A narrative review of factors associated with the development and progression of non‐alcoholic fatty liver disease

With the obesity pandemic, non‐alcoholic fatty liver disease (NAFLD) has become the most prevalent liver disease. NAFLD can progress to non‐alcoholic steatohepatitis (NASH), a potential cause of liver failure. It remains difficult to identify patients at risk for NASH, despite evolving insights in contributing factors, including genetic variance, hormones, adipokines, diet and body‐fat distribution. We aimed to present a broad perspective on these risk factors associated with NAFLD development and progression with a focus on their contribution in different age groups and susceptible high‐risk populations, hereby giving insight in the pathophysiology of NAFLD.

The heritability of NAFLD approaches 40% 3,8 and mutations in certain genes associated with lipid metabolism are known to cause NAFLD in patients with and without obesity. 3 However, the determinants and mechanisms driving progression of simple steatosis to more severe forms of NASH, sometimes resulting in liver fibrosis already in childhood, are incompletely understood.
To make the next step in clinical care for patients with NAFLD, a deep understanding of the pathogenesis of NAFLD and modifiers determining progression of liver disease is crucial. We set out to deepen this understanding by studying patients with severe forms of NAFLD and search for specific disease modifying factors. We considered children with NASH of particular interest since they have an early disease onset and are subject to large biological fluctuations in hormones, metabolism, insulin resistance (IR) and lifestyle. Likewise, severe phenotypes of NAFLD are seen with specific genetic variants providing insight in the contribution of affected pathways to NASH development. 9 Understanding the factors that pre-dispose to rapid disease progression can help to identify patients at increased risk for developing severe liver pathology and develop novel prevention and treatment strategies. In this narrative review, we will discuss the pathogenesis of NAFLD, elaborating on the relationship with IR and lipid metabolism. Subsequently, we will analyse the important factors in progression to NASH such as dysbiosis and TLR4 activation, the role of adipokines and how lipotoxicity is translated in oxidative and endoplasmic reticulum stress, leading to apoptosis. We will elaborate on risk factors that are subject to change during the transition from childhood to adulthood and on genetic variants that pre-dispose to NAFLD and NASH.

| ME THODS
PubMed electronic database from 2003 to 2018 was searched for relevant publications using the following terms: "NAFLD" OR "Non-alcoholic fatty liver disease" OR "NASH" OR "Non-alcoholic steatohepatitis" AND "pathogenesis" OR "mechanism" OR "pathology." This review focuses on risk factors in the transition from childhood to adulthood, for which we repeated the previous search combined with the following terms: AND "Child" OR "Childhood" OR "Children" Or "Adolescent" OR "Puberty" OR "Pubertal" OR "Paediatric" OR "Pediatric." The inclusion criteria were (a) Peer-reviewed academic journals published in English (b) research that focused on NAFLD development and factors associated with NAFLD development (c) articles with accessible abstracts and full text. Articles were read and assessed for relevance.
Hand searching of the references of retrieved literature was done.
Backward snowballing was used to complement the database search.

| PATHOG ENE S IS OF S TE ATOS IS AND INSULIN RE S IS TAN CE
Central to hepatic steatosis are TG accumulation and IR. Hepatocytes accumulate TGs through fatty acid (FA) acquisition from dietary fat, adipose tissue lipolysis and hepatic lipogenesis. They clear TGs through FA β-oxidation and TG export as very low-density lipoprotein (VLDL) via the endoplasmic reticulum (ER) ( Figure 1A). 3,10,11 Insulin influences these processes by inducing lipogenesis and antagonising tissue lipolysis.

| Obesity-induced systemic IR and FA release
Obese patients become less responsive to insulin, which is often attributed to chronic inflammation of adipose tissue. In obesity,

