The term ‘non-alcoholic steatohepatitis’ or NASH was first used by Ludwig et al. in 1980 to describe ‘the pathological and clinical features of non-alcoholic disease of the liver associated with the pathological features most commonly seen in alcoholic liver disease itself’.1 This description remains appropriate because non-alcoholic fatty liver disease (NAFLD) can range from simple steatosis, through NASH and fibrosis to cirrhosis with fat.2 In addition to fat, the histological diagnosis of NASH ideally requires evidence of (i) hepatocyte injury, manifest by swollen or ‘ballooned’ cells; (ii) an inflammatory infiltrate, predominantly neutrophils, with or without (iii) fibrosis, typically perivenular/pericellular in distribution. Each of these features can be graded to derive a score recently proposed by Brunt et al.3 Rigorous exclusion of alcohol as a cause of the histology is, of course required and this is best achieved by a combination of repeated questioning of patients and ideally friends/relatives, frequent random blood alcohol estimations and measurement of mean cell corpuscular volume (MCV). No pattern of liver blood test abnormalities is either specific or sensitive enough to distinguish between NAFLD and alcohol-related liver disease.
Why is the pathogenesis of non-alcoholic steatohepatitis important?
Interest in the pathogenesis of NASH has increased markedly in recent years for a variety of reasons. First, NASH is increasingly recognized to be a common condition, second only to viral hepatitis as a reason for referral in one study of urban North American office practise.4 A recent study in the United Kingdom demonstrated that NAFLD, either fatty liver alone or NASH, is the diagnosis in approximately two-thirds of patients presenting with abnormal liver function tests unrelated to viral hepatitis, immune disease or excessive alcohol intake.5 Some of this increase may be artifactual due to the increasing awareness of the condition; however, a ‘real’ increase seems likely in view of the increased prevalence of the conditions associated with NAFLD. In the various series reported thus far, 50% of patients are obese, particularly with a central or ‘male’ pattern and up to 40% of patients have type II diabetes mellitus (T2DM) or impaired glucose tolerance.2 Insulin resistance is an extremely common finding and may be a universal finding.6 Hypertension and hyperlipidemia are also common. These associations have led to the suggestion that NASH, or NAFLD in general, is the liver manifestation of the insulin resistance or ‘metabolic’ syndrome X.6,7
The interest in the pathogenesis of NASH has also increased in view of accumulating evidence that it may progress to advanced liver disease in at least some individuals. A review of the published biopsy studies of NASH reported up to 1998 showed that between 15 and 50% of patients had fibrosis or cirrhosis on their index biopsy.2 This has now been supplemented with a retrospective follow-up study of patients with NAFLD showing that over a median 8-year follow up, 25% of patients with NASH progressed to cirrhosis and 11% died of a liver-related cause. This is in contrast to only 3.4% of patients with simple fatty liver who developed cirrhosis and less than 2% who died a liver-related death.8 Further evidence that NASH may progress to cirrhosis has been provided by two studies suggesting that a large proportion, if not the majority, of patients with cryptogenic cirrhosis have the classical risk factors for NASH (obesity and T2DM), suggesting that NASH is probably the underlying cause of this disease in the majority of patients.9,10
The role of steatosis in the pathogenesis of non-alcoholic steatohepatitis
A growing body of evidence suggests that, rather than being an ‘innocent bystander’, steatosis per se may be a ‘guilty party’, playing a role in the progression of NAFLD to NASH and fibrosis.11 This evidence has come from studies in NAFLD as well as in alcoholic liver disease and hepatitis C. In all of these conditions, the severity of steatosis predicts either the risk of concomitant steatohepatitis or the risk of progression to fibrosis and cirrhosis.12–15 In addition, studies in alcoholic fatty liver have shown that the severity of fat correlates with the degree of hepatic stellate cell (HSC) activation.16 Hepatic stellate cells are the principal cells in the liver responsible for the production of extracellular matrix proteins and, accordingly, fibrosis.
Mechanisms of hepatocyte injury in non-alcoholic steatohepatitis
Explaining the association between the severity of steatosis and the risk of necroinflammation and fibrosis first requires an understanding of the mechanisms of hepatocyte injury thought to play a role in NAFLD. In looking for clues as to the nature of these mechanisms, it is pertinent to emphasize that the histological features of NASH are identical to those of alcoholic hepatitis. The similarity between the histology of these conditions suggests that common mechanisms of injury are likely to be involved. Currently at least three mechanistic pathways of liver injury are thought to play a role in the pathogenesis of alcoholic hepatitis: oxidative stress, endotoxin-mediated cytokine release and immunologically mediated mechanisms (reviewed in 17).
