A fresh look at NASH pathogenesis. Part 1: The metabolic movers


Professor Geoff Farrell MD FRACP, Gastroenterology and Hepatology Unit, The Canberra Hospital, Yamba Drive, Garran, ACT 2605, Australia. Email: geoff.farrell@act.gov.au


The strong relationship between over-nutrition, central obesity, insulin resistance/metabolic syndrome and non-alcoholic fatty liver disease (NAFLD) suggest pathogenic interactions, but key questions remain. NAFLD starts with over-nutrition, imbalance between energy input and output for which the roles of genetic predisposition and environmental factors (diet, physical activity) are being redefined. Regulation of energy balance operates at both central nervous system and peripheral sites, including adipose and liver. For example, the endocannabinoid system could potentially be modulated to provide effective pharmacotherapy of NAFLD. The more profound the metabolic abnormalities complicating over-nutrition (glucose intolerance, hypoadiponectinemia, metabolic syndrome), the more likely is NAFLD to take on its progressive guise of non-alcoholic steatohepatitis (NASH). Interactions between steatosis and insulin resistance, visceral adipose expansion and subcutaneous adipose failure (with insulin resistance, inflammation and hypoadiponectinemia) trigger amplifying mechanisms for liver disease. Thus, transition from simple steatosis to NASH could be explained by unmitigated hepatic lipid partitioning with failure of local adaptive mechanisms leading to lipotoxicity. In part one of this review, we discuss newer concepts of appetite and metabolic regulation, bodily lipid distribution, hepatic lipid turnover, insulin resistance and adipose failure affecting adiponectin secretion. We review evidence that NASH only occurs when over-nutrition is complicated by insulin resistance and a highly disordered metabolic milieu, the same ‘metabolic movers’ that promote type 2 diabetes and atheromatous cardiovascular disease. The net effect is accumulation of lipid molecules in the liver. Which lipids and how they cause injury, inflammation and fibrosis will be discussed in part two.


It is more than 30 years since alcoholic hepatitis-like lesions were recognized among over-weight or diabetic non-drinkers, for which Ludwig, in 1980, coined the term non-alcoholic steatohepatitis (NASH).1 Interest in this disorder has burgeoned recently.2–6 A decade ago it was known that fatty liver had many causes,2 and the present convention is to label cases with definable single etiologies as such, e.g. drug-induced steatohepatitis, fatty liver associated with parenteral nutrition, rather than ‘secondary NASH’. The term non-alcoholic fatty liver disease (NAFLD) should be reserved for those cases in which there is not one single cause. The latter are nearly always associated with overweight, particularly central obesity and insulin resistance, and often glucose intolerance/type 2 diabetes (T2D), dyslipidemia, hypertension and other features of the metabolic syndrome.2–5 For this reason, we proposed the term metabolic steatohepatitis,2 but the simpler term NAFLD infers an inextricable relationship between this type of fatty liver and the metabolic complications of over-nutrition.7

In most cases of NAFLD liver pathology is not known,7–9 and it may be best to reserve the term non-alcoholic steatohepatitis (NASH) for histologically confirmed cases that comply with recent pathological definitions,10,11 as recommended in the Asia-Pacific region.8 Here we refer to NAFLD/NASH when the discussion is about the pathologically more significant form of NAFLD, present in 20–30% of cases.3–5 In this review, we first consider the rationale for considering NAFLD as a distinct entity, where NASH fits into that concept, and the mechanistic implications of what appear to be inextricable connections between over-nutrition and insulin resistance; visceral adiposity and steatosis; adipose restriction, inflammation and failure and worsening insulin resistance. We then discuss newer aspects that now seem relevant to NASH pathogenesis, distinguishing between what is known and key questions that remain unanswered (Table 1). Since this field is now extensive, we will confine the first part of the review to metabolic factors, which we believe lead to steatohepatitis—not just steatosis. In Part 2 of the review, we will consider mechanisms whereby lipotoxicity leads to hepatocellular injury, inflammation and fibrosis, the pathological features of NASH.

Table 1.  The metabolic pathogenesis of NASH: key issues and outstanding questions
ProcessKey issuesWhat is knownWhat is not yet known/outstanding questions
  1. BMI, body mass index; ChREBP, carbohydrate-response element binding protein; CNS, central nervous system; FFA, free fatty acids; IR, insulin resistance; IL-6, interleukin-6; MCP-1, monocyte-chemoattractant protein-1; NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis; NPY, neuropeptide Y; SAT, subcutaneous adipose tissue; SREBP, sterol-response element binding protein; TNF-α, tumor necrosis factor-α; VAT, visceral adipose tissue; WAT, white adipose tissue.

Over-nutritionObesity genes: most locate to hypothalamus> 100, each has small effect, mostly on appetite regulationSummative or interactive effects of multiple genes?
Over-eating: CNS appetite regulationFailure to suppress by leptin; several possible pathways to ‘leptin resistance’Actual mechanisms of leptin resistance; other pathways important?
Energy intake: total, dietary compositionConflicting data: high saturated fats, simple sugarsDo NAFLD/NASH patients eat differently from BMI or family-matched controls?
Under-activityNASH/metabolic syndrome patients under-active, unfit (VO2); therapeutic benefits of reversal (> 150 min aerobic exercise/week)Behavioural determinants; primary or secondary to obesity, hypothalamic dysregulation?
Metabolic regulationPathways and mediatorsEndocannabinoid system in hypothalamus, liver and WAT; stress-response obesity in rodents related to sympathetic NS release of NPYRelative importance for lipid storage of liver/WAT pathways; small molecule modulators as therapy?
Alters energy equationThermogenesis from uncoupling of mitochondrial respiration in BAT in hibernation response; recent discovery of BAT in humans—less with obesityDoes this play any role in metabolic syndrome and NASH?
Hepatic lipid partitioningRelationship to visceral adipose (VAT)VAT correlates with NASH. VAT a source of FFA and cytokines (MCP-1, TNF-α, IL-6)How do genes, dietary and other factors determine VAT expansion?
Subcutaneous adipose (SAT) restriction/failureLarge, fat-filled adipocytes associated with FFA release, adipose inflammation (MCP-1, TNF-α, IL-6), decline of adiponectin secretionRelationships between SAT restriction and VAT expansion; conflicting data on SAT capacity and NAFLD/NASH; role of adiponectin polymorphisms
LipogenesisQuantitatively minor pathway in NASH; driven by insulin (via SREBP1c) and glucose (via ChREBP)Is this the earliest change in NAFLD? Does lipogenesis lead only to steatosis?
Hepatic lipid uptakeQuantitatively most important in NASH; secondary to unrestrained lipolysis (insulin resistance); FA uptake not only passive—CD36 is an active FA uptake protein?this is source of lipotoxic lipids (eg FFA) that distinguish NASH from steatosis; does NASH start at adipose, what are the key regulators of hepatic lipid uptake?
Hepatic lipid turnoverDecreased VLDL formation in NASH, probably insulin mediated; possible impairment of PPARα-regulated FA oxidation (see Fig. 5)Do polymorphisms (eg microsomal triglyceride transport protein) predispose to NASH?
Insulin resistanceHepatic insulin resistanceLikely secondary to FA accumulation; partial, not total abrogation of insulin-mediated effects (e.g., lipogenesis still activated via SREBP-1c)Starts in periphery (adipose, muscle)? -hyperinsulinemia stimulates lipogenesis; eventual hepatic IR; or in liver?
Adipose insulin resistanceLikely source of FFA taken up by liver in NASH; could be explained by PNPLA3 polymorphism; reversal correlates with improvement of NASH during pioglitazone treatmentGenetic predisposition or rapid expansion of adipocyte lipid stores? What are roles of stroma and blood supply, and macrophage recruitment?
Muscle insulin resistanceEarliest site after high carbohydrate feeding of over-weight young subjects (ref 142)Role of myokines (including IL-6) as well as adipokines?

