Targeting ferroptosis alleviates methionine-choline deficient (MCD)-diet induced NASH by suppressing liver lipotoxicity
1 | INTRODUC TION
NAFLD is a liver disorder characterized by liver abnormalities ranging from hepatic steatosis to steatohepatitis and fibrosis. While early-stage steatosis is generally considered benign, the later stage NASH is more aggressive, characterized by the presence of hepatocyte cell death and development of fibrosis, cirrhosis and HCC.1 Epidemiological studies have indicated that NASH affects 1.5%-6.45% of the general popu- lation.2,3 NASH is associated with poor prognosis, as 30% of NASH patients have advanced bridging fibrosis or cirrhosis2,4 that ultimately leads to liver failure and even HCC. The underlying mechanism in- volved in NASH progression is largely unknown, and there is currently no effective pharmacological therapy against NASH.5 A ‘two-hit’ theory was proposed several years ago, which suggested that in the setting of steatosis, a second ‘hit’ (eg oxidant stress) was required for the development of NASH.6 In the recent years this theory has been updated, although debates on pathogenesis of NASH still exists.7 In addition, the mechanism of hepatotoxicity in NASH remains unclear. Recent studies have demonstrated that inhibiting apoptosis, necrop- tosis or pyroptosis only partially improves liver injury in NASH,8-10 suggesting that other forms of cell death likely contribute to NASH associated hepatotoxicity.
Ferroptosis is a newly discovered form of regulated cell death characterized by iron dependent accumulation of lipid ROS, which is mor- phologically, biochemically and genetically distinct from other forms of cell death.11 Studies have linked ferroptosis to multiple cellular pro- cesses like iron homeostasis, redox homeostasis, lipid metabolism and glutaminolysis.12 Two long-chain PUFAs, AA and AdA, are prone to be oxidized to generate lipid peroxide when esterified into PE, thus representing the major inducers of ferroptosis.13 On the other hand, massive occurrence of ferroptosis is prevented by GPX4,14 an enzyme in mammalian cells that specifically eliminate lipid ROS or FSP1,15 which could reduce CoQ to RTA. Accordingly, Fer-1 and Lip-1, which are small molecule RTAs, are found to be effective inhibitors of fer- roptosis because of their ability of neutralizing lipid ROS. In the past few years, accumulating evidences have suggested that ferroptosis is implicated in several pathologies, such as tissue ischemia-reperfusion injury, neurodegeneration, stroke and cancer.16-18 It is important to understand whether ferroptosis, like other forms of cell death, is in- volved in other physiological and pathological processes.
The progression of NASH is accompanied by impaired lipid homeostasis and massive lipid accumulation in liver, which generates lipotoxic species, hence it could be speculated that ferroptosis may participate in the pathology of NASH. To test this hypothesis, we fed mice with MCD-diet to induce NASH. MCD-diet impairs hepatic production and secretion of VLDL and causes massive lipid accumulation in liver, which is considered to mimic the progression of steatosis and NASH.19 We performed RNA-seq in liver from MCD-diet mice to identify the potential mechanisms of NASH progression. We found that AA meta- bolic pathway was upregulated in mouse liver, suggesting a role of fer- roptosis in this process. Indeed, inhibition of ferroptosis significantly improved MCD-diet induced steatosis as well as subsequent patholog- ical indexes, including liver injury, inflammation and fibrosis.
Since NAFLD is characterized by aberrant accumulation of LDs in hepatocytes, the precise causes and roles of LDs in this process are still unclear.20 ROS has been reported to promote LDs accumulation following mitochondrial dysfunction in Drosophila glia cells, which promotes neurodegenerative disease.21 In other studies, however, LDs formation induced by hypoxia and ROS in a neural stem cell niche seemed to inhibit the oxidation of polyunsaturated fatty acids and protect glia and neuroblasts from stress.22 These observations led us to investigate the relationship between lipid ROS generated during ferroptosis and LDs accumulation in NASH mouse model. In fact, elim- ination of lipid ROS by ferroptosis inhibitors suppressed the produc- tion and accumulation of LDs in the liver cells. These results suggest that ferroptosis may represent a novel target for NASH treatment.
