Protective role of macrophage migration inhibitory factor in nonalcoholic steatohepatitis

Abstract

MIF is an inflammatory cytokine but is hepatoprotective in models of hepatotoxin-induced liver fibrosis. Hepatic fibrosis can also develop from metabolic liver disease, such as nonalcoholic fatty liver disease (NASH). We investigated the role of MIF in high-fat or methionine- and choline-deficient diet mouse models of NASH. Mif^−/− mice showed elevated liver triglyceride levels (WT, 53±14 mg/g liver; Mif^−/−, 103±7 mg/g liver; P<0.05) and a 2–3-fold increased expression of lipogenic genes. Increased fatty degeneration in the livers of Mif^−/−mice was associated with increased hepatic inflammatory cells (1.6-fold increase in F4/80⁺ macrophages) and proinflammatory cytokines (e.g., 2.3-fold increase in Tnf-α and 2-fold increase in Il-6 expression). However, inflammatory cells and cytokines were decreased by 50–90% in white adipose tissue (WAT) of Mif^−/− mice. Subset analysis showed that macrophage phenotypes in livers of Mif^−/− mice were skewed toward M2 (e.g., 1.7-fold and 2.5-fold increase in Arg1 and Il-13, respectively, and 2.5-fold decrease in iNos), whereas macrophages were generally reduced in WAT of these mice (70% reduction in mRNA expression of F4/80⁺ macrophages). The protective MIF effect was scrutinized in isolated hepatocytes. MIF reversed inflammation-induced triglyceride accumulation in Hepa1-6 cells and primary hepatocytes and also attenuated oleic acid-elicited triglyceride increase in 3T3-L1 adipocytes. Protection from fatty hepatocyte degeneration was paralleled by a 2- to 3-fold reduction by MIF of hepatocyte proinflammatory cytokine production. Blockade of MIF receptor cluster of differentiation 74 (CD74) but not of CXCR2 or CXCR4 fully reverted the protective effect of MIF, comparable to AMPK inhibition. In summary, we demonstrate that MIF mediates hepatoprotection through the CD74/AMPK pathway in hepatocytes in metabolic models of liver injury.—Heinrichs, D., Berres, M.-L., Coeuru, M., Knauel, M., Nellen, A., Fischer, P., Philippeit, C., Bucala, R., Trautwein, C., Wasmuth, H. E., Bernhagen, J. Protective role of macrophage migration inhibitory factor in nonalcoholic steatohepatitis.

Chronic liver diseases such as nonalcoholic liver disease (NAFLD) are a major health burden in the Western world (1). NAFLD is associated with lipid accumulation in the liver (steatosis). In advanced stages, steatosis is accompanied by inflammation and called nonalcoholic steatohepatitis (NASH). Further exacerbation leads to liver fibrosis, which can progress into cirrhosis and eventually hepatocellular carcinoma (HCC; refs. 2, 3). Triglyceride accumulation in the liver is a reversible process, offering therapeutic windows if the underlying mechanisms were understood. There are at least 3 mechanisms that can lead to a nonalcoholic fatty liver: increased uptake of dietary lipids or free fatty acids (FFAs) secreted by white adipose tissue (WAT); increased de novo synthesis of fatty acids in the liver; and impaired triglyceride export from the liver (1). This corresponds to the processes in patients with NAFLD caused by malnutrition, rapid weight loss, drugs, or metabolic diseases (4).

NAFLD is often associated with the development of a metabolic syndrome and is characterized by obesity, increased circulating triglyceride levels, decreased levels of high-density lipoprotein cholesterol, arterial hypertension, and insulin resistance (5). Thus, WAT is a tissue of interest (6), and in addition to FFAs released from WAT, several adipokines are secreted by WAT. These amplify systemic inflammation and contribute to inflammation in the liver. Adipokines such as adiponectin and leptin play a crucial role in the regulation of body weight, with adiponectin reduced and leptin elevated in obese patients. Moreover, proinflammatory cytokines such as tumor necrosis factor α (TNF-α) and interleukin-6 (IL-6) are secreted from WAT in obese patients (7–9).

Macrophage migration inhibitory factor (MIF) is a pleiotropic, proinflammatory cytokine with chemokine-like functions. MIF is stored in preformed intracellular pools from which it is rapidly secreted, e.g., by immune cells, on inflammatory and stress stimulation. Also, some parenchymal cells secrete MIF (10, 11). Inflammatory effects of MIF include inflammatory kinase pathways, phosphatidylinositol-4,5-bisphosphate 3-kinase/AKT- and p53-mediated macrophage survival, and accelerated leukocyte recruitment (11, 12). MIF signaling is mediated by noncognate interaction with the chemokine receptors CXCR2 and CXCR4 (12, 13) and through the surface form of an invariant chain, called cluster of differentiation 74 (CD74). Interaction with CXCR2/4 promotes leukocyte recruitment to inflammatory/atherogenic sites, whereas CD74 mainly regulates MIF-stimulated macrophage, B-lymphocyte, and tumor cell survival (14, 15). Interestingly, MIF/CD74 signaling also protects the myocardium from ischemia/reperfusion injury in an AMP-activated protein kinase (AMPK)-mediated fashion (16). Owing to its proinflammatory spectrum of action, MIF is a pivotal mediator of several inflammatory diseases including septic shock, atherosclerosis, and obesity (11, 12, 17).