NASH is not trash—in pursuit of defining a non-disease

To the extent that NASH has neither a single cause, unique and reproducible clinicopathological hallmarks or an accepted treatment, it is not a disease. But neither is high arterial blood pressure, cigarette smoking, expanded waistline nor hypercholesterolemia! Yet who would dispute the health implications of these pathophysiological measurements, behaviors or changes in body composition? When they are combined, the implications for cardiovascular health, T2D and cancer risk are strongly supported by epidemiological evidence, albeit there remains debate about the utility of combining them as a defined ‘metabolic syndrome’.12

Likewise, NAFLD is a condition in which hepatocytes, which normally contain only small amounts of storage lipid, contain supra-physiological amounts of fat. This can be observed by light microscopy as ‘steatosis.’10,11 With NAFLD, the amount of hepatic storage triglyceride varies from just above normal (5% liver mass) to greater than ten-fold normal levels.13–15 Whether there is more in NASH is important to establish, although with development of cirrhosis steatosis decreases. Variability in free fatty acids (FFA) and other lipid molecules is greater and may be more relevant to steatohepatitis pathogenesis, as discussed in part 2. The increased susceptibility of fatty livers to injury (after surgical resection,16 during ischemia-reperfusion,17 or with hepatitis C virus infection18) is one piece of evidence that NAFLD is not a healthy state, although with simple steatosis (SS), progression to cirrhosis or hepatocellular carcinoma (HCC) is rare.19,20 When hepatocytes leak their intracellular contents into serum, evident by a rise in serum alanine aminotransferase (ALT), ferritin, etc., the argument for liver injury is more persuasive, particularly when histological appearances include damaged hepatocytes (ballooning, Mallory bodies),10 lobular inflammation with polymorphs as well as macrophages and lymphocytes, and pericellular, perivenular (zone 3) fibrosis.11 These changes are used to designate NAFLD as ‘steatohepatitis’ (NASH), and there is expanding evidence that NASH with fibrosis can progress to cirrhosis or HCC over 5–20 years [reviewed in 4,7].

Thus, we do not accept that ‘NASH may be trash’, as proposed by Cassiman and Jaeken in their attempt to provoke more careful consideration about possible causes of fat in the liver.21 First, we do not accept that excluding toxic levels of alcohol exposure is ‘notoriously unreliable’ when a careful alcohol history is obtained by experienced gastroenterologists and hepatologists,22,23 and there is emphasis in contemporary medical and professional education on quantitative aspects, particularly life-long exposure.22–25 Second, neither international guidelines nor contemporary books on NAFLD use the guideline posited by Cassiman and Jaeken: ‘if most prevalent liver diseases are excluded and the biopsy shows fat accumulation, we convict the patient and drop them in the NAFLD trash bin’. Asia-Pacific Guidelines propose that: ‘diagnosis by abdominal ultrasonography, assessment of liver function and liver-related complications, exclusion of alcoholism and viral hepatitis B and C, and screening for insulin sensitivity and metabolic syndromeare required as initial assessment.’8,9

These authors go on to state that 10–20% of NAFLD/NASH patients do not have insulin resistance, and that NAFLD/NASH also occurs in type 1 diabetes; neither statements are referenced. A literature search indicates that > 95% of NASH patients in whom a measure of insulin sensitivity was obtained have insulin resistance [reviewed in 3–5,7], while steatosis in type 1 diabetes is associated with insulin use;26–28 few cases have been pathologically documented as NASH. The data on metabolic syndrome association cited by Cassiman and Jaeken (up to 50% cases) are also misleading. Studies reproducibly report > 85% of patients with histologically-proven NASH have metabolic syndrome [29,30; reviewed in 7], although the larger sub-population (two-thirds) of NAFLD patients without NASH are less likely to have metabolic syndrome, or not yet. If metabolic factors play a role in NASH pathogenesis, as we propose, and if NASH plays a causative role in metabolic syndrome, for which recent evidence lends increasing support,31,32 it is not surprising that the more severe the metabolic state the more severe the liver disease.30

Given the above considerations, we believe it is a huge leap of logic to infer that a substantial proportion of cases are ‘not NAFLD at all’, but, rather, undiagnosed instances of glycogen storage disease type VI, alpha-1 antitrypsin deficiency (routinely excluded in published NASH series),29,30 or ‘X-linked, dominant, recessive or mitochondrial inheritance pattern of liver or metabolic disease’.21 The possible involvement of mitochondria in causation of T2D and fatty liver disease remains intriguing,32–35 but we are not aware of congenital mitochondrial disorders (Alpers syndrome, mitochondrial DNA depletion syndrome) causing other than macrovesicular or microvesicular steatosis, cirrhosis or acute liver failure, not NASH.36 None-the-less, we do think consideration needs to be given to the fact that microvesicular steatosis is observed in some cases of NASH, and mitochondrial crystalline inclusions are commonly noted, particularly in severer cases;34,35 the implications will be discussed in Part 2. The list of causes of fatty liver that are not NAFLD/NASH presented in Cassiman and Jaekman's Table 1, as in a 2001 review,2 is less than 100. Individually they are exceedingly rare, < 1 per 10 000 population, versus 2000–4000 per 10 000 for NAFLD and 700–1300 per 10 000 for NASH. So one hundred of them could not account for even 5% of NAFLD cases. It also does not seem logical to us to exclude childhood monogenetic obesity syndromes (Bardet-Biedl, Alström, Prader-Willi syndromes) as causes of NASH when the associated metabolic factors (over-nutrition, obesity, insulin resistance, T2D, dyslipidemia) are identical to NASH, as discussed later.

While we think it unlikely that even a minority of cases presently diagnosed as NAFLD will turn out to be syndromes based on single gene mutations, we agree that only a minority of ‘the metabolically challenged’ (those with over-nutrition) will develop cirrhosis; individual susceptibility to NASH versus SS is a key issue in pathogenesis.2–5 However, we note with irony that the authors cite a review written by two of us as evidence in favor of ‘the magical two-hit hypothesis’ (sic) for progression from ‘NAFLD to NASH’ (sic, mis-using above terminology).21 In that review,4 we actually canvassed strongly, as we do here, the evidence against metabolic factors being self-limiting, and against the cytokine basis of a two-hit hypothesis. Like others in this field (including Day who proposed the two-hit hypothesis),[C Day, personal communication, EASL Single Topic Conference on NAFLD, Bologna, September 2009] we no longer think this is a helpful concept. This review will explore the evidence for what seems to us intuitively more plausible, the lipotoxicity concept of NASH pathogenesis.

Where NASH starts: overweight/obesity and NAFLD/NASH result from over-nutrition, much of which is genetically determined

While NAFLD is near universal among the obese (body mass index [BMI] > 30 kg/m2 in Europeans, > 25 kg/m2 in Asians), the interaction between obesity and NAFLD is more nuanced. The most striking correlates are with visceral fat accumulation and insulin resistance. As such, ‘metabolically obese, normal weight’ individuals may exhibit features of NAFLD in the absence of obesity but in association with an abnormal metabolic phenotype. But there appears to be a reproducible connection between NAFLD and over-nutrition—energy intake that exceeds energy utilization. Thus, a history of recent or progressive weight gain (which may not classify individuals as obese) is very common in NASH [reviewed in 5], and other disorders should be considered if waist circumference is normal. Obesity is associated with severer forms of liver pathology and fibrotic progression in NASH.37 While some controversy remains over the causes of the current obesity pandemic, there is strong evidence of increased caloric consumption in the last 20–30 years which correlates well with increasing weight and waist circumference.38 What drives over-nutrition is a complex interaction of biology (genes) and environment (behavior), but the following are all likely contributors: central appetite regulation, food/energy intake, energy expenditure and metabolic regulation (Fig. 1).

Figure 1.

Over-nutrition is an imbalance between energy (food) intake and energy expenditure, modified by metabolic regulation.