2 | METHODS
2.1 | Chemicals, plasmid and materials
Erastin (HY-15763), Fer-1(HY-100579), Lip-1(HY-12726), DFO(HY-B0568), α-tocopherol (HY-16686), troglitazone (HY-50935), atglistatin (HY-15859) and HSL-IN-2 (HY-102056) were purchased from MedChemExpress. Full length cDNA of GPX4 was amplified by PCR and cloned into indicated pLVX vector. William’s medium E (without L-methionine and choline chloride, WMP03) was purchased from caisson. All chemicals were purchased commercially.
2.2 | Animals and treatments
Male C57BL/6 J mice were purchased from SLAC, housed under pathogen-free conditions. Mice were placed on MCD-diet (A02082002B, Research Diets) or control diets for 2 to 4 weeks and were randomly assigned to vehicle or drug treatment groups. Mice in drug treatment group received intraperitoneal injection of Fer-1 or Lip-1 (dissolved in PBS) at 5 mg per kg body weight per day during MCD-diet feeding and were weighted daily. All animal-related experimental procedures were performed in ac- cordance with the National Institutes of Health guidelines and approved by the Laboratory Animal Ethical Committee of Fudan University.
2.3 | Tissue preparation
After treatment, mice were deeply anaesthetized with sodium pentobarbital (50 mg/kg). Just before sacrifice, blood was with- drawn directly from the heart, and the serum was separated and stored at −80°C. Liver tissues were rapidly harvested, weighed and then fixed in 4% paraformaldehyde in PBS or snap-frozen in liquid nitrogen and stored at −80°C for later assessment.
2.4 | RNA-sequencing
Whole-genome gene expression analysis was performed using the liver tissue from MCD and ND-diet mice (n = 3 per group) at 3 weeks. The total RNA was extracted using Trizol, and cDNA samples were sequenced using Illumina HiSeq 4000 (2 × 150bp read length) by Shanghai Majorbio Pharmaceutical Technology Co. Ltd. The reference Mus musculus ge- nome and gene information were downloaded from the National Center for Biotechnology Information database. All the differentially expressed genes were used for heat map analysis and KEGG ontology enrichment analyses. For KEGG enrichment analysis, a P < .05 was used as the thresh- old to determine statistical significant enrichment of the gene sets. 2.5 | Real-time polymerase chain reaction Total RNA from mouse livers was extracted using Trizol, and re- verse transcribed using TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotech, China). The resulting cDNAs were used for PCR using the TB Green® Premix Ex Taq™ (Takara). Relative abundance of mRNA was determined on a 7500 real-time PCR system (Thermo Fisher Scientific). Fold changes of RNA levels were calculated using the ΔΔCt method and were analysed by Student's t test. The primer pair sequences used for RT-PCR reactions for in vivo studies are listed in Table S1. 2.6 | Biochemical analysis Liver triglycerides were assayed by using a triglyceride assay kit (GPO-POD; Applygen Technologies Inc) according to the manu- facturer's recommended protocol. Serum ALT and AST levels were determined using commercial enzymatic kits (Roche) by automatic biochemical analyzer (Roche cobas C702). 2.7 | Transmission electron microscopy Livers were fixed with 2.5% glutaraldehyde in 0.1 mol/L of Sorenson's buffer (0.1 mol/L H2PO4, 0.1 mol/L HPO4 [pH 7.2]) for 4 hours, washed 3X with PBS and then treated with 1% OsO4 in 0.1 mol/L Sorenson's buffer for 2 hours at room temperature. After dehydration through an ethanol series and acetone (100%), livers were embedded in acetone and embed-812 (SPI). Thin sections were cut on an ultramicrotome (Leica), and the sections were stained with 2% uranyl acetate and lead citrate. Images were obtained using Hitachi TEM at 7000-fold magnification. 2.8 | Fluorescence staining Liver sections were fixed with 4% paraformaldehyde for 10 minutes and then permeabilized with 0.2% Triton X-100 in PBS for 10 min- utes. After washing in PBS for 5 minutes, slides were incubated with C11-BODIPY (Invitrogen™ D3861, 5 μmol/L) for 30 minutes at 37°C. After washing with PBS, DAPI was used to stain cell nuclei. Cells seeded on glass coverslips placed in 12-well plates at a den- sity of 1 × 105 were treated as indicated in specific experiments. After treatment, cells were fixed, permeabilized and incubated with C11-BODIPY, Lipid TOX (Invitrogen™ H34476) or FerroOrange for 30 minutes at 37°C, washed with PBS and stained with DAPI. Images were obtained on an Olympus BX61VS Microscope or a Leica TCS SP8 Microscope. ImageJ software was used for quantification and fluorescence signal analysis. 