Abstract

   Metabolic dysfunction–associated steatohepatitis is well accepted as a
   potential precursor of hepatocellular carcinoma. Previously, we
   reported that fibroblast growth factor 21 (FGF21) revealed a novel
   anti-inflammatory activity via inhibiting the TLR4–IL-17A signaling,
   which could be a potential anticarcinogenetic mechanism to prevent to
   MASH-HCC transition. Here, we set out to determine whether FGF21 has a
   major impact on Kupffer cells’ (KCs) ability during MASH-HCC
   transition. We found aberrant hepatic FGF21 and KC pool in human
   MASH-HCC. Lack of FGF21 up-regulated ALOX15, which converted the
   oxidized fatty acids to induce excessive KC death and mobilization of
   monocyte-derived macrophages (MoMFs) for KC replacement. Lack of FGF21
   oversupplied free fatty acids for sphingosine-1-phosphate (S1P) cascade
   synthesis to mediate MASH-HCC transition via S1P-YAP signaling and
   cross-talk between tumor cells and macrophages. In conclusion, lack of
   FGF21 accelerated MASH-HCC transition via the S1P-AP signaling.
   Compromised MoMFs could present as tumor-associated macrophage
   phenotype rendering tumor immune microenvironment for MASH-HCC
   transition.
     __________________________________________________________________

   FGF21 could prevent MASH-HCC transition via preservation of KC pool
   homeostasis and inhibition of S1P-S1PR2-YAP signaling.

INTRODUCTION

   Metabolic dysfunction–associated steatohepatitis (MASH), formerly known
   as nonalcoholic steatohepatitis (NASH), a severe form of metabolic
   dysfunction–associated steatotic liver disease (MASLD; formerly known
   as NAFLD), is well accepted as a potential precursor of hepatocellular
   carcinoma (HCC) ([42]1). Using a Markov model to forecast MASLD
   progression, MASH is predicted to increase 27%, while the incidence of
   HCC will increase 137% by 2030 ([43]2). MASH is becoming a main cause
   for end-stage liver disease, in particular malignancy, as a major
   clinical and economic burden ([44]3). Although MASH was first
   documented more than 30 years ago ([45]4), its
   pathogenesis/carcinogenesis is not fully elucidated. Because early and
   effective treatment may determine the long-term MASH prognosis ([46]5),
   there is an unmet need to identify the factor(s) that can be targeted
   therapeutically to prevent MASH progression.

   Previously, we reported that fibroblast growth factor 21 (FGF21)
   revealed a novel anti-inflammatory activity via inhibiting the
   Toll-like receptor 4 (TLR4)–interleukin-17A (IL-17A) signaling, which
   could be a potential anticarcinogenetic mechanism to prevent to
   MASH-HCC transition ([47]6). Regarding the tumor microenvironment in
   MASH liver, studies have emphasized the pathophysiological importance
   of TLR4 from hepatic resident Kupffer cells (KCs) ([48]7–[49]9) and
   stellate cells ([50]10). Of note, both hepatic resident KCs and
   infiltrated immune cells [monocyte-derived macrophages (MoMFs), natural
   killer (NK) cells, neutrophils, NKT cells, and T cells] are known as
   inflammation/immune mediators that contribute to MASH-HCC transition.
   KCs account for approximately 15% of the total liver cell population
   and are the largest population constituting 80 to 90% of the total
   number of tissue-resident macrophages in whole body. The maintenance of
   KC pool has been thought to involve two mechanisms, (i) replenishment
   by KC self-renewal and (ii) recruitment from bone marrow–derived
   monocytes (BMDMs) ([51]11). In MASH scenario, underlying self-renewal
   of hepatic resident KCs could be impaired, while the KC pool was
   gradually seeded by the MoMFs, which could be partly immature or less
   efficiently than the hepatic resident KCs, but exhibited a
   pro-inflammatory effect to exacerbate liver damage ([52]12). In MASH,
   metabolism of KCs/MoMFs can be reprogrammed to fulfill carcinogenetic
   processes, i.e., decreased glycolysis and fatty acid (FA) oxidation but
   increased FA synthesis ([53]13). KCs/MoMFs represent as one of the most
   abundant innate immune cells, while macrophage polarization has been
   associated with tumor growth, metastasis, angiogenesis, and poor
   prognosis ([54]14). Aberrant signaling from immature macrophages could
   be deleterious, but little is known about the KCs/MoMFs’ metabolic
   changes, which may determine functional fate of the macrophages during
   HCC development.

   FGF21, an endocrine FGF primarily produced in liver ([55]15, [56]16),
   elicits metabolic benefits protect liver against metabolic stress
   caused by MASH ([57]17). This major endocrine action of FGF21 results
   in a combination of effects, including control of lipolysis, clearance
   of excessive free FAs (FFAs), enhanced expenditure of the stored lipid
   energy by mitochondrial substrate oxidation, and catabolism and
   uncoupling, which negatively regulates hepatic or tissue steatosis and
   adiposity ([58]18, [59]19). Pharmacological application of FGF21 holds
   great promise as an effective therapeutic means for treating obesity,
   diabetes, and MASH ([60]20–[61]29). However, no previous study has been
   performed to investigate the role of FGF21 played on KCs/MoMFs, in
   particular, maintenance of the KC pool during MASH-HCC transition.

   Here, we set out to determine whether FGF21 has a major and
   long-lasting impact on KCs/MoMFs’ ability that in turn alters the
   response to lipotoxicity during MASH-HCC transition. Our findings
   reveal that compromised function of macrophages contribute to MASH
   progression to HCC, while FGF21 plays a critical role to maintain the
   KC pool to prevent MASH-HCC transition.