Central appetite regulation

Mammals are physiologically attuned to regulate food intake according to bodily energy needs. Such regulation is exerted by several hormones that either act rapidly to influence day-to-day food intake (reviewed in 39,40), or act more slowly to regulate adipose storage lipid. Long-term appetite regulators include insulin and leptin, which exert their effects on appetite centers in the hypothalamus and brainstem.40 Obesity resulting from over-eating (hyperphagia) involves defects in this control system. While insulin and adiponectin play some role in modulating appetite, discussion here will focus on the role of leptin, which several studies have shown has a more important role in the central nervous system (CNS) control of food intake and energy expenditure.

Originally identified as a major anorexigenic peptide,41 leptin arises from white adipose tissue (WAT), and serum levels increase in proportion to total body fat content to alert the brain to the state of body adiposity.42,43 Leptin crosses the blood-brain barrier via a saturable process and interacts, via liganding to the long form of the leptin receptor (LEPRb), with two distinct neuronal populations. The first synthesize and release orexigenic peptides, neuropeptide Y (NPY)44 and Agouti-related protein (AgRP).45 The second express the anorexigenic molecule, melanocyte stimulating hormone (α-MSH), derived from pro-opiomelanocortin (POMC),46 and cocaine and amphetamine-regulated transcript (CART).47 Thus, as shown in Fig. 2, leptin binds LEPRb on NPY/AgRP neurons to suppress these appetite-stimulating pathways, while simultaneously activating the appetite-suppressing POMC/α-MSH pathway (Fig. 2).

Figure 2.

Leptin receptor signaling on hypothalamic neuronal subpopulations suppresses appetite. AgRP, Agouti-related protein; CART, cocaine-amphetamine-regulated transcript; CRH, corticotropin releasing hormone; LEPR, leptin receptor; MC4R, melanocortin-4 receptor; MCH, melanin concentrating hormone; α-MSH, α-melanocyte stimulating hormone; NPY, neuropeptide Y; POMC, pro-opiomelanocortin.

Leptin resistance

In over-weight patients with NAFLD, leptin circulates in abundance and clearly fails to suppress appetite.48 Such CNS resistance to leptin action is now understood in terms of defective LEPRb signaling, involving several possible molecular mechanisms (Fig. 3). Leptin deficiency is the basis of obesity and NAFLD in ob/ob mice as well as rare cases of severe, monogenic childhood obesity (that can be corrected by exogenous leptin therapy).49–51 Mutations of the LEPRb occur in db/db mice and fa/fa (Zucker) rats, and are also rarely found in humans (Table 2).52,53 The molecular basis of tissue leptin resistance in multifactorial obesity is the subject of continuing research.54 It could be a consequence of altered blood-brain barrier, decreased expression of LEPRb, or decreased LEPRb signaling via Jak-2-Stat3 due to an imbalance between expression of negative regulators of leptin sensitivity (suppressors of cytokine signaling [SOCS] proteins, protein tyrosine phosphatase 1B [PTP1B] and SH domain phoshatase 2), and of cellular adaptor molecules, such as SH2B1, which facilitates leptin signaling (Fig. 3).54 Another possibility has recently come to light—requirement of Bardet-Biedl proteins for LEPR signaling, inferring a possible role of receptor scaffold assembly on neuronal primary cilia.55

Figure 3.

Pathways of leptin receptor signaling. The long form of the leptin receptor (LEPRb) signals via a Jak2/Stat3 phosphorylation pathway, with phospho-stat3 dimer (P-stat3) entering the nucleus to regulate transcription of responsive genes. Cellular phosphatases (protein tyrosine phosphataseB1 [PTPB1], SH domain phosphatase2 [SHP2]) and suppressor of cytokine signaling3 (SOCS3), itself induced by LEPRb signaling, dampen or inhibit this signaling pathway, while SH2 domain-containing protein 1B (SH2B1) facilitates signaling.

Table 2.  Selected human obesity genes
OMIM no.SyndromeLocusCandidate gene
  1. Modified from Raniken et al. 2005 [ref 74], which contains fuller details and primary reference sources.

122561; 602034Corticotropin-releasing hormone receptor 1; 217q12-q22; 7p14.3CRHR1; CRHR2
601751G-protein-coupled receptor 2422q13.2GPR24
601007Leptin receptor1p31LEPR
601665; 155541Melanocortin 3/4 receptor20q13.2-q13.3; 18q22MC3R; MC4R
600456Neurotrophic tyrosine kinase receptor type 29q22.1NTRK2
162150Pro-protein convertase subtilisin/kexin type 15q15-q21PCSK1
203800Alström syndrome2q13.1ALMS1
209901; 606151; 600151; 600374; 603650; 604896; 607590; 608132Bardet-Biedl syndromes 1-811q13.1; 16q13; 3p13-p12; 15q22.3-q23; 2q31; 20p12.2; 4q27; 4q32.1BBS1-5; MKKS; BBS7,8
269700; 606158Berardinelli-Seip congenital lipodystrophy 1; 29q34.3; 11q13AGPAT2; BSCL2
216550Cohen syndrome8q22.2COH1
601538Combined pituitary hormone deficiency5q35.3PROP1
139191; 139250Isolated growth hormone deficiency7p14; 17q22-q24GHRHR; GH1
138090; 604931Cortisone reductase deficiency1pter-p36.13; 1q32-q41H6PD; HSD11B1
600917Severe insulin resistance with obesity3p25; 7q31.1PPARG; PPP1R3A
103580; 103581Pseudopseudohypoparathyroidism20q13.2-q13.3; 15q11-qq13GNAS; AHO2
160980; 605244Carney complex with primary pigmented nodular adrenocortical disease and Cushing's syndrome (CNC1 and CNC2)17q24.3; 2p16PRKAR1A
604367; 151660Familial partial lipodystrophy type 3; type23p25; 1q23.1PPARG; LMNA
147670Insulin resistance syndromes19p13.3-p13.2INSR
131100Multiple endocrine neoplasia type 111q13MEN1
176270Prader-Willi syndrome15q11.2; 15q; 15q11-q12IPW; MKRN3; PWCR1; SNRPN; MAGEL2; NDN; GABRG3
603128Prader-Willi-like syndrome (chromosome 6q)6q16.3-q21SIM1
190160Thyroid hormone resistance syndrome3q24.1THRB
176270Prader-Willi-like syndrome, X-linkedXq23-q25PWLSX
309550Fragile X syndrome with Prader-Willi-like phenotypeXq28FMR1
300218Mental retardation X-linked, syndromic 7, 11, 16Xp11.3-q22.1; Xq26-q27; Xq28MRXS7; MRXS11; MECP2
309585Wilson-Turner syndromeXq21-q22WTS

Neuronal ciliopathies: relevance to obesity and metabolic syndrome

Neurons express a primary cilium, on which is assembled a range of receptors involved in development (e.g. Hedgehog), neurohormonal regulation (somatostatin receptor 3) and appetite (melanin-concentrating hormone receptor 1 [MC-1R], and LEPRb).55–59 Primary cilia are found on many cells, but not hepatocytes. They may act as sensory ‘cell antennae’, coordinating inter-cellular communications via receptor clustering and signalling.56–58 Bardet-Biedl syndrome (BBS) is the archetypical example of a ciliopathy with profound appetite dysregulation.59 Like children with leptin deficiency or LEPRb mutations,50–53 BBS children are unable to resist the drive to eat, becoming massively obese at an early age, and about half develop T2D and metabolic syndrome.59,60 Multiple mutations of the BBS gene have been described;55,59 in mice, at least four genotypes are associated with defective LEPRb signaling.55,59

Another childhood obesity syndrome that may be ascribed to a ciliopathy is Alström syndrome.61,62 In addition to their respective specific features (skeletal, retinal, renal and hepatobiliary fibrocystic abnormalities, hearing defects and male infertility), BBS and Alström syndrome are both associated with hyperphagic obesity, early onset of insulin resistance, T2D and (best described for Alström syndrome) severe fatty liver disease leading to cirrhosis.63 An animal model of Alström syndrome, the foz/foz mouse (which carries a mutation of the gene for basal body protein, Alms1), develops NAFLD.64 Further, environmental factors affect expression of liver pathology, so that mice fed chow develop only steatosis, whereas those fed a high fat diet develop NASH with fibrosis.64,65 An additional exciting finding is that pre-adipocytes also express primary cilia,66,67 and these play a role in their capacity to differentiate and form triglyceride-storing adipocytes and secrete adiponectin. Further studies should explore whether the problem of restricted adipose expansion in metabolic syndrome and NASH (discussed later) could actually be due to innate properties of pre-adipocytes/adipocyte differentiation rather than ‘adipose exhaustion’.