2.9 | PI staining PI (sigma P4170) was dissolved in PBS at a final concentration of 25 μg/mL. Necrotic hepatocytes are detected by injecting PI solution into tail vein. The livers were harvested, and frozen after sectioning. The samples were counterstained with DAPI. Images were captured on an Olympus BX61VS Microscope or a Leica TCS SP8 Microscope. 2.10 | HE, Sirius Red and Oil Red O staining Samples from the right lobe of all mouse livers were fixed in 4% para- formaldehyde for at least 24 hours at room temperature, embedded in paraffin and sectioned into 8 μm tissues. Slices were stained with Mayer's HE (Baso) as well as sirius red (Sigma) to assess fibrosis. For Oil Red O staining, the sections were incubated in 60% isopropanol, stained with Oil Red O staining solution (Sigma), and re-incubated in 60% isopro- panol. Then, the sections were washed with running water, and stained with HE after rinsing with PBS. Finally, the sections were washed with running water and mounted with buffer glycerin. Images were captured from three randomly selected fields using Olympus BX61VS. 2.11 | NAS scoring NAS score was calculated by a sum of scores of steatosis, lobular in- flammation and hepatocyte ballooning in HE stained liver sections. NAS score ranged from 0 to 8, which was calculated by a sum of scores of steatosis (0-3), lobular inflammation (0-3) and hepatocyte balloon- ing (0-2) according to criteria set forth by a well-validated grading sys- tem, NAS score of ≥5 strongly correlated with a diagnosis of ‘define NASH’. Samples were blindly scored by two experienced pathologists. 2.12 | Lentiviral shRNA production and infection ShRNA sequences were derived from shRNA bacterial glycerol stocks from Sigma Aldrich catalog. To generate SK-HEP-1 cells with GPX4 knockdown, lentivirus was produced by transfecting pLKO-vector or pLKO-GPX4 together with psPAX2 and pMD2G packaging plasmids into HEK293T cells. Lentiviral supernatant was harvested 36 hours after transfection, cleared by a 0.45-μmol/L filter, which was then used to infect SK-HEP-1 cells after supplementing with polybrene (8 μg/mL).Stable cell pools were selected with puromycin (1 μg/mL, Amresco) for 3 days. 2.13 | Cell culture, viability assay and transfection Human liver cancer cell line SK-HEP-1 was purchased from the ATCC and cultured in Dulbecco's modified Eagle's medium (Gibco) supple- mented with 10% fetal bovine serum (Gibco) and 50 μg/mL of peni- cillin/streptomycin. For cell viability assay, cells with the indicated treatments were collected for trypan blue staining. Cell viability was counted and calculated by using an automated cell counter (Count Star, IC 1000). Cell viability was reported as a percentage relative to the negative control treatment. Cells were transfected using Lipofectamine 2000 following manufacturer's protocol (Invitrogen). 2.14 | Iron measurement Iron and Fe2+ concentrations in serum and tissue were measured using the Iron Assay Kit (MAK025-1KT, Sigma) in accordance with the manufacturer's instructions. In brief, total iron and Fe2+ were measured in 10 µL of serum or 10 mg of tissue homogenate by measuring the absorption at 593 nm and comparing the value with a standard curve of known concentrations. 2.15 | Statistical analysis Results were analysed using GraphPad Prism 7.0. Statistical analy- ses were performed using two-tailed unpaired Student's t-test or one-way analysis of variance (ANOVA) with Bonferroni's test or Dunnett's test as indicated in corresponding figure legends. 3 | RESULTS 3.1 | MCD-diet induces upregulation of arachidonic acid metabolism in liver To explore the mechanisms underlying NASH, we fed C57/BL6 J mice with normal or MCD-diet for up to three weeks, and performed RNA-seq analysis of liver samples (Figure 1A). Mice fed on MCD-diet can develop steatohepatitis that is histologically similar to liver dis- ease observed in humans with NASH as early as two weeks.23,24 As expected, ALT and AST, two serum markers of hepatocyte death and liver injury, were significantly elevated by MCD-diet (Figure 1B). A total of 1877 genes showed differential expression between normal and MCD-diet livers (P < .05 and fold change >2) (Table S2), among which 1236 genes were upregulated and 641 genes were downregu- lated in the latter (Figure 1C).