RESULTS

Aberrant hepatic FGF21 and KC pool in human MASH-HCC

   In the specimens of patients with HCC with medical history of MASH, the
   pathological features of MASH-HCC were confirmed by hematoxylin and
   eosin (H&E) and Oil Red O staining. Much weaker staining of FGF21 was
   found in the tumor region in comparison with the adjacent benign region
   ([62]Fig. 1A). Consistently, significantly decreased protein and mRNA
   levels of FGF21 were found in the tumor tissues compared to the
   adjacent benign tissues ([63]Fig. 1B). Serological levels of FGF21 were
   determined by enzyme-linked immunosorbent assay (ELISA) in the
   postresection patients with HCC and followed up to 9 months. In
   comparison with preoperation, 8 of 10 patients (no relapse) showed
   increases of postresection FGF21 levels, while only 3 of 10 patients
   (relapse) showed increases of postresection FGF21 levels ([64]Fig. 1C).
   In 156 overweight [body mass index (BMI) ≥ 25 kg/m^2] patients with HCC
   from The Cancer Genome Atlas (TCGA) database, 108 patients were graded
   as G1–2, while 48 patients were graded as G3–4, according to the
   National Comprehensive Cancer Network severity scale. Analysis of
   survival rate indicated that low expression of FGF21 had a negative
   impact on the survival of high-BMI patients with advanced disease
   (G3–4) ([65]Fig. 1D). Cumulative survival of the overweight patients
   with HCC was further analyzed to associate with immune cells, including
   macrophage, neutrophil, dendritic cell, CD4^+ cell, CD8^+ cell, and B
   cell. By splitting 50% of the patients with high level of cell
   subpopulations and 50% with low level of cell subpopulations, the
   Kaplan-Meier curve showed a poor outcome for the patient with high
   level of macrophages (fig. S1A). When CD68 (a marker highly expressed
   by circulating macrophages) was normalized by FGF21, high expression of
   CD68 was negatively correlated with the percent survival in the
   overweight patients with HCC (fig. S1B). By coexpression analysis, a
   negative correlation was also found between CD68 and FGF21 (fig. S1C).
   To study the tumor immune microenvironment (TIME) in which both KCs and
   MoMFs were distributed in liver, imaging mass cytometry was performed
   in the human MASH-HCC tissues because this high-plex imaging allowed
   visualizing multiple markers to obtain a precise picture of cell
   subpopulations in the tissue architecture. To provide a detailed
   inventory of landscape of KC pool, we characterized the macrophages in
   tumor and adjacent benign regions of MASH-HCC samples using an
   optimized eight-marker panel for Hyperion Imaging. The cell
   subpopulations were identified in the regions of tumor and adjacent
   benign ([66]Fig. 1E) for spatial distribution analyses. On the basis of
   the expression of lineage markers, the subpopulation included CD68^+
   cells, CD68^+CD163^+ cells, and Cd68^+CD163^+CD14^+ cells. Extensively
   distributed CD68^+ macrophages were found in the regions either tumor
   or adjacent benign ([67]Fig. 1F), while relatively increased
   CD68^+CD163^+ macrophages (potential M2 polarization) and
   CD68^+CD163^+CD14^+ cells (MoMFs) were found in the tumor regions in
   comparison with that in the adjacent benign regions ([68]Fig. 1, G to
   I), which suggested the alternatively activated M2 macrophages and KC
   pool reseeding by MoMFs ([69]12) in TIME. Together, either decreased
   hepatic FGF21 or replacement of hepatic resident KCs by MoMFs
   negatively affected the outcome of patients with MASH-HCC. As the major
   metabolic benefits of FGF21 result in controlling lipid metabolism, it
   evoked the interest to explore the effect of FGF21 on the lipid
   metabolic signaling of KC pool in MASH liver.

Fig. 1. Aberrant hepatic FGF21 and KC pool in human MASH-HCC.

   [70]Fig. 1.
   [71]Open in a new tab

   (A) Representative images of MASH-HCC histology by H&E and Oil Red O
   staining and immunohistochemistry detection of GF21 protein
   distribution in the regions of tumor (white dash circle) tissue and
   adjacent benign tissue. (B) Protein and mRNA levels of FGF21 by Western
   blotting and qPCR in the tumor and adjacent tissues. GAPDH,
   glyceraldehyde-3-phosphate dehydrogenase. (C) Serum FGF21 protein
   levels by ELISA in the postresection patients with HCC (no relapse and
   relapse). (D) A cohort of 156 overweight (BMI ≥ 25 kg/m^2) patients
   with HCC from TCGA database were subgrouped into G1–2 patients and G3–4
   patients to analyze the survival rate, based on high/low expression of
   FGF21. (E) Representative images of Hyperion Imaging in
   paraffin-embedding tissues of patients with HCC. (F to H) Cell
   subpopulations of CD68^+ macrophages, CD68^+CD163^+ macrophages, and
   CD68^+CD163^+CD14^+ macrophages in tumor and adjacent tissues. (I) the
   ratio of CD68^+CD163^+CD14^+ macrophages in CD68^+ macrophages in tumor
   and adjacent tissues. *P < 0.05 and **P < 0.01.

KC pool and FA metabolism in FGF21KO-MASH mice

   To study the effect of FGF21 on lipid metabolic signaling in KCs upon
   MASH insult, two MASH mouse models were established in FGF21 knockout
   (KO) mice and wild-type (WT) littermates by feeding a high-fat
   methionine/choline-deficient diet (HFMCD) or a Western-style diet
   (WSHFD; high fat and high fructose). The mice with HFMCD showed high
   severity of MASH, while the mice with WSHFD displayed the MASH
   characteristics with clinical relevance, particularly lipid metabolic
   disorders (fig. S2). The WSHFD-MASH model was therefore used for
   further studies. The result of whole-body calorimetry showed
   significant decreases of respiratory exchange ratio and energy
   expenditure in the FGF21KO-MASH mice (fig. S3A), while rhFGF21
   treatment attenuated the changes (fig. S3B), suggesting that lack of
   FGF21 caused switch from glucose to FA for energy generation in the
   MASH liver. Macrophages were isolated from liver tissue for RNA
   sequencing (RNA-seq). Volcano plot showed that FGF21KO-MASH mice
   (versus WT-CD controls) had great numbers of differentially expressed
   genes (DEGs), including Dd93, Itgax, Mmp12, Gpnmb, Cd9, Fam20c, Emp1,
   Fabp5, Atp6v0d2, Anxa1, and Clec4e, which were uniformly expressed
   across all the macrophages ([72]Fig. 2A). The gene signatures of MoMF
   were up-regulated, but the gene signatures of KC were down-regulated in
   the FGF21-MASH mice ([73]Fig. 2B). These characteristics of gene
   expression and signature were agreed with the previous report in which
   MoMFs could gradually seed the KC pool in response to the death of
   embryo-derived KCs in MASH ([74]12). Running Enrichment Score showed
   that the gene signatures of FA metabolism were positively associated
   with the up-regulated genes of MoMF but negatively associated with
   genes of tissue-resident KC such as Timd4 and Cd207 genes in the
   FGF21KO-MASH mice ([75]Fig. 2C), suggesting that FA metabolism could
   play a critical role for seeding MoMFs to maintain the KC pool. As
   hepatic FA deposition has been reported to contribute to approximately
   60% of hepatic lipid accumulation ([76]30), the enzymes for FA
   metabolism were therefore analyzed by quantitative polymerase chain
   reaction (qPCR). Increases of FA synthesis (FASN, ACC1, ACC2, SREBP1c,
   and PPAR-r), FA update (Slc27a1, Slc27a2, Slc27a3, and CD36), TG
   synthesis (Dgat1 and Acat1), and export (Mttp and Apoa1) but decrease
   of FA oxidation (Cpt1-a, Acox1, Scacd, Mcad, and Ppar-γ) were found in
   the FGF21KO-MASH mice ([77]Fig. 2D). Because FASN and CD36 were noticed
   among the FA enzymes, their protein levels were further determined by
   Western blotting. High levels of FASN and CD36 were found in
   FGF21KO-MASH mice in comparison with other three groups, with
   statistical significance ([78]Fig. 2E). Restoration of FGF21 with
   rhFGF21 injection (125 μg/kg body weight) daily for 4 weeks resulted in
   significantly decreased FASN and CD36 ([79]Fig. 2F) in consistent with
   the alleviation of MASH, evidenced by the histological changes of
   steatohepatitis ([80]Fig. 2F) and decreased NAFLD activity score (NAS;
   fig. S3C). We next asked whether lack of FGF21 could increase FA
   synthesis and transport. The in vitro studies were performed using
   FL83B cells (a benign mouse cell line of hepatocyte) and FL83B-FGF21KD
   cells in which FGF21 gene was knocked down (KD) by short hairpin RNA.
   Significantly increased FASN and CD36 were found when the FL83B-FGF21KD
   cells were challenged with palmitic acid (PA) at 100 μM for 48 hours
   ([81]Fig. 2G), while PA-induced steatosis and increases of FASN and
   CD36 were significantly attenuated by rhFGF21 treatment (fig. S3, D and
   E). To elucidate whether lack of FGF21 could liberate FAs from adipose
   tissue, we measured the protein phosphorylation of hormone-sensitive
   lipase (HSL), which had broad substrate specificity to catalyze the
   hydrolysis of triacylglycerol, diacylglycerol, and
   1(3)-monoacylglycerol ([82]31). Significantly increased phosphorylation
   of HSL at Ser563 and Ser660, two major sites for the hydrolytic
   activity for triacylglycerol, was found in the adipose tissues from
   FGF21KO-MASH mice ([83]Fig. 2H). Significantly increased
   phosphorylation of HSL at Ser563 and Ser660 was also found in the
   FGF21KO mice compared to the WT mice ([84]Fig. 2H). All these results
   indicated the critical roles of FGF21 played in controlling of adipose
   tissue hydrolysis to liberate FAs, FA de novo synthesis, and FA
   accumulation in liver. To characterize the specificity of FA
   metabolites in FGF21KO-MASH mice, long-chain FAs (LCFAs) were analyzed.
   Two specific oxidized FAs, Hydroxyoctadecadienoic acids (HODE) and
   15-hydroxyeicosatetraenoic acids (15-HETE), which are converted by
   Alox15 from arachidonic acid and linoleic acid ([85]32, [86]33), were
   noticed in the liver tissue of FGF21KO-MASH mice ([87]Fig. 2I).
   Although the oxidized FAs such as HODE and HETE have been reported for
   MASH development ([88]34), there is no study to report the Alox15
   signaling in MASH progression. We thus investigated Alox15 expression
   in FG21KO-MASH mice thereby to understand FGF21-Alox15 signaling in
   disease progression and carcinogenetic transformation.