Highlights
• Non-alcoholic fatty liver disease (NAFLD) is the most prevalent liver disease, but factors contributing to progression to severe non-alcoholic steatohepatitis (NASH) are insufficiently known for targeted prevention or intervention strategies.
• Exploration of NAFLD development in various age groups, from childhood to late adulthood gives insight in specific disease contributing factors.
• Hormonal factors (androgens opposed to oestrogens) and dietary fructose appear most harmful in promoting NAFLD progression and the latter represents an easy target for intervention.
• The human body has different ways to handle excessive nutrients, but eventually each way leads to a burden such as NAFLD or cardiovascular disease.
• Stratification based on sex, body-fat distribution, diet, lifestyle, microbiome, adipokines, sex hormones, blood concentrations of liver enzymes, liver histology and genetic pre-disposition might help to identify patients at increased risk to develop NASH, and current genetic screening possibilities might reveal novel or more specific targets for prevention or intervention strategies. storage of excessive TGs leads to hypertrophic adipose tissue, which enhances the production of pro-inflammatory cytokines TNF-α, IL-1, IL-6 and monocyte chemoattractant (MCP-1). 11 These pro-inflammatory factors antagonise the lipogenic effects of insulin, thereby contributing to systemic IR, which increases adipose tissue lipolysis. Together, this leads to increased hepatic FA concentrations. 11 Furthermore, MCP-1 attracts macrophages, creating low-grade inflammation. These macrophages and other resident cells subsequently produce even higher levels of inflammatory cytokines, which promote leptin production, and reduce adiponectin production 10,11 and affect IR. 12 Adipose tissue inflammation, IR and increased FA influx in the liver stimulate each other, creating a positive feedback loop. More than half of hepatic fat is derived from lipolysis from adipose tissue 3 and this correlates with total fat mass in humans, resulting in increased FA release in obese patients. Normally, insulin inhibits adipose tissue lipolysis, but this is impaired in patients with systemic IR, resulting in elevated efflux of FAs from adipose tissue. 11,13 In the liver, the hyperinsulinaemia associated with IR inhibits β-oxidation and causes an upregulation of transcription factor SREBP-1c leading to enhanced lipogenesis. 3,11,13 These combined effects lead to hepatic steatosis. 11,13 Within this positive feedback loop, it is unclear whether IR precedes or follows steatosis. 3 Body fat distribution is important, as visceral adipose tissue produces more pro-inflammatory cytokines, pro-hyperglycaemic factors and less adiponectin than subcutaneous adipose tissue. 14 Hence, IR and NAFLD have a stronger correlation with visceral adiposity than with BMI BMI. 15 Also, IR seems to decrease with higher skeletal muscle mass, which can be attributed to insulin-mediated glucose utilisation. A 7-year follow-up of patients with NAFLD showed a correlation between increased skeletal muscle mass and improvement or prevention of NAFLD, which might contribute to a stratification for at-risk disease-susceptible individuals. 14

| FA influx in the liver and hepatic IR
In the liver, FA are metabolised through β-oxidation, exported as VLDL or stored as TGs ( Figure 1A 16 This in turn suppresses glycogen synthase activity, leading to decreased glycogen synthesis. In parallel, reduced IRS-2 activity causes translocation of Forkhead box protein O1 to the nucleus, where it enhances the expression of various gluconeogenic enzymes. 16 F I G U R E 1 Interpretation of the current literature on the pathogenesis of steatosis (A) and factors contributing to the progression to NASH (B). FA = fatty acid; SER = smooth endoplasmic reticulum; VLDL = very low-density lipoprotein; DAG = diacylglycerol; TG = triglyceride; IRS-2 = insulin receptor substrate 2, ROS = Reactive oxygen species, LPS = lipopolysaccharides. A. Increased FA influx from lipolysis in adipose tissue, diet and lipogenesis causes increased IR via DAG and decreased phosphorylation of IRS-2, which leads to elevated glucose via enhanced gluconeogenesis and reduced glycogen synthesis. Elevated glucose levels trigger more insulin release. Insulin resistance stimulates lipolysis of adipose tissue and de novo lipogenesis, while it decreases VLDL export, further contributing to increased FA concentrations and steatosis. B. Elevated β-oxidation of FAs generates ROS. FAs are incorporated in the ER membrane, leading to the unfolded protein response. Both pathways cause hepatocyte apoptosis and contribute to inflammation. Inflammation is further triggered by intestinal bacterial overgrowth and adipose tissue inflammation. Inflammatory cytokines increase IR. Adipose tissue ameliorates insulin sensitivity via increased production of leptin and adiponectin. Adiponectin production is decreased in NASH. Although leptin production is increased in NAFLD, this does not lead to lower insulin concentrations, suggesting leptin resistance. With this resistance, the protective effect of leptin against hepatic steatosis is lost Cytosolic DAG thus inhibits IRS-2, which leads to hepatic IR and enhanced gluconeogenesis and inhibited glycogen synthesis. 17

