COX-2 induction in mice with experimental nutritional steatohepatitis: Role as pro-inflammatory mediator

Authors

  • Jun Yu,

    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    Current affiliation:
    1. Department of Medicine and Therapeutics, Prince of Wales Hospital, Chinese University of Hong Kong, Hong Kong, China
    Search for more papers by this author
    • J. Yu is a Research Fellow of the Medical Foundation, University of Sydney. E. Ip is a Dora Lush Biomedical Research Scholar of the Australian NHMRC. A. dela Peña is the recipient of an Australian Postgraduate Award and Westmead Millennium Foundation support.

  • Emilia Ip,

    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    Search for more papers by this author
    • J. Yu is a Research Fellow of the Medical Foundation, University of Sydney. E. Ip is a Dora Lush Biomedical Research Scholar of the Australian NHMRC. A. dela Peña is the recipient of an Australian Postgraduate Award and Westmead Millennium Foundation support.

  • Aileen dela Peña,

    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    Search for more papers by this author
    • J. Yu is a Research Fellow of the Medical Foundation, University of Sydney. E. Ip is a Dora Lush Biomedical Research Scholar of the Australian NHMRC. A. dela Peña is the recipient of an Australian Postgraduate Award and Westmead Millennium Foundation support.

  • Jing Yun Hou,

    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    Search for more papers by this author
  • Jayshree Sesha,

    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    Search for more papers by this author
  • Natasha Pera,

    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    Search for more papers by this author
  • Pauline Hall,

    1. Division of Anatomical Pathology, Faculty of Health Sciences, University of Cape Town, Cape Town, South Africa
    Search for more papers by this author
  • Richard Kirsch,

    1. Division of Anatomical Pathology, Faculty of Health Sciences, University of Cape Town, Cape Town, South Africa
    Search for more papers by this author
  • Isabelle Leclercq,

    1. Laboratoire de Gastro-entérologie, Université Catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
  • Geoffrey C. Farrell

    Corresponding author
    1. Storr Liver Unit, Westmead Millennium Institute, University of Sydney at Westmead Hospital, Westmead, Australia
    • Professor of Hepatic Medicine and Director of Gastroenterology, Australian National University and The Canberra Hospital, Yamba Drive, Garran, ACT 2605 Australia
    Search for more papers by this author
    • fax: (612) 628 15179.


  • Potential conflict of interest: Nothing to report.

Abstract

The underlying mechanisms that perpetuate liver inflammation in nonalcoholic steatohepatitis are poorly understood. We explored the hypothesis that cyclooxygenase-2 (COX-2) can exert pro-inflammatory effects in metabolic forms of fatty liver disease. Male wild-type (WT) C57BL6/N or peroxisome proliferator–activated receptor α knockout (PPAR-α−/−) mice were fed a lipogenic, methionine- and choline-deficient (MCD) diet or the same diet with supplementary methionine and choline (control). COX-2 was not expressed in livers of mice fed the control diet. In mice fed the MCD diet, hepatic expression of COX-2 messenger RNA and protein occurred from day 5, continued to rise, and was 10-fold higher than controls after 5 weeks, thereby paralleling the development of steatohepatitis. Upregulation of COX-2 was even more pronounced in PPAR-α−/− mice. Induction of COX-2 was completely prevented by dietary supplementation with the potent PPAR-α agonist Wy-14,643 in WT but not PPAR-α−/− mice. COX-2 upregulation was preceded by activation of nuclear factor κB (NF-κB) and coincided with increased levels of tumor necrosis factor α (TNF-α), interleukin (IL)-6, and intercellular adhesion molecule 1 (ICAM-1). Selective COX-2 inhibitors (celecoxib and NS-398) protected against the development of steatohepatitis in WT but not PPAR-α−/− mice. In conclusion, induction of COX-2 occurs in association with NF-κB activation and upregulation of TNF-α, IL-6, and ICAM-1 in MCD diet–induced steatohepatitis. PPAR-α suppresses both COX-2 and development of steatohepatitis, while pharmacological inhibition of COX-2 activity ameliorates the severity of experimental steatohepatitis. COX-2 may therefore be a pro-inflammatory mediator in metabolic forms of steatohepatitis. (HEPATOLOGY 2006;43:826–836.)

Nonalcoholic fatty liver disease is a spectrum of liver disorders ranging from steatosis to nonalcoholic steatohepatitis to cirrhosis,1 but the molecular mechanisms of inflammatory recruitment that transform steatosis to steatohepatitis (with hepatocellular injury, lobular inflammation, and hepatic fibrosis) remain undefined.2, 3 Possible mediators include oxidative stress,4–6 which triggers lipid peroxidation in the steatotic liver, and cytokines such as tumor necrosis factor α (TNF-α) and interleukin (IL)-6. The products of lipid peroxidation can mediate inflammatory recruitment directly7, 8 or indirectly by activating nuclear factor κB (NF-κB) with downstream consequences that include expression of adhesion molecules, cytokines (TNF-α, IL-1β, IL-6), and cyclooxygenase-2 (COX-2).9, 10

COX catalyzes the conversion of arachidonic acid to prostanoids and thromboxanes, some of which are intensely pro-inflammatory.11 In hepatocytes, COX-1 is constitutively expressed, whereas COX-2 is present only at extremely low levels. COX-2 is rapidly induced by lipid peroxides, oxidant stress, NF-κB, cytokines (TNF-α, IL-1β, IL-6), and transforming growth factor β1 (TGF-β1).12–16 These mediators are increased in serum or liver from patients with nonalcoholic steatohepatitis and in experimental steatohepatitis.1, 4, 17–21 Although hepatic COX-2 expression is increased with experimental alcoholic steatohepatitis,22 this enzyme's potential role in the pathogenesis of metabolic steatohepatitis has not been studied.