Fig. 2. KC pool and FA metabolism in FGF21KO-MASH mice.

   [89]Fig. 2.
   [90]Open in a new tab

   (A) Volcano plot of DEGs (log2 fold change >1.5) by RNA-seq analysis in
   the isolated macrophages between FGF21KO-MASH mice and WT-CD controls.
   (B) Heatmap of the signature genes for monocytes and KCs by RNA-seq
   analysis between FGF21KO-MASH mice and WT-CD controls. (C) Running
   Enrichment Score of transcripts of genes of KCs/macrophages associated
   with fatty acid (FA) metabolism. (D) Heatmap of enzymes for FA
   metabolism by qPCR. Protein levels of FASN and CD36 by Western blotting
   in: (E) the WT/FGF21KO mice with WSHFD/CD diets, (F) the FGF21KO-MASH
   mice with rhFGF21 treatment compared to vehicle controls and WT-CD
   controls, and (G) FL83B/FL83B-FGF21KD cells treated with palmitic acid
   (PA), while 1% bovine serum albumin (BSA) was used as treatment
   control. (H) Protein levels of HSL and phosphorylated HSL at S563 and
   S660 by Western blotting in the WT/FGF21KO mice with WSHFD/CD diets.
   (I) Heatmap of long-chain FAs (LCFAs) detected by Waters ACQUITY UPLC
   Systems coupled with Waters Xevo TQ-S micro triple quadrupole mass
   spectrometer in the WT/FGF21KO mice with WSHFD/CD diets. 21KO, FGF21KO.
   *P < 0.05, **P < 0.01, and ***P < 0.001.

Up-regulated ALOX15 signaling in KCs from FGF21KO-MASH mice

   In the WSHFD-MASH model, significantly increased ALOX15 protein level
   was found in FGF21KO-MASH mice in comparison with other three groups
   ([91]Fig. 3A). The ALOX15 expression was almost undetected in the
   isolated hepatocytes but abundantly expressed in the isolated
   macrophages from liver ([92]Fig. 3B). Consistently, positive ALOX15
   staining was distributed in the regions where KCs were located (fig.
   S4A) and coexpression of F4/80 and ALOX15 was confirmed by
   immunofluorescent staining in the isolated macrophages ([93]Fig. 3C).
   To determine whether the oxidized FA could induce KC death, an in vitro
   study was performed using the isolated mouse KCs from WT and FGF21KO
   mice. As expected, 13(S)-HODE challenging caused significant increases
   of cleaved caspase-3 (C-casp3) and FAS, while treatment with
   N-acetyl-l-cysteine (NAC), a commonly used antioxidant against reactive
   oxygen species, attenuated levels of C-casp3 and FAS ([94]Fig. 3D). In
   FGF21KO-MASH mice, pronounced cell death was also found, evident as
   significantly increased apoptotic cells (fig. S4B), while the increased
   C-casp3 and FAS in FGF21KO-MASH mice were attenuated by restoration of
   FGF21 ([95]Fig. 3E). We next asked whether MoMFs could be mobilized for
   KC replacement. The study of thioglycollate medium–induced peritoneal
   macrophages showed much high mobilization ability of MoMFs and
   up-regulated monocyte chemoattractant protein-1/C-C motif chemokine
   ligand 2 (MCP-1/CCL2), the key chemokine to regulate migration and
   infiltration of monocytes/macrophages), from the FGF21KO mice (versus
   WT) (fig. S4, C and D). All these results suggested that overexpression
   of ALOX15 caused increases of oxidized FAs, which induced excessive KC
   death, while mobilization ability of MoMFs was increased in FGF21KO
   mice. As the ALOX15-converted oxidized FAs could accelerate MASH
   progression ([96]35), we wondered whether knockout of ALOX15 would
   prevent MASH development/progression. ALOX15KO mice were therefore
   chosen to establish MASH models by feeding WSHFD or HFMCD. As expected,
   lack of ALOX15 protected the MASH insult, evidenced by decreased NAS,
   which was counted by the decrease of inflammation other than steatosis
   or ballooning ([97]Fig. 3F), confirming that ALOX15 signaling
   contributed to inflammation for MASH progression. To investigate
   whether ALOX15 signaling could affect the balance of KCs and MoMFs in
   the KC pool, cytometry by time-of-flight (CyTOF) was performed to
   precisely analyze the KC and MoMF subpopulations. KCs were defined as
   F4/80^hiTim4^+CX3CR1^− cells, while MoMFs were defined as
   CD11b^+Tim4^−CX3CR1^+ cells. In the WSHFD-MASH model, significantly
   decreased KCs but significantly increased MoMFs were found in both
   WT-MASH mice and FGF21KO-MASH mice compared to WT-CD control mice.
   Treatment with either PD146176 (an inhibitor of ALox15) or rhFGF21
   significantly attenuated the imbalance of KCs/MoMFs in KC pool
   ([98]Fig. 3G), suggesting that inhibition of ALox15 or restoration of
   FGF21 could preserve homeostasis of KC pool against MASH insulant. To
   further understand that lack of FGF21 could render inflammation/immune
   microenvironment contributing to MASH progression, RNA-seq data were
   analyzed to associate the metabolic pathway with immune response. The
   Kyoto Encyclopedia of Genes and Genomes heatmap of genes in metabolic
   pathway showed great changes in FGF21KO-MASH mice compared with WT-CD
   mice (fig. S4E), while up-regulated genes of immune response signaling
   and cytokine-cytokine interaction signaling were found in consistent
   with the up-regulated genes in metabolic pathway ([99]Fig. 3H).
   Together, FGF21 was essential against ALOX15-converted oxidized FAs to
   keep homeostasis of KC pool, while lack of FGF21 could render
   inflammation/immune microenvironment to accelerate MASH-HCC transition.