Human obesity genes: relationship to diabetes, metabolic syndrome and NAFLD/NASH

A family clustering study of 33 children with NAFLD and 11 overweight controls without NAFLD found the heritability of NAFLD was 1.0 (that for liver fat was less, at 0.39), after adjusting for age, gender, race and BMI.68 Fatty liver disease often runs in families and is more common in certain ethnic groups.69–71 In south-western United States of America (USA), the prevalence of increased hepatic triglyceride content by magnetic resonance spectroscopy (MRS) varies from ∼20% in Afro-Americans and European women, through ∼30% in European men to ∼40% in Hispanics.70 Further, rates of T2D and cardiovascular disease are highest among Asian Indians, followed by Chinese and Europeans.72 Populations which until recently lived hunter-gatherer lifestyles, like Pima Indians, Malays, Australian aboriginals and Pacific Islanders, now have exceptional rates of obesity and its metabolic complications—T2D, atherosclerosis, gallstones and NAFLD/NASH (reviewed in 7). Thus, although life-style factors provide the setting for over-nutrition/obesity,73 ethnic (genetic) differences are explained by differential expression of genes that influence appetite control, food choices and bodily lipid distribution. Likewise, family clustering, adoption and twin studies have usually calculated the heritability of obesity to be between 0.6 and 0.7.69,74,75 This does not mean that environmental factors are not critical in pathogenesis of overweight and NASH,73 simply that in the present socio-economic conditions of energy abundance (cheap processed foods) and sedentary lifestyles that prevail in most countries, people with obesity genes are those most likely to succumb (Fig. 4).

Figure 4.

Interactions between genes and environment determine whether a person becomes overweight, bodily distribution of adiposity and development of both steatosis and metabolic complications, including diabetes, atherosclerosis and non-alcoholic steatohepatitis (NASH).

What are the crucial obesity genes, and how do they exert their effects?

To date, about 100 genes have been associated with obesity, but few individually account for more than a few percent of even severe obesity (BMI > 40 kg/m2).69,74–76 It has therefore been proposed that combinations of perhaps 10–30 genes may be required for expression of the obese phenotype.75 Alternatively, defects in common regulatory processes (such as basal body/cilial function) may involve several genes.74 Because obesity is physiologically complex, genes might act at various levels. However, among those identified to date, more than 100 act on the hypothalamus and brainstem to influence brain sensing of fat stores.74–76

During the last 3 years, genome-wide association studies (GWAS) have been adopted to identify stretches of genomic DNA (single nucleotide polymorphisms, SNPs) which correlate significantly with phenotype. Determining the structure of the DNA regions linked to the phenotype allows the potentially implicated genes to be identified.75,77–79 A particularly strong association has been found between the A allele of rs9939609 on chromosome 16 and adiposity.75,77–80 The frequency of the A alleles is 0.45 in Europeans, 0.54 among West Africans and 0.14 in Chinese, while the odds ratio (OR) for A allele and obesity is 1.31, and 1.18 for overweight; respective population attributable risks are 20% for obesity and 13% for overweight. In Scottish children, Cecil et al. (2008) found that the A allele had a major effect on energy intake, particularly preference for high energy foods, but no effect on resting energy expenditure; physical activity was actually increased.80 The A allele is also strongly associated with T2D,81 linking over-nutrition/obesity with its metabolic complications. Studies in NAFLD/NASH will now be of interest.

The A allele contains the first intron of fat mass and obesity-associated gene [FTO] and another gene, fantom (FTM). FTO and FTM are expressed in the hypothalamus and suppressed by fasting, implicating roles in appetite suppression.82 Like Alms1, FTM is a structural component of basal bodies, indicating the potential relevance to cilial function. In support of multiple gene interactions, Loos et al. found summation in the effects of FTO and MC4R with fat mass and risk of obesity.77 Likewise, in Han Chinese, the combined effects of three genes, one in the estrogen receptor, two in peroxisome-proliferation activator receptor-gamma (PPAR-γ), was greater than any individual gene; collectively, they conferred > 5-fold (OR 5.3) increased risk of severe obesity (Table 3).83

Table 3.  SNPs and genes influencing adiposity and fat distribution identified by genome-wide association scan (GWAS) studies
SNP (effect allele)Gene(s) (chromosomal locus)PhenotypeReferences
  • *

    Expressed at high levels in brain and hypothalamus.

  • Role in leptin signaling—see Figs 2 and 3.

  • AST, aspartate aminotransferase; BMI, body mass index; FTO, fat tissue and obesity; FTM, fantom; MC4R, melanocortin-4 receptor; SNP, single nucleotide polymorphism; WC, waist circumference; WHR, waist-hip ratio.

rs9939609FTO,* FTM*BMI, obesity75,78,79,81
rs17782313MC4R*BMI, obesity77
rs987237 (G/A)TFAP2B (6p12)WC (total and central adiposity)79
rs7826222 (G/C)MSRA (8p23.1)WC (total and central adiposity)79
rs2605100 (G/A)LYPAL1 (1q41)WHR (central adiposity), females79
rs738409 (G allele; met148Ile); rs2072907 (C/G)PNPLA3Decreased obesity; among obese subjects, BMI and WC higher with Ile-carriers, as is adipose insulin resistance. G allele of rs738409 correlates with liver fat content and serum AST; most common in Hispanics, least in Afro-Americans84,85,87,88


Robust studies linking individual genes to development of fatty liver have been lacking until recently. The patatin-like phosphatase family consists of nine genes, five collectively designated as the adiponutrin family (PNPLA1-5).84,85 The proteins are expressed in WAT and liver, and their action is believed to complement hormone-sensitive lipase (HSL), a key enzyme involved with adipocyte lipolysis.84 Genetic variations in HSL have been previously linked to obesity, glucose intolerance and dyslipidemia—all relevant to NASH.86 In 2008, Romeo et al. used GWAS to identify a SNP, rs738409, within PNPLA3 that was strongly associated with increased hepatic fat content.87 In subjects drawn from the multiethnic Dallas Heart Study mentioned earlier,70PNPLA3 (rs738409 G allele) was present with highest frequency in Hispanics, intermediate in whites and lowest in blacks.87 Hepatic fat content was two-fold greater for G allele homozygotes than for non-carriers of this allele, accounting for all the ethnic differences in MRS-determined hepatic triglyceride content. Another PNPLA3 allele (rs6006460 [T]) was associated with lower hepatic fat content and is most common in African Americans, the group with lowest prevalence of NAFLD.70 The association between PNPLA3 (rs738409 G allele) and hepatic fat content has been confirmed in Finnish and Argentinian cohorts,88,89 and is independent of age, gender, BMI and insulin resistance. In data presented at a recent meeting, PNPLA3 polymorphisms were associated with fibrotic severity of NASH.90,91

