DNA methylation

Among epigenetic mechanisms, DNA methylation arises as a potent regulator of gene expression, whose intervention has been strongly related to carcinogenesis⁽ Reference Franco, Schoneveld and Georgakilas ^{40 )}. DNA methylation is a modification that occurs at cytosine nucleotides, in particular in those which precede a guanine, usually referred as CpG islands. These CpG islands are typically located in promoter regions of oncogenes, DNA repair genes or tumour-suppressor genes⁽ Reference Lahtz and Pfeifer ^{41 )}, thus affecting gene expression in several cancers. In this respect, it has been shown that while hypermethylation blocks gene transcription, hypomethylation induces gene activation. This degree of DNA methylation depends on the activity of DNA methyltransferases (DNMT).

Nutrition is strongly related to DNA methylation since this epigenetic mechanism depends on the availability of S-adenosylmethionine (SAM), which needs methyl donors from food (folate, choline and betaine) for its synthesis⁽ Reference Niculescu and Zeisel ^{42 )}. In this respect, methyl-deficient diets were reported to induce liver injury in a similar way as that observed in human NASH patients⁽ Reference Anderson, Sant and Dolinoy ^{43 )}. Moreover, this diet was shown to affect DNA methylation as a consequence of an important decrease in liver SAM concentration. At the same time, it resulted in the differential expression of more than a hundred genes associated with DNA damage and repair, lipid and glucose metabolism, fibrosis, and liver tissue remodelling. In this regard, the number of hypomethylated genes was significantly greater than the number of hypermethylated ones⁽ Reference Lillycrop, Phillips and Torrens ^{44 )}. Another study reported that in a model of intra-uterine growth restriction, using rats exposed to a low-protein diet during pregnancy, the placenta showed hypomethylation of the wingless-type MMTV integration site family member 2 (Wnt2) promoter, which is associated with fetal growth⁽ Reference Reamon-Buettner, Buschmann and Lewin ^{45 )}. Moreover, it was evidenced that a low-protein diet during pregnancy was conducive to DNA hypermethylation of H19/insulin-like growth factor II (IGF2) genes that resulted in up-regulation of DNMT1 and DNMT3 in the liver of male offspring⁽ Reference Gong, Pan and Chen ^{46 )}.

As was explained before, DNA methylation can be inherited from parents. It has been shown that exposure to a high-fat diet (HFD) during pregnancy can induce hypomethylation of cyclin-dependent kinase inhibitor 1a (CDKN1a), a liver cell cycle inhibitor, that acts during the early postnatal life of the offspring⁽ Reference Dudley, Sloboda and Connor ^{47 )}. Given that CDKN1a up-regulation has been related to hepatocyte growth in liver disease, this result suggests that a HFD in this period induces an early liver dysfunction, which may later predispose to NAFLD. Another study demonstrated that consumption of a Western diet during pregnancy and early postnatal life is conducive to an augmented susceptibility to NAFLD in male offspring⁽ Reference Pruis, Lendvai and Bloks ^{48 )}.

Liver steatosis plays a key role in NAFLD pathophysiology. It requires coordination between PPARα and PPARγ for balance between fatty acid synthesis and oxidation⁽ Reference Varga, Czimmerer and Nagy ^{49 )}. While PPARα has been shown to be involved in the modulation of peroxisomal and mitochondrial β-oxidation in liver and skeletal muscle, PPARγ, on the other hand, has been found to be central in the regulation of adipogenesis and to decrease the expression of proinflammatory cytokines. It has been reported that PPARα expression is down-regulated in liver steatosis, thus contributing to lipogenesis⁽ Reference Sun, Fan and Qiao ^{50 )}. This effect can be aggravated by the up-regulation of PPARγ, as a result of PPARγ promoter hypomethylation⁽ Reference Giby and Ajith ^{51 )}. In addition, the epigenetic alteration of liver PPARγ in NAFLD patients has been shown to contribute to insulin resistance. Furthermore, methylation levels of PPARγ coactivator 1α (PGC1α) promoter correlated with the homeostatic model assessment of insulin resistance (HOMA-IR) and fasting insulin, while those of mitochondrial transcription factor A (TFAM) promoter were inversely correlated with fasting insulin⁽ Reference Sookoian, Rosselli and Gemma ^{52 )}.

Mitochondrial DNA methylation has also been involved in the development of NAFLD⁽ Reference Chen, Broséus and Hergalant ^{53 )}. It has been reported that mitochondrial NADH dehydrogenase 6 (MT-ND6) is strongly methylated, which leads to MT-ND6 mRNA down-regulation in NASH patients. Moreover, hepatic methylation of MT-ND6 was shown to be associated with NAFLD⁽ Reference Pirola, Gianotti and Burgueño ^{54 )}.

