent with the aforementioned data obtained in DIO mice, we found that Ad-shFGF9 treatment increased the expression of several genes involved in lipogenesis such as ChREBP, Fasn, and PPARγ in primary hepatocytes. The expression of lipid-coated proteins such as cidea and plin2 was also increased. However, the expression of genes related to fatty acid oxidation such as Cpt1a, CYP4a10, and Cyp4a14 was down-regulated in primary hepatocytes infected with Ad-shFGF9. The Western blotting analysis confirmed real-time PCR results, as shown in Figures 3C,D. These results suggested that FGF9 regulating hepatic lipid metabolism is a cell-autonomous effect and FGF9 acts as an autocrine factor.

Given that FGF9 knockdown increased cellular Lipid accumulation, we next explored whether an increased expression of FGF9 in hepatocytes can reduce cellular lipid contents. To this end, the primary hepatocytes were infected with either Ad-FGF9 or Ad-GFP and followed by treatment with palmitic acid and oleic acid (OA&PA). As expected, cellular lipid accumulation induced by OA&PA was suppressed by Ad-FGF9 treatment (Figures 4A,B).

Correspondingly, Ad-FGF9 treatment decreased the expression of several genes involved in lipogenesis such as ChREBP, Fasn, and PPARγ, and fatty acid transport, including CD36, in primary hepatocytes, whereas it increased the expression of genes related to fatty acid oxidation such as Cpt1a, CYP4A10, and Cyp4a14 (Figure 4C). Meanwhile, the Western blotting analysis further confirmed these results (Figure 4D). The results suggested that the overexpression of FGF9 in primary hepatocyte reduced cellular lipid accumulation induced by OA&PA.

To further explore the effect of FGF9 overexpression on hepatic lipid metabolism, we injected Ad-FGF9 or Ad-GFP (as a control) into chow or high-fat diet-fed C57BL/6J mice via the tail vein. As shown in Supplementary Figure S2, FGF9 overexpression in chow diet-fed mice did not significantly influence the systemic glucose metabolism or hepatic lipid metabolism.

However, the injection of Ad-FGF9 into DIO mice decreased the ratio of liver weight to body weight (Figure 5A). In contrast to the results obtained in AAV-shFGF9-infected DIO mice, Ad-FGF9 injection improves glucose intolerance as revealed by GTT experiments (Figure 5B), and enhanced insulin sensitivity, as revealed by ITT experiments (Figure 5C), and enhanced hepatic insulin signaling (Figure 5D). Notably, histological analysis of liver sections (H&E and Oil Red O staining) indicated that the fatty liver phenotype was markedly alleviated in DIO mice injected with Ad-FGF9 (Figure 5E). The biochemical analysis confirmed the reduced hepatic triglyceride content in Ad-FGF9-treated DIO mice (Figure 5F). At the same time, the decreased ALT and AST levels in DIO mice injected with Ad-FGF9 indicate that the overexpression of FGF9 alleviated the liver injury induced by HFD (Figures 5G,H).

Consistent with the results obtained in vitro, Western blotting and real-time PCR analysis confirmed that FGF9 overexpression in the liver of DIO mice inhibited the expression of genes involved in lipogenesis and increased the expression of genes related to fatty acid oxidation (Figures 5I,J). These results clearly suggest that FGF9 regulates the hepatic lipid metabolisms in DIO mice.