Contribution of dietary factors

Environmental factors, in particular dietary composition, undoubtedly play a role in facilitating the imbalance between energy intake and consumption that underlies the present pandemic of over-nutrition. In atherosclerosis there is strong evidence to support the proposal that dietary factors such as high saturated fat and cholesterol intake, as well as low polyunsaturated fat intake, increase the risk of disease and poor outcome.92 In NAFLD/NASH the picture is less clear, as reviewed.93 While it is clear that caloric excess is associated with obesity, it does not discriminate between those with NASH, simple steatosis or normal liver histology; when matched for body weight, most studies report similar calorie intake. In a Japanese study, where patients with NASH and SS were compared to a large, healthy population, increased energy intake was observed, particularly in younger patients, but failed to distinguish steatohepatitis from steatosis.94 Differences in total fat content have been noted in a single study.95 Another, in non-obese individuals with NASH found significant differences only in saturated fatty acid intake,96 but this has not been a consistent finding, with others finding differences in simple carbohydrate content (substrate for lipogenesis), not fat.97,98 In the latter study in morbidly obese individuals, carbohydrate content correlated with hepatic inflammatory response, and fat content appeared to have a protective effect.98 In agreement with this, the potential importance of fructose intake is highlighted by the association of sugar-sweetened beverages and NAFLD.99 Studies of polyunsaturated fatty acid (PUFA) content have also produced divergent results; some have shown lower n-3 PUFA content,94,97 others comparable n-3 PUFA but higher intake of n-6 PUFA in subjects with NAFLD.95,99

Re-analysis of data derived from a population-based survey identified a significant association between dietary cholesterol and cirrhosis, irrespective of cause.100 Finally, lower intake of anti-oxidant vitamins A, C and E have been variably reported in patients with NASH, as have decreased zinc and iron intake.93–99 Clearly, more extensive investigation is required to determine if certain dietary factors do predispose to NASH, or whether more subtle diet/genotype interactions alter an individual's susceptibility to development of liver injury and inflammation. The latter has been suggested for metabolic syndrome, in respect of adiponectin promoter polymorphisms.101

Physical activity

Several studies have shown that increasing aerobic exercise on a regular basis improves the metabolic indices strongly associated with NASH (waist circumference, serum insulin, hyperglycemia, serum lipids); this is associated with correction of liver test abnormalities.102–104 Histological studies have confirmed improvement of NASH following institution of an organized lifestyle program that combines physical activity with dietary advice.102 Few studies have compared a combined lifestyle approach delivered in a motivating (behavior changing) context with caloric restriction alone. However, available data indicate that the combined approach may be more successful, and there have been no reports of declining liver function or worsening hepatic fibrosis,102–104 such as have been reported with catastrophic weight reductions induced by drastic caloric restriction [reviewed in 4].

While proof that physical inactivity contributes importantly to pathogenesis of NASH is lacking, patients with metabolic syndrome, including those with NASH, have very low levels of muscle fitness (VO2) [reviewed in 103,104]. In a recent study, Newton and colleagues found that fatigue in patients with NAFLD (often profound) correlated inversely with physical activity, not with insulin resistance and disease severity.105 Further, NAFLD patients had lower levels of physical activity than liver disease and healthy controls, walking ∼15% fewer steps each day. The contribution of this to such pathogenically key variables as insulin resistance, mitochondrial function and cellular ATP levels is unclear and deserves greater study.

Could metabolic regulation be disordered in NASH?

Many humans have a poor ability to reduce food intake when consuming energy-dense foods, and this failure to maintain isocaloric intake stems from ineffective satiety signals in addition to poor food choices (a behavior).38,53,75,80,92 In this respect, over-nutrition and NASH can be regarded as a consequence of an unhealthy behavior. In support of this, commonly used antidepressants and anti-psychotic agents cause weight gain as unwanted effects of stabilizing mood and behavior.106,107 The likely explanation is that regions of the brain concerned with appetite regulation interact closely with those subserving pleasure, mood and physical activity; as discussed above, patients with NASH are generally inactive.103–105 Central regulators of food intake also influence metabolic regulation. For example, in hibernating mammals the drive to eat diminishes concurrently with the start of the resting state and induction of thermogenesis108—activation of uncoupling protein-1 (UCP-1) in brown adipose tissue [BAT] which enables fuel to be oxidized without ATP generation. The recent identification of BAT in humans has revived the concept that activation of thermogenesis could be a target for therapeutic intervention in obesity and its related metabolic disorders.109

The hibernation response is a dramatic example of metabolic regulation. An increasing number of metabolic regulators has now been identified, including peptides produced from adipose and muscle, respectively, adipokines and myokines.40,110 These metabolic regulators can act both centrally, where they alter appetite, basal metabolic rate and energy expenditure (including physical activity), and peripherally, through direct ligand-receptor interactions. The endocannabanoid system is an example of metabolic regulation involving both CNS and peripheral pathways.111,112 Thus, cannabinoid type 1 receptors (CB1R) appear functional in the hypothalamus, where they are co-expressed with melanin concentrating hormone (MCH) and CART, in WAT, where they decrease adiponectin and increase lipogenesis in vitro, and in liver.111–114

In the hypothalamus, leptin reduces endocannabinoid levels via a LEPRb-independent mechanism.115 After fat feeding, CB1R and anadamide (an endocannabinoid) increase in liver where they activate lipogenesis via sterol regulatory element binding protein-1c (SREBP1c)-induction of lipogenic gene expression and inhibition of the AMP kinase ([AMPK] (which suppresses lipogenesis by phosphorylating acyl-CoA carboxylase [ACC]).113 This can lead to diet-induced steatosis, dyslipidemia, and both insulin and leptin resistance.114 Rimonabant, the prototypic CB1R antagonist, reduced hepatic steatosis and improved dyslipidemia in fa/fa diabetic rats, disproportional to effects on food intake.116 This preliminary observation led to the suggestion that pharmacological approaches to metabolic regulation could have beneficial effects in NAFLD/NASH.116,117 Unfortunately, clinical development of Rimonabant as an anti-obesity agent was discontinued because of a high frequency of depression.

Other systems involved in both CNS and peripheral metabolic regulation include NPY and melanocortin.118,119 Thus, stress in rodents releases NPY from sympathetic nerves. In turn, this up-regulates NPY and its Y2 receptor in abdominal fat by a glucocorticoid-dependent mechanism, resulting in abdominal obesity.118 There is also evidence that blockade of CNS melanocortin receptors (MCR) triggers mobilization of lipid uptake, triglyceride synthesis and fat accumulation in WAT, changes that are independent of food intake.119 This indicates that loss-of-function mutations in MC4R, which have been associated with human obesity, may affect both CNS and peripheral metabolic regulation in favor of adiposity.

From over-nutrition to steatosis: hepatic lipid partitioning is the next step in NASH pathogenesis

Links between visceral and subcutaneous adipose and steatosis

One reason for earlier controversies about NAFLD and NASH is that not all affected patients are obese, although we contend that most are either over-weight or ‘metabolically obese, normal weight’.120 From a burgeoning literature [reviewed in 7,121], the most consistent relationships with NASH have been between central obesity, reflecting visceral adiposity, and insulin resistance. Anthropometric indicators of visceral (central) obesity (VAT), such as waist circumference, have been bolstered by determination of hepatic triglyceride stores using MRS.70,122 Among morbidly obese individuals, steatosis correlates directly with VAT,122 while correlations with subcutaneous adipose tissue (SAT) stores are less clear, with inconsistent results between studies (some positive, others no correlation).123–126 The relationship between central obesity and NAFLD/NASH is consistent with the proposal that metabolically unhealthy fat is what leads to insulin resistance and cardiovascular disease,127–129 as supported by the strong relationship between central obesity and cardiovascular death and even all-cause mortality.130,131

In non-obese subjects, the relationship between waist circumference and NAFLD/NASH is less clear. Musso and colleagues compared non-obese, non-diabetic NASH patients to controls; there was no difference in waist circumference or waist-hip ratio.132 In contrast, these patients showed markers of insulin resistance, consistent with a recent study which concluded that, in early obesity, the strongest factor correlating with steatosis was insulin resistance, rather than waist circumference.127 This apparent discrepancy may be explained by a ‘critical VAT threshold’ which could vary greatly between individuals; once the threshold is reached, insulin resistance arises together with its complications, including NASH.128 It has recently been reported that intrahepatic fat is a better marker than visceral adiposity for metabolic derangements associated with obesity,133 and this may explain less than perfect correlations with VAT in other studies.