MIF is expressed in the liver, with hepatocytes and Kupffer cells identified as origins of hepatic MIF production (18). Also, MIF is up-regulated in toxin-induced, inflammatory, autoimmune, and fibrotic liver disease in animal models and human disease (19–21), but functional studies to examine the role of MIF in liver disease have been limited. In ethanol-induced liver injury, Mif-deficient mice exhibit reduced steatosis, inflammation, and steatohepatitis (22). We recently studied the role of MIF in mouse models of chronic (carbon tetrachloride-induced) and toxin (thioacetamide)-induced liver fibrosis. Unexpectedly, MIF exhibited a prominent hepatoprotective effect in both models that was traced to an attenuation of the profibrotic capacity of hepatic stellate cells mediated by the CD74/AMPK pathway (23). The mechanism shares similarities with that observed in MIF/CD74/AMPK-mediated cardioprotection (16). Thus, MIF, despite its overall proinflammatory activity profile and its reported exacerbating role in ethanol-induced steatohepatitis, is a potent antifibrotic factor in the liver. Yet there have been no functional studies on MIF's role in NAFLD.

Here we have investigated the role of MIF in NASH using high-fat diet (HFD)- and methionine- and choline-deficient (MCD) diet-induced models of experimental NAFLD in conjunction with Mif gene deficiency. The consequence of Mif absence was studied in liver and WAT by assessing lipid metabolism and inflammatory parameters. In vivo studies were accompanied by mechanistic in vitro experiments in hepatocytes. The data show for the first time that MIF has a potent hepatoprotective role in models of NAFLD.

MATERIALS AND METHODS

Proteins and other reagents

Murine MIF was prepared as described elsewhere (13). Oleic acid, metformin, AMPK inhibitor compound C, and CXCR4 inhibitor AMD3100 were purchased from Sigma-Aldrich (Munich, Germany). CXCR2 inhibitor SB225002 was obtained from Tocris Bioscience (Bristol, UK). Neutralizing anti-mouse CD74 antibody (clone In-1) was obtained from BD Pharmingen (Heidelberg, Germany). Polyclonal antibodies against CXCR2 (EB11362) and CXCR4 (AB2074) were purchased from Everest Biotech (Upper Heyford, UK) and Abcam (Cambridge, UK), respectively.

Murine in vivo experiments and determination of hepatic steatosis

Animal experiments were approved by the Animal Welfare Committee of Bezirksregierung Köln (Cologne, Germany). C57BL/6 Mif^−/− (13) and wild-type (WT) mice (n=7/group) were fed an HFD (60% fat/calorie) for 16 wk. Weight gain was measured weekly. At the end of the diet treatment, a glucose tolerance test was performed on mice after overnight food withdrawal. Glucose (40%, 25 μl/10 g body weight i.p.) was administered, and glucose levels were measured from tail blood before and every 30 min up to 180 min with the Contour-blood glucose monitoring system (Bayer Diabetes Care, Leverkusen, Germany). In a second model, Mif^−/−and WT mice (n=10/group) were fed an MCD diet for 8 wk.

Liver steatosis was assessed histologically and by enzymatic assay using routine procedures (24). Briefly, histological staining was performed with oil-red O as described previously (23). Intrahepatic triglyceride content was also quantitated by enzymatic assay following lipase-mediated hydrolysis in homogenized liver tissue (buffer: 10 mM Tris-HCl, 2 mM EDTA, 0.25 M sucrose, pH 7.5). Triglyceride determination was done following the manufacturer's protocol of Triglycerides Liquicolor Mono-Assay (Human Diagnostics, Wiesbaden, Germany).

Expression analysis of lipogenesis-related and inflammatory genes

Total RNA was isolated from livers and epididymal WAT (EWAT) and reversely transcribed using Maxima First Strand cDNA Synthesis Kit for quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR; Thermo Fisher Scientific, Rockford, IL, USA).