Oxidative stress in alcoholic liver injury
In heavy drinkers oxidative stress arises as a result of alcohol metabolism either via cytosolic alcohol dehydrogenase or the microsomal cytochrome P450 CYP2E1. This oxidative stress leads to lipid peroxidation, damaging plasma and intracellular membranes, resulting in cell death due to necrosis and/or apoptosis.18 In addition, the end products of lipid peroxida-tion (malondialdehyde (MDA) and hydroxynonenal (HNE)) are capable of inducing all of the classical features of alcoholic hepatitis. Both are chemotactic for neutrophils. They also stimulate the transcription of extracellular matrix-encoding genes in HSC, and, via induction of the transcription factor nuclear factor (NF)κB, they can lead to the increased transcription of cytokines and other pro-inflammatory genes in both hepatocytes and Kupffer cells.
Endotoxin and cytokines in alcoholic liver injury
The importance of endotoxin-mediated release of cytokines as an important pathway of liver injury in alcoholic liver disease has been stressed by the animal work of Thurman et al (reviewed in 17). In this model, chronic alcohol consumption first leads to increased intestinal permeability to endotoxin. The resulting portal endotoxemia then activates hepatic Kupffer cells to produce the pro-inflammatory cytokine tumor necrosis factor-α (TNF-α). Tumor necrosis factor-α subsequently induces mitochondrial oxidative stress in hepatocytes, leading to necrosis and apoptosis, and in endothelial cells leads to increased transcription of adhesion molecules such as intercellular adhesion molecule-1 (ICAM-1), leading to the recruitment of inflammatory cells. Most of the evidence for this pathway has come from the Tsukamoto–French continuous intragastric feeding rodent model of alcoholic liver injury, although there are some supportive data from studies in humans. Alcohol certainly increases gut permeability in humans,19 and the serum level of TNF-α is increased in patients with alcoholic hepatitis and correlates with abnormal liver function and mortality.20
Immune mechanisms in alcoholic liver injury
Acetaldehyde, the principal metabolite arising from alcohol metabolism by alcohol dehydrogenase, and MDA and HNE and other products of alcohol metabolism including hydroxyethyl radicals are capable of binding covalently to ‘self’ proteins to form adducts that are capable of stimulating an immune response. These adducts have been demonstrated in the blood and liver of animals fed alcohol, and humans with alcoholic liver disease. Importantly, the adducts are present on the surface of hepatocytes. Anti-adduct and autoantibodies have been demonstrated in the serum of heavy drinkers, and titers of these alloantibodes and autoantibodies are highest in patients with advanced alcoholic liver disease.21,22 Finally, the anti-adduct antibodies have been shown to be capable of mediating antibody-dependent cytotoxicity (ADCC) in hepatocytes isolated from alcohol-fed rats.23
Evidence of alcohol-related disease mechanisms in non-alcoholic steatohepatitis
Given the histological similarity between NASH and alcoholic hepatitis, is there any evidence that these mechanisms of alcoholic liver injury play a role in the liver injury occurring in NASH? First, with respect to oxidative stress, several studies using immunohistochemical methods to look for proteins adducted to MDA and HNE have provided evidence that lipid peroxidation occurs in the livers of patients with NAFLD.11 Moreover, recent work has demonstrated a variety of potential sources of oxidative stress in patients with NAFLD. At least one study has demonstrated evidence of increased iron in the liver of patients with NASH.24 Studies in animal models of NASH and humans with NASH have shown induction of cytochrome P450 enzymes CYP2E1 and CYP4A, both of which are capable of generating reactive oxygen species (ROS) during the metabolism of fatty acids and ketones.25 Most recently, Sanyal et al. have provided evidence that increased mitochondrial β-oxidation of free fatty acids (FFA) due to hepatic as well as peripheral insulin resistance may be an important source of ROS in NASH.26
As regards endotoxin-induced cytokine release, this is considered to be the important mechanism of the NASH occurring in patients following jejuno-ileal bypass surgery for obesity, because liver injury can be prevented by antibiotics aimed at reducing growth of Gram-negative bacteria in the ‘blind loop’ and the resulting portal endotoxemia.27 Non-alcoholic steatohepatitis has also been reported to occur in patients presenting with small intestinal bacterial overgrowth, and most recently bacterial overgrowth has been demonstrated in patients with apparently ‘primary’ NASH.28 Further support for a role for endotoxin in NASH has come from the leptin-deficient (Ob/Ob) mouse model of NASH, in which mice with extensive steatosis have been shown to be extremely sensitive to small doses of intraperitoneally administered endotoxin.29 Further support for a role for this pathway in human NASH has come from our recent work demonstrating that a gain-of-function mutation in the CD14 endotoxin receptor on Kupffer cells is significantly more common in patients with NASH than in patients with simple non-alcoholic fatty liver.30 A role for TNF-α in human NASH has been supported by data from a recent study demonstrating the overexpression of both TNF-α mRNA and TNF receptor p55 mRNA in the liver of patients with NASH.31
Finally, with respect to immune mechanisms and NASH, our group has some preliminary evidence that patients with NASH have antibodies to proteins adducted to lipid peroxidation products in their serum. As in alcoholic liver disease, further work is required to determine whether this immune response plays any role in the liver injury in NASH.