Although there are more data from animal models of NAFLD, DNA methylation alterations have also been reported in NAFLD patients. Furthermore, nearly 70 000 methylated CpG sites in genes related to tissue repair and metabolic regulation have been found in these patients⁽ Reference Murphy, Yang and Moylan ^{55 )}. Another study revealed the methylation of nine genes related to intermediate metabolism and insulin-like signalling, which are strongly associated with NAFLD⁽ Reference Ahrens, Ammerpohl and von Schönfels ^{56 )}. In addition, it has been demonstrated that those who experienced famine in utero, as a result of the Dutch ‘Hunger Winter’, presented an altered methylation of the insulin-like growth factor-2 (IGF2) gene in adult life⁽ Reference Heijmans, Tobi and Stein ^{57 )}.

Histone modifications

Epigenetic alterations related to NAFLD are not restricted to DNA methylation. In fact, aberrant histone modifications have been reported to promote the development of insulin resistance, thereby playing a key role in NAFLD⁽ Reference Ling and Groop ^{58 )}. Histone tails can be post-transcriptionally modified by acetylation, methylation, ubiquitinylation, phosphorylation and SUMOylation, thus determining if chromatin will be active or not and, as a result, affecting the expression of genes within that portion of chromatin⁽ Reference Jang and Serra ^{38 )}.

Acetylation on lysine residues has been shown to be the most frequent histone modification. While acetylation is related to activation of gene transcription and is catalysed by histone acetyltransferases (HAT), deacetylation is associated with gene inactivation and is catalysed by histone deacetylases (HDAC). In fact, it has been evidenced that the imbalance between HDAC and HAT affects histone acetylation, thereby altering gene expression in NAFLD⁽ Reference Tian, Wong and Chan ^{59 )}. Among HDAC, sirtuins (SIRT) have been reported to be involved in multiple cellular events related to energy homeostasis. In particular, SIRT1, the most studied member of the family, has been shown to regulate metabolic processes such as fatty acid oxidation, gluconeogenesis, lipolysis and mitochondrial activity⁽ Reference Feige and Auwerx ^{60 )}. Given that SIRT1 inhibits NF-κB action and, in doing so, decreases the inflammatory response, its liver deletion augments steatosis and obesity-induced inflammation. On the other hand, SIRT1 overexpression exerts a beneficial effect against steatosis and insulin resistance⁽ Reference Purushotham, Schug and Xu ^{61 )}. The deletion of SIRT3 in mice, which is localised in mitochondria, has been associated with alterations compatible with NAFLD⁽ Reference Hirschey, Shimazu and Jing ^{62 )}. In addition, a HFD during pregnancy was found to induce the reduction of fetal liver HDAC1, suggesting that this diet can affect fetal chromatin through histone modifications⁽ Reference Aagaard-Tillery, Grove and Bishop ^{63 )}. Another study reported that the HAT activator p300, which is related to NAFLD progression, modulates carbohydrate-responsive element-binding protein (ChREBP), inducing its hyperacetylation, thus promoting its transcriptional activity. Moreover, this epigenetic modification in p300 led to liver lipogenesis in mice and was associated with insulin resistance⁽ Reference Bricambert, Miranda and Benhamed ^{64 )}. In addition, it has been proven that alterations in the regulation of circadian clock genes, as a result of HDAC3 action, can induce inadequate lipid metabolism in the liver, thereby helping in the development of NAFLD⁽ Reference Feng, Liu and Sun ^{65 ,} Reference Mazzoccoli, Vinciguerra and Oben ^{66 )}.

Histone methylation is different between cell types and has been related to fetal development. This histone modification can occur in lysine as well as in arginine residues in H3 and H4 histones, and can be reversed by histone demethylases. The histone methyltransferase PRMT1 has been shown to induce PPARγ co-activator 1α (PGC1α) methylation, thus altering fatty acid oxidation and energy metabolism⁽ Reference Gueant, Namour and Gueant-Rodriguez ^{67 )}. On the other hand, the histone demethylase JHDM2A has been reported as a regulator in energy homeostasis and fat deposition⁽ Reference Portha, Fournier and Kioon ^{68 )}.

MicroRNA

Among epigenetic mechanisms, miRNA have been strongly related to the development of NAFLD. miRNA are non-coding single-stranded RNA that modulate mRNA degradation or inhibition of translation, thus affecting the expression of different genes. It has been shown that one miRNA can alter the expression of multiple genes and, on the other hand, multiple miRNA can affect a single gene. Moreover, miRNA can modulate DNA methylation and histone modifications⁽ Reference Lynn ^{69 )}.

