Introduction
Liver cirrhosis poses a substantial disease burden worldwide due to its increased probability of developing serious complications, such as massive variceal bleeding, refractory ascites, and hepatorenal syndrome.1 Liver cirrhosis is the end-stage of progressive fibrosis caused by persistent chronic liver injury. Chronic injuries lead to hepatocyte death, which subsequently initiates and exacerbates liver fibrosis.2 Unlike other types of cell death, such as apoptosis, necrosis, and autophagy, ferroptosis is an iron-dependent form of regulated cell death characterized by impaired mitochondrial morphology, iron overload, disturbed lipid peroxidation, and excessive oxidative stress.3 Ferroptosis has been implicated in several liver diseases, such as drug-induced liver injury, ischemia-reperfusion injury, chronic liver disease, and hepatocellular carcinoma.4
Cyclooxygenase (COX) is a key rate-limiting enzyme in the conversion of arachidonic acid (AA) to prostaglandins (PTGS), thromboxane, and prostacyclin.5 There are three isomers of COX: COX-1, COX-2, and COX-3.6 Among these isomers, COX-2 is widely believed to participate in inflammation and tumorigenesis.7,8 Furthermore, our previous studies have revealed that COX-2 is involved in and regulates the progression of liver fibrosis/cirrhosis via various mechanisms.9–11 However, whether COX-2 regulates ferroptosis to impact liver fibrosis has yet to be thoroughly investigated.
Many transcription factors, including nuclear factor erythroid 2-related factor 2 (Nrf2), regulate ferroptosis.12,13 Nrf2 is pivotal for regulating various pathophysiological processes, such as cellular redox status, detoxification, and anti-inflammatory responses.14 Considering that both Nrf2 and COX-2 regulate inflammation, Nrf2 may act as a link between COX-2 and ferroptosis in the context of chronic liver injury. Therefore, this study aimed to investigate whether COX-2 regulates ferroptosis via the Nrf2 signaling pathway during the progression of chronic liver injury.
Methods
Animal model
The animal experiments were approved by the Animal Use and Nursing Committee of West China Hospital of Sichuan University. Wild-type, EllaCre, and COX-2flox/flox C57BL/6 mice were purchased from GemPharmatech (Nanjing, China) for this study. All the mice were maintained under constant temperature and humidity with standard 12-h light-dark cycles and had free access to diet in the experimental animal center of Sichuan University (Chengdu, China). EllaCre mice were crossed with COX-2flox/flox mice to generate conventional COX-2 knockout mice (COX-2–/–) and littermate controls (COX-2+/+).15 Age-matched male mice (six to eight weeks old, weighting 20–25 g) were selected for further experiments. Thioacetamide (TAA), which induces lipid peroxidation and oxidative stress—two components of ferroptosis—was employed to establish a chronic liver injury model in this study.16 Mice were randomly allocated to different groups, with six mice in each group. COX-2+/+ mice in the control (CO) normal saline (NS) group and COX-2–/– mice in the knockout (KO) NS group were injected intraperitoneally with NS at a dosage of 0.2 mL/d every three days for eight weeks. COX-2+/+ mice in the CO TAA group and COX-2–/– mice in the KO TAA group were injected intraperitoneally with 4% thioacetamide (TAA, Sigma-Aldrich, USA) dissolved in NS at a dosage of 200 mg/kg every three days for eight weeks.15 Mice were anesthetized with sodium pentobarbital and sacrificed three days after the last intraperitoneal injection to harvest liver samples. Liver samples were fixed for light or transmission electron microscopy (TEM) evaluation or stored at −80°C for quantitative real-time polymerase chain reaction (qRT-PCR), Western blot, lipid metabolomics, and measurements of ferrous ions, hepcidin, glutathione (GSH), and malondialdehyde (MDA).