qRT-PCR analysis was performed using 2× SensiMix SYBR No-ROX Kit (Peqlab, Erlangen, Germany) in a Rotor-Gene 6000 (Corbett Life Science/Qiagen, Hilden, Germany) with specific murine primer pairs: liver x receptor α (Lxra), forward 5′-CGACAGAGCTTCGTCCACAA-3′ and reverse 5′-GCTCGTTCCCCAGCATTTT-3′; sterol element binding protein 1 (Srebp-1), forward 5′-GGCAAAGGAGGCACTACAG-3′ and reverse 5′-AGATAGCAGGATGCCAACAG-3′; acetyl-CoA-carboxylase (Acc), forward 5′-TGTCCGCACTGACTGTAACCA-3′ and reverse 5′-TGCTCCGCACAGATTCTTCA-3′; fatty acid synthase (Fas), forward 5′-GGTGTCTGACACTGGCAATCTG-3′ and reverse 5′-CGCAGCTCCTTGTATACTTCTCC-3′; diglyceride acyltransferase 1 (Dgat1), forward 5′-TCACCACACACCAATTCAGG-3′ and reverse 5′-GACGGCTACTGGGATCTGA-3′; Dgat2, forward 5′-GAAGATGTCTTGGAGGGCTG-3′ and reverse 5′-CGCAGCGAAAACAAGAATAA-3′; natural killer 1.1 (NK1.1), forward 5′-CGAATCCCATCCTGATTATTTTAC-3′ and reverse 5′-TCTTCTCTTAGAATGGCACAACTG-3′; F4/80, forward 5′-CATTTGCACACATGGAATCAG-3′ and reverse 5′-TGGATGTGCTGGAGGGATTAG-3′; Tnf-α, forward 5′-ACCACGCTCTTCTGTCTACTGA-3′ and reverse 5′-TCCACTTGGTGGTTTGCTACG-3′; Il-1β, forward 5′-AGAGCATCCAGCTTCAAATCTC-3′ and reverse 5′-CAGTTGTCTAATGGGAACGTCA-3′; Il-6, forward 5′-TCAGGAAATTTGCCTATTGAAAA-3′ and reverse 5′-CCAGCTTATCTGTTAGGAGAGCA-3′; chemokine (C-C motif) ligand 2 (Ccl2), forward 5′-ATTGGGATCATCTTGCTGGT-3′ and reverse 5′-CCTGCTGTTCACAGTTGCC-3′; arginase 1 (Arg1), forward 5′-GAACCCAACTCTTGGGAAGAC-3′ and reverse 5′-GGAGAAGGCGTTTGCTTAGTT-3′; inducible nitric oxide synthase (iNos), forward 5′-CTCTACAACATCCTGGAGCAAGTG-3′ and reverse 5′-ACTATGGAGCACAGCCACATTGA-3′; Il-4, forward 5′-AAACTCCATGCTTGAAGAAGAACT-3′ and reverse 5′-CTACGAGTAATCCATTTGCATGAT-3′; Il-13, forward 5′-TGGATTCCCTGACCAACATCTC-3′ and reverse 5′-GGTTACAGAGGCCATGCAATATC-3′; β-actin, forward 5′-CTGACTGACTACCTCATGAAGATCCT-3′ and reverse 5′-CTTAATGTCACGCACGATTTCC-3′.

qRT-PCR data were analyzed via Rotor-Gene 6000 Series 1.7 software (Corbett Life Science/Qiagen, Hilden, Germany). Relative mRNA levels were calculated utilizing β-actin as a housekeeping gene and were normalized to an untreated control sample.

qRT-PCR was carried out for Il-10 with Assay on Demand from Applied Biosystems (Darmstadt, Germany).

Adiponectin and leptin ELISAs

Serum levels of adiponectin and leptin were analyzed by murine ELISA (R&D Systems, Minneapolis, MN, USA) following the manufacturer's instructions.

Immunohistochemical staining of murine liver tissue

Staining of immune cells was performed on paraffin-embedded sections of liver tissue and EWAT. Liver sections were incubated with avidin-biotin-HRP with DAB as chromogen in an autostainer-system (Dako, Hamburg, Germany). Macrophages were analyzed using a monoclonal rat anti-mouse F4/80 antibody (clone BM8; Dianova, Hamburg, Germany).

Western blotting analysis

Western blotting against CD74, CXCR2, and CXCR4 was performed from total cell lysates. Lysates were subjected to blotting as described previously (13), and blots were developed using the above-described anti-CD74, anti-CXCR2, or anti-CXCR4 primary antibodies followed by peroxidase-conjugated secondary antibody and chemoluminescence detection.

Hepatic immune cell isolation and flow cytometric analysis

Single-cell suspensions from freshly harvested liver and EWAT were isolated by standard mechanical/enzymatic digestion, and cells were prepared as described previously (25). For flow cytometric analysis, cells were stained with fluorochrome-conjugated antibodies against CD45, CD3, CD4, CD8, and NK1.1 (eBioscience, San Diego, CA, USA). NK cells were gated as CD45⁺, NK1.1⁺, CD3⁻ cells, while NKT cells were gated as CD45⁺, NK1.1⁺, and CD3⁺ cells. Relative numbers were quantified using a FACS Canto-II (BD Bioscience, Heidelberg, Germany) and data analyzed with FlowJo 8.8.6 (Tree Star, Ashland, OR, USA).

Surface-expressed receptors CD74, CXCR2 and CXCR4 were analyzed by using FITC-/PE-conjugated antibodies or using appropriate isotype controls (BD Bioscience; R&D Systems).

In vitro experiments with murine hepatocytes and adipocytes

The hepatoma cell line Hepa1-6 was starved for 16 h in DMEM (PAA Laboratories, Freiburg, Germany) with 0.5% FCS and stimulated with oleic acid (11 mM), mouse IL-1β (10 ng/ml), and recombinant mouse MIF (200 ng/ml). For blockade experiments, cells were preincubated with 12 μg/ml anti-CD74 antibody or 25 μM compound C for 60 min. After 24 h of incubation, cells were harvested, and the triglyceride content was determined as described above.

Primary hepatocytes were isolated from C57BL/6 WT mice at the age of 12 wk, and after 4 h of incubation were treated as described above (25).

The murine preadipocyte cell line 3T3-L1 was cultured for 6–7 d until cells were confluent. After an additional 48 h incubation time, cells were differentiated with differentiation medium (Zenbio, Research Triangle Park, NC, USA) for 3 d. After differentiation, mature adipocytes were stimulated with oleic acid (11 mM) and recombinant mouse MIF (200 ng/ml). After 24 h of incubation, cells were harvested, and triglyceride content was determined as described above.

Statistical analysis

Data are given as means ± sem. Continuous variables were compared by 2-sided t test with Welch's correction in case of unequal variances. Values of P < 0.05 were considered significant. Statistical tests were performed using GraphPad Prism 5.0 (GraphPad, San Diego, CA, USA).