The ‘two hit’ model of non-alcoholic steatohepatitis pathogenesis
These mechanisms of injury in NASH provide several potential explanations for the previously described correlation between the severity of steatosis and the risk of inflammation and necrosis. First, with respect to oxidative stress, studies in animals with NAFLD32 and humans with alcoholic fatty liver33 have shown that the degree of lipid peroxidation in the liver correlates with the severity of steatosis. Therefore, if lipid peroxidation is indeed a key mechanism involved in the pathogenesis of NASH, then the more fat there is present in the liver, the more this will lead to lipid peroxidation in response to oxidative stress, resulting in more inflammation, necrosis and fibrosis.11 With respect to endotoxin and cytokines, as discussed in the previous section, work in the Ob/Ob mouse has shown that fatty livers are more sensitive to endotoxin. This has been attributed to their hepatic cytokine profile (low interleukin (IL)-10, high γ-interferon) that will sensitize the liver to the effects of TNF-α. Recent work from Diehl's group has suggested that this cytokine profile may result from the depletion of hepatic CD4+NKT cells resulting in fatty livers having a predominantly T-helper cell-1 (Th1) ‘cell-mediated’ immune phenotype.34 With respect to immune mechanisms, this observation also suggests that fatty livers would be more sensitive to Th1-mediated immune damage, perhaps initiated by an immune response to lipid peroxidation adducts.
Further work in the Ob/Ob mice has shown that their fatty livers have an increased basal expression of uncoupling protein 2 (UCP2).35 This mitochondrial protein leads to dissipation of the proton gradient that normally exists across the inner mitochondrial membrane and results in the generation of heat during the re-oxidation of NADH and FADH2 rather than synthesis of adenosine triphosphate (ATP). Consistent with this observation in mice, a recent preliminary study in humans has shown that patients with NASH cannot regenerate ATP in response to a fructose challenge.36 Accordingly, the presence of steatosis would also impair the response of the liver to any situation requiring an increase in the synthesis of ATP. This presumably explains, in part, the decreased tolerance of fatty livers to ischemic injury37 and their impaired regenerative response after partial hepatectomy.38 In other tissues, particularly cardiac muscle, the accumulation of triacylglycerol (TAG) has been associated with an accelerated rate of apoptosis.39 These various deleterious effects of lipid accumulation in non-adipose tissue, which also include insulin resistance40 (see the following section), have been collectively termed ‘lipotoxicity’.
These observations have led to the concept of NASH as a two-hit phenomenon with the first hit, steatosis, sensitizing the liver to the second hits of oxidative stress, endotoxin, adduct formation and any condition requiring an increased supply of ATP11 (Fig. 1).
Obesity, insulin resistance and non-alcoholic steatohepatitis: the role of free fatty acids
Does this, now widely accepted, ‘two-hit’ model, explain the close association between obesity, insulin resistance and NASH? Several studies have demonstrated convincingly that the degree of obesity predicts the risk of advanced liver disease in both alcoholic liver disease41 and, more importantly, in NAFLD.12,42,43 Most of the NAFLD studies have also found the presence of T2DM or the severity of insulin resistance to be an independent predictor of NASH12,44 or fibrosis.42 In terms of the two-hit model, this link between obesity, insulin resistance/T2DM and progressive disease may be thought to reflect the known association between these conditions and the severity of fatty liver: the first hit. However, the post-mortem study by Wanless and Lentz demonstrated that obesity and T2DM were associated with an increased frequency of NASH independent of the severity of steatosis.12 This suggests that obesity and T2DM/insulin resistance increase the risk of NASH over and above their effects on the severity of steatosis. An increasing body of evidence now suggests that the link between obesity, insulin resistance and the risk of NASH is explained by an increased release of FFA from adipose tissue.