We recently captioned the VAT/NASH connection,134 as well as emerging evidence that the converse may be equally important, that is, if SAT stores become saturated and unable to further expand in response to continued demands for storing excess energy as lipid, ectopic lipid accumulation arises—potentially increasing VAT mass in concert with steatosis. Evidence for this comes from human imaging studies and two animal models. First, in human studies there is fairly consistent correlation between visceral adiposity and steatosis,123–125 but the relationship between SAT mass and steatosis is less consistent. Some studies report positive or negative correlations, and others show no predictive value of SAT for steatosis [126–128; reviewed in 121,134]. Second, in ob/ob mice which develop obesity, diabetes and steatosis, forced over-expression of an adiponectin transgene reduced steatosis and reversed diabetes in association with massive proliferation and expansion of SAT.135 Last, in the foz/foz mouse model, diabetes, hypercholesterolemia and marked steatosis (which progresses to steatohepatitis) develops secondary to a plateau in adipose expansion indicating restricted adipose storage capacity.65 Thus, there is now strong evidence for a link between impaired adipose function, or adipose restriction, which predisposes to abnormal hepatic lipid partitioning;136 the consequences lead into the NAFLD spectrum.

Pathways of hepatic lipid accumulation

The sources and causes of hepatic lipid accumulation have been extensively reviewed.137–140 Recent evidence has confirmed roles for enhanced de novo lipogenesis, increased delivery of FFA (and other lipids) from the diet but more significantly from the periphery, increased hepatocellular lipid uptake, and impaired catabolism and export (Fig. 5). Insulin and glucose both drive lipogenesis, by the respective transcription factors SREBP1c and ChREBP.137 Thus, lean, sedentary young men with insulin resistance, when fed a high-carbohydrate diet, develop striking post-prandial hyperinsulinemia which drives hepatic lipogenesis.141 Further evidence to support the concept that hyperinsulinemia present in the early stages of insulin resistance contributes to hepatic lipid accumulation comes from animal models.142 In SREBP1c-over-expressing mice, hepatic lipogenesis leads to steatosis but not NASH.142–144 In foz/foz mice, development of steatosis is associated with early rises in serum insulin and glucose levels and increased fatty acid synthase (FAS) activity,65 the rate-limiting step in lipogenesis. Later, development of steatohepatitis (NASH) is associated with increased expression of cluster differentiation protein-36 (CD36),65 a pathway of active hepatic fatty acid uptake.145,146

Figure 5.

Sources and fates of fatty acids in liver cells. CD36, cluster differentiation protein-36; ChREBP, carbohydrate regulatory element binding protein; CYP, cytochrome P450; ER, endoplasmic reticulum; FFA, free fatty acids; SREBP1c, sterol regulatory element binding protein; TG, triglyceride; VLDL, very low density lipoprotein.

The significant contribution of lipid uptake to hepatic lipid pools is supported by tracer studies in obese humans with NASH. Thus, Donnelly et al. demonstrated that ∼60% of hepatic triglyceride arises from non-esterified (free) fatty acids (FFA), predominantly derived from adipose, compared to only ∼25% from de novo lipogenesis.147 Increased hepatic levels of FFA have been implicated in NASH pathogenesis [148–150; reviewed in 140] and may be a distinguishing feature from simple steatosis (this will be discussed in Part 2). With insulin resistance, serum FFA levels increase because of failure of insulin to suppress HSL-mediated lipolysis in adipose. From the liver perspective, this is particularly relevant to VAT stores, partly because these adipose pads drain directly to the liver, but also because adipocytes in these sites exhibit greater lipolysis and are less responsive to insulin.151–153

The increased delivery of lipids to the liver can be exacerbated by active fatty acid uptake. Previous concepts of fatty acid uptake as a predominantly passive (or facilitated diffusion) event have been challenged by studies demonstrating that CD36 can induce steatosis,146 and insulin increases its expression.145,146 Hepatocellular expression of CD36 is up-regulated in several experimental forms of NAFLD,65,146 and the dynamic nature of such expression—whether it is responsive to dietary fatty acids, the hormonal changes of metabolic syndrome (high serum insulin, low adiponectin), or to altered expression of nuclear transcription factors, such as liver X receptor (LXR), PPAR-γ (reproducibly up-regulated in experimental NASH),65,154 is an important subject for future research.

In addition to stimulated uptake and synthesis, impaired lipid export can also exacerbate steatosis. Decreased secretion of very low density lipoprotein (VLDL) in obese patients with NASH has been reported.155 More recently, dysfunctional VLDL synthesis and secretion has been identified in steatohepatitis compared to simple steatosis.156 High insulin levels also suppress VLDL secretion.157 Finally, mitochondrial beta-oxidation of long chain fatty acids may also be suppressed by insulin,137,139 as well as by impaired tissue responsiveness to PPAR-α, the master fatty acid oxidation-governing transcription factor whose function appears to be impaired in experimental NASH.64,158

Is NASH a disorder of the liver or adipose? Role of adipokines

Just as the initial steps in pathogenesis of T2D have little to do with the pancreatic beta cell, NAFLD/NASH may not be related to intrinsic defects in liver cells. Instead, we conceptualize the early stages of NAFLD/NASH as a failure of whole-body lipid partitioning and metabolic regulation in response to prolonged over-nutrition and development of insulin resistance. A major contributor to this failure is likely to be the adipose tissue. An insufficient response could initiate a cascade of events including rapid hypertrophy of adipocytes without compensatory proliferation, leading to ectopic lipid deposition in muscle and liver. This worsens insulin resistance, further impairing adipocyte proliferation and reinforcing the cycle of impaired metabolic regulation (Fig. 6). In this autopropagative scenario, key adipocyte proteins are likely to play a role, including CD36 which also governs fatty acid uptake in fat tissue and muscle,133,159 phospholipases, such as members of the adiponutrin family mentioned earlier,84–86 and HSL.

Figure 6.

Interactions between adipose tissue in differing sites, muscle and liver in the development of insulin resistance. FFA, free fatty acids; IR, insulin resistance; SAT, subcutaneous adipose tissue; VAT, visual adipose tissue. TG, triglyceride.

Adipokines are important players in this process:160 increased expression and secretion of pro-inflammatory adipokines, such as tumor necrosis factor (TNF)-α and interleukin (IL)-6,161 worsens insulin resistance, while anti-inflammatory and anti-lipotoxic adipokines, including adiponectin and leptin, are dysregulated. Thus, leptin levels rise but tissue leptin resistance develops,48,54 thereby impairing the ability of leptin to decrease food intake, increase energy expenditure and prevent partitioning of lipid into ectopic stores such as muscle and liver (where leptin physiologically activates AMPK and suppresses stearoyl Co-A desaturase-1 [SCD1]).

In contrast, adiponectin levels fall in both metabolic syndrome and NASH (reviewed in 7,138,160), attenuating the anti-inflammatory and pro-proliferative effects of this adipokine on adipose.162 Low serum adiponectin levels also alter lipid partitioning in hepatocytes, where adiponectin switches the metabolic profile by inhibiting lipogenesis and activating fatty acid oxidation through effects on AMPK and PPAR-α.163,164 As evidenced by the adiponectin transgenic ob/ob mouse,135 enhancing subcutaneous fat stores can reverse steatosis and insulin resistance by restoring ‘metabolically healthy’ whole-body lipid distribution. Likewise, treating NASH patients with thiazolidinedione PPAR-γ agonists decreases hepatic lipid content while body weight increases because more fat is stored subcutaneously.14,165 Thus, Harrison and colleagues noted that the most impressive pathophysiological change after institution of pioglitazone therapy in NASH was reversal of adipose insulin resistance,166 thereby restituting HSL-mediated suppression of fasting lipolysis so as to interrupt the unmitigated flow of FFA from adipose to liver.