RESULTS

Mif deficiency promotes weight gain and development of metabolic syndrome after high-fat feeding

The functional role of MIF in NASH was studied by comparing HFD-fed Mif^−/− and WT mice. Body weight was measured weekly and showed a significantly higher gain in Mif^−/− mice over the first 10 wk, which gradually leveled off thereafter (Fig. 1A and Supplemental Fig. S1A). The higher body weight of Mif^−/− mice was reflected by an elevated absolute liver weight (Fig. 1B) and relative liver-to-body weight ratio (Supplemental Fig. S1B) compared to WT as determined after 16 wk of HFD. Levels of liver transaminases (alanine transaminase, aspartate transaminase) were slightly lower in the Mif^−/− mice, but this difference did not reach statistical significance (Supplemental Fig. S2A, B).

Figure 1.

Mif^−/− mice show higher body and liver weight and impaired glucose tolerance after HFD. A) Mif^−/−mice have a higher relative body weight gain over the course of a 16-wk HFD. B) The increase in body weight in Mif^−/− mice was mirrored by higher absolute liver weights in Mif^−/− mice. C, D) Mif^−/− mice have impaired glucose tolerance: blood glucose levels were higher in Mif^−/− mice over a course of 3 h on glucose challenge compared to WT-mice (C); at the end of the HFD, fasting blood glucose levels are significantly higher in Mif^−/− mice (D). E, F) Impaired glucose tolerance in Mif^−/− mice is paralleled by altered adipocytokine levels. Serum adiponectin levels are decreased (E) and serum leptin levels increased (F) in Mif^−/−mice. Data are expressed as means ± sem of 7 mice/group. *P < 0.05, **P < 0.01, ***P < 0.001.

Elevated liver and body weights of Mif^−/− mice were paralleled by significantly impaired glucose tolerance compared to WT (Fig. 1C). Of note, fasting glucose levels after HFD and before glucose challenge were significantly increased in Mif^−/− mice (Fig. 1D).

Elevated body/liver weights combined with an impaired glucose tolerance indicated that Mif^−/− mice were more prone to develop a metabolic syndrome during HFD. Adipokines are typically dysregulated in metabolic syndrome. Thus, we assessed the levels of adiponectin and leptin. Adiponectin levels were significantly decreased (Fig. 1E) and leptin levels increased (Fig. 1F) in the serum of Mif^−/− mice. As both adipokines are secreted from adipose tissue and are involved in appetite regulation, we analyzed food intake in both mouse strains throughout the HFD but observed no differences (Supplemental Fig. S3).

Increased hepatic steatosis in Mif^−/− mice under HFD and MCD

After 16 wk of HFD, the degree of hepatic steatosis was examined in Mif^−/− vs. WT mice. First, the hepatic lipid and triglyceride contents were compared. As expected from the liver weight analysis, liver sections from Mif^−/− mice exhibited increased oil-red O staining compared to WT (Fig. 2A). Increased oil-red O intensity in Mif^−/− livers was accompanied by a substantially higher triglyceride content of the liver as analyzed by lipase assay (Fig. 2B). Elevated lipid levels in the livers of Mif^−/− mice were paralleled by enhanced circulating cholesterol and triglyceride levels in these mice, although this difference was not statistically significant (Supplemental Fig. S2C, D).

Figure 2.

Mif^−/− mice fed HFD exhibit enhanced hepatic fatty degeneration. A, B) Enhanced lipid accumulation in the livers of Mif^−/− mice fed with HFD for 16 wk compared to WT mice: representative oil-red O stainings of liver tissue areas (A); quantification of hepatic triglycerides (B). C, D) Mif^−/− mice fed HFD exhibit increased hepatic expression of lipogenic transcription factors; relative mRNA-levels of Lxra (C) and Srebp-1 (D) in Mif^−/− vs. WT-mice. E) Mif^−/− mice fed HFD exhibit increased expression of lipogenic enzymes; relative mRNA-levels of Acc, Fas, Dgat1, and Dgat2 in Mif^−/− vs. WT mice. Data are expressed as means ± sem of 7 mice/group. *P < 0.05, **P < 0.01.

Exacerbated steatosis in Mif^−/− livers was accompanied by an increased expression of lipogenesis-related genes. Messenger RNA levels of the transcription factors Lxra and Srebp-1 were up-regulated in Mif^−/− mice compared to WT counterparts (Fig. 2C, D).

To confirm the enhancing effect of Mif deletion on lipogenesis, we measured the expression of enzymes involved in liver fatty acid synthesis and triglyceride formation. Figure 2E shows that all lipogenic enzymes studied, i.e., Acc, Fas, Dgat1, and Dgat2, were significantly up-regulated in Mif^−/− mice fed HFD.

To confirm the effect of Mif gene deletion for steatosis, we fed mice an MCD diet for 8 wk. Hepatic triglyceride content was found to be significantly increased in Mif^−/− mice (Supplemental Fig. S4A, B). Similarly, mRNA levels of Lxra, Srebp-1, Acc, Fas, Dgat1, and Dgat2 were significantly up-regulated in Mif^−/− mice compared to WT (Supplemental Fig. S4C–E).

Thus, Mif^−/− mice showed increased liver steatosis in two independent murine in vivo models.