Non-alcoholic steatohepatitis is associated with factors that increase the supply of free fatty acids to the liver
When the adipose tissue depot is expanded, plasma FFA become elevated, most likely due to an increased release from the expanded adipose mass.45,46 In obese patients, factors that further augment the supply of fatty acids such as sudden weight loss or increasing insulin resistance increase the risk of development of NASH. Several studies have shown that the risk of NASH seems to be particularly associated with central rather than peripheral obesity.6 It is known that visceral adipocytes are more lipolytically active than peripheral adipocytes. This has been variously attributed to them having a greater complement of adrenergic receptors,47 an increased production of the anti-insulin hormone resistin,48 and/or an increased production of cortisol due to their higher activity of the enzyme 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD-1) that converts inactive cortisone to active cortisol.49 Furthermore, any FFA released from central adipose sites will drain directly to the liver via the portal vein.
Free fatty acids, tumor necrosis factor-α and insulin resistance
Advances in our basic understanding of the mechanisms of insulin resistance associated with obesity have provided potential explanations for how an increased release of FFA from adipose tissue could lead to the generation of the second ‘hits’ in the liver that, along with steatosis, could result in NASH. Recent studies have shown that FFA activate a kinase cascade involving protein kinase Cθ (PKCθ), IκB kinase (IKK) and NFκB activation that results in insulin resistance via serine/threonine phosphorylation of the insulin receptor substrate-1 (IRS-1).50,51 This blocks the IRS-1 tyrosine phosphorylation normally initiated by insulin receptor binding and blocks insulin signaling.52 This activation of IKK/NFκB by FFA will also enhance the production of TNF-α, because its transcription is also dependent on NFκB activation. The TNF-α, in turn, will lead to further IKK activation resulting in a self-perpetuating cycle of increasing insulin resistance, lipolysis and the release of FFA and TNF-α from adipose tissue.51 In support of this hypothesis, a recent study in patients with NASH has shown them to have an increased expression of adipose tissue TNF-α mRNA.31
The resulting increased supply of FFA to the liver may explain, at least in part, the hepatic insulin resistance observed by Sanyal et al. in humans with NAFLD.26 Polyunsaturated fatty acids (PUFA), in particular, have been shown to suppress the upregulation of lipogenic and glycolytic enzymes by insulin. They appear to do this by reducing the nuclear abundance and DNA binding activity of key transcription factors mediating the effect of insulin on genes encoding these enzymes, including sterol regulatory element binding protein-1 (SREBP-1), nuclear factor Y (NF-Y) and possibly hepatic nuclear factor-4 (HNF-4).53 Whether these effects of PUFA are exerted through the IKK/NFkB pathway in hepatocytes is unknown at present. Whatever the precise mechanism of the inhibitory effects of FFA on insulin's downstream effects, this presumably explains why lipid accumulation in non-adipose tissue leads to insulin resistance as part of general ‘lipotoxicity’.54 Importantly, recent evidence suggests that, by stimulating FFA oxidation,55 the principal role of the adipocyte-derived hormone, leptin, in obesity may be to limit the accumulation of fat in non-adipose tissue and prevent lipotoxicity.56 Given these observations, in obese patients the development of overt hepatic insulin resistance probably first requires the accumulation of TAG, which will depend on the balance between FFA supply and insulin secretion/sensitivity on the one hand (favoring storage), and leptin secretion/sensitivity on the other (favoring oxidation). The gradual switch from insulin sensitivity to resistance with increasing steatosis may be a mechanism whereby the liver protects itself from continuing enlargement in the face of continuing excessive substrate supply.