Fig. 3. Up-regulated ALOX15 signaling in KCs from FGF21KO-MASH mice.

   [100]Fig. 3.
   [101]Open in a new tab

   (A) Protein levels of Alox15 by Western blotting in the liver tissues
   of WT/FGF21KO mice with WSHFD/CD diets. (B) Alox15 mRNA levels by qPCR
   analysis in the isolated hepatocytes and KCs and the protein levels of
   Alox15 by Western blotting in KCs from the WT/FGF21KO mice with
   WSHFD/CD diets. (C) Immunofluorescent staining using the antibodies of
   anti-ALOX15 and anti-F4/80 to detect the positive ALOX15 macrophages.
   Green, ALOX15-positive staining cells; red, F4/80-positive staining
   cells; blue, 4′,6-diamidino-2-phenylindole (DAPI) staining to detect
   the nuclei as a counterstain. (D) Protein levels of FAS and cleaved
   caspase-3 by Western blotting in the isolated mouse KCs challenged with
   13-HODE and treated with N-acetyl-l-cysteine (NAC). (E) Protein levels
   of FAS and cleaved caspase-3 by Western blotting in the liver tissues
   from FGF21KO-MASH mice with rhFGF21 treatment compared to vehicle
   controls or WT-CD controls. (F) Histology-based NAFLD active score
   (NAS) in the WT/ALOX15KO mice with HFMCD/WSHFD diets. (G) CyTOF
   analysis in the isolated macrophages from mice to detect the KCs
   defined as F4/80^hiTim4^+CX3CR1^− cells and the MoMFs defined as
   CD11b^+ Tim4^−CX3CR1^+ cells in KC pool. The FGF21KO-MASH mice were
   treated daily with PD146176 or rhFGF21 for 4 weeks. (H) Analysis of the
   RNA-seq data for the mRNA fold changes of immune response signaling
   pathway and cytokine-cytokine interaction signaling. 21KO, FGF21KO. *P
   < 0.05 and **P < 0.01.

Lack of FGF21 accelerates MASH-HCC transition

   To validate the role of FGF21 against MASH-HCC transition, four
   MASH-HCC models were established in WT and FGF21KO mice with
   diethylnitrosamine (DEN) treatment at 2 weeks old, following with (i)
   control diet (CD), (ii) HFMCD, (iii) WSHFD, and (iv) HFD ([102]Fig.
   4A). On the basis of the appearance of tumor nodule by ultrasound (fig.
   S5A), the mice were euthanized at 16 weeks for DEN + HFMCD, 26 weeks
   for DEN + WSHFD, 28 weeks for DEN + HFD, and 36 weeks for DEN + CD
   ([103]Fig. 4A). Histological examination by H&E staining, Oil Red O,
   and Sirius Red staining confirmed the MASH, which was characterized as
   steatohepatitis (wildly distributed lipid drops, infiltration of
   inflammatory cells in the acinar zone, and hepatocyte ballooning) and
   HCC (showed the cytological features of cancerous cells ranging from
   well to poorly differentiated distributed in parenchyma showing an
   abnormal hepatic architecture) (fig. S5, B and C). The MASH-HCC in
   FGF21KO mice showed an aggressive tumor growth pattern, evident as the
   significantly increased number and maximal diameter of tumor nodule
   compared to the MASH-HCC in WT mice ([104]Fig. 4B). NAS system was
   applied in combination of histological and serological findings to
   further evaluate the characteristics of MASH-HCC. Significantly
   increased NAS was found in FGF21KO-MASH-HCC mice (versus WT-MASH-HCC
   mice) induced by either DEN + HFMCD or DEN + WSHFD ([105]Fig. 4C).
   Consistently, significantly increased levels of alanine transaminase
   (ALT) and hepatic/serum TG were found in the FGF21KO-MASH-HCC mice in
   comparison with WT-MASH-HCC (fig. S5D). Restoration of FGF21 was
   performed in the FGF21KO mice with rhFGF21 injection (125 μg/kg body
   weight) daily for 4 weeks in both DEN + HFMCD model and DEN + WSHFD
   model. Although tumor nodules were detected in the mice with rhFGF21
   treatment, the tumor number was smaller and tumor size was lower, with
   statistical significance compared to the vehicle controls ([106]Fig.
   4D). Then, we compare the KC/MoMF subpopulations between MASH mice and
   MASH-HCC mice. By CyTOF analysis, significantly decreased KCs but
   significantly increased MoMFs were found in both MASH mice and MASH-HCC
   mice compared to the CD control mice. No statistical significance was
   found between MASH mice and MASH-HCC mice; however, decreased KCs but
   increased MoMFs were found in the FGF21KO-MASH mice (versus WT-MASH
   mice) and FGF21KO-MASH-HCC mice (versus WT-MASH-HCC mice) with
   statistical significance ([107]Fig. 4E), suggesting that lack of FGF21
   worsened the KCs’ depletion and the KC pool could be seeded by MoMFs
   during MASH-HCC transition. To characterize the specificity of FA
   metabolites in MASH-HCC mice, LCFAs were analyzed. Consistently,
   markedly increased ALOX15-converting productions such as HODE and
   arachidonic acid, a precursor of bioactive lipid mediators
   (prostaglandin production), were found in the liver tissues from
   FGF21KO-MASH-HCC mice ([108]Fig. 4F). Significantly increased protein
   levels of ALOX15 and the key lipolytic enzymes, cytosolic phospholipase
   A 2 (cPLA2), which catalyzes the hydrolysis of phospholipids to
   arachidonic acid (cleave and release FAs), and monoacylglycerol lipase
   (MAGL), which hydrolyzes intracellular TG to FAs and glycerol, were
   found in the FGF21KO-MASH-HCC mice ([109]Fig. 4G), suggesting the
   important role of bioactive FA mediators in contribution to the
   MASH-HCC transition. To investigate the role of FGF21 on FAs release,
   FFA levels were determined in three MASH-HCC models (DEN + CD, DEN +
   HFMCD, and DEN + WSHFD). Significant increases of FFA in the serum were
   found in FGF21KO mice (versus WT mice) in all three MASH-HCC models
   ([110]Fig. 4H), suggesting that lack of FGF21 caused increase of FFA
   release during MASH-HCC transition. Previous studies revealed that
   cPLA2 was regulated by sphingolipids ([111]36). Sphingosine-1-phosphate
   (S1P), one of the most active sphingolipids, had been reported to
   promote liver fibrosis and cancer ([112]37). Thus, S1P levels in both
   serum and liver tissue were investigated. As expected, significantly
   increased S1P levels were noticed in FGF21KO mice (versus WT mice) in
   all three MASH-HCC models ([113]Fig. 4I). Together, these results
   indicated that lack of FGF21 accelerated MASH-HCC transition, while
   abnormal FA metabolism and increased S1P were noticed in the FGF21KO
   mice during MASH-HCC transition.