Mif^−/− mice fed HFD show increased hepatic inflammatory cell infiltration and cytokine expression

NASH is characterized by a notable inflammatory phenotype. To study the effect of Mif deletion on liver inflammation, we determined the degree of leukocyte recruitment into the liver after 16 wk of HFD. We first analyzed infiltrated immune cells via flow cytometry. Total CD45⁺ leukocytes were significantly increased in Mif^−/− livers (Fig. 3A). This increase was not due to differences in CD3⁺, CD4⁺, or CD8⁺ T cells, NKT cells, or NK cells, as these cells did not differ between Mif^−/− and WT livers (Fig. 3B, C, top panel). In contrast, liver macrophages were strongly enhanced in Mif^−/− mice, as evidenced by quantitative PCR (qPCR; Fig. 3C, bottom panel) and immunohistochemical staining (Fig. 3D) of macrophage marker F4/80.

Figure 3.

Mif^−/− mice show increased immune cell infiltration and expression of inflammatory mediators in the liver after 16 wk of HFD. A) CD45⁺ leukocyte infiltration is increased in Mif^−/− mice (flow cytometry). B, C) No difference in hepatic T-cell and NK/NKT-cell infiltration between WT and Mif^−/− mice as analyzed by flow cytometry (B) and qPCR (C, top panel). C, D) Increased hepatic monocyte/macrophage infiltration in Mif^−/−mice (qPCR; expression of monocyte/macrophage marker F4/80; C, bottom panel) and immunohistochemistry for F4/80 (representative F4/80 stainings of WT and Mif^−/− mice; D). E) Increased mRNA expression of proinflammatory cytokines in the livers of Mif^−/−mice as analyzed by qPCR. F) Increased mRNA expression of the M2 marker arginase-1 (Arg1) and anti-inflammatory cytokines and decreased expression of the M1 marker iNos in the livers of Mif^−/−mice (qPCR). Data are expressed as means ± sem of 7 mice/group. *P < 0.05, **P < 0.01.

Enhanced hepatic macrophage accumulation in Mif^−/− livers was accompanied by an increase in mRNA of inflammatory cytokines. Messenger RNA expression levels of Tnf-α and Il-6 were significantly up-regulated in Mif^−/− mice. In addition, mRNA levels of Ccl2 and Il-1β were elevated in Mif^−/− mice, although these effects did not reach significance (Fig. 3E). We next analyzed the livers for M1 and M2 markers. Livers from Mif^−/− mice after HFD exhibited significantly elevated mRNA levels of the M2-inducing cytokine Il-13 and the M2 marker Arg1, while mRNA expression of M2-inducing cytokines Il-10 and Il-4 was not different between Mif^−/− and WT strains (Fig. 3F). Consistently, the level of the M1 marker iNos was reduced in Mif^−/− livers. Thus, elevated macrophage counts in Mif^−/− livers likely represent the M2 subset, a cell category with known profibrotic effects. The observed increased Tnf-α and Il-6 levels are not contradictory to this notion, as not only M1, but also M2 macrophages, as well as hepatocytes, produce these cytokines; elevated levels might thus be a reflection of an overall aggravated inflammatory microenvironment in Mif^−/−mice.

Mif^−/− mice fed HFD show decreased inflammatory cell infiltration and cytokine expression in WAT

We asked whether the inflammatory phenotype in the liver was paralleled by changes in WAT. The comparison of EWAT specimens from Mif^−/− and WT mice revealed an inverse inflammatory profile compared to the liver phenotype. Compared to WT, Mif^−/− mice after HFD treatment showed significantly reduced CD3⁺ T cells (Fig. 4A), lower levels of NK/NKT cells (Fig. 4B), and strongly reduced macrophages (Fig. 4C, D). Similarly, the cytokine profile was inverted, with mRNA levels for Tnf-α, Il-6, and Ccl2 significantly decreased in Mif^−/− mice (Fig. 4E). Due to drastically reduced macrophage numbers in Mif^−/− EWAT, both the M1 marker iNos and M2 marker Arg1 were down-regulated in Mif^−/− mice (Fig. 4F). Anti-inflammatory Il-4 and Il-10 were not down- but up-regulated in Mif^−/− EWAT (Fig. 4F), but these products could also be adipocyte derived.

Figure 4.

Mif^−/− mice show decreased immune cell infiltration and expression of inflammatory mediators in EWAT after 16 wk of HFD. A) Decreased CD3⁺ NK1.1⁻ T-cell infiltration into EWAT of Mif^−/−mice (flow cytometry). B, C) Decreased monocyte/macrophage and NK/NKT-cell infiltration in EWAT of Mif^−/−mice (qPCR): expression of the NK/NKT-cell marker NK1.1 (B); expression of the monocyte/macrophage marker F4/80 (C). D) Decreased monocyte/macrophage cell infiltration in EWAT of Mif^−/−mice (immunohistochemistry for F4/80; representative F4/80 stainings of WT and Mif^−/− mice). E) Decreased mRNA expression of proinflammatory cytokines in EWAT of Mif^−/−mice (qPCR). F) Decreased mRNA expression of M2 marker Arginase and M1 marker iNos and increased expression of anti-inflammatory cytokines in EWAT of Mif^−/− mice (qPCR). Data are expressed as means ± sem of 7 mice/group. *P < 0.05, **P < 0.01.