| PATHOG ENE S IS OF NA S H
Elevated hepatic FA concentrations increase the susceptibility of the liver to injury and are seen as the first hit. Multiple hits including IR, nutritional factors, adipokines, gut microbiota and genetic factors are now thought to cause the liver to progress from steatosis to NASH ( Figure 1B). 10,13,18

| Inflammation
Pro-inflammatory cytokines play a crucial role in the progression from steatosis to NASH. 3 First, it has been demonstrated in rodents that specific cytokines can elicit the same response in liver tissue as seen in NASH, including neutrophil chemotaxis, hepatocyte apoptosis/necrosis, Mallory body formation and stellate cell activation. 11 Second, the enhanced expression of TNF-α and IL-6 in adipose tissue was detected in obese subjects before liver inflammation was present, 19 suggesting that adipose tissue mediated cytokine release precedes, and might lead to liver inflammation. Third, inhibition of TNF-α led to amelioration of IR and histological parameters of NASH in an obese mouse model on a high-fat diet. 20 Cytokines lead to NASH development through the activation of the nuclear factor κB (NF-κB) and C-Jun NH 2 -terminal kinase (JNK) pathway. FAs, particularly saturated FA, were found to activate the toll-like receptor 4 (TLR4), activating both the NF-κB and JNK pathway. 21,22 In the liver, TLR4 is expressed by hepatocytes, Kupffer cells and stellate cells. Upon phagocytosis of cholesterol, Kupffer cells are activated and TLR4 expression is upregulated. 23 Activated Kupffer cells produce transforming growth factor β (TGF-β), which triggers the fibrogenic state of stellate cells. 24 The importance of TLR4 signalling in the pathogenesis of NASH is illustrated by TLR4 mutant mice, that neither develops extreme adiposity and IR in response to a high saturated fat diet nor NASH in response to a methionine/choline-deficient diet. 25,26 Thus, inflammation through TLR4 receptor activation appears to be a major factor in murine models in the progression from steatosis to NASH. In humans, these mechanisms still require confirmation.

| Adipokines
In addition to being a major source of FAs and cytokines (IL-6 and TNF-α), adipose tissue releases adipokines, such as leptin and adiponectin. 10 Leptin and adiponectin enhance hepatic insulin sensitivity 12 and appear to be the important factors in the pathogenesis of NASH.
Leptin acts centrally to decrease food intake and has anti-hyperglycaemic effects. 27 In response to lower glucose concentrations, insulin concentrations decrease, reducing hepatic lipogenesis and increasing lipolysis. Similarly, in the liver, leptin stimulates FA β-oxidation and suppresses lipogenesis in vitro, decreasing the fat accumulation and lipoapoptosis. 10,27 Leptin is markedly increased in obesity and hepatic steatosis, and even more in NASH. 27 High leptin levels do not decrease insulin concentrations or IR in obese humans. Potentially, the protective effects of leptin are lost in obesity due to excessive leptin exposure and leptin resistance.

This might be caused by the suppressor of cytokine 3 (SOCS-3)
which is activated by leptin and inhibits leptin signalling when overexpressed. SOCS-3 expression is also stimulated by insulin and SOCS-3 overexpression leads to IR, whereas the inhibition of SOCS-3 ameliorates insulin sensitivity and hepatic steatosis. 27 In obesity, leptin enhances the pro-inflammatory signalling cascades via activation of NF-κB. 28 Moreover, prolonged hyperleptinaemia may lead to fibrosis via activated stellate cells. 27  Adiponectin promotes insulin sensitivity and reduces inflammation, thereby protecting against NAFLD progression. Total body fat is inversely correlated with adiponectin, and adiponectin blood concentrations are lower in patients with NASH than with steatosis. 19,28 Adiponectin increases FA β-oxidation and decreases lipogenesis, thereby reducing FA concentrations. 18,28 Adiponectin upregulates IRS-2 in an obese mouse model, thereby promoting insulin sensitivity. 29 Insulin sensitivity is further increased by adiponectin via antagonising TNF-α and IL-6 effects, and enhancing the secretion of anti-inflammatory cytokines IL-10 and IL-1 receptor antagonists in human leucocytes in vitro. 30 Adiponectin downregulates TLR4-induced NF-κB activation and leads to reduced TNF-α production in rat Kupffer cells in vitro and in mice upon exposure to lipopolysaccharides (LPS). 31 Finally, adiponectin has been shown to inhibit stellate cell proliferation in vitro, possibly via inhibition of ROS production 30 ( Figure 2). Summarising, adiponectin promotes insulin sensitivity, protects from liver inflammation and limits fibrosis ( Figure 2). Decreased adiponectin levels might pre-dispose NAFLD patients to progress to NASH.