Isolation and treatments of primary mouse hepatocytes
Primary mouse hepatocytes were isolated from eight-to-twelve-week-old male C57BL/6 mice using the two-step collagenase perfusion method, as previously described.17 Briefly, hepatocytes were isolated by digestion with type II collagenase (17101015, Gibco, Grand Island, USA) and then separated by Percoll (P1644, Sigma-Aldrich, USA) for centrifugation. The isolated hepatocytes were cultured in serum-free medium (DMEM-F12, Gibco™) for 12 h, and then treated with 0.1% dimethyl sulfoxide (DMSO) (Solarbio, Beijing, China), erastin (20 µM, ferroptosis inducer, S7242, Selleck, Shanghai, China) + DMSO, etoricoxib (10 µM, selective COX-2 inhibitor with the best affinity to COX-2, HY-15321, MedChem Express, Shanghai, China) + erastin (20 µM) + DMSO, and tBHQ (50 µM, Nrf2 agonist, HY-100489, MedChem Express, Shanghai, China) + erastin (20 µM) + DMSO for 8 h. Erastin, etoricoxib, and tBHQ were dissolved in DMSO. After interventions, primary hepatocytes were fixed on coverslips for immunofluorescence staining, or scraped from culture dishes and fixed for TEM evaluation, or scraped and stored at −80°C for qRT-PCR, Western blot, and GSH and prostaglandin E2 (PGE2) measurements.
Light microscopy and TEM
Liver specimens from mice were fixed in 4% formalin, embedded in paraffin, and sectioned into 4-µm-thick slices. Hematoxylin and eosin (HE) and Sirius red staining were used for histological evaluation and assessment of collagen deposition, respectively.9 TEM (JEM1200EX, JEOL, Japan) was performed to evaluate changes in mitochondrial morphology following a standard protocol.18 In brief, liver tissues and primary mouse hepatocytes were collected, fixed in 4% glutaraldehyde (Solarbio, Beijing, China) for 2 h, and then fixed in 1% osmium tetroxide for 1 h at 4°C. The samples were dehydrated, embedded in epoxy resin, and sectioned into 70-nm-thick slices for microscopic examination. These slices were stained with 5% uranyl acetate and examined under a TEM. All slides were observed and evaluated by two technicians independently, who were blinded to the sample identification information. Their conclusions were compared, and if discrepancies were found, a third technician was included to make a final determination.
Immunofluorescence staining
Primary mouse hepatocytes were cultured on coverslips at the bottom of 24-well plates. After treatment, the samples were fixed with 4% paraformaldehyde, rinsed with phosphate buffer solution, and then permeabilized with 0.1% Triton-X-100.19 The samples were blocked with donkey serum and incubated with a primary antibody against mouse glutathione peroxidase 4 (GPX4) (dilution: 1:200, 67763-1-Ig, Proteintech, Wuhan, China) and secondary antibodies. The slides were coverslipped with anti-fading medium containing 4′, 6-diamidino-2-phenylindole and scanned using a fluorescence microscope.
Ferrous ion, hepcidin, GSH, MDA, and PGE2 measurements
Liver tissues were processed in NS to create tissue homogenates in an ice-water bath and then centrifuged. The supernatants were collected for the measurement of various indicators. Primary mouse hepatocytes were collected through ultrasonication with phosphate buffer solution in an ice-water bath after the intervention, and the resulting centrifugal supernatants were gathered. Concentrations of ferrous ion, hepcidin, GSH, MDA, and PGE2 were measured according to the manufacturer’s instructions (Jiancheng, Nanjing).20 Ferrous ion was determined using a colorimetric assay kit (A039-2-1, Jiancheng, Nanjing), where the centrifugal supernatants were mixed with chromogenic reagents for colorimetric detection at 520 nm using a spectrophotometer. Hepcidin and PGE2 were measured using enzyme-linked immunosorbent assay kits (Hepcidin: H252, Jiancheng, Nanjing; PGE2: H099-1-2, Jiancheng, Nanjing). Briefly, processed liver tissues and primary hepatocytes were added to enzyme pre-coated wells with antibodies, and recognition antigens were incubated for 30 m at 37°C to form immune complexes, which were then detected at 450 nm. GSH was measured with a spectrophotometric kit (A006-1-1, Jiancheng, Nanjing), quantitatively determined at 420 nm. MDA was measured using a thiobarbituric acid (TBA) assay kit (A003-1-2, Jiancheng, Nanjing), where TBA reagents were added to the centrifugal supernatants to generate the MDA-TBA complex, detected at 532 nm.