Free fatty acids, tumor necrosis factor-α and the ‘second hits’
As alluded to here, in addition to their potential role in the evolution of hepatic insulin resistance, the increased delivery of FFA and TNF-α to the liver in obesity/insulin resistance is capable of producing the second ‘hits’ required for the development of NASH, principally oxidative stress. Hepatic insulin resistance, particularly in the face of high concentrations of leptin, will favor the entry of FFA into the mitochondria via reduced acetyl coenzyme A (CoA) carboxylase activity, which will lead to lowered concentrations of malonyl-CoA and activation of carnitine palmitoyl transferase-I (CPT-I), which transfers fatty acyl-CoA into the mitochondria. In addition, FFA and their metabolites are ligands for the transcription factor, peroxisomal proliferator-activated receptor-α (PPAR-α). Peroxisomal proliferator-activated receptor-α regulates the transcription of a variety of genes encoding enzymes involved in mitochondrial and peroxisomal β-oxidation, as well as cytochrome P450 enzymes involved in the ω-oxidation of FFA including CYP4A family members. CYP2E1 is not regulated by PPARα, but it is normally downregulated by insulin57 and therefore hepatic insulin resistance potentially explains its upregulation in NASH.25 Hepatic insulin resistance and upregulation of PPARα-regulated genes by FFA will therefore result in increased FFA oxidation by at least three different pathways, all of which are capable of generating ROS that can contribute to the development of oxidative stress.58 The increased delivery of TNF-α to the liver, as well as initiating the expression of other pro-inflammatory cytokines and adhesion molecules, will further increase oxidative stress during mitochondrial FFA oxidation by impairing electron flow along the respiratory chain.59 The relative importance of mitochondrial compared to extra-mitochondrial FFA oxidation in the pathogenesis of NASH is suggested by the contrasting effect of defects in these pathways. Mice lacking the gene encoding fatty acid acyl-CoA oxidase, the initial enzyme in the peroxisomal β-oxidation system, develop severe NASH,60 presumably reflecting increased FFA oxidation via the other pathways. In contrast, NASH is not a feature of children with inborn errors of mitochondrial β-oxidation. These children develop severe microvesicular steatosis, fasting hypoglycemia and liver failure thought to be due to impaired energy production.
In addition to inducing hepatic oxidative stress, studies by Cortez-Pinto et al. have shown that FFA and TNF-α are capable of inducing UCP2 in hepatocyte mitochondria which, as aforementioned, will impair the ability of the liver to synthesize ATP during stress.61 Dicarboxylic acids (DCA) arising from the ω-oxidation of FFA by CYP4A enzymes are also toxic to mitochondria and are capable of uncoupling oxidative phosphorylation.
Free fatty acids and non-alcoholic steatohepatitis: A unifying mechanism
In summary therefore the increased release of FFA from adipose tissue in obesity can produce both the first and the second hits required for NASH (Fig. 2). In the periphery, the increased FFA lead to insulin resistance and the transcription of TNF-α which initiates a self-perpetuating cycle of lipolysis and TNF-α release from adipose tissue. The resulting increased supply of FFA to a still relatively insulin-sensitive liver will initially result in increased FFA esterification and TAG storage. As the severity of steatosis increases the liver becomes more insulin-resistant, and the incoming FFA will be diverted into the mitochondria and oxidized by enzymes whose genes are unregulated by FFA-induced activation of PPARα. The increased delivery of TNF-α to the liver will increase the generation of ROS during mitochondrial β-oxidation of FFA by impairing the flow of electrons along the mitochondrial respiratory chain. The upregulation of peroxisomal and microsomal FFA oxidation by PPARα will also contribute to oxidative stress. This oxidative stress in the presence of steatosis will result in lipid peroxidation and potentially lead to the development of hepatocyte death and associated necroinflammation. Circumstantial evidence suggests that hepatocytes laden with fat may be particularly sensitive to apoptotic cell death induced by oxidative stress and/or TNF-α. In addition the upregulation of UCP2 by FFA and TNF-α, along with dicarboxylic acids derived from microsomal FFA oxidation, leads to the uncoupling of oxidative phosphorylation and an impairment in the liver's ability to synthesize ATP in response to stress.
Individual susceptibility to non-alcoholic steatohepatitis
Given this scenario, perhaps the most intriguing aspect of NASH at present is why all obese patients with fatty liver do not develop NASH. It seems likely that a variety of environmental and, in particular, genetic factors determine an individual's response to an increased supply of FFA and TNF-α. Studies have recently begun in this regard30,62 and undoubtedly will yield important and exciting results over the next few years.
The pathogenesis of NASH is complex, involving an interplay between FFA acid metabolism and cytokines, initially resulting in the deposition of fat in the liver and later in increased oxidation of fat resulting in oxidative stress. Precisely how the liver switches from a relatively insulin-sensitive organ storing fat to an insulin-resistant organ preferentially burning fat remains to be clarified but seems likely to reflect a protective mechanism against the further accumulation of fat in an already fatty liver. Althoughe much detail remains to be clarified, these mechanisms already suggest a variety of targets for therapy, including reducing the severity of obesity and steatosis, treatment of insulin resistance, anti-oxidant therapy, and anti-TNF treatment.