To further confirm whether the overexpression of FGF9 inhibits NAFLD development induced by HFD, we generated liver-specific FGF9 transgenic mice (FGF9 ^alb-cre ) by crossing Alb-Cre mice and FGF9 Rosa26 knockin mice. Alb-Cre mice express the Cre-recombinant gene under the control of the albumin gene promoter. The FGF9 Rosa26 knockin mice were generated by the insertion of FGF9 cDNA downstream of the Rosa26 promoter and a loxP-stop-loxP cassette. The Western blotting analysis confirmed the FGF9 overexpression in the liver of FGF9 ^alb-cre (Figure 6A). FGF9 Rosa26 knockin mice were used as control mice. Again, when mice were fed a chow diet, FGF9 transgene in the liver did not affect markedly the systemic glucose metabolism and hepatic lipid metabolism (Supplementary Figure S3). However, when mice were fed a high-fat diet, we observed that the ratio of liver weight to body weight in FGF9 ^alb-cre mice was lower than that of control mice (Figure 6B). Moreover, FGF9 transgene enhanced the insulin sensitivity and improved glucose intolerance in diet-induced obese FGF9 ^alb-cre mice when compared to the control mice (Figures 6C–E). Furthermore, lipid accumulation in the liver was reduced from FGF9 ^alb-cre mice, as revealed by H&E and Oil Red O staining analysis (Figure 6F), and biochemical analysis (Figure 6G). Consistently, FGF9 transgene decreased the levels of serum ALT and AST, further confirming that the overexpression of FGF9 can alleviate the liver injury induced by HFD (Figure 6H).

We also examined the expression of genes involved in lipid metabolism in the liver of FGF9 transgenic mice. Western blotting and real-time PCR data further confirmed that FGF9 transgene inhibited the expression of lipogenic genes, while it enhanced the expression of genes related to fatty acid oxidation (Figures 6I,J). Collectively, these results clearly suggested that liver-specific FGF9 transgenic mice were resistant to HFD-induced NAFLD.

Male C57BL/6J, ob/ob, db/db, and db/m mice at 8 weeks of age were purchased from the Model Animal Research Center of Nanjing University (Nanjing, China) and housed and maintained in 12 h light and dark photoperiods with a regular unrestricted diet.

For adenovirus infection, 8-week-old male C57BL/6J mice were fed a chow diet or HFD (D12492, Jiangsu Xietong Bio-engineering Co., Ltd., China) for 3 months, followed by injection with 1–2 × 10⁹ PFU per recombinant virus (Ad-GFP or Ad-FGF9) via the tail vein. Mice were fasted for 6 h and sacrificed for further analysis, 15 days after infection.

For AAV infection, mice were injected with 5×10¹¹ vg of the AAV-shFGF9 or AAV-shCon via the tail vein, followed by feeding a chow diet or HFD for 8 weeks. Then, mice were sacrificed for further analysis.

FGF9 Rosa26 knockin mice were generated by the insertion of FGF9 cDNA downstream of the Rosa26 promoter and a loxP-stop-loxP cassette at Beijing Biocytogen Co., Ltd., and Alb–Cre mice (express the Cre-recombinant gene under the control of the albumin gene promoter) were provided by Hongbing Zhang (Institute of Basic Medical Sciences, Peking Union Medical College), and then liver-specific FGF9 transgenic mice were obtained by crossing Alb–Cre mice to FGF9 Rosa26 knockin mice. FGF9 Rosa26 knockin mice were used as controls. All the mice were fed on a chow diet or HFD for 3 months.

All the protocol related with animals was approved by the Animal Care and Use Committee of Tianjin Medical University (TMUaMEC) and conformed to criteria outlined in the National Institutes of Health (NIH; Bethesda, MD) Guide for the Care and Use of Laboratory Animals.

Briefly, 8-week-old male mice were anesthetized with bromethol and perfused with 0.5 mg/ml type II collagenase (Sigma-Aldrich) via the inferior vena cava to obtain hepatocytes as previously described (Wang et al., 2010). The cells were seeded at 6 well collagen-coated plates, and cultured with RPMI 1640 medium (H10394; Invitrogen) containing 10% (v/v) FBS (ExCell Bio, Shanghai, China), 50 units/ml penicillin and 50 μg/ml streptomycin (Penicillin–Streptomycin Solution, 100X, Solarbio Life Sciences, Beijing, China). After cell attachment for 4 h, a fresh medium was added after the unattached cells were washed away with PBS. For the adenovirus infection, primary hepatocytes were treated with 0.5 mM OA (CAS No. 112-80-1, Selleck) and 0.25 mM PA (CAS No. 57-10-3, Selleck) for 24 h before (Ad-shCon and Ad-shFGF9, MOI = 100) or after (Ad-GFP and Ad-FGF9, MOI = 100) adenovirus particle infection for 12 h.