Inhibition of hepatic steatosis by MIF is mediated by hepatocyte CD74/AMPK

We explored the molecular mechanism underlying the increased fatty degeneration and inflammation in Mif^−/− livers, hypothesizing that this effect would be predominantly mediated through pathways in hepatocytes, as they are the essential cell type involved in NASH pathogenesis. We first tested this notion by performing in vitro experiments in the cell line Hepa1-6. Cells were stimulated with oleic acid (OA) and IL-1β to simulate steatosis and inflammation (26). Fatty degeneration was analyzed by quantification of triglycerides and staining for lipid-droplets. OA/IL-1β treatment of Hepa1-6 led to substantially enhanced triglyceride and lipid-droplet contents, while coincubation with rMIF reduced the hepatocyte lipid content by a statistically significant margin. Of note, coapplication of compound C, an AMPK inhibitor, reversed the attenuating effect of rMIF, indicating that MIF counterregulates hepatocyte lipid accumulation through an AMPK-mediated pathway (Fig. 5A, C). This finding was confirmed in primary murine hepatocytes. In these cells, MIF even fully reversed OA/IL-1β-triggered lipid accumulation (Fig. 5B). We also checked whether this role of MIF in reducing lipid accumulation was unique to hepatocytes and therefore tested MIF's effect on adipocyte lipid uptake. Differentiated 3T3-L1 adipocytes were incubated with OA in the presence vs. absence of rMIF. Of note, OA increased lipid content of 3T3-L1 by 50%, and this effect was halved by rMIF coincubation (+30%; Fig. 5D).

Figure 5.

Protection from hepatic fatty degeneration is mediated by MIF/AMPK-triggered signaling. A, B) MIF attenuates OA/IL-1β-stimulated triglyceride accumulation in hepatocytes in an AMPK-mediated manner: Hepa1-6 (A); primary murine hepatocytes (B). C) Representative oil-red O stainings in Hepa1-6 as indicated. D) MIF reduces fat accumulation in adipocytes. MIF attenuates OA-stimulated triglyceride accumulation in the adipocyte cell line 3T3-L1. E) MIF directly elicits AMPK activation in Hepa1-6. AMPK activation was measured by Western blotting of phospho-AMPK. F) MIF attenuates OA/IL-1β-stimulated mRNA-expression of proinflammatory cytokines in Hepa1-6. Relative mRNA-expression levels of Tnf-α, Il-1β, Il-6, and Ccl2. Note that metformin, an AMPK inducer, has similar effects as rMIF. Data are expressed as means ± sem. P = 0.07 for Tnf-α in F. *P < 0.05, **P < 0.01, ***P < 0.001.

Next, the role of the AMPK pathway in MIF-induced counterregulation of hepatocyte lipid accumulation was tested more directly. Hepa1-6 cells were incubated with rMIF and rapid AMPK activation followed by analysis of phospho-AMPK levels. MIF triggered AMPK phosphorylation in a dose-dependent manner, and the effect of MIF was comparable to that of metformin, an AMPK inducer, at a MIF dose of 200 ng/ml (Fig. 5E).

Hepatocytes express proinflammatory cytokines during steatosis (27). We tested whether MIF would affect hepatocyte cytokine expression. Hepa1-6 cells were stimulated with OA/IL-1β, and mRNA expression of Tnf-α, Il-1β, Il-6, and Ccl2 was determined by qPCR. OA/IL-1β treatment significantly up-regulated Tnf-α, Il-6, and Ccl2. Coincubation with rMIF attenuated the stimulatory effect of OA/IL-1β on these cytokines, although statistical significance was only reached for Ccl2 and with the effect for Tnf-α close to significance (P=0.07). Treatment with metformin together with OA/IL-1β confirmed the attenuating capacity of the AMPK pathway (Fig. 5F).

Next, we asked which MIF receptors were responsible for the protective effect of MIF in steatosis. We first determined the expression level of the MIF receptors CD74, CXCR2, and CXCR4. As indicated by flow cytometry, CD74 is the only receptor expressed on the cell surface of Hepa1-6 (Fig. 6A), while all receptors were principally present at appreciable levels (Western blotting of lysates; Fig. 6B). We next added neutralizing antibodies to CD74 or pharmacological inhibitors of CXCR2 and CXCR4 to Hepa1-6 cultures treated with OA/IL-1β and rMIF. Only anti-CD74 but not an isotype control or blockade of CXCR2 or CXCR4 was able to reverse the attenuating effect of MIF (Fig. 6C).

Figure 6.

Protection from fatty degeneration of hepatocytes is mediated by CD74. A, B) FACS analysis reveals surface expression of MIF-receptor CD74 but not CXCR2 and CXCR4 in Hepa1-6 cells (A), while intracellularly, these cells express measurable levels of all 3 MIF receptors (B), as indicated by Western blot analysis from lysates of Hepa1-6. C) Protection by MIF of OA/IL-1β-stimulated triglyceride accumulation in Hepa1-6 is mediated by CD74, but not CXCR2 or CXCR4, as tested by blockade with neutralizing antibody (CD74) or pharmacological inhibitors (CXCR2, CXCR4). Data are expressed as means ± sem. *P < 0.05, **P < 0.01.

This suggests that the hepatoprotective effect of MIF in steatosis is mediated, at least in part, by the CD74/AMPK pathway in hepatocytes (Fig. 7).

Figure 7.