We performed KEGG pathway enrichment analysis based on the RNA-seq data. Interestingly, the pathway of AA metabolism showed a significant change on the basis of affected genes and statistical significance, which was followed by the pathways of osteoclast dif- ferentiation, cytokine-cytokine receptor interaction and retinol me- tabolism (Figure 1D,E). We verified the RNA-seq results by RT-PCR and confirmed that MCD-diet led to a remarkable upregulation of genes involved in AA metabolism, such as Lpcat2 and Pla2g4a, which are responsible for incorporating AA into membrane and catalysing the hydrolysis of membrane phospholipids to release AA respec- tively.25 In addition, the cytochrome P450 family enzymes that par- ticipate in AA cytochrome metabolism, Cyp2a22, Cyp2b13, Cyp2b9, Cyp2a4, Cyp4f18, Cyp2c55, Ace and Cyp4a31, were also upregulated in MCD-diet mice (Figure 1F). These results suggest that AA me- tabolism is significantly affected during MCD-diet induced NASH progression in mice.
3.2 | Ferroptosis occurs in MCD-diet fed mouse livers
Earlier studies have suggested that the pathological progression of NASH is governed by multiple types of cell death, and oxidative stress induced by the abnormal accumulation of lipid is considered to be an important initiation factor of this pathogenesis.26 Interestingly, AA derived lipid peroxidation products have been claimed as the proximate executioners of ferroptosis,12 a recently defined new form of non-apoptotic cell death. This prompted us to hypothesize that AA mediated ferroptosis participates in the progression of NASH.
Consistent with previous studies,27 MCD-diet led to body weight loss during 4-week feeding period, but did not affect liver/ body weight ratio of mice (Figures S1A,B). Notably, accumulation of lipid ROS products, which were detected by specific fluorescent probe C11-BODIPY (581/591), was observed after two weeks of MCD-diet feeding, and further increased during prolonged feed- ing (Figure 2A,B). In contrast, no lipid ROS accumulation was seen in control mouse livers. To further elucidate the role of ferroptosis in this process, we treated mice with ferroptosis inhibitor Fer-1 by daily injection. Administration of Fer-1 had no additional influence on body weight or liver/body ratio of mice in both normal and MCD- diet conditions (Figures S1A,B). The general condition of the animals remained good and their behaviour appeared normal throughout the experimental period, indicating the lack of adverse effects of Fer-1 on mice. Consistent with its role as RTA, Fer-1 treatment resulted in remarkable decrease of lipid ROS signals at three weeks of MCD- diet feeding (Figure 2C,D). We also verified this phenomenon by in- jecting mice with Lip-1, another RTA and ferroptosis inhibitor, and observed similar effect as Fer-1 treatment (Figure 2C,D). These data support the notion that MCD-diet induces lipid ROS accumulation and ferroptosis in mouse livers.
To further detect ferroptosis in MCD-fed mouse liver, we used TEM to analyse the mitochondria morphology in hepatocytes. Morphological change of mitochondria is one of the features of ferroptotic cells.11 Shrunk mitochondria with increased membrane density were observed in MCD-diet mouse livers (Figure 2E). Consistent with the effect in reducing lipid ROS, Fer-1 treatment re- markably ameliorated MCD-diet induced mitochondrial abnormality (Figure 2E). These data provide convincing evidence of the occur- rence of MCD-diet induced ferroptosis in mouse liver.