Peroxisome proliferator–activated receptor α (PPAR-α) regulates both hepatic lipid turnover and pathways of tissue inflammation.23–25 We recently showed that the potent PPAR-α agonist Wy-14,643 prevents and reverses steatohepatitis induced by feeding mice a lipogenic, methionine- and choline-deficient (MCD) diet.26, 27 In accordance with this protective role of PPAR-α against steatohepatitis (shown also for alcoholic hepatitis28, 29), the severity of liver injury was greater in PPAR-α–deficient (PPAR-α−/−) mice fed the MCD diet than in wild-type (WT) mice.26, 27 Other evidence indicates that PPAR-α may be a negative regulator of hepatic inflammation,28, 30, 31 and there is a cell type-specific relationship between COX-2 expression and PPAR-α. Thus, Wy-14,643 upregulates COX-2 expression in epithelial cells32 but suppresses it in aortic smooth muscle cells.33

In the present study, we first tested whether hepatic COX-2 expression is normally downregulated by PPAR-α. We then examined the proposal that intake of a lipogenic, MCD diet overcomes this protective suppression by NF-κB–mediated feed-forward upregulation of COX-2. The MCD model of metabolic steatohepatitis exhibits all the pathological characteristics of nonalcoholic steatohepatitis.1, 6, 34–36 We have previously shown that NF-κB is activated9, 10, 17 in MCD diet–induced steatohepatitis independently of TNF-α and TNF receptor 1.37 We have now characterized hepatic expression of COX-2 in relation to steatohepatitis development, activation of NF-κB, and expression of TNF-α, IL-1β, and IL-6, as well as intercellular adhesion molecule 1 (ICAM-1). In addition to male WT (C57BL6/N) mice, experiments were repeated in their PPAR-α−/− counterparts. Having found that COX-2 is upregulated during the more pronounced inflammatory phase of this model, we conducted intervention studies to clarify the pathogenic significance of COX-2 upregulation, first with the potent and specific PPAR-α agonist Wy-14,643, then with the selective COX-2 inhibitors celecoxib and NS-398.38

Abbreviations

COX, cyclooxygenase; WT, wild-type; PPAR-α, peroxisome proliferator–activated receptor α; MCD, methionine- and choline-deficient; NF-κB, nuclear factor κB; TNF-α, tumor necrosis factor α; IL, interleukin; ICAM-1, intercellular adhesion molecule 1; TGF-β1, transforming growth factor β1; Cyp, cytochrome P450; mRNA, messenger RNA; ALT, alanine aminotransferase.

Materials and Methods

Animal Treatments.

Male 8- to 10-week-old C57BL/6N mice or PPAR-α−/− mice were bred and housed as previously described.26 Mice were fed the MCD (cat. no. 960439; ICN, Aurora, OH) or control diet (MCD diet supplemented with L-methionine [3 g/kg] and choline chloride [2 g/kg]) (cat. no. 960441; ICN). Where indicated, dietary additions included Wy-14,643 (0.1% wt/wt; Chemsyn Laboratories, Lenexa, KS) or celecoxib (1,500 ppm; Searle & Co., Caguas, Puerto Rico). Another group of MCD diet–fed mice received daily intraperitoneal injections of NS-398 (2 mg/kg38, 39) in dimethyl sulfoxide diluted with phosphate-buffered saline (Cayman Chemical, Ann Arbor, MI) for 3 weeks. At the end of the experiments, animals were killed under anesthesia and tissues were obtained and stored as previously reported.26, 37 All protocols and procedures were approved by the Western Sydney Area Health Service Animal Ethics Committee in accordance with National Institutes of Health guidelines.

Isolation of Hepatocytes and Nonparenchymal Cells.

In separate experiments, the liver was perfused in situ with collagenase type 2 (Worthington, Lakewood, NJ) to dissolve the extracellular matrix. Hepatocytes in the cell suspension were separated from hepatic nonparenchymal cells (including Kupffer, endothelial, sinusoidal, and inflammatory cells) via differential centrifugation.40 Contamination of the nonparenchymal cell fraction was less than 5%, as determined via phase contrast microscopy and cell counting.

Quantitation of Hepatic Messenger RNA Expression Levels.

Total RNA was extracted from whole liver or cell pellets using TRIzol (Invitrogen, Carlsbad, CA). Hepatic complementary DNA (prepared as previously described27) was used to quantify COX-1, COX-2, TNF-α, IL-1β, IL-6, TGF-β1, ICAM-1, acyl coenzyme A oxidase, cytochrome P450 (Cyp) 4a10, and Cyp4a14 messenger RNA (mRNA) levels via real-time polymerase chain reaction using SYBRGreen Master Mix (Applied Biosystems, Foster, CA). Mouse GAPDH served as an internal control to assess the overall complementary DNA content. The specific primers used are listed in Table 1.

Table 1. Primers for Quantitative Real-Time Polymerase Chain Reaction Analysis
GeneGenbank Accession NumberPrimer Sequence
  1. Abbreviations: ACO, acyl coenzyme A oxidase; COX, cyclooxygenase; Cyp, cytochrome P450; ICAM-1, intercellular adhesion molecule 1; IL, interleukin; TGF-β1, transforming growth factor β1; TNF-α, tumor necrosis factor α.