Western blot
Proteins were extracted from frozen liver tissues and primary mouse hepatocytes using a protein extraction kit (KGP2100, KeyGEN BioTECH, Jiangsu, China). Protein concentrations were measured using a bicinchoninic acid assay kit (P0010, Beyotime Biotechnology, Shanghai, China). Proteins (50 µg from liver tissues, 30 µg from hepatocytes) were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes. The membranes were blocked in 5% defatted milk in TBST and incubated with specific primary antibodies overnight at 4°C,10 including mouse GPX4 (dilution: 1:1,000, 67763-1-Ig, Proteintech, Wuhan, China) and rabbit NRF2 (dilution: 1:2,000, 16396-1-Ap, Proteintech, Wuhan, China). After binding with secondary antibodies (dilution: 1:20,000, ZB-2301, ZB-2305, Santa Cruz, USA), the target protein blots were visualized using an enhanced chemiluminescence detection kit (P0018AM, Beyotime Biotechnology, Shanghai, China) and exposed on a chemiluminescence imaging system (ChemiScope6100, Clinc, Shenzhen, China).
qRT-PCR
Total RNA was extracted from liver tissues and primary mouse hepatocytes using TRIzol (RE-03011, RE-03113, FOREGENE, Chengdu, China). RNA was reverse-transcribed into cDNA using a first-strand cDNA kit (K1691, Thermo Scientific, Shanghai, China). qRT-PCR was conducted using SYBR Green probes (Thermo Scientific, Shanghai, China).21 The sequences of primers used in this study are listed in Table 1. The results were normalized to β-actin using the 2−ΔΔCt method.
| Gene | Forword | Reverse |
|---|
| β-actin | 5′-AAGGCCAACCGTGAAAAGAT-3′ | 5′-GTGGTACGACCAGAGGCATAC-3′ |
| Gpx4 | 5′-GCAGGAGCCAGGAAGTAATC-3′ | 5′-GGCTGGACTTTCATCCATTT-3′ |
| Tfr | 5′-AAGCCAGATCAGCATTCTCT-3′ | 5′-CGGCATTTTCTTCT CATCT-3′ |
| Fpn | 5′-TTGCAGGAGTCATTGCTGCTA-3′ | 5′-TGGAGTTCTGCACACCATT-3′ |
| Ptgs2 | 5′-ATTCCAAACCAGCAGACTCATA-3′ | 5′-CTTGAGTTTGAAGTGGTAACCG-3′ |
| Slc7a11 | 5′-AACCAACAAGCAGGTGGTTGA-3′ | 5′-CCTTGTAGATGTCCACAGCCAGAGT-3′ |
Lipid metabolomics
Liver tissues were pretreated, and hydrophilic or hydrophobic metabolites were extracted. Liquid chromatography-tandem mass spectrometry (MS/MS) was used to detect lipid metabolites. Ultraperformance liquid chromatography (UPLC, ExionLC AD system, SCIEX, Shanghai, China) and tandem mass spectrometry (MS/MS, QTRAP® System, SCIEX, Shanghai, China) were applied for lipid information collection.22 Based on the MetWare database, secondary data were analyzed qualitatively and quantitatively according to the retention time, precursor ions, and product ions of the detected substances.
Statistical analysis
All quantitative variables were expressed as mean ± SD and analyzed using SPSS 22.0 software (IBM, Inc., Chicago, IL, USA). One-way analysis of variance was applied for comparisons among multiple groups. Graphs were generated using GraphPad Prism 8.0 (GraphPad, Inc., La Jolla, CA, USA). A two-tailed p-value of <0.05 was considered statistically significant.