Fig. 4. Lack of FGF21 accelerates MASH-HCC transition.

   [114]Fig. 4.
   [115]Open in a new tab

   (A) Schematic diagram for establishing MASH-HCC models and the gross
   anatomy of tumor mass in four MASH-HCC models with CD, HFMCD, WSHFD,
   and HFD, respectively. (B) Maximal diameter of tumor nodules and number
   of tumor nodules in four MASH-HCC models. (C) Histology-based NAS in
   the MASH-HCC mice with HFMCD and WSHFD diets. (D) Schematic diagram for
   rhFGF21 treatment in the MASH-HCC mice with HFMCD and WSHFD diets. (E)
   CyTOF analysis of the KCs (F4/80^hiTim4^+CX3CR1^− cells) and MoMFs
   (CD11b^+Tim4^−CX3CR1^+ cells) from the MASH mice and MASH-HCC mice, as
   well as the control mice. (F) LCFAs in the FGF21KO-MASH-HCC mice
   compared to normal WT controls. (G) Protein levels of cPLA2, Alox15,
   and MAGL by Western blotting in the liver tissues of FGF21KO-MASH-HCC
   mice compared to normal WT controls. (H) The serum FFA levels in
   FGF21KO-HCC mice in comparison with the WT-HCC mice, with three diets
   (CD, HFMCD, and WSHFD). (I) S1P levels in serum and liver tissue from
   FGF21KO-HCC mice in comparison with the WT-HCC mice, with three diets
   (CD, HFMCD, and WSHFD). 21KO, FGF21KO. *P < 0.05, **P < 0.01, and ***P
   < 0.001.

S1P-S1PR2-YAP signaling mediated MSAH-HCC transition in FGF21KO mice

   Given that the FFA and S1P levels were increased in FGF21KO-MASH-HCC
   mice, we next asked whether lack of FGF21 would activate the S1P
   signaling to promote the HCC initiation ([116]37). We established
   premalignant models (preHCC) in FGF21KO mice and WT littermate, with
   DEN-HFMCD for 14 weeks or with DEN-WSHFD for 24 weeks. In preHCC mice,
   no tumor was detected but the preHCC mice were characterized by
   increased plasma ALT activity, increased α-fetoprotein level, and
   elevated hepatic TG content (fig. S6A). The premalignant models
   recapitulated the tumor microenvironment for MASH-HCC transition; thus,
   we could get insight into the metabolic features of FA/S1P and
   carcinogenetic signaling. A qPCR assay was performed to profile gene
   expression for the enzyme cascade of S1P synthesis. Fourfold changes
   (FGF21KO-preHCC mice versus WT-preHCC mice) of up-regulated gene
   expression were found, including the enzymes for FFA uptake [CD36 and
   FABP (1, 3, and 4)] and FA de novo synthesis (ACC1 and FASN), S1P
   cascade synthesis [SPTLC1, CERS (1, 5, and 6), DES2, and GCS (ceramide
   de novo pathway), ASMase, and NSMase (sphingomyelin pathway)],
   sphingosine synthesis [SGPP1, ACER (1–3), and ASAH1], and S1P synthesis
   (SPKK1) ([117]Fig. 5A). This result suggested the oversupplied FFA and
   the substrates for S1P cascade synthesis in FGF21KO mice during
   MASH-HCC transition. Consistently, significant increase of
   phosphorylated SPHK1 protein along with the increased S1P level was
   found in liver tissues from FGF21KO-preHCC mice, while treatment with
   rhFGF21 (125 μg/kg body weight, daily for 4 weeks) significantly
   attenuated the increased S1P level in liver tissues from FGF21KO-preHCC
   mice ([118]Fig. 5B). Previously, we reported that YAP had a negative
   impact on the survival of patients with HCC ([119]6). In the
   FGF21KO-preHCC mice, decrease of phosphorylated YAP ([120]Fig. 5C) and
   increased expression of YAP-targeting genes (Ctgf and Cyr61) (fig. S6B)
   were found, suggesting that lack of FGF21 could initiate the
   SPHK1/S1P-YAP signaling. To address this point, in vitro studies were
   performed. Challenged with PA, significantly increased S1P levels
   ([121]Fig. 5D) and phosphorylated SPHK1 ([122]Fig. 5E) were found in
   both FL83B-FGF21KD cells and Hepal-6-FGF21KD cells. Treatment with
   rhFGF21 attenuated the increased levels of S1P and phosphorylated SPHK1
   ([123]Fig. 5E). To validate that phosphorylated YAP was induced by S1P,
   the Hepal-6 and two human HCC cell lines, Huh7 and Hep3B, were treated
   with S1P. Time/dose-dependent effects were found to be associated with
   YAP phosphorylation (fig. S6C), while S1P-induced YAP nuclear
   translocation in Hepal-6 cells was detected by fluorescent staining
   (fig. S6D). The SPHK1/S1P-YAP signaling transduction was further
   studied. Inhibition of SPHK1 by SKI-II attenuated the decreased levels
   of phosphorylated YAP in Hepal-6 cells with PA challenging ([124]Fig.
   5F), while inhibition of the S1P receptor 2 (S1PR2) by JTE013
   attenuated the decreased levels of phosphorylated YAP in the Hepal-6
   cells with S1P challenging ([125]Fig. 5G). We reasoned that such a
   S1P-S1PR2-YAP signal would be potential to favor the MASH-HCC
   transition in FGF21KO mice. To test this hypothesis, the expression of
   S1P receptors was determined, and the significantly increased mRNA and
   protein levels of S1PR2 were found in the FGF21KO-DEN-WSHFD mice (fig.
   S6, E and F). This result encouraged us to further investigate
   S1P-S1PR2-YAP signaling in contribution to tumor growth in vitro and in
   vivo. Inhibition of S1PR2 by JTE013 or inhibition of YAP by verteporfin
   significantly prevented the colony growth in the Hepal-6 cells with S1P
   challenging ([126]Fig. 5H), while significantly suppressing tumor
   growth was consistent with the increased phosphorylation of YAP in the
   tumor tissues from in MASH-HCC mice (DEN-WSHFD model) with JTE013
   treatment ([127]Fig. 5, I to J). Together, all these results suggested
   a negative feedback of FGF21 on the S1P-S1PR2-YAP signaling. The next
   question is whether S1P signaling could affect the macrophages to rend
   an immunocompromised microenvironment in liver.