ACONM_015729F 5′-GGAAGACTTCCAATCATGCGATAG-3′
  R 5′-GACAACAAAGGCATGTAACCCG-3′
COX-1NM017043F 5′-CTTTGCACAACACTTCACCCACC-3′
  R 5′-AGCAACCCAAACACCTCCTGG-3′
COX-2XM192868F 5′-TGCTGGAAAAGGTTCTTCTACGG-3′
  R 5′-GAACCCAGGTCCTCGCTTATG-3′
Cyp4a10BC051049F 5′-CTGTCCCAGGCATTGTCAG-3′
  R 5′-GAGTGTGACCTGGACACCTTTG-3′
Cyp4a14NM_007822F 5′-GCCAGAATGGAGGATAGGAACA-3′
  R 5′-ATGAATGTGTCCACCTCTGCAC-3′
GAPDHXM132897F 5′-AACGACCCCTTCATTGAC-3′
  R 5′-TCCACGACATACTCAGCA-3′
ICAM-1NM010493F 5′-CCTGCCTCTGAAGCTCGGAT-3′
  R 5′-ACAGGAACTTTCCCGCCACC-3′
IL-1βBC011437F 5′-GACAGTGATGAGAATGACC-3′
  R 5′-CTCCACTTTGCTCTTGAC-3′
IL-6NM031168F 5′-GCCTATTGAAAATTTCCTCTG-3′
  R 5′-GTTTGCCGAGTAGATCTC-3′
TGF-β1NM011577F 5′-ATGCTAAAGAGGTCACCC-3′
  R 5′-CAAAAGACAGCCACTCAG-3′
TNF-αNM013693F 5′-CACGTCGTAGCAAACCACCAA-3′
  R 5′-CCCATTCCCTTCACAGAGCAA-3′

Western Blot Analysis of Hepatic Proteins.

Liver tissue was homogenized in Tris-HCl (pH 7.4) buffer containing a protease inhibitor cocktail (Roche, Indiapolis, IN), and total protein as determined using the DC protein assay (Bio-Rad, Hercules, CA). Twenty-five micrograms of protein were separated via 12% SDS-PAGE and Western blotting performed as previously described41 using specific antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) against COX-1 (dilution 1:1,000), COX-2 (1:1,000), or β-actin (1:5,000). Proteins were detected via enhanced chemiluminescence (ECL, Amersham Corporation, Louisville, KY), and bands were quantified via scanning densitometry using the SCAN Control (Scanco∼1.lnk) imaging system. Individual levels of COX-1 and COX-2 protein expression were normalized to β-actin.

Nuclear protein was isolated from 100 mg of fresh liver tissue.41 Fifty micrograms of nuclear protein were electrophoresed through a 12% polyacrylamide gel, and immunoblots were incubated with primary antibodies against the NF-κB subunits p65 and p50 (1:500) (Santa Cruz Biotechnology).

Immunohistochemistry for COX-1 and COX-2.

COX-1 and COX-2 were detected in paraffin-embedded liver sections using the above-mentioned specific antibodies and an avidin–biotin complex immunoperoxidase method (Santa-Cruz Biotechnology). Briefly, endogenous peroxidase activity was blocked by treating sections with 3% hydrogen peroxide. After blocking with 10% nonimmunized goat serum, the primary goat antiserum was applied and incubated overnight at 4°C for COX-1 (1:100 dilution) and 2 hours at room temperature for COX-2 (1:150 dilution). Primary antibodies were omitted, and nonimmunized goat serum was used for negative controls. After rinsing, the biotinylated secondary antibody, avidin–biotin complex, and horseradish peroxidase (Dako A/S, Glostrup, Denmark) were applied. Peroxidase activity was visualized by applying diaminobenzidine to the sections, which were then counterstained with hematoxylin.

Electrophoretic Mobility Gel Shift Assay.

Electrophoretic mobility gel shift assays using oligonucleotide-containing specific consensus binding sites for NF-κB (Promega, Madison, WI) were performed on hepatic nuclear protein extracts as previously described.9, 37, 41 Supershift analysis was undertaken by adding specific p65 and p50 antibody (Santa Cruz Biotechnology) to the reaction.9, 37, 41

Assessment of Liver Pathology.

Hematoxylin-eosin–stained sections of paraffin-embedded liver tissue were examined by 2 independent liver pathologists blinded to the study (P. H., R. K.) and were scored for hepatic steatosis (0, none; 1, 1%-25%; 2, 26%-50%; 3, 51%-75%; 4, 76%-100% hepatocytes affected) and necroinflammation (0, none; 1, mild; 2, moderate; 3, severe).

COX-2 Activity.

COX-2 activity of liver homogenate was assayed by monitoring the appearance of oxidized N,N,N',N'-tetramethyl-ρ-phenylenediamine (TMPD) at 590 nm using a commercial kit (Cayman Chemical). COX-1 activity was suppressed using the specific inhibitor SC560.

Biochemical Assays.

Serum alanine aminotransferase (ALT) levels were measured using automated techniques (Department of Clinical Chemistry, Westmead Hospital), liver triglycerides were measured using the triglyceride E-test kit (Wako Pure Chemical Industries, Osaka, Japan), and hepatic lipoperoxide levels were measured as thiobarbituric acid-reactive substances.6

Statistical Analysis.

The results are presented as the mean ± SD. Comparisons between 2 groups were analyzed via Student's t test, and comparisons between more than 2 groups were analyzed via 1-way ANOVA to identify differences among means. Associations between COX-2 mRNA and COX-2 protein expression were analyzed via Pearson's rank correlation analysis. A 2-sided P value of less than .05 was considered statistically significant.

Results

COX-2 is Expressed in Livers of MCD Diet–Fed Mice.