We also tested the expression of important genes involved in ferroptosis process, Hmox1, Acsl4, Alox5ap, Gpx4, Ptgs1, Ptgs2 and Nox2. All these genes were upregulated in MCD-diet mouse livers (Figure 2F). However, Fer-1 treatment did not reverse the mRNA expression level of these genes (Figure 2F), in accordance with its role in directly scavenging lipid hydroperoxides. Finally, we exam- ined ferroptotic cell death in vivo by PI staining,28 which showed increased cell death in MCD-diet mouse livers was suppressed by Fer-1 treatment (Figure 2G). Collectively, we conclude that ferropto- sis participates in MCD-diet induced NASH in mice.
3.3 | Iron is accumulated in the liver and serum of MCD-diet induced mice
Previous studies have reported local iron overload in MCD-diet model of rats29 and disrupted iron homeostasis in NAFLD pa- tients.30 Consistently, we observed increased level of ferrous ion in mouse livers fed on MCD-diet as indicated by FerroOrange staining (Figure 3A). Treatment with DFO, a commonly used iron chelator, eliminates the fluorescent signal, confirming that the signal detected was indeed ferrous iron (Figure 3A). Next, we measured ferrous ions and total iron in mouse livers and in the serum. A slight but signifi- cant increase in both ferrous and total iron levels was observed in MCD-fed mouse livers and in the serum. These results were not af- fected by Fer-1 treatment as Fer-1 is a direct radical-trapping reagent (Figure 3B).
We also measured the expression of key regulators of iron me- tabolism in our mouse model. Hepatic Bmp6 gene expression was upregulated in MCD-fed mice livers, which is consistent with the previous study.31 MCD diet also induced upregulation of hepatic genes, Hamp1 and Hamp2. In addition, we also observed increased expression of hepatic genes involved in iron transport and storage that includes Trf1, Fth, Ftl and Fpn, which encode transferrin receptor 1, ferritin H, ferritin L and ferroportin respectively. Notably, Fer-1 did not have much effect on the expression of these genes (Figure 3C). As Fe2+ is an essential factor for ferroptosis, increased Fe2+ level may cooperate with increased AA metabolism to generate lipid peroxide and to promote NASH pathology. Collectively, these data show iron overload in the liver of MCD-diet induced mice.
3.4 | Ferroptosis inhibitors reduce MCD-diet induced liver injury, inflammation and fibrogenesis
We then set out to investigate the role of ferroptosis on pathological progression of MCD-diet induced NASH in mice. It is notable that
MCD-diet induced serum ALT and AST elevation was reversed by Fer-1 treatment (Figure 4A,B), indicating that suppression of ferrop- tosis alleviates mouse liver injury by MCD-diet.
To analyse the progression of steatohepatitis in mice, we ex- amined the degree of liver fibrosis and inflammation. Mice fed on MCD-diet for 3 weeks had greater area of sirius red-stained fibrils demonstrated by liver morphometry, while a significant reduc- tion in hepatic fibrosis was observed in mice treated with Fer-1 (Figure 4C). We further examined the expression of liver inflamma- tory genes (Tnfα, Mcp-1) and fibrogenic genes (Tgfβ, Collagen α1). As shown in Figure 4D, expression of these genes were upregulated in MCD-diet mice compared to those fed on normal diet (Figure 4D). Fer-1 treatment reduced the expression of these genes (Figure 4D), further demonstrating reduced liver injury by ferroptosis inhibition. Consistently, Lip-1 treatment resulted in similar effect on reducing MCD-diet induced ALT and AST elevation (Figure 4E) as well as in- flammatory and fibrogenic gene expression (Figure 4F). Hence, we conclude that ferroptosis participates in pathological progression of MCD-diet induced NASH in mice.