| Lipotoxicity
Previously, it was thought that accumulation of TGs in hepatocytes caused hepatic inflammation and IR. However, storage of FAs as inert lipids is not toxic 10,13 and the conversion of FAs to TGs functions as a protective mechanism. 16 This is supported by mouse models with deficient TG synthesis, resulting in decreased hepatic steatosis but increased liver injury and fibrosis. 32 Similarly, genetic defects that prevent the removal of TG from the liver cause steatosis, but not IR. 3,16 Thus, intrahepatic TG content is a marker for increased FA exposure-rather than a cause of hepatic IR.
FA excess leads to hepatocyte dysfunction and ultimately apoptosis. Specifically, saturated FAs cause toxicity. 33 Mono-unsaturated FAs cause TG accumulation without leading to cell damage.
Potentially, the ratio between mono-unsaturated and saturated FAs determines whether hepatocytes are damaged by high FA concentrations. 18 This ratio is determined by stearoyl-CoA desaturase-1 (SCD1), which converts saturated FA to mono-unsaturated FA. 34 Genetic inhibition of SCD1 in mice leads to less steatosis, but more apoptosis, supporting the theory that saturated FAS are lipotoxic. 33 Saturated FA influx in the hepatocyte leads to concentrationdependent β-oxidation of FAs in the mitochondria. 35  Saturated FAs exert further lipotoxic effects by changing the critical free cholesterol-to-phospholipid ratio of the ER membrane, affecting ER membrane fluidity and calcium homeostasis. 18,24 Disruption of this homeostasis leads to accumulation of misfolded F I G U R E 2 Schematic representation of the effects related to NAFLD/NASH of oestrogens, androgens, adiponectin and leptin. A. Combined effects of oestrogens reduce the chance of progression to NASH. B. Androgens promote visceral adipose tissue distribution and inhibit production of adipokines, thereby increasing the risk of progression to NASH. In obesity, androgen conversion to oestrogen is increased, but inhibition of leptin production by androgens is lost. C. Adiponectin inhibits several risk factors for progression to NASH and promotes hepatic lipolysis and FA β-oxidation. D. Leptin has similar effects to adiponectin, but these are partly lost in obesity, putatively related to leptin resistance. Moreover, high leptin levels trigger stellate cell activation and stimulate stellate cell proliferation, instigating a positive feedback loop with activated stellate cells producing leptin, thereby contributing to progression to NASH. -: effects under normal conditions; ---: pronounced or changed effects in obesity; ↑: stimulation; ⊥: inhibition

| Dysbiosis
The microbiome composition of patients with obesity or NAFLD is different from controls 23,38 and small intestinal overgrowth is more common in NAFLD. 39 43 Additionally, NAFLD patients have higher serum ethanol levels, which is attributed to more fermenting bacteria in their microbiome. 44 This increases the translocation of endotoxins, TLR4 activation, generation of ROS and TNF-α. 44 Concurrently, probiotic treatment of paediatric NAFLD leads to an improved lipid profile and decreased hepatic steatosis. [45][46][47] These data suggest that high-fat and high-fructose diets contribute to altered gut microbiome and translocation of LPS into the circulation, which stimulates hepatic inflammation via TLR4.

| PAED IATRI C NAFLD
NAFLD already occurs in early childhood, 15 but is more common in adolescents. 8 Insight in paediatric NASH can be considered of specific relevance to identify factors affecting fast progression of NAFLD. In addition, paediatric patients offer the opportunity to evaluate the effects of lifestyle, growth, anthropometry and hormonal factors on the pathogenesis of NAFLD, since these factors all highly fluctuate during the transition from childhood to adulthood.