COX-1 mRNA and protein levels after 3 weeks of feeding were detected in WT mice fed the control diet but were not altered by MCD dietary feeding (Fig. 1; Supplementary Fig. 1B). COX-2 mRNA was barely detectable in livers from mice fed the control diet but accumulated from day 5 during intake of the MCD diet. By 10 days, COX-2 mRNA was twice as high as in controls and was 11-fold higher by weeks 3 and 5 (Supplementary Fig. 1A). COX-2 protein expression also increased following administration of the MCD diet, reaching values at 3 and 5 weeks that were 4.8-fold and 6.5-fold higher, respectively, than corresponding dietary controls (Fig. 1). There was a significant correlation between COX-2 mRNA and protein expression (r = 0.868; P < .001). Analysis of COX-2 mRNA levels after 3 weeks of feeding showed that COX-2 expression is increased following MCD dietary feeding in both hepatocytes and hepatic nonparenchymal cells (which include inflammatory cells) when compared with mice fed the control diet (Supplementary Fig. 2).

Figure 1.

Effects of the MCD diet and treatment with Wy-14,643 on hepatic COX-1 and COX-2 protein expression. WT mice were fed the MCD diet (black bars), control diet (white bars), or MCD diet with Wy-14,643 (0.1% wt/wt) (gray bars) for 3 or 5 weeks. Hepatic COX-1, COX-2 and β-actin protein levels were determined via Western blotting (A), bands were quantitated, and COX-1 and COX-2 protein levels were expressed as relative amounts normalized to β-actin (B). Data are expressed as the mean ± SD (n = 3-4 per group). *P < .05, **P < .01 (MCD vs. control). #P < .01 (MCD + Wy-14,643 vs. MCD diet for same total time). Abbreviations: MCD, methionine- and choline-deficient; COX, cyclooxygenase.

The temporal profile of increased COX-2 expression paralleled the histological development of steatohepatitis.6, 9, 10, 26 In livers from control diet–fed mice, COX-2 protein was noted in occasional hepatocytes (Fig. 2A). In sections from MCD diet–fed mice, strong and dense COX-2 immunoreactivity in hepatocytes was present at 3 weeks (Fig. 2B) and affected a larger proportion of hepatocytes at 5 weeks (Fig. 2C). COX-2 immunostaining was primarily in the cytoplasm and appeared to be more intense in fat-laden hepatocytes (Fig. 2B-C). As expected,42 inflammatory cells were also immunopositive for COX-2.

Figure 2.

Immunohistochemistry for COX-2 in livers of mice fed the MCD or control diet parallels steatohepatitis severity; effect of Wy-14,643. (A) WT mice fed the control diet for 5 weeks show normal liver histology; few COX-2–positive cells are seen. (B) When fed the MCD diet for 3 weeks, macrovesicular fat droplets (asterisks) and small groups of inflammatory cells (arrowheads) are observed. The number of COX-2–positive cells (arrows) is increased. (C) After 5 weeks of MCD dietary feeding, the severity of steatohepatitis is greatly increased, and COX-2 immunopositive cells are widespread. (D) Treatment of MCD diet–fed WT mice with Wy-14,643 for 5 weeks prevented steatohepatitis and COX-2 expression. Representative slides obtained from 3 to 5 animals per group. (Original magnification ×200.)

Evidence that PPAR-α Could Modulate COX-2 Expression in MCD Diet–Induced Steatohepatitis.

Others have shown that COX-2 expression may be influenced by PPAR-α.32, 33 Our earlier work showed that coadministration of the potent and specific PPAR-α agonist Wy-14,643 prevented the development of MCD diet–induced steatohepatitis.26 Consistent with this protective effect, Wy-14,643 coadministration suppressed induction of COX-2 (Fig. 1; Supplementary Fig. 1A) so that few hepatocytes were immunopositive for COX-2 (Fig. 2D). Conversely, in PPAR-α−/− mice fed the MCD diet, hepatic COX-2 mRNA (Fig. 3A) and protein levels (Fig. 3B) were not only significantly enhanced compared with their corresponding controls, but after 5 weeks of MCD dietary feeding, hepatic expression levels of COX-2 mRNA (18 ± 2.0 vs. 12 ± 3.0, P < .05) and protein (9.2 ± 1.0 vs. 6.5 ± 1.0, P < .05) were higher in PPAR-α−/− mice compared with WT mice (Fig. 3; Supplementary Fig. 1A-B). Wy-14,643 failed to reduce steatohepatitis-associated increases in COX-2 expression in mice lacking PPAR-α (Fig. 3A).

Figure 3.

Effects of the MCD diet and treatment with Wy-14,643 on hepatic COX-1 and COX-2 expression in PPAR-α−/− mice. PPAR-α−/− mice were fed the control diet (white bars), MCD diet (black bars), or MCD diet with Wy-14,643 (gray bars) for 3 weeks or 5 weeks. (A) Relative values of COX-1 and COX-2 mRNA were obtained via real-time polymerase chain reaction (normalized to GAPDH). (B) COX-1 and COX-2 protein levels were measured via Western blotting (normalized to β-actin). Data are expressed as the mean ± SD (n = 3-5 per group) and are expressed relative to values obtained in mice fed the control diet for 3 weeks, which were arbitrarily assigned a value of 1.0. *P < .05, **P < .01 (MCD diet with or without Wy-14,643 vs. control diet). There were no differences between groups fed the MCD diet with or without Wy-14,643. Protein levels of COX-2 were not determined in the Wy-14,643 intervention experiment in PPAR-α−/− mice because of insufficient material. Abbreviations: COX, cyclooxygenase; mRNA, messenger RNA; MCD, methionine- and choline-deficient.

Relationships Between NF-κB Activation, ICAM-1, and COX-2 Regulation in MCD Diet–Induced Steatohepatitis.