3.5 | Ferroptosis inhibitors attenuate MCD-diet liver steatosis
The above results suggest that targeting ferroptosis may reduce liver injury in NASH. In fact, liver under NASH is featured by dra- matic lipid accumulation, namely steatosis. However, previous stud- ies have reported a more complicated relationship between liver injury and steatosis. For instance, it has been shown that inhibiting triglyceride synthesis improves hepatic steatosis but exacerbates liver damage and fibrosis,28 which indicates that improvement in liver function is not necessarily accompanied by reduction in stea- tosis. Thus, it is intriguing to elucidate how suppression of ferrop- tosis in NASH model may affect steatosis in addition to reducing liver injury. We observed that mouse liver tissue appeared pale and yellow after feeding with MCD-diet for 3-4 weeks, indicating lipid accumulation in the liver. Fer-1 treatment restored the slimy- smooth, dark-red exterior of healthy liver (Figure 5A), possibly through removal of excessive lipid. Histological analysis showed accumulation of LDs in mouse livers under 2-weeks of MCD-diet, which was exaggerated during prolonged feeding (Figure 5B). Similar to its effect on liver injury, Fer-1 treatment largely reduced MCD-diet induced LDs accumulation (Figure 5B). We then used NAS score to evaluate the degree of NASH in mouse livers. Mice fed on MCD-diet for 3 weeks (with NAS score>5) developed to the grade of NASH while Fer-1 treatment substantially improved the grade of steatohepatitis by reducing NAS score to <4 in MCD-diet fed mouse livers (Figure 5C). To demonstrate lipid accumulation more clearly, we used Oil Red O to stain LDs in the liver. As reported recently, MCD-diet resulted in generation of giant LDs in mouse livers,32 which was significantly suppressed by Fer-1 treatment (Figure 5D). To elucidate whether re- duced LD size is correlated with a decrease in total amount of neu- tral lipid, we measured the content of hepatic triglyceride, one of the major component of LDs. Indeed, Fer-1 treatment also lowered triglycerides elevation induced by MCD-diet (Figure 5E). Moreover, Lip-1 treatment resulted in similar improvement as Fer-1 on gross liver morphology as well as TG accumulation in MCD-diet mouse liv- ers (Figure 5F,G). Taken together, our data indicate that inhibition of ferroptosis attenuates MCD-diet induced hepatic steatosis through both suppressing giant LD formation and reducing neutral lipid level. 3.6 | Suppression of ferroptosis related lipid ROS ameliorates LDs accumulation in cell culture model of hepatocytes ROS has been reported to promote LDs accumulation, which in- trigued us to look into the relation between lipid ROS generation and LDs accumulation in NASH model. SK-HEP-1, a human liver can- cer cell line, was treated with erastin, which induces lipid ROS and ferroptosis by inhibiting the cystine transporter system Xc-. Erastin induced ferroptosis in SK-HEP-1 cells was confirmed by PI staining, which could be suppressed by Fer-1 and Lip-1 (Figures S2A,B). Upon erastin treatment, LDs increase was accompanied by lipid ROS accu- mulation. Similarly, erastin induced LDs accumulation was reduced by treatment with Fer-1 or DFO, an iron chelator that could effi- ciently inhibit ferroptosis (Figure 6A,B). Vitamin E and troglitazone are used as medications for NASH that reduces oxidative stress and act as PPARγ agonist respectively.33 Interestingly, troglitazone is also known to inhibit lipid peroxidation. We observed that α-tocophrol (a liposoluble form of vitamin E) and troglitazone efficiently sup- pressed erastin induced death in SK-HEP-1 cells (Figures S3A,B), implying that targeting of ferroptosis may be a mechanism for these clinically used medications in NASH therapy. Then we mimicked the liver cells of MCD-diet mice by culturing SK-HEP-1 cells in MCDM.34 We observed mild LDs accumulation as well as lipid ROS elevation in cells cultured in MCDM for 12 hours, which was reversed by Fer-1 treatment (Figure 6C,D). GPX4 is the only enzyme in mammalian that could eliminate lipid ROS using re- duced glutathione as substrate. To better investigate the relationship between lipid ROS and LDs accumulation, we knocked down