Fig. 5. S1P-S1PR2-YAP signaling mediated MSAH-HCC transition in FGF21KO mice.

   [128]Fig. 5.
   [129]Open in a new tab

   (A) Schematic diagram of S1P biosynthetic cascade enzymes with the fold
   changes (FGF21KO-preHCC versus WT-preHCC mice). ND, no detection; NC,
   no change; red, >2-fold up-regulated; blue, <2-fold down-regulated. (B)
   The S1P levels and protein levels of phosphorylated SPHK1 in liver
   issues from WT/FGF21KO-preHCC mice compared to that from WT/FGF21KO
   mice. (C) The protein levels of phosphorylated YAP and YAP in
   WT/FGF21KO-preHCC mice. (D) S1P content in FL83B/FL83B-FGF21KD cells
   and Hepal-6/Hepal-6-FGF21KD cells challenged with PA and treated with
   rhFGF21. (E) The protein levels of phosphorylated SPHK1 in
   FL83B/FL83B-FGF21KD cells and Hepal-6/Hepal-6-FGF21KD cells challenged
   with PA and treated with rhFGF21. (F) The protein levels of
   phosphorylated YAP in Hepal-6/Hepal-6-FGF21KD cells challenged with PA
   or PA + S1P and treated with SKI-II, an inhibitor of SPHK1. (G) The
   protein levels of phosphorylated YAP in Hepal-6/Hepal-6-FGF21KD cells
   challenged with S1P and treated with inhibitors of S1P receptors
   (JTE013 for S1PR2 and W146 for S1PR1). (H) Colony-forming assay to
   detect the number of cell colonies in Hepal-6 cells challenged with S1P
   and treated with JTE013 or verteporfin, an inhibitor of YAP. (I)
   Schematic diagram for JTE013 treatment in the MASH-HCC mice with WSHFD
   diets. (J) The protein levels of phosphorylated YAP in MASH-HCC mice
   with JTE013 treatment compared to vehicle controls. 21KO, FGF21KO. *P <
   0.05, **P < 0.01, and ***P < 0.001.

Tumor-associated macrophage polarization mediated NSAH-HCC transition in
FGF21KO mice

   As an important component of TIME, KCs are believed to be essential to
   control the MASH insults. However, it is not fully elucidated how the
   TIME contributed to MSAH-HCC transition via impaired macrophages
   ([130]38). To provide insight into the transcriptome of macrophages/KCs
   during MASH-HCC transition, the DEN-WSHFD–induced MASH-HCC,
   WSHFD-induced MASH, and normal controls were selected to perform
   RNA-seq in the isolated KCs/MoMFs. The top 1000 highly variant genes
   are shown in heatmap (fig. S7A). Defined as false discovery rate < 0.05
   and fold change ≥ 2, the volcano plots showed 1826 DEGs (MASH versus
   MASH-HCC) (fig. S7B). Standardized by normal controls, there are 1296
   overlapping DEGs between 2773 DEGs (MASH versus controls) and 2468 DEGs
   (MASH-HCC versus controls). Among the 1296 overlapping DEGs, 680
   up-regulated, 584 down-regulated, and 32 DEGs reversed ([131]Fig. 6A),
   suggesting that these shared genes could be critical for comprised
   function of macrophages. By pathway enrichment analysis (MASH versus
   MASH-HCC), peroxisome proliferator–activated receptor (PPAR) signaling
   pathway and inflammation/cytokine pathway were noticed in the top 15
   pathways ([132]Fig. 6B). It is well known that all PPAR family members
   are sensitive to FAs, while PPARβ/δ is not only expressed in
   macrophages but also presented in liver ([133]39). Further analysis
   indicated that PPARβ/δ signaling was up-regulated in MASH-HCC compared
   to MASH ([134]Fig. 6C). IL-10 was found as one of the top five
   up-regulated cytokines in the inflammation/cytokine pathway ([135]Fig.
   6D). Although the role of IL-10 in carcinogenesis is currently
   controversial, the macrophage-produced IL-10 has been reported to
   importantly contribute to TIME, in particular, macrophage polarization
   into tumor-associated macrophages (TAMs) ([136]40). In the tumor tissue
   from FGF21KO-MASH-HCC mice, M2 polarization was noticed by
   significantly increased F4/80^+CD206^+ cells by fluorescent staining
   ([137]Fig. 6E), while TAM phenotype was demonstrated by qPCR analysis
   ([138]Fig. 6F). As the mutual interactions between tumor cells and
   stromal microenvironment could contribute to phenotypically
   polarization of TAMs, BMDMs of mouse were therefore isolated to
   demonstrate whether cross-talk between macrophages and tumor cells
   could be mediated by PPARβ/δ and IL-10 signaling. Treated with the
   culture medium of Hepal-6 cells (HCM) challenged with S1P, the M0 BMDMs
   switched their expression profile, showing TAM phenotype (M2d), along
   with significantly increased protein expressions of PPARβ/δ and IL-10
   ([139]Fig. 6G). To elucidate the mechanism of S1P in regulating the TAM
   polarization, we investigated the expressions of IL-4 and
   phospho–signal transducers and activators of transcription 6 (p-STAT6),
   which were reported as an important pathway to activate the
   PPAR-δ/IL-10 signaling for macrophage M2 polarization by FA ([140]41).
   Treatment with HCM-S1P induced p-STAT6 translocation to nucleus (fig.
   S7C), while the increased protein levels of IL-4 and p-STAT6 were
   significantly attenuated by inhibition of SPHK1 ([141]Fig. 6H),
   suggesting the potential mechanism of S1P mediated TAM polarization by
   PPAR-δ/IL-10 activation via IL-4/STAT6 pathway. To determine whether
   the oversupply of FA could contribute to the TIME in term of TAM
   polarization, the BMDMs were treated with HCM, HCM + PA, and HCM + S1P.
   Significantly increased lipid accumulation by Oil Red O staining and
   significantly increased S1P level by ELISA were detected in the BMDMs
   with HCM + PA treatment and HCM + S1P treatment (fig. S7, D and E).
   Both HCM + PA treatment and HCM + S1P treatment induced phenotypically
   polarization of TAMs in BMDMs, while use of sulfosuccinimidyl oleate
   (SSO), an inhibitor of CD36 to prevent FA uptake, attenuated the
   polarization of TAMs (fig. S7F). Together, up-regulation of PPAR-δ and
   IL-10 in KCs/MoMFs of MASH-HCC mice could be the potential signaling
   pathway of TAM polarization. S1P induced PPAR-δ/IL-10 activation via
   IL-4/STAT6 pathway to mediate TAM polarization. Oversupply of FA
   contributed, at least partly, to the increase of S1P production to
   render TIME for MSAH-HCC transition.

Fig. 6. TAM polarization mediated NSAH-HCC transition in FGF21KO mice.