Induction of COX-2 transcription by proinflammatory cytokines and lipoperoxides is mediated, at least in part, by activation of NF-κB signaling.14, 15 Furthermore, activation of NF-κB is an early event in the evolution of MCD diet–induced steatohepatitis, being detected at 3, 5, and 10 days of feeding (Fig. 4A) and occurring in both hepatocytes and hepatic nonparenchymal cell fractions.37 Consistent with such activation, hepatic nuclear protein levels of p65 and p50 increased significantly after 5 days of MCD dietary feeding compared with controls, and this difference was maintained until 3 weeks; thereafter, nuclear p65 and p50 expression in control diet–fed mice increased so that the difference between MCD diet–fed and control diet–fed animals diminished (Fig. 4B-C). Consistent with the proposed pro-inflammatory role of NF-κB in this model, hepatic ICAM-1 was induced in a similar pattern to that of COX-2 during the evolution of steatohepatitis and after administration of Wy-14,643 (Supplementary Fig. 3).

Figure 4.

Activation of NF-κB in MCD diet–fed WT mice. (A) Electrophoretic mobility gel shift assay was performed on hepatic nuclear extracts from WT mice fed the MCD or control diet for 3, 5, and 10 days using a NF-κB oligonucleotide consensus binding sequence. Supershift assays were used to confirm the identity of bands, as described in Materials and Methods (data not shown). (B) Nuclear protein expression of NF-κB subunits, p65 and p50, was determined by Western blotting in mice fed the control (white bars) or MCD diet (black bars) for 3, 5 and 10 days and (C) 3 and 5 weeks. Data were normalized to internal control (β-actin) and are expressed as the mean ± SD (n = 3-4 per group). *P < .05, **P < .01 (MCD vs. control diet). Abbreviation: MCD, methionine- and choline-deficient.

Hepatic Expression of COX-2 Correlates With TNF-α and IL-6, But Not With IL-1β or TGF-β1.

To determine whether the observed upregulation of COX-2 in MCD diet–induced steatohepatitis could be mediated via changes in the cytokines known to modulate COX-2 expression,12–16, 43, 44 levels of corresponding mRNA species were determined. Concomitant with initial induction of COX-2, TNF-α and IL-6 mRNA levels were significantly elevated after 5 days of MCD dietary feeding, and further increases were noted thereafter (Table 2). In PPAR-α−/− mice, MCD dietary feeding also increased mRNA levels of TNF-α and IL-6 compared with corresponding controls (Table 2). Coadministration of Wy-14,643 almost entirely abolished induction of TNF-α and IL-6 mRNA in WT but not PPAR-α−/− mice (Table 2). There were no significant differences in hepatic mRNA levels of IL-1β and TGF-β1 between groups (Table 2).

Table 2. Effect of MCD Diet With or Without Wy-14,643 on Hepatic ICAM-1, TNF-α, IL-6, TGF-β, and IL-1 mRNA Expression Levels in Wild-Type and PPAR-α−/− Mice
DietWild-TypePPARα−/−
3 Days5 Days10 Days3 Weeks5 Weeks5 Weeks
  • NOTE. Specific mRNA values were normalized to the expression of GAPDH as described in Materials and Methods. Data are expressed as the mean ± SD (n = 3–5/group) relative to the values obtained in mice fed the control diet, which were arbitrarily assigned a value of 1.0.

  • Abbreviations: MCD, methionine- and choline-deficient; ICAM-1, intercellular adhesion molecule 1; TNF-α, tumor necrosis factor α; IL, interleukin; TGF-β, transforming growth factor β; mRNA, messenger RNA; PPAR-α, peroxisome proliferator–activated receptor α.

  • *

    P < .05,

  • **

    P < .01 (MCD vs. control diet for same genotype mice fed diet for same total period of time).

  • #

    P < .05,

  • ##

    P < .001 (MCD vs. MCD + Wy-14,643 for WT mice fed diet for same total time [Wy-14,643 treatment did not significantly affect the expression of any genes in PPAR-α−/− mice]).

ICAM-1      
 Control1.0 ± 0.11.0 ± 0.11.0 ± 0.11.0 ± 0.41.0 ± 0.31.0 ± 0.4
 MCD1.0 ± 0.21.1 ± 0.11.6 ± 0.2**2.5 ± 0.8*1.6 ± 0.5*3.6 ± 0.8**
 MCD + Wy-14,643NDNDND1.0 ± 0.1##0.8 ± 0.1*3.8 ± 1.7
TNF-α      
 Control1.0 ± 0.11.0 ± 0.21.0 ± 0.11.0 ± 0.81.0 ± 1.01.0 ± 0.8
 MCD1.3 ± 0.41.4 ± 0.2*1.5 ± 0.1**5.2 ± 0.7*12 ± 5.1**14 ± 4.8**
 MCD + Wy-14,643NDNDND0.9 ± 0.4##1.0 ± 0.3#13 ± 2.7
IL-6      
 Control1.0 ± 0.31.1 ± 0.11.0 ± 0.01.0 ± 1.41.0 ± 1.51.0 ± 0.9
 MCD1.3 ± 0.31.8 ± 0.3**2.4 ± 0.3**1.2 ± 1.16.3 ± 4.1*19 ± 9.2*
 MCD + Wy-14,643NDNDND1.2 ± 2.60.9 ± 0.0#18 ± 5.2
TGF-β      
 Control1.0 ± 0.11.2 ± 0.20.9 ± 0.21.0 ± 0.21.2 ± 0.41.0 ± 0.1
 MCD1.0 ± 0.21.2 ± 0.40.8 ± 0.21.2 ± 0.51.6 ± 0.51.5 ± 0.2
 MCD + Wy-14,643NDNDND0.6 ± 0.30.6 ± 0.22.2 ± 1.4
IL-1β      
 Control1.0 ± 0.11.3 ± 0.11.2 ± 0.11.0 ± 0.31.0 ± 0.71.0 ± 0.9
 MCD1.0 ± 0.21.4 ± 0.11.1 ± 0.20.8 ± 0.61.5 ± 0.51.2 ± 0.4
 MCD + Wy-14,643NDNDND0.47 ± 0.10.9 ± 0.51.7 ± 0.6

Effect of Selective COX-2 Inhibition on MCD Diet–Induced Steatohepatitis.