Results
COX-2 deteriorated hepatocyte morphology under both TAA and erastin treatments
Under intraperitoneal TAA challenge, the liver structure of COX-2+/+ mice, as revealed by HE staining, underwent significant changes, characterized by inflammatory cell infiltration and fibrotic septa formation in the liver tissue. In contrast, liver inflammation and fibrosis were milder in TAA-treated COX-2–/– mice (Fig. 1A). The fibrotic area of the liver shrank by 63.1% in the KO TAA group compared to the CO TAA group (CO TAA vs. KO TAA: 6.5 ± 2.2 % vs. 2.4 ± 0.2 %, p = 0.002). These findings are consistent with our previous studies on the role of COX-2 in chronic liver injury.15
TEM revealed smaller and ruptured mitochondria, with fewer cristae in the livers of the CO TAA group compared to those in the CO NS group (Fig. 1A). Additionally, high-density substances were deposited in the hepatocyte mitochondria of the CO TAA group. These mitochondrial morphological changes are consistent with the characteristics of ferroptosis. However, the hepatocyte mitochondria of the KO TAA group did not show remarkable shifts compared to those in the CO TAA group. Furthermore, erastin induced ferroptosis in primary mouse hepatocytes, with smaller mitochondria and fewer cristae, whereas etoricoxib, a selective COX-2 inhibitor, partially reversed the mitochondrial abnormalities induced by erastin (Fig. 1B).
COX-2 affected iron metabolism in TAA-induced chronic liver injury
Given that COX-2 was involved in mitochondrial changes related to ferroptosis, iron metabolism was the next focus of this study. Under TAA challenge, ferrous ion levels were examined by a colorimetric assay, which revealed a two-fold increase from 0.22 ± 0.11 mg/gram protein in the CO NS group to 0.48 ± 0.12 mg/gram protein in the CO TAA group (p = 0.017) (Fig. 2A). In contrast, the ferrous ion level in the livers of the KO TAA group (0.18 ± 0.04 mg/gram protein) was only 37.5% of that in the CO TAA group (p = 0.004) (Fig. 2A). Several molecules are involved in maintaining iron homeostasis. Transferrin (Tfr) facilitates iron transport into cells, while ferroportin (Fpn) exports iron out of cells. Hepcidin interacts with and degrades Fpn to inhibit iron efflux.23 In our study, although hepcidin, Tfr, and Fpn did not show significant differences between the different phenotypes of COX-2, these biomolecules still tended to be affected by COX-2, thereby altering iron metabolism (Fig. 2B–D).
COX-2 modulated lipid peroxidation and oxidative stress in TAA-induced chronic liver injury and erastin-induced hepatocyte ferroptosis
Lipid peroxidation is pivotal for initiating ferroptosis, and polyunsaturated fatty acids are the most susceptible lipids in peroxidation, such as arachidonic acid (AA, C20:4), adrenic acid (AdA, C22:4), linoleic acid (Lin, C18:2), and C22:6.24 Furthermore, phosphatidylethanolamine, phosphatidylcholine, and phosphatidylserine have been previously reported as the most relevant phospholipids in ferroptosis.25 Liquid chromatography–MS/MS-based lipidomic data from our study revealed that free fatty acids (FFAs) in the liver differed between the CO TAA and KO TAA groups, while similar levels of phosphatidylcholine, phosphatidylethanolamine, and phosphatidylserine were detected in these two groups (Fig. 3A–D). Specifically, AA, AdA, Lin, and FFA (C22:6) all decreased in the livers of the KO TAA group by 36.1%, 39.4%, 30.1%, and 37.4%, respectively [CO TAA vs. KO TAA: AA (11.34 ± 1.93) × 105vs. (7.25 ± 1.31) × 105, p = 0.038; AdA (10.78 ± 1.64) ×106vs. (6.53 ± 0.66) × 106, p = 0.014; Lin (1.55 ± 0.08) × 108vs. (1.08 ± 0.09) × 108, p = 0.003; FFA (C22:6) (2.93 ± 0.17) × 107vs. (1.83 ± 0.16) × 107, p = 0.001; Fig. 3E–H].