An important ‘missing link’ in the chain from over-nutrition to NAFLD/NASH and other metabolic disorders, is why some individuals expand VAT at the expense of (or in addition to) SAT expansion. One possibility is innate differences in adipose tissue depots.167 In some individuals, these differences may be genetically exacerbated or compromised. Mature adipose also requires appropriate vascular and stromal support;168,169 angiogenesis may be ineffective during periods of rapid weight gain, leading to a state of hypoxic or other stress (endoplasmic reticulum [ER] stress has been suggested) that initiates inflammatory recruitment.168,169 The SAT depots can be viewed physiologically as a rapidly expandable reservoir of small, insulin-sensitive adipocytes that are ready to absorb excess circulating FFA and TG in the postprandial state.151 The insulin responsiveness of this tissue enables lipid-laden adipocytes to be supplemented by proliferation and maturation of pre-adipocytes. However, if this response is compromised, the subcutaneous lipid store may become replete, with the spill-over accumulating in visceral adipocytes and non-adipose tissues. In contrast to subcutaneous adipocytes, visceral adipocytes are generally larger, store greater amounts of lipid and are less responsive to insulin; this leads to increased (and chronic) lipolytic activity.151,152,167–173]

Another important difference between VAT and SAT is the adipokines released; the VAT depot releases more pro-inflammatory cytokines compared to SAT, while SAT releases more leptin.151–153,170 There is less consensus on which depot is the major source of serum adiponectin, possibly because of different in vitro techniques used to study this aspect.171–175 It is therefore not clear whether reduced secretion of adiponectin by de-differentiated SAT or inflamed VAT is responsible for the drop in adiponectin observed in metabolic syndrome. Likewise, while some workers have found that increased adiponectin levels secondary to thiazolidinedione treatment are due to increased VAT secretion, others have reported that SAT contributes more to serum adiponectin.171,175 Further studies are required to clarify the role of SAT and VAT in regulating serum adiponectin levels in NASH.

In contrast to leptin and adiponectin, the majority of pro-inflammatory cytokines released from adipose tissue come from the non-adipocyte fraction, and VAT is an abundant source of this fraction.170–174 Thus, recruited macrophages play a key role in obesity-associated inflammation.176 VAT secretes more pro-inflammatory cytokines, including TNF-α, IL-6 and monocyte chemoattractant protein-1 (MCP-1),170,172 and this, coupled with direct drainage to the liver, emphasizes the ability of visceral adipose to directly impair hepatic insulin signaling and promote inflammation. TNF-α and IL-6 can activate nuclear factor-kappaB (NF-κB) and c-jun N-terminal kinase (JNK), promoting serine phosphorylation of the insulin receptor substrate so as to directly impair insulin signal transduction.178 Furthermore, while MCP-1 can activate inflammatory pathways, it can also promote hepatocyte triglyceride accumulation directly.177 The coupling of adipose inflammation to hepatic insulin resistance is one of many possible connections between adipose and liver in NASH, as addressed next.

The insulin riddles: where does resistance start, does steatosis cause or result from insulin resistance, and what is partial insulin resistance?

Insulin resistance can be defined as a reduced tissue response to the protean effects of insulin, but operationally it is usually detected by site-specific impairment of insulin's effects on glucose regulation (suppression of hepatic glucose production, impairment of peripheral glucose uptake). Unless pancreatic beta cell production of insulin fails (as in established diabetes), one consequence of insulin resistance is hyperinsulinemia, which may exert multiple effects on hepatic metabolism and growth. One example is stimulation of hepatic lipogenesis via both SREBP1c, for which insulin increases nuclear expression, and ChREBP if glucose intolerance occurs, as well as suppression of fatty acid beta oxidation and VLDL production.155–157 As mentioned earlier, insulin also stimulates expression of the fatty acid uptake pathway, CD36,145,146 thereby exacerbating and potentially altering the pattern of hepatic lipid accumulation.

Clinical significance of insulin resistance in NAFLD/NASH

For some time there was debate about the reproducibility of relationships between NAFLD and insulin resistance (IR). Part of the confusion related to liver pathology (NASH versus SS), and definition of insulin resistance. The usual static estimate is by homeostatic model assessment (HOMA-IR), a computation from fasting serum glucose and insulin values, usually referenced to an arbitrary ‘normal value’ (often 2.0) that ideally should be validated in the local normal population. HOMA-IR is inappropriate when diabetes is associated with declining serum insulin levels. Studies avoiding these pitfalls have tended to find close concordance (> 95%) between HOMA-IR and NASH,4,7,29,30,138 although normal values may be found in a few patients with less severe forms of NAFLD. Dynamic tests of insulin sensitivity, the ‘gold standard’ for which is the euglycemic insulin clamp method, would be more valuable, particularly those using isotope methods for determining peripheral glucose uptake and hepatic glucose output.179,180 One such study found that the peripheral compartment, particularly adipose, contributed most to insulin insenstivity.180 Recent data with conduct of 75G oral glucose tolerance tests (OGTT), coupled with 120 min serum insulin measurements, indicate tighter associations of NAFLD with post-prandial hyperinsulinemia and hyperglycemia than previously appreciated.132,181,182

An emerging theme in NASH pathogenesis is that metabolic abnormalities occur post-prandially. These include a series of changes in response to an oral fat load, such as post-prandial hypertriglyceridemia,96,132 as well as hypoadiponectinemia.121 Such responses could have a genetic basis; for example, subjects carrying certain polymorphisms of microsomal triglyceride transfer protein (MTP), which lipidates apolipoprotein beta to form VLDL, exhibit greater degrees of hypertriglyceridemia, higher serum FFA levels and steatosis.183 Other polymorphisms relevant to post-prandial lipemia include those within the adiponectin gene,184 and the transcription factor 7-like 2 gene.185 The latter, is part of the wnt signalling pathway involved in glucose homeostasis.187 The possibility that polymorphisms in adiponectin or other genes that influence lipid turnover and storage (such as PPAR-α, PPAR-γ and estrogen receptor) could contribute to NASH pathogenesis, perhaps by worsening insulin resistance, has been reviewed recently.187

Which comes first?

Temporal and therefore etiopathogenic relationships between steatosis and insulin resistance remain difficult to unravel. As we previously reviewed,138 both states can potentiate the other and it remains unclear whether insulin resistance or steatosis arises first. This is compounded by the identification of partial, or selective insulin resistance, which can occur where one tissue but not another becomes refractory to the effects of insulin, or at the cellular level when some signaling cascades downstream of the insulin receptor are interrupted while others remain responsive to insulin. At the whole body level, hepatic insulin resistance may develop, while peripheral tissues remain sensitive to the effects of insulin. One example is the methionine and choline deficient (MCD) model of steatohepatitis where peripheral insulin sensitivity is enhanced (by weight loss),188,189 but defects in hepatic insulin receptor signaling develop in association with FFA accumulation and induction of cytochrome P4502E1.190 As mentioned above, there is more evidence to support a peripheral site of insulin insensitivity with NAFLD,141,166,180 with the resultant hyperinsulinemia driving lipogenesis (Fig. 6).