If COX-2 is a pro-inflammatory pathway in MCD diet–induced steatohepatitis, as has been proposed for alcoholic hepatitis,22 selective COX-2 inhibition should dampen or abrogate the development of this disorder. We first showed that celecoxib (1,500 ppm added to the diet for 3 weeks) suppressed MCD diet–induced COX-2 activity in WT mice by 42% (P < .05) (Fig. 5A). COX-2 mRNA and protein were also reduced (Fig. 5B-C). In mice fed the MCD diet with supplementary celecoxib, serum ALT levels were significantly less than in corresponding positive controls (Fig. 6A), and the severity of liver necroinflammation was significantly reduced (Fig. 7; Table 3). Celecoxib also significantly improved steatosis (Table 3), a finding supported by reduction in hepatic triglyceride content (Fig. 6B) and lowered lipoperoxide levels (Fig. 6C).

Figure 5.

The effect of celecoxib on COX-2 activity, mRNA, and protein expression in livers of WT and PPAR-α−/− mice fed the MCD diet for 3 weeks. WT or PPAR-α−/− mice were fed the control diet (white bars), MCD diet (black bars), or MCD diet with celecoxib (1,500 ppm) (gray bars) for 3 weeks. Hepatic (A) COX-2 activity, (B) COX-2 mRNA (normalized to GAPDH), and (C) COX-2 protein levels (normalized to β-actin) were measured as described in Materials and Methods. Values were normalized to internal control and expressed relative to control diet–fed mice, which were arbitrarily assigned a value of 1.0. Data are expressed as the mean ± SD (n = 4-6 per group). *P < .01, **P < .001 (MCD vs. control diet–fed mice). #P < .05, ##P < .01 (MCD + celecoxib vs. MCD diet alone). Abbreviations: COX, cyclooxygenase; mRNA, messenger RNA; PPAR-α, peroxisome proliferator–activated receptor α; MCD, methionine- and choline-deficient.

Figure 6.

Effects of the MCD diet and celecoxib on serum ALT, hepatic triglyceride and lipoperoxide content in WT and PPAR-α−/−mice. Mice were fed the control diet (white bars), MCD diet (black bars), or MCD diet treated with celecoxib (1,500 ppm) (gray bars) for 3 weeks and (A) serum ALT levels, (B) hepatic triglyceride content, and (C) hepatic lipoperoxide levels, measured as thiobarbituric acid–reactive substances, were assessed. Data are expressed as the mean ± SD (n = 4-6 per group). *P < .05, **P < .01, ***P < .001 (MCD vs. control diet–fed mice). #P < .05, ##P < .01 (MCD vs. MCD + celecoxib). The concentration of thiobarbituric acid–reactive substances in PPAR-α−/− mice fed the control diet was very low and thus cannot be represented. Abbreviations: ALT, alanine aminotransferase; PPARα, peroxisome proliferator–activated receptor α; TBARS, thiobarbituric acid–reactive substances.

Figure 7.

Effect of treatment with celecoxib on MCD diet–induced steatohepatitis. (A) Liver sections from WT mice fed the control diet show normal liver histology. (B) In MCD diet–fed mice, numerous necroinflammatory foci (arrows) and macrovesicular fat droplets (asterisks) were present. (C) Treatment of MCD diet–fed mice with celecoxib (1,500 ppm) for 3 weeks largely ameliorated steatohepatitis. Slides are representative of 4 to 6 separate experiments. (Hematoxylin-eosin stain; original magnification ×100.)

Table 3. Effect of MCD Diet and Treatment With Celecoxib on Scores for Hepatic Steatosis and Necroinflammatory Lesions in Wild-Type and PPAR-α−/− Mice Fed the MCD or Control Diet for 3 Weeks
 Wild-TypePPAR-α−/−
ControlMCDMCD + CelecoxibControlMCDMCD + Celecoxib
  • NOTE. Values of hepatic steatosis and necroinflammation are expressed as the mean ± SD (n = 8–10/group).

  • Abbreviations: MCD, methionine- and choline-deficient; PPAR-α, peroxisome proliferator–activated receptor α.

  • *

    P < .05,

  • b

    P < .001 (MCD vs. control diet in same genotype mice.

  • #

    P < .05,

  • ##

    P < .01 (MCD vs. MCD with celecoxib treatment in same genotype mice).

Steatosis0.1 ± 0.43.6 ± 0.7*1.5 ± 1.1##0.0 ± 0.04.0 ± 0.0*2.8 ± 1.0#
Necroinflammation0.3 ± 0.52.6 ± 0.7*1.2 ± 0.9##0.0 ± 0.02.8 ± 0.4*2.0 ± 0.9

In contrast to these impressive anti-inflammatory and cytoprotective effects in WT mice, celecoxib failed to significantly lower serum ALT or hepatic triglyceride levels in PPAR-α−/− mice (Fig. 6A-B). Although there was a trend toward reduced lipoperoxide levels following celecoxib treatment, the data were too variable to permit conclusions (Fig. 6C). Furthermore, in PPAR-α−/− mice, hepatic COX-2 enzyme activity and mRNA expression were not significantly reduced by celecoxib (Fig. 5). In accordance with the proposal that celecoxib could activate endogenous PPAR-α in WT mice, transcript levels of PPAR-α–regulated genes Cyp4a10 and Cyp4a14 were significantly increased in celecoxib-treated mice compared with untreated mice fed the MCD diet (Cyp4a10, 12 ± 2.7 vs. 6.6 ± 2.8 [P < .05]; Cyp4a14, 21 ± 8.7 vs. 9.1 ± 2.5 [P < .05]). The expression of acyl coenzyme A oxidase, another PPAR-α–regulated gene, was unaffected by celecoxib administration (P = .82).