As a downstream event of lipid oxidation, oxidative stress fuels ferroptosis and contributes to cell damage. GPX4, together with GSH, plays a pivotal role in countering oxidative stress during ferroptosis. MDA, a metabolite of lipid peroxidation, reflects the intensity of lipid peroxidation and oxidative stress. In the CO TAA group, GSH and MDA levels, as determined by spectrophotometry, were significantly depleted in the liver, while GSH and MDA levels were restored in the liver of the KO TAA group (CO TAA vs. KO TAA: GSH, 2.24 ± 0.24 µmol/g protein vs. 3.63 ± 0.16 µmol/g protein, p < 0.001; MDA, 0.62 ± 0.26 nmol/mg protein vs. 0.39 ± 0.04 nmol/mg protein, p = 0.015) (Fig. 4A, B). The loss of COX-2 downregulated GPX4 in the liver after TAA treatment (CO TAA vs. KO TAA: 0.47 ± 0.26 vs. 1.23 ± 0.19, p = 0.027) (Fig. 4C). Regarding primary mouse hepatocytes treated with erastin, Ptgs2 mRNA expression and PGE2 levels were higher than those in DMSO treatment (DMSO vs. erastin+DMSO: Ptgs2 mRNA, 1.00 ± 0.06 vs. 2.75 ± 0.36, p = 0.001; PGE2, 37.01 ± 5.12 ng/L vs. 55.40 ± 6.60 ng/L, p = 0.005), while etoricoxib reversed these effects (erastin+DMSO vs. etoricoxib+erastin+DMSO: Ptgs2 mRNA, 2.75 ± 0.36 vs. 1.91 ± 0.31, p = 0.037; PGE2, 55.40 ± 6.60 ng/L vs. 40.35 ± 6.49 ng/L, p = 0.017) (Fig. 4D, E). GSH decreased in the erastin+DMSO group, and etoricoxib reversed the GSH decrement induced by erastin (erastin+DMSO vs. etoricoxib+erastin+DMSO: 10.72 ± 0.91 µmol/g protein vs. 21.62 ± 4.02 µmol/g protein, p = 0.002) (Fig. 4F), which was even higher than that induced by DMSO treatment (13.46 ± 1.87 µmol/g protein). Etoricoxib also inhibited the production of MDA in primary mouse hepatocytes (erastin+DMSO vs. etoricoxib+erastin+DMSO: 1.05 ± 0.08 nmol/mg protein vs. 0.91 ± 0.08 nmol/mg protein, p = 0.045) (Fig. 4G), which was close to the results of the DMSO treatment (0.82 ± 0.09 nmol/mg protein). Slc7a11, a membrane transporter for GSH synthesis, was restored in the etoricoxib+erastin+DMSO group compared with the erastin+DMSO group (erastin+DMSO vs. etoricoxib+erastin+DMSO: 0.61 ± 0.07 vs. 3.13 ± 0.71, p = 0.004) (Fig. 4H). In addition, both GPX4 mRNA and protein levels were also recovered in the etoricoxib+erastin+DMSO group (erastin+DMSO vs. etoricoxib+erastin+DMSO: GPX4 mRNA, 0.65 ± 0.01 vs. 0.84 ± 0.06, p = 0.004; GPX4 protein, 0.45 ± 0.04 vs. 0.89 ± 0.24, p = 0.034) (Fig. 4I–K).
Inhibition of COX-2 upregulates the Nrf2 signaling pathway to attenuate ferroptosis
The aforementioned results suggested that COX-2 was implicated in ferroptosis. Since the Nrf2 signaling pathway is involved in antioxidation, it is important to explore whether this pathway plays a bridging role between COX-2 and ferroptosis. Data from primary mouse hepatocytes showed that the Nrf2 agonist tBHQ upregulated the Nrf2 signaling pathway to reverse the decline of GPX4 induced by erastin (erastin+DMSO vs. tBHQ+erastin+DMSO: Nrf2, 0.21 ± 0.02 vs. 0.55 ± 0.17, p = 0.026; GPX4, 0.61 ± 0.11 vs. 1.16 ± 0.25, p = 0.025) (Fig. 5A). These results also suggested that the inhibition of COX-2 by etoricoxib could upregulate the Nrf2 signaling pathway to ameliorate erastin-induced ferroptosis (erastin+DMSO vs. etoricoxib+erastin+DMSO: Nrf2 protein, 0.32 ± 0.08 vs. 0.85 ± 0.09, p = 0.002) (Fig. 5B).