GPX4 in SK-HEP-1 cells using shRNAs against GPX4 (Figure 6E). Since GPX4 is an essential gene for cell survival,35 Fer-1 was supplemented in the medium to keep cell viability during normal cell culture. To induce lipid ROS accumulation, control or shGPX4 cells were cultured in MCDM for 48 hours without Fer-1 supplementation. Accumulation of lipid ROS was observed in control cells, which was exaggerated by GPX4 knockdown (Figure 6E,F). Interestingly, LDs accumulation showed consistent change with lipid ROS, thus providing a new evidence of LDs accumulation induced by lipid ROS (Figure 6E,F). Furthermore, we also transiently overexpressed GFP-tagged GPX4 in SK-HEP-1 cells (Figure S4) and found that MCDM induced LD accumulation was significantly suppressed by GPX4 overexpression compared to adjacent cells without GPX4 overexpression (Figure 6G,H). Taken to- gether, these data reinforce the notion that lipid ROS accumulation is related to LDs, and that GPX4 may play an important role to LDs homeostasis in NASH progression. 3.7 | Lipid ROS induces LDs accumulation by promoting lipogenesis LDs are highly dynamic organelles, which are created de novo from neutral lipids synthesis, fuse together to form bigger LDs and are degraded via lipolysis.36 Since increasing lipid ROS increases LDs ac- cumulation and inhibiting lipid ROS accumulation decreases the level of LDs, we assessed how lipid ROS affects LDs dynamics. Important genes involved in LDs dynamics were tested by RT-PCR in liver tis- sues. Interestingly, expression of lipogenesis genes, Srebp1c, Elovl6 and Lipin1, were upregulated in MCD-diet fed mouse liver, which could be reversed by Fer-1 treatment (Figure 7A). CIDE family proteins (CIDEA, CIDEB and CIDEC) are essential factors in mediating LDs fusion. It is notable that expression of Cidec is dramatically increased in MCD- diet fed mouse liver, suggestive of giant LDs formation.37 However, Fer-1 treatment did not reverse its upregulation (Figure 7B). We also observed upregulation of coat proteins38 Plin2 and Plin5 in MCD-fed mice that could be reversed by Fer-1 treatment (Figure 7B), which cor- related with LDs level in MCD-fed and Fer-1 treated mice. In contrast, lipolysis genes, Pnpla2 (atgl), Hsl and Abhd5, did not show consistent correlation in MCD-diet fed mice and Fer-1 treated mice (Figure 7C). In order to further clarify the role of lipid ROS in lipogenesis process, we treated SK-HEP-1 cells with lipolysis inhibitors (ATGL- targeting atglistatin39 and HSL-targeting HSL-IN-240). Treatment of cells with both lipolysis inhibitors resulted in lipid accumulation to a level similar to erastin treatment. Notably, co-treatment with eras- tin further increased LDs accumulation, indicating erastin promotes LDs level upstream of lipolysis (Figure 7D,E). Consistently, additional Fer-1 treatment reduced LDs to the level similar to lipolysis inhibi- tors-only treatment (Figure 7D,E). Collectively, these results indicate that lipid ROS generation during ferroptosis may target the process of lipogenesis, but not lipolysis. 4 | DISCUSSION During NASH progression, caspase-dependent apoptosis has long been considered as the major form of cell death in hepatocytes,41 al- though necrotic and pyroptotic cell death have also been implicated in NASH.10,42 A recent study reported that in the choline-deficient, ethionine-supplemented diet induced NASH, which triggers severe inflammation and cell death in mouse liver within two days, ferrop- tosis was suggested as a trigger for this acute pathological process.28 In our study, we first uncovered ferroptotic phenotypes in MCD- diet induced NASH, including lipid ROS accumulation, morphologi- cal change of mitochondria, and upregulation of ferroptotic genes (Figure 2). Notably, suppression of ferroptosis efficiently alleviated pathological injury as well as steatosis in liver during a prolonged lipotoxic stress, demonstrating its key role in chronic progression of NASH. Interestingly, we first discovered that AA metabolism was one of the most upregulated processes during MCD feeding (Figure 1), which is consistent with the role of AA in promoting ferroptosis, and it indicates that an altered lipid metabolism may be an important