Medaka as a model for human nonalcoholic steatohepatitis

Nonalcoholic fatty liver disease (NAFLD) is the general term for fatty liver diseases that are not the result of a history of alcohol consumption. NAFLD is the most common cause of human liver dysfunction, and it is estimated that about 30% of the general population in the USA suffers from excessive fat accumulation in the liver (Browning et al., 2004). The incidence of NAFLD, which has been increasing in recent years, is closely tied to obesity, diabetes, hyperlipidemia and insulin resistance. Indeed, NAFLD is considered to be a manifestation of metabolic syndrome in the liver (Powell et al., 1990; Sanyal, 2002). NAFLD can be broadly divided into two subgroups: (1) non-progressive simple steatosis, and (2) nonalcoholic steatohepatitis (NASH) with ballooning degeneration and fibrosis (Schaffner and Thaler, 1986; Younossi et al., 1998; Brunt et al., 1999; Matteoni et al., 1999). In cases of progressive chronic liver disease, NASH may progress from steatohepatitis to liver cirrhosis, and may eventually lead to hepatocellular carcinoma.

There have been many attempts to create NASH models in rodents through the use of either genetic mutation, or dietary or pharmacological manipulation (Anstee and Goldin, 2006). The medaka (Oryzias latipes) is a small freshwater fish found in Japan and Asia. This member of the ricefish family has a history of use as an animal model in Japan and a number of purebred strains exist (Masahito et al., 1989). Medaka compare favorably to rodents as experimental animals for drug screening because medaka have a high reproductive rate, mature rapidly, and cost little in terms of rearing space and daily maintenance owing to their small size. Sequencing of the medaka genome has been completed and techniques for producing transgenic and knockout animals have been established (Kasahara et al., 2007; Yamauchi et al., 2000). Physiologically, medaka are omnivores and metabolize sugars and lipids in a manner analogous to that of mammals (Sheridan, 1988; Brown and Tappel, 1959). However, despite their metabolic similarities to humans, medaka have not been used previously for NASH research. In this study, we fed a HFD to medaka to induce NASH and determined whether the progression of liver disease in these fish was similar to the progression of human NASH. In addition, we used our medaka NASH model to verify the efficacy of the administration of the n-3 polyunsaturated fatty acid (PUFA) eicosapentaenoic acid (EPA) as a NASH therapy. Our results have shed much-needed light on the mechanisms potentially underlying NASH in humans, and may point to new avenues of therapy for this debilitating disorder.

For an animal model to be useful for determining the molecular and cellular basis of NASH, it must show the same metabolic abnormalities as the human disease. These anomalies include obesity and insulin resistance; liver damage owing to ‘fatty liver’ and hyperlipidemia; macrovesicular fat deposition in the liver; infiltration by inflammatory cells; and ballooning degeneration of hepatocytes. Like mammals, medaka have livers, gall bladders, digestive tracts and other internal organs in the peritoneum (Fig. 1A,B). In normal medaka liver tissue, the hepatocytes are arranged in sheets that are separated by a sinusoidal mesh. Histological examination of normal medaka liver shows that the portal vein, hepatic artery and bile duct are independent, as they are not organized in portal triads (Fig. 2A).

To induce NASH in medaka, 8-week-old fish (n=14/group) were fed either a control diet or a high-fat diet (HFD) for 12 weeks, and monitored for gross anatomical and histological changes at 4, 8 and 12 weeks after HFD initiation. Control medaka grew normally over the 12-week period, from 95±17 mg to 241±32 mg. By contrast, all 14 fish that were fed the HFD exhibited significantly greater body weights, reaching 368±80 mg by the end of the 12 weeks (P<0.05). HFD-medaka also showed distention of the abdomen, owing to increased fat deposition in the viscera, and the presence of a white, swollen liver (Fig. 1C–E). Edematous degeneration of the liver was observed in five out of 14 (36%) HFD-medaka (Fig. 1F). The mean liver weight (Fig. 1G) and the mean liver weight:body weight ratio (Fig. 1H) of HFD-medaka were increased by 2.8-fold and 1.8-fold, respectively, compared with controls, consistent with the observed hepatomegaly in HFD-medaka. Lastly, the weight of intrahepatic total lipids in the HFD-medaka was 6.2-fold greater than in controls (Table 1).