Discussion
Ferroptosis, an alternative form of programmed cell death, is implicated in various disease processes, such as tissue injury, cancer, and neurodegenerative diseases.26,27 In the chronic liver injury mice of this study, damaged mitochondria in hepatocytes and increased iron concentrations, along with enhanced lipid peroxidation and oxidative stress, suggest that chronically injured livers undergo ferroptosis. Ferroptosis is thought to play a dual role in liver fibrosis. On the one hand, ferroptosis regulates the autophagy signaling pathway in hepatic stellate cells to alleviate liver fibrosis.28 On the other hand, ferroptosis promotes liver fibrosis in global and hepatocyte-specific gene-modified mice.12,29,30 As the liver is home to various types of cells, with hepatocytes accounting for the majority (∼80%),31 ferroptosis in the large population of hepatocytes would be potent enough to outweigh ferroptosis in hepatic stellate cells, thus determining the fate of the liver. Our study also revealed that TAA-treated mouse livers underwent both ferroptosis and fibrosis. Liver fibrosis was mitigated when ferroptosis was simultaneously inhibited by global knockout or inhibition of COX-2, supporting the notion that ferroptosis of hepatocytes is detrimental to the progression of chronic liver disease.
COX-2 is a versatile protein involved in the pathogenesis of deteriorating liver fibrosis and cirrhosis. COX-2 induces hepatocyte–cholangiocyte transdifferentiation,15 promotes hepatocyte apoptosis by inhibiting endoplasmic reticulum stress,9 and inhibits the epithelial–mesenchymal transition of hepatocytes.32 In this study, inhibition of COX-2 not only restored the damaged mitochondria in hepatocytes and reduced iron concentrations in TAA-treated liver tissue and erastin-treated primary mouse hepatocytes, but also decreased the levels of polyunsaturated fatty acids, including Lin, AA, AdA, and C22:6, and downregulated lipid peroxidation and oxidative stress. These findings indisputably suggest that COX-2 accelerates ferroptosis, influencing the progression of chronic liver disease. Other investigators have reported that COX-2 regulates ferroptosis in renal and cerebral conditions.33–35 These results share some similarities with our findings. In contrast, there is also a study suggesting that COX-2 is a downstream molecule of ferroptosis.36 Yang et al. reported that erastin- or (1S, 3R)-RSL3-induced ferroptotic cell death was not affected by indomethacin, and that PTGS2, the gene encoding COX-2, is a downstream marker of ferroptosis.36 However, our data revealed a different relationship between COX-2 and ferroptosis. Our findings indicated that inhibition of COX-2 activated the Nrf2 signaling pathway to relieve ferroptosis in the liver, positioning COX-2 upstream in the regulation of ferroptosis in liver fibrosis. There are two possible reasons for this distinction compared to the published work.36 First, the BJeLR cells used by Yang et al. were derived from human skin and were engineered to express the catalytic subunits of human telomerase, SV40 large T and small T antigens, and an oncogenic allele of HRAS. In contrast, our study utilized mouse livers and primary mouse hepatocytes. Second, Yang et al. focused on malignancy, while we focused on chronic liver injury, which is a benign disease by nature. Additionally, the indomethacin used by Yang et al. is a non-selective COX inhibitor that can bind both COX-1 and COX-2 simultaneously to exert its pharmacological effects. In contrast, we employed COX-2 knockout mice and a selective COX-2 inhibitor (etoricoxib) to conduct a series of assays, ensuring target specificity for COX-2. The differences in cell origins, disease nature, and COX target specificity help explain the divergent conclusions regarding the relationship between COX-2 and ferroptosis across studies.
The transcription factor Nrf2 was previously verified to regulate the cellular adaptive response to excessive oxidative stress and inflammation, thereby inhibiting ferroptosis in iron overload-induced liver injury.12 Similarly, in our primary mouse hepatocyte assay, Nrf2 was downregulated in erastin-induced ferroptosis, and tBHQ stimulated Nrf2 to combat ferroptosis, implying that the Nrf2 signaling pathway was involved in the regulation of ferroptosis and that COX-2 affected this regulation. In combination with our in vivo assays, inhibition of COX-2 upregulated the Nrf2 signaling pathway to inhibit ferroptosis, ultimately impeding the progression of chronic liver diseases. This study discovers that ferroptosis is a new target of COX-2, adding to the mechanisms by which COX-2 exacerbates chronic liver diseases. Currently, several COX-2 inhibitors have been successfully developed, and selective COX-2 inhibitors are widely applied in rheumatic diseases, acute inflammation, and pain control. This study suggests that selective COX-2 inhibitors could be a potentially effective treatment for chronic liver diseases, shedding light on the potential therapeutic strategies for liver fibrosis.