| Histology
Paediatric NASH differs in histological characteristics from adult NASH. Whereas adult NASH is characterised by perisinusoidal changes including steatosis, lobular inflammation, hepatocellular ballooning and fibrosis in the absence of portal changes (defined as type 1), 4 paediatric NASH predominantly includes portal steatosis, inflammation and fibrosis (defined as type 2), 48,49 or an overlap between both types. The portal pathology is most striking in boys, severe metabolic syndrome and severe obesity. 1,7,50,51 The portal area is exposed to blood first, receiving the highest concentrations of nutrients, oxygen, hormones and inflammatory or toxic substances, and harbours the highest concentration of Kupffer cells.
Portal hepatocytes are specialised in oxidative liver functions such as gluconeogenesis, FA β-oxidation and cholesterol synthesis. 52 One might thus hypothesise that increased oxidative stress or high concentrations of harmful nutrients such as fructose result in portal injury and might accelerate progression from steatosis to NASH in children. This is supported by the finding of increased steatosis in the portal area in rats fed with both sucrose and fructose-enriched diets. 50 Similarly, Nobili et al recently showed that high-fructose intake in children correlated with increased portal steatosis, inflammation and hypothesis. 53

| Lifestyle and diet
During childhood, children change their lifestyle habits, which influence NAFLD progression in a gender dependent manner. During adolescence, physical activity decreases, while sugar and fat intake increase. 54 Girls are less active than boys and already have reduced levels of physical activity before the onset of puberty, 55,56 which does not correspond with the observation of lower NASH prevalence in girls. The high NASH prevalence in boys correlates with the high-carbohydrate intake in boys. 56 High-carbohydrate intake in humans has been linked to increased IR and portal FA concentrations, upregulation of lipogenesis and inhibition of β-oxidation, all contributing to increased FA load and steatosis. 57 More specifically, fructose intake has emerged as a determinant in NAFLD. 58 The average yearly consumption of fructose has significantly increased over the years. 59 Fructose consumption is higher in children from 9 to 18 years compared to adults 60 and two to threefold higher in NAFLD patients compared to controls. 58,61,62 While glucose mainly serves as a substrate for glycogenesis, fructose mainly serves as a substrate for lipogenesis. 63 High fructose instead of glucose intake was found to increase dyslipidaemia, intrahepatic TG accumulation, IR, hepatic inflammation and fibrosis in humans. 10,64,65 High-fructose diet elicited inflammatory changes in the JNK pathway 66 and enhanced the SOCS-3 expression inducing leptin resistance in rats. 59 High-fructose diet was further found to induce low-grade endotoxaemia via enhanced-intestinal permeability, thereby pre-disposing to NASH. 43,61 Overall, the effects of fructose on NAFLD are clearly illustrated by the decrease in liver transaminases induced by a low fructose and low-glycaemic index diet without caloric restriction in children with NAFLD. 67 Similarly, obese children showed reduced hepatic fat, visceral fat, lipogenesis and IR already after 9 days of fructose restriction. 63

| Growth (anabolism)
Insulin sensitivity fluctuates with sex and age 66,68 (Figure 3). At the beginning of puberty, IR is similar in both sexes and gradually rises during puberty. 66,68 Growth hormone (GH) secretion at night leads to increased IR. 69 To compensate for IR, insulin secretion increases two to threefold during puberty. 69 GH and its downstream primary mediator insulin-like growth factor 1 (IGF-1) play an important role in the regulation of lipid and glucose metabolism. GH stimulates adipose tissue lipolysis, increasing the hepatic influx of FAs. GH also enhances the hepatic lipogenesis through the activation of SREBP-1c. 70 GH signalling stimulates the hepatic TG secretion via increased VLDL export and enhances FA oxidation. 70,71 Overall, GH lowers fasting free FAs. 69 GH stimulates the hepatic glucose production, and triggers IR specifically in adipose tissue and muscle, leading to hyperinsulinaemia. Hyperinsulinaemia synergises with GH and IGF-1 to stimulate protein anabolism supporting growth.
The increase in GH, IGF-1 and insulin concentrations during adolescence ensure anabolism and thereby protect against steatosis.
Circulating GH and IGF-1 levels are lower in NAFLD patients than in controls. 72,73 Patients with GH deficiency have an increased NAFLD prevalence. 72 Moreover, GH secretion is reduced in obesity, 74 without a proportional decrease in IGF-1 concentrations, 70 maintaining a negative feedback loop for GH. This consequently enhances the NAFLD progression. GH replacement therapy in a patient with GH deficiency reversed NASH. 70,74 Furthermore, treatment with GH reduced visceral adiposity and liver fat. 75 All these studies support the anabolic role for GH and IGF-1, protecting against IR and hepatic steatosis during puberty. 69,70