Compared with medaka that were fed a control diet, medaka fed the HFD for 4 weeks exhibited considerably more macrovesicular fat deposition around the hepatic veins (data not shown). By 8 weeks, fat deposition had spread over the entire liver (Fig. 2B; compare with Fig. 2A). By 12 weeks, focal ballooning degeneration and inflammatory cell infiltration in the livers of HFD-medaka were prominent (Fig. 2C,D), consistent with the edematous liver degeneration noted upon gross examination (Fig. 1F). Histopathologically, nine out of 14 (64%) and four out of 14 (29%) HFD-medaka displayed steatohepatitis and simple steatosis, respectively. Liver fibrosis (Fig. 2E,F) and fat accumulation (Fig. 2G) were also observed in these fish. Electron microscopic examination of liver cells from HFD-medaka revealed the presence of lipid droplets (Fig. 2H,I), and the ballooning hepatocytes showed cytoplasmic vacuolation and enlarged lysosomes (Fig. 2H,J).

A two-hit theory has been proposed as the mechanism underlying the onset of NASH (Day and James, 1998). The first hit involves increased expression of fatty acid transport-related and lipogenic genes that alter the lipid metabolism inside the liver, such that the liver becomes steatotic. The second hit takes the form of inflammatory cytokines and oxidative stress that drive inflammation and fibrosis. To determine whether a classical first hit was occurring in our HFD-medaka, we analyzed serum components of these fish. Triglyceride (TG) levels in HFD-medaka serum were significantly elevated over those in controls by 4 weeks, and were sevenfold greater by 12 weeks (Fig. 3A). When we analyzed these serum TGs by high-performance liquid chromatography (HPLC), we found that very low-density lipoprotein triglyceride (VLDL-TG) was increased by 15-fold in the serum from HFD-medaka at 12 weeks (2362 mg/dl) compared with controls (160 mg/dl) (Fig. 3B). These results suggested that the HFD-medaka were experiencing an increase in endogenous fatty acid synthesis by hepatocytes. As shown in Fig. 3C, plasma glucose levels were also significantly elevated in HFD-medaka at 8 weeks, and were fourfold greater than in controls by 12 weeks. Serum alanine aminotransferase (ALT) levels were elevated the most (3.7-fold) in HFD-medaka at 4 weeks, but continued to be significantly increased compared with controls at 12 weeks (Fig. 3D). These altered serum parameters parallel our histological examinations of HFD-medaka, in which steatosis was seen at 4 weeks, macrovesicular fat deposition was widespread at 8 weeks, and ballooning degeneration was observed at 12 weeks. Taken together, these data suggest that a first hit of hyperlipidemia, hyperglycemia and hepatic steatosis occurred in HFD-medaka, followed by a second hit of liver dysfunction owing to ballooning hepatocyte degeneration. Thus, our observations are consistent with the twohit theory of human NASH development, in which a metabolic abnormality sets the stage for the development of NASH symptoms.

To investigate, at a molecular level, the changes in the fatty acid metabolism of medaka liver during fatty acid loading periods, we used reverse transcription (RT)-PCR to analyze the mRNA expression levels of genes associated with fatty acid synthesis and oxidation. These genes included the main regulator of fatty acid synthesis, sterol regulatory element-binding protein-1c (SREBP-1c); the target genes of SREBP-1c, fatty acid synthase (FAS) and acetyl-CoA carboxylase 1 (ACC1); the main regulator of fatty acid β-oxidation, peroxisome proliferator-activated receptor α (PPARA); the mitochondrial β-oxidation marker carnitine palmitoyltransferase 1 (CPT1); and the peroxisome β-oxidation marker acyl-CoA oxidase 1 (ACO1). Compared with controls, the mRNA expression levels of SREBP-1c, FAS and ACC1 were all elevated in HFD-medaka liver at 4 weeks (Fig. 3E, left), confirming that fatty acid synthesis had accelerated in these animals. Conversely, we observed decreased expression of PPARA and CPT1, but increased expression of ACO1 (Fig. 3E, right), indicating decreased mitochondrial β-oxidation and accelerated peroxisomal β-oxidation in the HFD-medaka liver. Taken together, these data suggest that HFD-medaka suffer from increased fatty acid synthesis accompanied by decreased fatty acid β-oxidation and inflammation, accounting for their swollen, fatty livers.