Scheme summarizing the protective role of MIF in NASH and adipose tissue inflammation and proposed underlying mechanisms. MIF reduces the triglyceride content in the liver through CD74/AMPK signaling in hepatocytes. Hepatic inflammation is decreased by MIF, but there is a switch toward an M2 macrophage phenotype. This phenotype may be due to a direct effect of MIF on infiltrating macrophages (arrow 1) and/or on macrophage polarization (arrow 2). Alternatively, macrophage infiltration/polarization may be indirectly affected through hepatocyte responses, which in turn are instigated by MIF (arrow 3). Observed effects of MIF on the hepatocyte encompass an up-regulation of AMPK, a down-regulation of relevant lipogenic transcription factors and enzymes. In adipose tissue, MIF promotes the infiltration of macrophages (arrow 4), and this is associated with increased levels of Tnf-α and IL-6. In serum, triglycerides, cholesterol and leptin levels are decreased, while adiponectin is increased. These mediators likely have a direct effect on fatty degeneration in the liver and possibly on WAT. It is not known if and how infiltrated macrophages regulate the serum levels of adipocytokines or free fatty acids.

DISCUSSION

This study identifies a hepatoprotective effect of the cytokine MIF in steatosis. Evidence was obtained in two independent murine steatosis models. Moreover, in vitro experiments indicated that this effect is prominently mediated by the CD74/AMPK pathway in hepatocytes. While fatty degeneration and associated inflammation was higher in the livers of Mif^−/− mice, an inverse correlation was seen in WAT.

Hepatic steatosis is associated with inflammation in the liver. This link was initially conceptualized by Day (28) as the “2-hit hypothesis,” and has now been recognized to be more complex. Overall, our data are in keeping with the 2-hit model. When fed HFD, Mif^−/− mice gained more weight, had lower glucose tolerance, including dysregulation of adipokine levels, and exhibited elevated hepatic lipid levels compared to WT mice. Accordingly, Mif^−/− mice showed an increased inflammatory state in the liver.

We verified that food intake was equal between the Mif^−/− and WT mouse strains. Thus, the increased triglyceride levels in the liver of Mif^−/− mice could be caused by an increased rate of hepatic lipid uptake. Our in vitro experiments are in support of this possibility. Interestingly, in vitro data with differentiated 3T3-L1 adipocytes incubated with oleic acid (but not with Il-1β) confirmed this notion and indicated that MIF may also attenuate triglyceride incorporation in adipocytes. Of note, an earlier study (29) showing that Mif siRNA knockdown in 3T3-L1 preadipocytes led to a reduced triglyceride content suggested that MIF regulation of adipocyte fat storage may be cell stage dependent. Moreover, this may be associated with an altered MIF receptor expression signature, as preadipocytes express CXCR7 and CD74, while differentiated adipocytes express CXCR2 (30). A contributing effect of MIF to adipocyte lipid content is also supported by the observed overall weight gain in Mif^−/− mice (∼4 g), which is only partly accounted for by an increase in liver weight (1.5 g). In hepatocytes, increased de novo fatty acid synthesis and triglyceride formation could alternatively be responsible for the observed increased triglyceride levels in Mif^−/− livers. In NASH, WAT releases increased levels of FFAs that are then stored in the liver as triglycerides (7–9). We measured increased expression levels of the responsible enzymes for incorporating fatty acids into triglycerides. However, we also detected an up-regulation of enzymes and transcription factors in the livers of Mif^−/− mice involved in de novo fatty acid synthesis. A mechanistic explanation for the up-regulated lipogenic genes in Mif^−/− mice could be the observed increased Tnf-α expression. High Tnf-α expression has been found to correlate with an increased expression of Srebp1 in murine NASH models (31), which then up-regulates Acc and Fas (32). Thus, our findings are in keeping with Donnelly et al. (4), who showed that hepatic triglyceride in patients with NAFLD is generated by de novo synthesis and could be due to increased fatty acid uptake.

Increased fatty liver degeneration of Mif^−/− mice was confirmed in an independent murine NASH model, i.e., the MCD diet model. MCD mice develop an elaborated state of steatohepatitis, but in contrast to the HFD model and human NASH, animals show an inverse metabolic profile (33), losing weight and not developing impaired glucose tolerance, and have an impaired secretion of hepatic very low density lipoprotein (VLDL; ref. 34).

Progression of fatty liver disease is often associated with the development of metabolic syndrome (35). Mif^−/− mice showed increased weight gain, impaired glucose tolerance, and increased triglycerides compared to control mice fed the same HFD.

Thus, Mif^−/− mice are prone to develop a metabolic syndrome, a notion corroborated by analysis of adipocytokines. Adiponectin levels are reduced in obese patients (36), while those of leptin are increased in the serum of obese patients (37), who show leptin resistance. Adiponectin down-regulates Srebp1 and could thus protect the liver from steatosis (7–9, 38), whereas leptin up-regulates Srebp1. Therefore, our findings are in line with the notion that Mif-deficiency promotes metabolic syndrome progression.

The hepatoprotective effect of MIF in NASH was accompanied by anti-inflammatory activities. We observed significant increases in the hepatic recruitment of macrophages in Mif^−/− mice combined with an increase in gene expression of proinflammatory cytokines, such as Tnf-α or Il-6. Curiously, an inverse inflammatory phenotype was detected in WAT, with strongly decreased macrophage content and reductions in CD3⁺ T and NK/NKT cells in WAT of Mif^−/− mice fed HFD. Similarly, the cytokine expression profile was the inverse of that in the liver.