initiator in NASH. This observation is supported by previous findings that knocking out of Pla2g4a or Alox, two genes involved in AA metabolism, alleviates steatosis and inflammation in obese mice models,43,44 emphasizing the importance of AA pathways in NASH. Since hepatocytes play key roles in iron metabolism through iron storage and recycling, we examined the role of iron in NASH progression and found that the liver and serum iron levels are both increased in MCD-diet induced mice (Figure 3), which is consistent with previous studies.31,45 In the iron overloaded mice liver induced by MCD-diet, the expression of hepcidin was found to be upregulated, likely because of iron-mediated Bmp6 upregulation, or because of inflammation since hepcidin is an acute-phase protein.46 It is known that TFR1 and ferritins are reciprocally regulated by IRPs at post-transcriptional level, and TFR1 is suppressed by iron overload.47 However, we observed that the transcriptional expression of Tfr1, Ftl and Fth, and Fpn is exclusively upregulated in MCD model (Figure 4D). It could be speculated that the transcriptional upregulation of these genes is likely because of increased inflammation and oxidative stress response in MCD livers,48 which is known to stimulate the expression of iron-metabolism genes.49-51 Taken together, these findings indicate that iron overload and AA metabolism are both important in promoting NASH progression. An intriguing observation in our study is that inhibition of fer- roptosis significantly reduced LDs in liver, while lipid ROS induced by erastin, MCDM or GPX4 knockdown triggered lipid accumula- tion in SK-HEP-1 cells. In fact, it has been reported that ROS pro- motes LDs accumulation, which in turn limits the level of ROS and inhibits the oxidation of PUFA, and thus play an antioxidant role in many different biological contexts.22 Although mild LDs formation is protective against lipotoxicity, massive LDs accumulation would result in inflammation and liver injury.20 It seems that impaired in- terplay between lipid ROS and LDs formation may be an important cause for NASH progression. We further demonstrated that lipid biosynthetic genes, Srebp1c, Elovl6 and Lipin1, are upregulated in MCD-diet mouse liver, and eliminating lipid ROS reverses their ex- pression level (Figure 7A-C). In addition, we showed that lipid ROS promotes LDs accumulation under condition of lipolysis suppres- sion (Figure 7D-F). Collectively, these results support a model that lipid ROS affects lipid dynamics by promoting lipogenesis. A key issue in unraveling the mechanism fueling NASH devel- opment is to identify the pro-inflammatory factors on the basis of steatosis.26,52 The accumulated inflammatory changes exacerbate liver injury and activate immune cells, which are the major sources of fibrogenesis.53 Hepatic fibrosis is the main prognostic deter- minant of liver-related mortality.54 Targeting steatosis, inflam- mation or fibrosis can alleviate NASH progression. Our results demonstrate that inhibition of ferroptosis significantly reduces inflammation and fibrogenesis related genes expression, Sirius Red-stained fibrils, and improved steatosis condition, indicating the potential therapeutic effects on deterring NASH progression. Currently, four main pathways are targeted in NASH therapy33: (a) hepatic fat accumulation and metabolic stress (eg PPAR agonists, bile acid-farnesoid X receptor axis, inhibitors of de novo lipogenesis), (b) oxidative stress and inflammation (eg antioxidants, TNF-α pathway and immune modulators), (c) targets in gut (eg gut microbiome modula- tors), (d) fibrotic process. Despite strategic progress, pharmacological effects of these existing therapies are still limited as the histological improvement of NASH has not exceeded 50% with the current ther- apies.33 Our work shed light on the pathophysiology of NASH and demonstrates ferroptosis as a novel mechanism leading to NASH re- lated liver pathology (Figure 7F). With the development of more ef- ficient ferroptosis inhibitors, targeting ferroptosis, especially the AA metabolism pathway, may be exploited as a viable and effective ther- apeutic strategy for both Erastin2 hepatic steatosis and steatohepatitis.