To investigate the abnormal fatty acid deposition observed in the HFD-medaka liver, we compared the composition of the fats in control livers with HFD-medaka livers (Table 1). Oleic acid (C18:1n-9) levels in HFD-medaka were clearly higher than in controls, whereas linoleic acid (C18:2n-6), arachidonic acid (C20:4n-6), α-linolenic acid (C18:3n-3), EPA (C20:5n-3) and docosahexaenoic acid (DHA) (C22:6n-3) levels were lower. Although concentrations of both n-3 PUFAs and n-6 PUFAs had declined, the n-3:n-6 ratio was reduced in HFD-medaka because the decrease in n-3 PUFAs was greater than the decrease in n-6 PUFAs. These results are consistent with features of human NASH, in which PUFA levels and n-3:n-6 PUFA ratios are reportedly low in the liver (Araya et al., 2004; Puri et al., 2007). Interestingly, levels of mead acid (C20:3n-9), a marker of essential fatty acid deficiency, were increased in HFD-medaka liver.

Because administration of the n-3 PUFA EPA has been investigated as a therapy for human NASH, we explored whether EPA treatment could slow disease progression in our medaka NASH model. Medaka were fed a HFD with or without EPA for 12 weeks. The HFD+EPA medaka showed increased fat deposition in the internal organs and a white, swollen liver similar to that observed in the HFD group (Fig. 4A). However, histological examination revealed that the degree of fat deposition in the HFD+EPA group appeared to be less than that in the HFD group during the same period (Fig. 4B). Indeed, analysis of total lipids showed that the fat levels in the HFD+EPA liver were 48% lower than in the HFD liver (Table 1). Moreover, unlike HFD-medaka, the HFD+EPA medaka were not hyperglycemic after fasting (Fig. 4C). HLPC analysis showed that serum TG and VLDL-TG levels were 16% (3132 vs 2643 mg/dl) and 21% (2362 vs 1873 mg/dl) lower in HFD+EPA livers compared with HFD livers, respectively (Fig. 4D). EPA treatment increased EPA (C20:5n-3) and DHA (C22:6n-3) levels, as well as the n-3:n-6 PUFA ratio, in the liver (Table 1). Finally, mead acid (C20:3n-9) levels were decreased in HFD+EPA livers compared with HFD livers (Table 1), suggesting that any essential fatty acid deficiencies that were present had been mitigated.

With respect to gene expression, the levels of SREBP-1c, FAS and ACC1 mRNAs were reduced in HFD+EPA medaka compared with HFD-medaka, and the accelerated fatty acid synthesis caused by HFD consumption was abrogated (Fig. 4E, top). By contrast, the levels of PPARA and CPT1 expression were increased, ACO1 expression was decreased, and markers of mitochondrial β-oxidation became predominant in the HFD+EPA liver (Fig. 4E, bottom). Thus, the pattern of gene expression in livers of HFD+EPA medaka was essentially the same as that in medaka fed the control diet. These data indicate that de novo fatty acid synthesis in HFD-medaka liver decreases upon EPA treatment, and that normal fat utilization is restored.

It has been reported that liver PUFA levels and n-3:n-6 PUFA ratios are low in NASH patients (Araya et al., 2004; Puri et al., 2007). We therefore investigated whether a deficiency in n-3 PUFA could induce NASH in medaka. We prepared an n-3 PUFA-deficient (n-3PUFA⁻) diet with about the same energy content as the control diet, and used it to feed medaka for 8 weeks. Compared with controls, the n-3PUFA⁻ medaka showed moderate increases in fat deposition in internal organs, and their livers, although pink, were swollen (Fig. 5A). By gross observation, five out of ten (50%) n-3PUFA⁻ medaka showed edematous degeneration of the liver, and histopathological examination revealed that steatohepatitis was present in seven out of ten (70%) of these animals (Fig. 5B). The mean serum ALT value in n-3PUFA⁻ medaka was 442±228 IU/l (Fig. 5C), confirming the presence of liver damage. Thus, NASH is induced in medaka by conditions of n-3PUFA deficiency.