Liver tissue of 100 mg was homogenized in 1 ml of 5% Nonidet P-40 dissolved in water, heated to 95°C for 5 min, centrifuged for 2 min at 13,000×g. A fluorometric assay kit was used to measure triglyceride levels according to the manufacturer’s guidelines (Applygen Technologies, Inc., Beijing, China). Finally, the levels of triglyceride levels were normalized to the weight of liver and expressed as mg of triglyceride/g of tissue weight.

For the FGF9 knockdown studies, AAV gene delivery vectors were constructed by cloning FGF9 shRNA sequences into an AAV-shRNA-Ctrl plasmid (Addgene #85741, Watertown, MA), and AAV-shRNA-Ctrl encoding a nontargeting shRNA, was used as a control virus (AAV-shCon). For AAV-shFGF9, we generated the shRNA that would target FGF9 by using the Dharmacon siDESIGN center (http://www.dharmacon.com) as described (Yu et al., 2002), the antisense-loop-sense oligonucleotides synthesized, annealed, and subcloned into the Bbs1 and Xba1 sites of the AAV-shRNA-Ctrl plasmid. The AAV2/8 virus was generated by transfecting HEK-293T cells with pAAV2 insert containing either shRNA control or shRNA-FGF9 under the control of the mouse U6 promoter, pAAV2/8 (Addgene #112864, Watertown, MA) packaging plasmid expressing Rep/Cap genes, and pAdDeltaF6 (#112867).

The three plasmids were used for viral production with a triple-transfection, and the virus was purified according to the modified published methods (Challis et al., 2019). Briefly, triple plasmids were co-transfected to HEK-293T cells, and the cell culture medium was harvested at 72, 96, and 120 h. After purification by an iodixanol step gradient, the fraction containing the virus was desalted on 100 K concentrators with PBS as the diluent. AAV titers were determined by qPCR using primers targeting CMV. Each virus titer was calculated in genomes/ml with AAV-shRNA-Ctrl Luciferase at 2.4×10¹³ vg/ml, AAV-shFGF9 at 2.6×10¹³ vg/ml. The viruses were aliquoted into siliconized tubes and stored at −80°C.

The primers used for constructing AAV-shFGF9 are as follows:

shRNA FGF9 F: 5′-GAT​CCG​CAG​GAC​TGG​ATT​TCA​TTT​AGT​TCA​AGA​GAC​TAA​ATG​AAA​TCC​AGT​CCT​GCT​TTT​TTC​TCG​AGG-3’;

shRNA FGF9 R: 5′-AAT​TCC​TCG​AGA​AAA​AAG​CAG​GAC​TGG​ATT​TCA​TTT​AGT​CTC​TTG​AAC​TAA​ATG​AAA​TCC​AGT​CCT​GCG-3’.

Adenoviruses expressing FGF9 was prepared as previously described (Luo et al., 2007). Briefly, the full-length mouse FGF9 gene was first cloned into the pAd-Track-CMV shuttle vector. The resultant plasmid was linearized by digesting with restriction endonuclease PmeI and subsequently transformed into competent cells, which were BJ5183 derivatives containing the adenoviral backbone plasmid pAdEasy-1. Recombinants were selected for kanamycin resistance, and recombination was confirmed by restriction endonuclease analyses. Overall, the confirmed recombinant adenovirus plasmids were digested with PacI to liberate both inverted terminal repeats and transfected into HEK-293A cells. Recombinant adenoviruses were typically generated within 14–20 days. Ad-shCon and Ad-shFGF9 were purchased from Hanbio Biotechnology Co., Ltd.