Moreover, an analysis of M1 and M2 macrophage subsets indicated that the inflammatory phenotype in Mif^−/− mice is likely more complex. Although the data so far rely only on qPCR, they suggest that Mif deficiency skews the macrophage phenotype toward M2 in the liver. This is likely due to the observed elevated levels of hepatic Il-13 and Il-6, and a trend toward increased levels of Il-10 and Ccl2, that all favor M2 polarization, although the observed increase in Tnf-α levels appears counterintuitive (39, 40). A switch from M1 to M2 macrophage polarization often characterizes the transition from early to chronic phases of disease, and M1 polarization has been associated with steatosis, while liver fibrosis is correlated with M2 phenotypes (40). However, macrophage plasticity has been recognized to be highly complex, with M2a, 2b, and 2c and regulatory macrophage subtypes identified (39, 40). In fact, although a recent publication has directly implicated MIF as an M2-inducing factor in the tumor microenvironment (41), MIF has generally been viewed to favor M1 polarization because it is an upstream inflammatory mediator eliciting Tnf-α and IFNγ production, suggesting that the contribution of MIF in macrophage polarization responses may be disease context and stage dependent. Although the 16-wk HFD-NASH model applied in our current study represents a relatively early NASH model, with various models spanning 6–72 wk, it can thus not be excluded that the phenotype and/or M1/M2 profile of the Mif^−/− mice might be different at much earlier time points, i.e., 4–12 wk, in our models of NASH pathogenesis.

In WAT, the macrophage and inflammatory profile appeared more straightforward. While the overall observation of an inverse phenotype compared to the liver is surprising, marker analysis unanimously argues for an anti-inflammatory phenotype in Mif^−/− WAT with total macrophage counts, including M1 and M2 markers, and all inflammatory cytokines down-regulated. This indicated that Mif deficiency strongly impaired monocyte recruitment into adipose tissue, consistent with a previous report (42). In summary, Mif^−/− livers show increased macrophage accumulation with apparent M2 skewing, whereas in Mif^−/− WAT, macrophage infiltration is strongly reduced, and the cytokine profile is anti-inflammatory. These observations invite future investigations to clarify how the inflammatory phenotype relates to the hepatoprotective effect of MIF in NASH.

Hepatocytes are the main cell type in the liver and the main site of fatty acid synthesis and storage. MIF was shown to modulate hepatocyte functions (22), and we hypothesized that the hepatoprotective effect of MIF in steatosis was mainly mediated through MIF action on hepatocytes. The main MIF receptor on hepatocytes is CD74, which is also critical in hepatic stellate cells (23). We showed that stellate cells are able to shed CD74 to produce a soluble receptor fragment, which, at least in models of autoimmune hepatitis, has protective effects (19). However, the observed inflammatory and macrophage recruitment profile in Mif^−/− livers and, particularly, WAT could also be due to activation/impaired activation of the CXCR2 or CXCR4 MIF/receptor axis (42).

To explore the mechanism of presumed MIF/hepatocyte-mediated effects in NASH, we applied a hepatocyte cell model of fatty degeneration (26, 43), mimicking inflammatory steatosis conditions. Recombinant MIF significantly reduced fatty hepatocyte degeneration in both Hepa1-6 and primary murine hepatocytes. The mechanistic experiments revealed that CD74, but not CXCR2 or CXCR4, was responsible for the attenuating the MIF effect. CD74 has been found to exert tissue-protective effects in the ischemic myocardium and the fibrotic liver through activating the AMPK signaling pathway (16, 23). Consistently, an involvement of AMPK in MIF's attenuating effect on hepatocyte lipid accumulation was seen. We assume that the protective effect of the MIF/CD74/AMPK axis in hepatocyte fatty degeneration is mediated through an effect on lipid synthesis/triglyceride formation. AMPK phosphorylates phosphofructokinase-2 (PFK2) and ACC, which results in a reduced activity of this enzyme essential in fatty acid synthesis (44). The mechanistic studies provide evidence that the protective effect of MIF in steatosis is mediated through hepatocyte targeting.

It is interesting to compare our results on MIF's role in NASH with those of a previous study in which the role of MIF in alcoholic steatohepatitis was examined. Overall, Barnes et al. (22) found that MIF contributed to ethanol-induced steatosis and hepatocyte damage. As a main underlying mechanism, they found that MIF promotes chemokine production and immune cell infiltration in the liver during ethanol feeding, indicating that under conditions of alcohol feeding, a different pattern of fatty degeneration of the liver develops. In alcoholic liver disease, an increased fatty acid synthesis rate, a decreased degree of fatty acid oxidation, and impaired secretion of VLDL from the liver appear to represent the critical underlying mechanisms. The last effect is reminiscent of the hepatic VLDL transport defect after MCD feeding. In NAFLD, an increased delivery rate of fatty acids to the liver as well as an increased SREBP-1-mediated signaling response is predominant (45).

In summary, we have demonstrated a protective effect of MIF in NASH. This effect was shown in two independent NASH models in vivo. While the assignment of the responsible cell source producing MIF in these conditions will have to await the generation of corresponding cell-specific conditional Mif-knockout mouse models, and while the inherent nature of the applied chronic in vivo models of NASH does not exclude the possibility that hepatoprotection by MIF may be indirectly mediated by other mediators, in vitro evidence argues for a role of the MIF/CD74/AMPK axis in hepatocytes in mediating, at least in part, the protective effect. Despite MIF's spectrum of proinflammatory activity, these results identify MIF-agonistic agents as potential therapeutic options for the treatment of NASH.