Partial insulin resistance

The cellular divergence of insulin signaling, while still poorly understood, is likely to underlie the up-regulation of hepatic de novo lipogenesis observed with hyperinsulinemia, indicating continued sensitivity to one action of insulin, compared to impaired suppression of hepatic gluconeogenesis (‘classical’ insulin resistance).191 With complete hepatic insulin resistance, achieved experimentally by liver-specific knockout of the insulin receptor, steatosis does not develop. This indicates that steatosis which arises during hepatic insulin resistance requires a functional insulin receptor and is secondary to hyperinsulinemia.129 Some evidence suggests that the divergence may occur at the level of the insulin receptor substrate (IRS) molecules.191 In models of insulin resistance with hyperinsulinemia, IRS2 levels decrease in association with persistent expression of gluconeogenic genes, while nuclear translocation of SREBP1c is enhanced.141 IRS2 mediates gluconeogenesis by a signaling cascade involving Akt and FOXO-1; activity of these molecules is decreased in selective insulin resistance.129,161 However, the mechanism(s) by which insulin continues to enhance SREBP1c activity remains unclear. Alternatively, insulin-stimulated SREBP1c activation may indeed be impaired, and non-classical pathways may contribute to enhanced SREBP1c activity and subsequent steatosis. For example, endoplasmic reticulum (ER) stress was recently identified as a possible mediator of insulin-independent SREBP1c activation.192

In addition to controlling effects on carbohydrate and lipid metabolism, the insulin receptor also signals via JAK-STAT and mitogen-activated protein (MAP) kinases.193 In addition to the PI3 kinase/Akt/S6 kinase pathway, these signaling pathways have roles in cell growth and survival, cell proliferation and opposition to cell death that could contribute to inflammatory recruitment, fibrogenesis (for example, via connective tissue growth factor) and hepatocarcinogenesis with NAFLD/NASH. For example, the increasing evidence that a high-fat diet might predispose to HCC, both directly and by contributing to obesity,194,195 is consistent with the known effects of high dietary fat on reducing insulin sensitivity in liver and elsewhere. Understanding the effects of insulin resistance on these pro-proliferative pathways may help unravel the relationships between obesity, metabolic disease and carcinogenesis.

Tissue resistance to the hormone/cytokine actions is not confined to insulin; leptin resistance is commonly recognized in obesity,41,43,48,54 while other signaling and regulatory pathways may also be impaired in metabolic disease. For example, we have demonstrated hepatic adiponectin resistance in the MCD model of steatohepatitis, in which high serum adiponectin levels activate AMPK in muscle, but fail to activate AMPK or PPAR-α in liver.154 In the foz/foz model (metabolic syndrome-associated steatohepatitis), there also appears to be hepatic refractoriness to activation of PPAR-α, despite accumulation of fatty acids (including those derived from de novo lipogenesis) which usually activate PPAR-α.64 The failure of these homeostatic (or adaptive) pathways leads to worsening metabolic disease.

Amplifying loops—from injury to steatosis

In his recantation of the ‘two-hit’ hypothesis, Dr Chris Day emphasized the importance of ‘injury mechanisms’ themselves perturbing hepatic lipid homeostasis.[C Day—verbal communication, 26 September 2009; and reviewed 196] Pathways such as those activated by MCP-1 and ER stress have already been mentioned here, while TNF-α, oxidative stress and mitochondrial injury may all lead to hepatic accumulation of fatty acids and/or triglyceride, particularly by impairing fatty acid oxidation (Table 4). While we are not convinced of the primacy of these pathways for causing steatosis, they are likely to play roles in steatohepatitis transition by facilitating accumulation of FFA and other potentially toxic lipid molecules, as will be discussed in Part 2 of this review.

Table 4.  Metabolic-inflammatory ying-yang in the liver: inflammatory mediators that influence hepatic lipid metabolism
MediatorEffects on inflammationEffects on hepatic lipid metabolismNet effect on hepatic lipidsEffects on insulin sensitivity
  1. AMPK, AMP protein kinase; ER, endoplasmic reticulum; IL, interleukin; JNK, c-Jun N-terminal protein kinase; MCP-1, monocyte-chemoattractant protein-1; MTTP, microsomal triglyceride transport protein; PPAR-α, peroxisome proliferation activated receptor-α; PKC, protein kinase C; ROS, reactive oxygen species; SIBO, small intestine bacterial overgrowth; SOCS3, suppressor of cytokine signaling-3; SREBP, sterol-response element binding protein; TLR, toll-like receptor; TNF-α, tumor necrosis factor-α; VLDL, very low density lipoprotein.

AdiponectinBlocks TNF-α synthesis and releasevia AMPK: opposes lipogenesis (ACC phosphorylation)
via PPAR-α: stimulates oxidation pathways (see Fig. 5)
DecreaseIncreased sensitivity (opposes insulin resistance)
PPAR-αAnti-inflammatory (blocks NF-κB activated gene transcription)Stimulates oxidation pathways (see Fig. 5)DecreasesIncreased sensitivity
IL-1βInflammatory mediatorIn apoE knockout mice and cultured hepatocytes, activates SREBP2 with downstream effects on cholesterol turnover favouring accumulationNot shown—potentially increases cholesterolNot shown
MCP-1Activates macrophage recruitment, TNF-α synthesis and releaseStimulates lipogenesis (via SREBP-1c)Possible increase (shown in vitro)Not shown
TNF-αMediates, propogates inflammation (via NF-κB, IL-6); suppresses adiponectin/ adiponectin receptor 2 expressionStimulates lipogenesis (via SOCS3-induced activation of SREBP1c); impairs mitochondrial β-oxidation, MTTP-dependent formation of VLDLNot shownSuspected mediator of insulin resistance
IL-6Pro-inflammatory (time/dose dependent), but hepatoprotective/proproliferative on hepatocytesStimulates lipogenesis (via SOCS3-induced activation of SREBP1c)IncreasesSOCS3 induction impairs insulin receptor signaling
Oxidative stressActivates redox-sensitive pathways: JNK, NF-κB, PKCDecreases VLDL formation/secretionExpected to increaseImpaired via JNK—serine/ threonine phosphorylation of insulin receptor
ER stressROS generation; JNK activationStimulates lipogenesis (via SREBP-1c)Causal association not shownImpaired via JNK—serine/ threonine phosphorylation of insulin receptor
Endotoxin (SIBO)TLR4-dependent activation of NF-κBPossible downstream effects via TNF-α and IL-6 (see above)Possible worsening of steatosis severityNot shown

Summary and future directions: NASH is a metabolic disease

The modern context of abundant, cheap high-energy food together with sedentary lifestyle favors over-nutrition, including among children and young adults. NASH has its origins in such early-onset over-nutrition and insulin resistance, and better characterization of which diets, lifestyles and socio-economic factors are most detrimental is an important direction in future research. Genetic factors that drive the urge to eat high density foods and those that affect lipid storage and adipose biology are likely to play important roles. Abnormal lipid partitioning favoring visceral (central) adiposity is central to understanding NASH pathogenesis; fundamental studies of the adipose and factors which regulate its expansion and contraction, inflammatory recruitment and decreased adiponectin secretion (adipose failure) should provide insights into NASH as well as metabolic syndrome. Steatosis may have its origins in hyperinsulinemia and hyperglycemia driving hepatic lipogenesis, and this may cause hepatic insulin resistance that is ‘exported’ to peripheral sites (muscle, adipose) by inflammatory mediators like TNF-α and IL-6. Alternatively, these cytokines might arise first from stressed and inflamed, failing adipose tissue, particularly VAT, causing both adipose and hepatic insulin resistance. Once systemic insulin resistance is established, hepatic uptake of the continuous stream of FFA arising from post-prandial lipolysis in adipose seems to be what augments hepatic lipid to critical levels and/or favors a molecular lipid profile that causes tissue injury (lipotoxicity). Since adipose de-differentiation is pharmacologically reversible (e.g. by PPARγ agonist ‘glitazones’), a better understanding of these processes could be harnessed to halt progression of steatosis to steatohepatitis. Ultimately, the liver, too, has a finite reserve capacity for lipid storage. Why this appears to ‘hold firm’ in those with simple steatosis but becomes insufficient in those with NASH, a failure of adaptive mechanisms, is the critical issue in NASH pathogenesis. Explanations could lie in the types of lipid molecules that accumulate, ways in which they are packaged into safer storage sites (or not), effects of lipid molecules on critical organelles such as the ER, mitochondria and plasma membrane, and differential innate immune responses—in which the gut microflora may play a role. These issues will be addressed in the next part of this review.