Medaka compare favorably to rodents as experimental animals for drug screening because medaka have a high reproductive rate, mature rapidly, and cost little in terms of rearing space and daily maintenance owing to their small size. Our results show that NASH can be induced in medaka by feeding a HFD, and that the development of NASH can be largely prevented by EPA administration. Importantly, we have also demonstrated that n-3 PUFA deficiency can induce NASH in medaka.

We were able to induce steatohepatitis in medaka through feeding, thus creating a NASH model that is easily established in a relatively short period of time. The features of our model include the efficient induction of ballooning degeneration, a key feature in the diagnosis of human NASH (Brunt et al., 1999; Matteoni et al., 1999; Neuschwander-Tetri and Caldwell, 2003). The frequency of ballooning degeneration in normal medaka is 10–20% (Bunton, 1990; Boorman et al., 1997; Brown-Peterson et al., 1999), whereas our HFD-medaka showed a ballooning degeneration frequency of greater than 60%. By contrast, Deng et al. have shown that 46% of C57BL/6 mice fed a HFD through an implanted gastrostomy tube develop steatohepatitis (Deng et al., 2005). Thus, our medaka model of NASH induction is slightly superior to the mouse model in terms of the efficiency of ballooning degeneration induction (64% vs 46%).

The mechanisms underlying the digestion and absorption of lipids; the transport of exogenous and endogenous lipids; and fatty acid oxidation are almost identical in fish and mammals (Brown and Tappel, 1959; Sheridan, 1988). However, there are some differences in how fatty acids are released from intestinal epithelial cells into blood vessels or lymph ducts (Sheridan, 1988). In mammals, short chain fatty acids (SCFAs) and medium chain fatty acids (MCFAs) that are absorbed by intestinal epithelial cells are released directly into the portal vein without esterification. Most long chain fatty acids (LCFAs) are resynthesized into TGs and released into lymph ducts as chylomicron particles, but some are released into the portal vein without esterification. In fish, LCFAs are rapidly processed into free fatty acids (FFAs), and only later are resynthesized as TGs and released as chylomicron particles into the blood (Robinson and Mead, 1973; Sheridan, 1988). Our medaka NASH model was created by feeding a HFD based on LCFAs, and we believe that many of the consumed LCFAs arrive in the liver as FFA-albumin complexes. In obese humans that develop NASH, insulin signaling in adipocytes is disrupted such that hormone-sensitive lipases hydrolyze TGs into fatty acids (Hotamisligil et al., 1995). These fatty acids are subsequently released from the adipocytes and enter the liver through the portal vein, exacerbating NASH development (Hotamisligil et al., 1995; Day, 2002). We believe that a similar scenario occurs in our HFD-medaka, such that FFAs derived from consumed LCFAs flow into the liver. After obesity develops in medaka, FFAs from fat tissues may also join this flow, perhaps explaining why NASH develops more easily in medaka than in other animal models.

In contrast to the lipids of terrestrial vertebrates, fish lipids contain a high proportion of PUFAs, particularly the n-3 PUFAs whose function is to maintain lipid liquidity at low temperature. However, as with terrestrial mammals, fish are unable to synthesize n-6 and n-3 PUFAs, making these molecules essential fatty acids that must be consumed. The HFD used in our study included 64.9% oleic acid, 12.8% palmitic acid (C16:0), 7.6% stearic acid (C18:0), 10.3% linoleic acid and 0.2% α-linolenic acid. Monounsaturated fatty acids (MUFAs) and saturated fatty acids (SFAs) accounted for most of the fatty acids, with only an extremely small proportion of αlinolenic acid being present. The synthesis of highly unsaturated fatty acids (HUFAs), such as arachidonic acid, EPA, DHA and other fatty acids with a carbon chain length of greater than 20, depends on the action of delta-6 and delta-5 desaturases (Sprecher, 1981). These enzymes preferentially desaturate n-3PUFAs then n-6PUFAs and then n-9PUFAs in a process that is controlled by HUFAs (Cho et al., 1999). Thus, EPA and DHA levels and n-3:n-6 PUFA ratios declined in our HFD-medaka owing to their greatly reduced consumption of α-linolenic acid, EPA and DHA. By contrast, the levels of mead acid increased in HFD-medaka owing to the desaturation of oleic acid. In HFD-medaka treated with EPA, the increased levels of EPA and DHA in the liver probably inhibited the activity of delta-6 and delta-5 desaturases. As a result, the conversion of linoleic acid to arachidonic acid, as well as that of oleic acid to mead acid, was inhibited.