To confirm that the effects observed with celecoxib were not due to nonspecific effects of this drug, the above experiment was repeated with NS-398, another selective and specific COX-2 inhibitor.38 After first determining the lowest dose, within a range used by others,39 that produces maximal inhibition of COX-2 activity in vivo (data not shown), we showed in WT mice that NS-398 attenuated MCD diet–induced steatohepatitis as indicated by histology (data not shown) and extent of hepatocellular injury (by ALT) (Table 4). NS-398 also abrogated MCD diet–induced upregulation of hepatic COX-2 activity, mRNA, and protein (Table 4). As expected, NS-398 had no effect on liver histology (data not shown) or COX-2 expression and activity in PPAR-α−/− mice (Table 4).

Table 4. Effect of NS-398 on Serum ALT Levels, Hepatic COX-2 Activity, Protein and mRNA, and Cyp4a mRNA Expression in Wild-Type and PPAR-α−/− Mice Fed the MCD or Control Diet for 3 Weeks
DietWild-TypePPAR-α−/−
ControlMCDMCD + NS-398ControlMCDMCD + NS-398
  • NOTE. Specific mRNA values were normalized to the expression of GAPDH as described in Materials and Methods. Data are expressed as the mean ± SD (n = 3–5/group) relative to the values obtained in mice fed the control diet at 3 weeks, which were arbitrarily assigned a value of 1.0.

  • Abbreviations: ALT, alanine aminotransferase; COX, cyclooxygenase; Cyp, cytochrome P450; mRNA, messenger RNA; PPAR-α, peroxisome proliferator–activated receptor α; MCD, methionine- and choline-deficient.

  • **

    P < .01,

  • ***, **

    P < .001 (MCD vs. same genotype as control diet–fed mice.

  • #

    P < .05,

  • ##

    P < .01,

  • ###

    P < .001 (MCD vs. same genotype as MCD–fed mice treated with NS-398.

  • P < .05,

  • ††

    P < .01,

  • †††

    P < .001 (wild-type vs. PPAR-α−/− mice fed the MCD diet.

  • P < .05,

  • ‡‡

    P < .01,

  • ‡‡‡

    P < .001 (wild-type vs. PPAR-α−/− mice fed the MCD diet and treated with NS-398).

ALT (IU/L)58 ± 40695 ± 236***360 ± 68#37 ± 311067 ± 91***,1254 ± 576
COX-2 activity9.8 ± 5.233 ± 11**15 ± 8.6#8.0 ± 6.741 ± 13**35 ± 9.4
COX-2 protein0.2 ± 0.10.9 ± 0.2**0.3 ± 0.2##0.3 ± 0.11.0 ± 0.2***0.8 ± 0.06‡‡
COX-2 mRNA1.0 ± 0.5912 ± 2.2***4.2 ± 2.8###1.0 ± 0.8619 ± 8.2**12 ± 11
Cyp4a10 mRNA1.0 ± 0.212.3 ± 0.36***4.9 ± 2.1#1.0 ± 1.10.33 ± 0.06†††0.41 ± 0.22‡‡
Cyp4a14 mRNA1.0 ± 0.177.4 ± 2.2***13 ± 4.1#0.11 ± 1.30.57 ± 0.10††0.65 ± 0.61‡‡‡

Discussion

We report that COX-2 (but not COX-1) is markedly upregulated during the evolution of a nutritional form of steatohepatitis induced by feeding mice a lipogenic, MCD diet. The temporal profile of COX-2 induction appeared to coincide with increased inflammatory changes in the steatotic liver and paralleled induction of ICAM-1, TNF-α, and IL-6. COX-2 expression appeared to be most pronounced in fat-laden hepatocytes and inflammatory cells, indicating that metabolic changes in hepatic parenchyma could, at least partly, account for upregulation of a pathway that enhances synthesis of pro-inflammatory and vasoactive eicosanoids. Other evidence links hepatic upregulation of COX-2 to tissue injury in alcoholic hepatitis.22 The present data indicate that a broader role for COX-2 activation in the transition of steatosis to steatohepatitis is plausible.

We considered whether enhanced hepatic COX-2 expression in MCD diet–induced steatohepatitis could be due to loss of PPAR-α–mediated suppression of COX-2 in the normal liver. Consistent with this proposal, COX-2 upregulation was greater in PPAR-α−/− mice than in WT mice. Furthermore, coadministration of the potent and specific PPAR-α agonist Wy-14,643 with the MCD diet completely prevented hepatic COX-2 expression induced by MCD dietary feeding in WT but not PPAR-α−/− mice. PPAR-α has not been shown to directly affect COX-2 transcription at the promoter level, so it is unlikely that these effects are due to direct actions of PPAR-α on the COX-2 gene. However, others have shown that, in certain cell types, PPAR-α agonists prevent induction of COX-2 expression,33 which is at least partly attributable to inhibition of NF-κB activation by PPAR-α,31 and thus of NF-κB–mediated transcription of COX-2.43, 44 Because we have previously reported that Wy-14,643 prevents development of steatohepatitis, in part by reducing hepatic lipoperoxide levels,26 an alternative explanation may be that PPAR-α activation limits the accumulation of lipoperoxides, which in untreated MCD diet–fed mice act to increase COX-2 expression via activation of NF-κB and/or release of pro-inflammatory cytokines (Fig. 8).

Figure 8.