For the GTTs, mice were injected with glucose (25% glucose, 1 g/kg) intraperitoneally after a 16-h fasting. For the ITTs, mice were fasted for 6 h and injected with insulin (0.75 U/kg; Novolinhuman insulin) intraperitoneally. Blood glucose levels were recorded using a glucose monitor (OneTouch; LifeScan, Inc., Milpitas, CA) at 0, 15, 30, 60, 90 and 120 min after glucose or insulin injection.

Mice were fasted for 16 h and intraperitoneally injected with insulin (0.75 U/kg). After twenty minutes, mice were euthanized, and the liver tissues were quickly excised, snap-frozen in liquid nitrogen, and stored at −80°C until use. For the evaluation of insulin signaling, primary antibodies against Ser-473-AKT (Cat No. AP0655, ABclonal) and total AKT (Cat No. A3145, ABclonal) were used for Western blotting analysis.

The livers were mounted and frozen in Tissue-Tek O.C.T and sectioned at 5 μm. The frozen sections were air-dried, fixed in 10% neutral-buffered formalin, rinsed in tap water followed by 60% isopropanol, and stained in Oil Red O solution for 15 min. Then, the sections were further rinsed in 60% isopropanol, and the nuclei were stained with hematoxylin followed by aqueous mounting and cover slipping. For primary hepatocytes, the cells were fixed in 10% neutral-buffered formalin, rinsed in tap water followed by 60% isopropanol, and stained in the Oil Red O solution for 15 min.

All the sections were imaged using a color Axiocam105 camera with Zen 2 software attached to a Zeiss Axioplan microscope. The images were analyzed using ImageJ software.

For H&E staining, liver tissues were fixed in 10% neutral-buffered formalin, embedded in paraffin, and cut into 5 μm sections. The sections were stained with hematoxylin and eosin followed by the manufacturer’s instructions (Polysciences, #24901). Briefly, the sections were deparaffinized and rehydrated in distilled water, followed by staining. Finally, the samples were dehydrated, cleared, and mounted, followed by imaging using light microscopy.

The homogenate of liver tissue or cells was prepared in lysis buffer (20 mM Tris–Cl pH7.5, 140 mM NaCl, 1 mM CaCl₂ and MgCl₂, 10 mM NaF, 1% NP-40, 10% glycerol, 2 mM Na-Vanadate, and 1 mM PMSF) supplemented with complete Protease Inhibitor Cocktail (cOmplete™, Sigma-Aldrich, Dallas, TX) for 30 min, centrifuged at 12,000 rpm at 4°C for 15 min. The protein samples were resolved by SDS-polyacrylamide gel electrophoresis and electrophoretically transferred to PVDF membranes. The membranes were blocked at room temperature for 2 h in 5% defatted milk dissolved in TBST (10 mM Tris–HCl, pH 7.4, and 150 mM NaCl), incubated at 4°C for overnight with the following antibodies: PPARγ (#AF6284, Affinity, 1:1000), FGF9 (#A6374, ABclonal, 1:500 and #ab206408, Abcam, 1:500), ChREBP (#A7630, ABclonal, 1:1000), Fasn (#DF6106, Affinity, 1:1000), or β-Actin (#AF7018, 1:10000), washed in TBST (0.1%Tween 20) for 15 min (repeated three times) and incubated with a goat anti-mouse (# S0002, Affinity) or goat anti-rabbit (# S0001, Affinity) IgG (H + L) HRP secondary antibody (1:5000 dilution in TBS) for 1 h at RT. Immunoreactivity was visualized and quantified by infrared scanning using the Odyssey system (LI-COR Biosciences).

Total RNA from either the mouse livers or the primary hepatocytes were extracted using the TRIzol-based method (Invitrogen). cDNA was prepared using the Applied Biosystems’ High-Capacity cDNA Reverse-Transcription Kit. Quantitative real-time reverse-transcriptase PCR (qRT-PCR) was performed using the SYBR Green I Q-PCR kit (TransGen) on a Bio-Rad IQ5 system. All gene expression data were normalized to 36B4 expression levels. All primer sequences are shown in Supplementary Table S1.