   [142]Fig. 6.
   [143]Open in a new tab

   (A) The volcano plots for the profile of DEGs in isolated
   macrophages/KCs from the liver tissues (MASH versus controls and
   MASH-HCC versus control) and the overlapping DEGs. (B) Fifteen enriched
   pathways were identified in MASH and MASH-HCC. (C and D) Up-regulated
   PPARδ in the lipid metabolism pathway and up-regulated IL10 in the
   inflammation/cytokine pathway were identified. (E) Immunofluorescent
   staining using the antibodies of anti-F4/80 and anti-CD206 to detect
   the M2 polarization of macrophages in FGF21KO-MASH-HCC mice,
   FGF21KO-MASH, and WT control mice. Red, F4/80-positive cells; green,
   CD206-positive cells; blue, positive DAPI staining to detect the nuclei
   as a counterstain. (F) qPCR detection for the markers of M2/TAM
   polarization in the isolated macrophages/KCs from FGF21KO-MASH-HCC
   mice, FGF21KO-MASH mice, and WT control mice. (G) Schematic diagram of
   mouse BMDMs treated with HCM, HCM + S1P, and Dulbecco’s modified
   Eagle’s medium (DMEM). qPCR detection for the markers of M2/TAM
   polarization in BMDMs induced by HCM and HCM + S1P. The protein levels
   of PPARβ/δ and IL-10 in BMDMs with treatments of HCM and HCM + S1P. (H)
   The mRNA fold changes of IL-4 by qPCR and the protein levels of IL-4
   and phosphorylated STAT6 in BMDMs treated with HCM + S1P and HCM + S1P
   + SKI-II. BMDMs, bone marrow–derived macrophages. HCM, culture medium
   of Hepal-6 cells. *P < 0.05, **P < 0.01, and ***P < 0.001.

DISCUSSION

   Liver is the major target of lipotoxicity. Hepatic lipotoxicity
   resulting from an excessive FFA influx has been accepted as a crucial
   cellular event clearly involved in the progression from MASH to HCC
   ([144]42, [145]43). As the largest population of tissue-resident
   macrophages, KCs have been reported to be essential for controlling the
   inflammatory process in MASH liver ([146]44); however, accumulation of
   “harmful lipids” has a long-lasting impact on KCs and, in turn, alters
   the liver immune microenvironment contributing to MASH-HCC transition.
   FGF21, as a liver safeguard ([147]45), has been widely reported to
   alleviate the hepatic fat stress ([148]46–[149]48), but the action of
   FGF21 on macrophage heterogeneity in particular for their distinct
   ontogeny (embryonic KCs versus BMDMs/MoMFs) is largely unknown. In this
   study, we report that lack of FGF21KO renders liver immunosuppressive
   microenvironment via compromised macrophages during MASH-HCC
   transition. The major signaling components mediated alteration of KC
   homeostasis and carcinogenetic transition are shown as a schematic
   diagram in [150]Fig. 7.

Fig. 7. Schematic diagram of working hypothesis.

   [151]Fig. 7.
   [152]Open in a new tab

   The major signaling components mediated alteration of KC homeostasis
   and carcinogenetic transition polarization mediated NSAH-HCC transition
   being studied. HFMCD, high-fat methionine/choline-deficient diet;
   WSHFD, Western-style diet (high-fat and high-fructose); M1, macrophage
   M1 polarization; M2, macrophage M2 polarization; TAM, tumor-associated
   macrophage.

   Previously, TLR-4 signaling and the essential role of KCs in the
   development of MASH were reported ([153]49). KCs could promote hepatic
   steatosis via IL-1β–dependent suppression of PPAR-α activity ([154]50).
   Recruitment of KCs via CC-chemokine receptor 2 contributed to the
   progression of MASH by recruiting BMDMs ([155]51). KCs are known to
   self-renew locally to maintain themselves in the long term, while
   steatohepatitis may cause exhaustion of KC pool, resulting in
   mobilization of MoMFs for KC replacement ([156]12). However, the impact
   of FAs on KC during MASH-HCC transition remains undefined. Here, we
   asked first whether the lipid metabolites such as oxidized FAs in MASH
   liver could be the leading cause of KC death to affect the homeostasis
   of KC pool. We found that lack of FGF21 causes severe aberrant FA
   signaling, leading to not only liberation of FAs from adipose tissue
   but also increased FA de novo synthesis. Up-regulating Alox15
   expression in KCs, other than hepatocytes, resulted in the increases of
   HODE and HETE to induce KC death. Increase of CD11b^+ Tim4^−CX3CR1^+
   cells in FGF21KO-MASH liver indicated the replacement of KC pool by
   MoMFs. ALOX15 signaling affecting KC homeostasis for the MASH
   development was further supported by following data: (i) NAS was
   significantly decreased in ALOX15KO mice and (i) inhibition of ALox15
   significantly attenuated the increased MoMFs in MASH liver. Supported
   by multiple lines of evidence, our findings indicated that aberrant FA
   metabolism via KC-ALOX15 signaling played a causal role for
   mobilization of MoMFs to replenish KC pool in FGF21KO-MASH liver. Of
   note, FGF21 may directly inhibit the ALOX15 activity to save KCs.
   Further study is needed to address this important issue.

   We next investigated the cross-talk between macrophages and tumor
   cells, in particular the S1P-mediated cross-talking before and during
   MASH-HCC transition. Although existing data indicate the carcinogenetic
   role of S1P/SPHK signaling in HCC, it is largely unknown how the
   S1P-mediated macrophage polarization contributes to the TIME for
   MASH-HCC transition. Here, we found that oversupplied FAs and the
   substrates for S1P cascade synthesis and activated SPHK1/S1P-YAP
   signaling in the preHCC-FGF21KO liver. RNA-seq analysis in the isolated
   KCs/MoMFs indicated that about 50% shared genes (MASH versus MASH-HCC),
   which could be critical for comprised function of macrophages. On the
   basis of pathway enrichment analysis, up-regulation of PPARβ/δ and
   IL-10 was noticed. PPAR-δ/IL-10 activation via IL-4/STAT6 pathway was
   identified as the key signals to mediate the cross-talking between
   tumor cells and macrophages. Up-regulation of PPAR-δ and IL-10 in
   macrophages/KCs by S1P induction could be potential mechanism of TAM
   polarization, and this was supported by the following: (i) M2d
   polarization was found in the BMDMs treated with HCM + S1P via
   activation PPAR-δ/IL-10 signaling, and (ii) blockage of FA supply
   attenuated S1P production thereby prevented M2d polarization of BMDMs.
   All these findings demonstrated that S1P production mediated not only
   the YAP signaling to initiate carcinogenetic cells but also TAM
   polarization, contributing to the TIME for MASH-HCC transition.

   In conclusion, FGF21 is essential against ALOX15-converted oxidized FAs
   and thereby preserve homeostasis of KC pool in liver. Lack of FGF21
   accelerates MASH-HCC transition via the SPHK1/S1P-S1PR2-YAP signaling.
   Compromised macrophages presented as TAM phenotype contribute to TIME
   for MASH-HCC transition.