Proposed scheme for regulation of COX-2 and interactions between lipid peroxidation, NF-κB, cytokines, and COX-2 as pro-inflammatory factors and PPAR-α as an anti-inflammatory mediator in nutritional steatohepatitis. In these experiments, Wy-14,643 completely abolished COX-2 induction in WT (but not in PPAR-α−/− mice), directly via PPAR-α activation and/or indirectly by inhibiting (or combating) lipid peroxidation and effects on NF-κB activation. As a result, PPAR-α activation causes regression of pro-inflammatory stimuli that convert steatosis to steatohepatitis. In addition to the demonstrated direct effects of inhibiting hepatic COX-2 activity in WT (but not PPAR-α−/− mice), celecoxib could partly suppress MCD diet–induced steatohepatitis through activation of endogeneous PPAR-α. Abbreviations: PPARα, peroxisome proliferator–activated receptor α; MCD, methionine- and choline-deficient; NF-κB, nuclear factor κB; ICAM-1, intercellular adhesion molecule 1; TNF-α, tumor necrosis factor α; IL-6, interleukin 6; COX-2, cyclooxygenase 2, PMNs, polymorphonuclear leukocytes.

In addition to NF-κB response elements, the promoter region of the COX-2 gene contains binding sites for TNF-α, IL-6, IL-1β, and TGF-β1.14, 43, 45 In the present study, hepatic mRNA levels of TNF-α and IL-6, but not IL-1β or TGF-β1, were increased following MCD dietary feeding. Furthermore, the temporal profile of their upregulation was similar to that of COX-2. These findings are consistent with either a direct mechanistic role for TNF-α and IL-6 as regulators of COX-2, or with coregulation of COX-2 and these cytokines by 1 or more factors operating in the MCD model. We envisage a model in which activation of NF-κB increases synthesis of TNF-α, IL-6, ICAM-1, and COX-2, all of which are downstream targets of NF-κB (Fig. 8).37, 44 The cytokines could also effect further increases in COX-2 expression via a feed-forward mechanism.

The present finding that hepatic COX-2 expression and activity are upregulated in a metabolic form of steatohepatitis does not prove that COX-2 is a mechanism for inflammatory recruitment in fatty liver diseases. To test this, we conducted intervention experiments using 2 selective and specific COX-2 inhibitors. When celecoxib was added to the diet at a dose that selectively inhibits COX-2 but not COX-1 in extrahepatic tissues,46 development of MCD diet–induced steatohepatitis was ameliorated, as indicated by ALT levels and liver histology. Similar protection was shown with NS-398. The protection against steatohepatitis transition afforded by these 2 selective COX-2 inhibitors was associated with reduced hepatic triglyceride and lipoperoxide levels, as well as inhibition of hepatic COX-2 transcriptional, translational, and enzyme activities. However, these effects were only partial in WT mice. Furthermore, PPAR-α−/− mice were refractory to the inhibitory effects of celecoxib and NS-398 on hepatic COX-2 activity, and still developed severe steatohepatitis during MCD dietary feeding without amelioration by celecoxib.

One possible explanation for the beneficial effects of celecoxib and NS-398 on MCD diet–induced steatohepatitis in WT but not PPAR-α−/− mice is that these COX-2 inhibitors activate endogenous PPAR-α, either directly or through their effects on arachidonic acid metabolites. Other nonsteroidal anti-inflammatory drugs have been shown to act as ligands for PPAR-α.47, 48 Additional evidence that COX-2 selective inhibitors could exert antisteatotic and anti-inflammatory effects against steatohepatitis by stimulating PPAR-α was derived by studies in male WT mice, in which MCD dietary feeding upregulated expression of Cyp4a14 and Cyp4a1026; celecoxib further increased the hepatic expression of these genes. Whereas hepatic acyl coenzyme A oxidase expression was not affected by coadministration of celecoxib with the MCD diet, Cyp4a enzymes are known to be more sensitive to transcriptional induction by PPAR-α agonists than acyl coenzyme A oxidase and liver fatty acid binding protein.49 Hence, the differential effect of celecoxib on expression of these PPAR-α–regulated genes can be interpreted in terms of celecoxib being a relatively weak activator of PPAR-α.

In summary, we have demonstrated that hepatic COX-2 expression is increased in a metabolic form of steatohepatitis produced by feeding mice a lipogenic, MCD diet. In this model, induction of COX-2 may be a consequence of NF-κB activation, but the simultaneous increases in TNF-α, IL-6, and ICAM-1 could each contribute to the sustained upregulation of COX-2. Pharmacological activation of PPAR-α suppressed COX-2 expression and development of steatohepatitis. Conversely, hepatic COX-2 expression was induced to a greater extent in PPAR-α−/− mice fed the MCD diet compared with WT mice, concordant with histologically more severe steatohepatitis. Finally, evidence for a causal role of COX-2 enzyme activity in steatohepatitis perpetuation in this model was provided by showing that both celecoxib and NS-398 ameliorated liver injury during intake of the MCD diet, though these agents failed to completely correct steatohepatitis. We therefore propose that COX-2 is among several injurious and proinflammatory pathways in metabolic steatohepatitis. Should similar findings apply in human nonalcoholic fatty liver disease, blockade of COX-2 and/or the factors responsible for upregulation of this pro-inflammatory enzyme would be logical targets for new therapeutic approaches to prevent or treat nonalcoholic steatohepatitis.

Acknowledgements

The authors are grateful to Frank J. Gonzalez (National Cancer Institute, National Institutes of Health, Bethesda, MD) for supplying breeding pairs of C57BL6/N and PPAR-α−/− mice, Sandy Bierach for coordination of animal care and breeding, and Jacqueline Field for technical assistance.

Ancillary