| Hormonal sex differences during puberty
The role of hormonal influences in the development of NASH is implied by the studies showing that NAFLD is more common in boys than girls, 7 nearly reaching a ratio of 2:1 in obese children 6 and further diverges during adolescence, increasing in boys and decreasing in girls. 76 In obese children, liver steatosis increases from 40% to 51% during puberty in boys, while it decreases from 17% to 12% in girls (reaching 25% mid-puberty; Figure 4). 76 During puberty, oestrogen secretion increases steadily in girls and only slightly in boys ( Figure 3). Oestrogens reduce gluconeogenesis and glycogenolysis. 77 Consequently, human subjects with deficient oestrogen synthesis exhibit increased IR, which can be resolved by oestrogen supplementation. 77 Oestrogens have anti-oxidant properties, as they inhibited ROS generation, lipid peroxide levels, hepatic inflammation, stellate cell proliferation and activation, apoptosis and fibrosis in rats 28 ( Figure 2). Oestrogens stimulate the use of lipids as energy source and promote nonvisceral adipose tissue distribution 78 (Figure 2).
Leptin levels also show a sex-dependent pattern: in girls, leptin concentrations increase steadily during puberty, while decreasing from mid-puberty onwards in boys (Figure 3). Androgens inhibit leptin production, but this association is lost in obesity as leptin levels increase 79 ( Figures 2 and 4). This might be explained by enhanced conversion of androgens to oestrogens in obesity. 77 Moreover, oestrogens are thought to increase the sensitivity to central leptin. 78 In obese boys and girls, leptin concentrations are high throughout puberty, with a with an increased prevalence of hepatic steatosis, increased ALT concentrations, severe inflammation and fibrosis. 82 The PNPLA3 protein has lipase activity towards TGs and is highly expressed in hepatocytes and stellate cells. 83  Another strong genetic predictor for NAFLD in children and adults is the glucokinase regulator (GCKR) SNP (rs1260326) that inhibits GCKR's response to fructose-6-phosphate and associates with liver fat content in NAFLD. 89,90 Another common SNP that is associated with NAFLD is the  I  I I  III  IV  V I  II  III  IV  V I  II  III  I V  V   48 I  I I  III  IV  V I  II  III  IV  V I  II  III  I 91 Consequently, hepatic TG content increases and circulating lipid concentrations decrease, leading to an improved overall cardiovascular risk profile, but increased risk of NASH. 92,93 The association between increased hepatic steatosis in patients with the TM6SF2 SNP and advanced liver disease has been found both in adults and children. 94 The association of this gene with NAFLD suggests a significant role for VLDL export in preventing NASH.
Other candidate genes associated with NAFLD have been pro- FAs are stored in the liver and adipose tissue as TGs, lowering toxic FA intermediates. However, TG storage leads to adipose tissue inflammation, which increases the risk of developing the metabolic syndrome. Accordingly, elevated FA β-oxidation reduces FA concentrations, but increases oxidative damage, accelerating progression to NASH. Alternatively, FAs can be exported from the liver as VLDL, decreasing the risk of progression to NASH, but worsening the cardiovascular risk profile. Strain on specific pathways will determine the symptoms in individual patients. Because each strategy to handle excess nutrients comes at a price, optimising the balance between energy intake and expenditure still seems the most effective therapeutic option.

E TH I C S S TATEM ENT
The authors confirm that the ethical policies of the journal, as noted on the journal's author guidelines page, have been adhered to. No ethical approval was required as this is a review article with no original research data.