MATERIALS AND METHODS

Human studies

   To study FGF21 expression (by immunohistochemistry, qPCR, and Western
   blotting) and cell subpopulations in KC pool (by Hyperion Imaging), the
   human liver tissue samples were collected from 19 patients with
   MESH-HCC along with corresponding adjacent benign tissues. To study the
   serological levels of FGF21 (by ELISA), the human serum samples were
   collected from 20 patients with MESH-HCC tissue who had undergone HCC
   nodule resection and were followed up to 9 months. The liver tissue
   samples and serum samples were from the patients who were recruited
   between 2010 and 2020 from the James Graham Brown Cancer Center
   Bio-Repository at the University of Louisville. A microscope
   examination of the cellular composition of hepatic tissue confirmed the
   diagnoses of MASH-HCC and benign on these liver tissues, reviewed by
   two pathologists independently, blinded to the subject’s clinical
   history. All patients and their legally authorized representatives had
   given informed consent for the release of information and use of
   available tissue samples. The human sample collection procedures at the
   University of Louisville for this study were approved by the
   Institutional Review Board for Human Study of the University of
   Louisville (IRB no. 2.496). Gene Expression Profiling Interactive
   Analysis 2.0, an online tool, was used for survival analyses based on
   TCGA database. In the TCGA database, 156 overweight (BMI ≥ 25 kg/m^2)
   patients with HCC were graded as G1–G4 according to the National
   Comprehensive Cancer Network severity scale. The gene expression higher
   than median was defined as high expression group, while the gene
   expression lower than median was defined as low expression group. Using
   a log-rank test for hypothesis evaluation, survival rate of patients
   graded as G1–F2 and G3–G4 were analyzed on the basis of high/low gene
   expression levels of FGF21. To explore the correlation between
   survivals and six immune cells (B cells, CD4^+ T cells, CD8^+ T cells,
   neutrophils, macrophages, and dendritic cells), the split percentage of
   patients is set as 50% of the patients with high/low level of cell
   subpopulations to generate the Kaplan-Meier curve. CD68 was further
   normalized by FGF21 expression to estimate the survival probability
   over time. The Pearson’s correlation analysis shows that the expression
   levels of CD68 (x axis) and FGF21 (y axis) are negative correlation, as
   indicated by the correlation coefficient of −0.109 and P value of 3.52
   × 10^−2. The result was analyzed by TIMER2.0 database.

Animals and in vivo studies

   Male FGF21 KO (FGF21KO) mice with C57 BL/6J background were granted by
   S. Kliewer (University of Texas Southwestern Medical Center). C57BL/6J
   (strain no. 000664) WT mice and 12/15-LOX (Alox15) KO mice (strain no.
   002778) were purchased from Jackson Laboratories (Bar Harbor, ME). All
   animals were housed four per cage, given rodent chow and tap water ad
   libitum, and maintained at 22°C and on a 12-hour light/dark cycle. The
   WT C57BL/6J mice were crossed to FGF21KO or Alox15KO mice to generate
   heterozygous F1 offspring. The F1 heterozygous mice were crossed to
   generate homozygous FGF21^−/− mice, homozygous Alox15^−/− mice, and WT
   littermates, which were used for establishment of animal models. Two
   MASH model were established by feeding (i) HFMCD (A06071302, l-amino
   acid diet with 60% kcal fat, 0.1% methionine, and no added choline,
   Research Diet, Inc., NJ) or (ii) WSHFD (D16030909, 60% kcal fat and 19%
   kcal fructose, Research Diets Inc., NJ). At 3 months, the mice were
   euthanized for advanced MASH models according to our previous report
   ([157]52). Three MASH-HCC models were established by administration of
   DEN in mice aged 2 weeks at 40 mg/kg body weight (i.p.) and fed with
   the following: (i) HFMCD, (ii) WSHFD, and (iii) high-fat diet (D12492,
   60% kcal% fat, Research Diets Inc., NJ). The diet (D12450B, 10% kcal%
   fat, Research Diets, Inc. NJ) was used as CD. Ultrasound was performed
   to monitor the tumor nodules to determine HCC growth “windows” in all
   three MASH-HCC models. Treatments in animals were as follows:
   recombinant human FGF21 (rhFGF21) (no. 100-42; PeproTech) was injected
   subcutaneously every day at 125 μg/kg body weight. JTE013 (no. 1866,
   Axon Medchem) was injected intraperitoneally every day at 4 mg/kg body
   weight. PD146176 (no. NC1641198, Cayman Chemical) was injected
   subcutaneously every day at 5 mg/kg body weight. All animal procedures
   were approved by the Institutional Animal Care and Use Committee of the
   University of Louisville, which is certified by the American
   Association for Accreditation of Laboratory Animal Care (IACUC no.
   19528).

Cell lines and in vitro studies

   The cell lines for in vitro studies include a mouse hepatocyte line
   FL83B (ATCC CRL-2390) and a mouse hepatoma cell line Hepal-6 (ATCC
   CRL-1830), as well as two human HCC cell lines, Huh7 (GBTC-099H) and
   Hep3B (ATCC HB-8064). FL83B cells were cultured in the F12K medium
   (ATCC), Hepal-6, Huh7, and Hep3B cells in Dulbecco’s modified Eagle’s
   medium, with 10 to 15% fetal bovine serum. Establishment of
   Hepa1–6–FGF21KD cell and FL83B-FGF21KD cell and isolation of
   macrophages are described in the Supplementary Materials. Palmitate
   (PA) media (no. P9767, Sigma-Aldrich) was made by dissolving 2% bovine
   serum albumin (no. A1311, US Biologicals) in cell culture medium, and
   the 100 μM PA working solution was prepared from a high concentration
   (20 mM) stock PA solution made by dH[2]O heated to 70°C. The in vitro
   treatments were the following: rhFGF21 (no. 100-42; PeproTech) at 100
   ng/ml, 13-HODE (no. 38600, Cayman Chemical) at 1 μM, NAC (no. A9165,
   Sigma-Aldrich) at 50 mM, S1P (no. 567727, Sigma-Aldrich) at 10 μM,
   SKI-II (no. S5696, Sigma-Aldrich) at 15 μM, W146 (no. 10009109, Cayman
   Chemical) at 10 μM, JTE013 (no. 1866, Axon Medchem) at 10 μM, and SSO
   at 50 μM to treat cells. All the antibodies and primers are listed in
   the Supplementary Materials.

Statistical analysis

   Statistical analysis and graphics were performed by using GraphPad
   Prism 7.00 software (San Diego, California). Statistical significance
   was determined by analysis of variance (ANOVA). The post hoc Tukey’s
   test was used for analysis of any differences between groups. The
   collected data from experiments were presented as mean ±SD. For in
   vitro studies, experiments were performed with a minimum of triplicate
   samples and triplicate repetition of experiments. Group difference was
   considered significant for P < 0.05 (*), P < 0.01 (**), and P < 0.001
   (***).

Acknowledgments