The n-3 PUFAs are ligands for PPARα, and PPARα activation increases the expression of fatty acid oxidation enzymes such that fatty acid oxidation is promoted (Gottlicher et al., 1992; Forman et al., 1997; Ren et al., 1997). Therefore, when PUFAs are deficient, fatty acid oxidation declines and livers accumulate fat (Fukazawa et al., 1971). In our medaka NASH model, expression of the mRNA encoding the lipogenic transcription factor SREBP-1c was induced, and expression levels of the fatty acid synthesis pathway enzymes ACC1 and FAS were increased. Conversely, expression levels of the lipolytic transcription factor PPARα, as well as those of CPT1 (the rate-limiting enzyme in mitochondrial β-oxidation), were reduced. Concomitantly, the expression of ACO1, which is involved in peroxisome β-oxidation, was increased. The observed reduction in CPT1 expression may be attributable to the production of malonyl CoA, an intermediate product of fatty acid synthesis. Malonyl CoA interferes with CPT1 expression, and it is thought that compensatory peroxisome β-oxidation is induced (McGarry et al., 1977).

In humans, n-3 PUFA deficiency has been linked to neuropathy and immune system impairment (Holman, 1998). Variations in n-3 PUFA concentrations in the liver may thus indirectly regulate the inflammation associated with NASH. The notion that n-3 PUFAs are involved in the pathology seen in our medaka NASH model is supported by the observation that the expression levels of genes involved in fatty acid metabolism were restored to control levels when HFD-medaka were treated with EPA.

In conclusion, our study shows that the induction of fatty liver and the reproduction of NASH pathology in medaka can be achieved by feeding a diet that is low in EPA and DHA, but that has about the same energy content as a control diet. Moreover, our model is easily, cheaply and rapidly established in the laboratory, allowing ample latitude for exploration of its usefulness. We anticipate that our medaka NASH model will indeed be helpful for clarifying human NASH pathology and for assisting in the development of novel therapeutics aimed at preventing or alleviating this burgeoning health problem.

Himedaka strain Cab (an orange-red variety of medaka, Oryzias latipes) fish that were 8 weeks old were used for most experiments. Fish were maintained at a stocking level of ten fish/tank in tap water with aeration. The ten fish in a given tank received a daily ration of 200 mg of the diet prescribed for that group, an amount that was consumed completely within 14 hours. All fish were maintained in accordance with the Animal Care Guidelines of Yamaguchi University. During experiments, fish were kept in plastic tanks covered with plastic covers and illuminated with fluorescent light from 8:00 a.m. to 10:00 p.m. The tank water temperature was maintained at 26±1°C.

The proportions of protein, fat and carbohydrate, as well as the fatty acid compositions, of the control, HFD and n-3 PUFA-deficient diets that were used in this study are shown in Tables 2 and 3. The energy content of the control diet (Hikari Crest; Kyorin Co. Ltd, Hyogo, Japan) was 3.3 kcal/g, with 25.3% of the calories from fat, 62.5% of calories from protein, and 13.8% of calories from carbohydrate, plus vitamins and minerals as recommended. The energy content of the high-fat diet (HFD32; CLEA Japan Inc., Tokyo, Japan) was 5.1 kcal/g, with 56.7% of calories from fat, 20.1% of calories from protein, and 23.2% of calories from carbohydrate, plus vitamins and minerals as recommended. The n-3 PUFA-deficient (EPA- and DHA-deficient) diet, which consisted of fish meal-free powder (F1 diet; Funabashi Farm Co. Ltd, Chiba, Japan), had an energy content of 3.6 kcal/g, with 11.0% of calories from fat, 23.9% of calories from protein, and 65.1% of calories from carbohydrate, plus vitamins and minerals as recommended. For EPA treatment, eicosapentaenoic acid ethyl ester (EPA-E, purity >99%; Mochida Pharmaceutical Co. Ltd, Tokyo, Japan) was mixed with HFD32 to a concentration of 5% by weight.

Euthanized fish were slit open from the anal vent to the gills, and the entire body was fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (Muto, Tokyo, Japan). The liver was dissected, dehydrated in alcohol, and embedded in paraffin according to standard procedures. Serial sections (3–5 μm thick) were cut and stained with hematoxylin and eosin (H&E). Liver fibrosis was assessed by Gitter staining. Intracellular lipids were stained with Oil Red O to analyze fat accumulation in the liver.

Chunks of medaka liver were fixed with 2% glutaraldehyde and 4% paraformaldehyde in 0.1 M phosphate buffer for 30 minutes at 4°C. Fixed liver pieces were washed thoroughly in 0.1 M phosphate buffer, post-fixed for 30 minutes with 1% osmium tetraoxide in 0.1 M phosphate buffer, and then block-stained with 2% aqueous uranyl acetate for 30 minutes to enhance the contrast for electron microscopy. Samples were dehydrated through a graded ethanol series, infused with propylene oxide, and embedded in epoxy resin. Ultra-thin sections were collected on copper grids, stained with 2% aqueous uranyl acetate for 8 minutes and 0.5% lead citrate for 8 minutes, followed by examination under a Hitachi H7500 electron microscope.

Blood samples were obtained from medaka following a 12-hour fast. Fish were kept on ice for 1–2 minutes and then bled by cutting a ventral portion of the tail fin. Blood was collected in a microcapillary tube and the volume measured. Blood samples were kept at room temperature for 1 hour before centrifugation at 1200 × g for 30 minutes at 4°C. Serum ALT concentrations were measured using a Fuji Dry-Chem 3500 (Fuji Film Co. Ltd, Tokyo, Japan). Plasma glucose levels were determined using a Glucocard G-meter (Arkray Co. Ltd, Kyoto, Japan). Cholesterol and TG profiles in total lipoproteins were analyzed using a dual-detection HPLC system (Skylight Biotech, Akita, Japan) with two tandemconnected TSKgel LipopropakXL columns (300 ×7.8 mm; Tosoh, Japan).

The fatty acid composition of homogenized liver tissue (20 mg tissue/ml saline) was determined by capillary gas chromatography. Total lipids were extracted by Folch’s procedure (Folch et al., 1957), and fatty acids were methylated with boron trifluoride and methanol. Methylated fatty acids were analyzed using a Shimadzu GC-17A gas chromatograph (Shimadzu Co. Ltd, Kyoto, Japan) and a BPX70 capillary column (0.25 mm internal diameter ×30 mm; SGE International Ltd, Melbourne, Australia). Tricosanoic acid (C23:0) was used as the internal standard. The minimum detectable fatty acid concentration detected by this assay is 0.5 μg/ml.

To avoid any acute effects of food intake, fish were fasted overnight prior to sacrifice. Livers were isolated and total RNA was extracted and purified using the RNeasy kit (Qiagen, Hilden, Germany). cDNAs were synthesized using purified RNA plus random hexamers and the Transcriptor First Strand cDNA synthesis kit (Roche, Indianapolis, IN). Semi-quantitative RT-PCR was performed using a PCR master mix (Promega, Madison, WI) and the primers listed in Table 4. Gene expression levels were normalized against β-actin (endogenous control).

Numerical data are expressed as the mean ± S.D. The Student’s t-test was performed to assess statistical significance among groups of medaka. P values less than 0.05 were considered to be significant.

This study was supported by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (Nos. 18590737, 18659209, 19590693, 19390199, 20659116 and 21659189); the Knowledge Cluster Initiative; the Ministry of Health, Labour and Welfare; and Japan Aerospace Exploration Agency (JAXA). We also thank Dr Masaaki Soma and Dr David Tosh for critical reading of the manuscript.