The aberrant expression of miRNA has been related to insulin resistance, obesity and NAFLD⁽ Reference Lakner, Bonkovsky and Schrum ^{70 )}. It has been reported that almost 100 miRNA are differentially expressed in NASH patients⁽ Reference Cheung, Puri and Eicken ^{71 )}. In another study, it was shown that forty-six miRNA were distinctively expressed in NASH patients, with twenty-three of them being found up-regulated, including miRNA-21, miRNA-100 and miRNA-34a, and the rest down-regulated, such as miRNA-126 and miRNA-122⁽ Reference Cheung, Puri and Eicken ^{71 )}. A maternal HFD during gestation and lactation has also been proven to alter the expression of twenty-three miRNA in offspring liver⁽ Reference Zhang, Zhang and Didelot ^{72 )}. In another study, it was shown that the differential expression of miRNA was related to the progression from liver steatosis to HCC, suggesting that each pathological state involves diverse miRNA⁽ Reference Tessitore, Cicciarelli and Del Vecchio ^{73 )}.

In particular, miRNA122, the principal hepatic microRNA, was found to be down-regulated in NAFLD patients compared with a control group⁽ Reference Zhang, Zhang and Didelot ^{72 )}. Moreover, the deletion of this liver-specific miRNA in mice induced hepatic steatosis, inflammation and HCC⁽ Reference Hsu, Wang and Kota ^{74 )}. As a result, this miRNA has been proposed to be a major contributor to the regulation of lipid metabolism in the liver, thus having a tumour-suppressor role in this organ⁽ Reference Wen and Friedman ^{75 )}.

Apart from miRN-122, there are other miRNA that have been related to NAFLD. The miRNA 146b, 143, 34a and 23a were reported to be significantly increased in NAFLD⁽ Reference Tian, Wong and Chan ^{59 )}. Other miRNA have been shown to modulate lipid and cholesterol regulatory genes, such as miRNA 370, 33, 103 and 104, thus ultimately contributing to the development of NAFLD.

Kupffer cells

Liver-resident macrophages, known as KC, have an important role in the immune response. Given that they are the body’s primary line of defence against micro-organisms or endogenous signals⁽ Reference Smedsrod, De Bleser and Braet ^{81 )}, their location is critical for KC to carry out their different functions in liver (phagocytosis, antigen processing and presentation and secretion of different proinflammatory mediators such as cytokines, prostanoids, NO and reactive oxygen species)⁽ Reference Bilzer, Roggel and Gerbes ^{82 )}. KC feature with a vast array of pattern recognition receptors (PRR)⁽ Reference Su, Klein and Aminlari ^{83 )}, complement receptors⁽ Reference Schieferdecker, Schlaf and Jungermann ^{84 )} and Fc receptors⁽ Reference van Egmond, van Garderen and van Spriel ^{85 )}, and are able to respond to cytokine, Toll-like receptor (TLR)⁽ Reference Wu, Meng and Jiang ^{86 )}, RIG-like receptor and NOD-like receptor (NLR) signalling⁽ Reference Miura, Yang and van Rooijen ^{87 )}. However, to maintain the steady state, KC can mount opposite responses to exogenous signals, polarising to M1 (classical) or M2 (alternative) phenotypes⁽ Reference Sica and Mantovani ^{88 )}. The M1 phenotype is characterised by the increment in the production of inflammatory cytokines, reactive nitrogen species and reactive oxygen species, and is related to microbicidal and tumoricidal activities. On the other hand, M2 macrophages exhibit immunomodulatory functions and are important in immunity against parasites and tissue remodelling⁽ Reference Sica and Mantovani ^{88 )}.

KC can promote the pathogenesis of liver disease, augmenting the production of proinflammatory cytokines⁽ Reference Sica and Mantovani ^{88 ,} Reference Biswas and Mantovani ^{89 )}. It has been demonstrated that TNF-α and IL-6 are closely related to steatosis, insulin resistance and inflammatory disorders⁽ Reference Paz, Hemi and Le Roith ^{90 )}. Furthermore, IL-6 expression correlates with the degree of hepatic inflammation, and fibrosis⁽ Reference Abiru, Migita and Maeda ^{91 ,} Reference Haukeland, Damas and Konopski ^{92 )}. Contrariwise, adiponectin shifts KC polarisation to the M2 anti-inflammatory phenotype⁽ Reference Bonizzi and Karin ^{93 ,} Reference Lawrence and Gilroy ^{94 )}, preventing NASH development⁽ Reference Hagemann, Lawrence and McNeish ^{95 )}. The lack of adiponectin induces hepatic steatosis progression, fibrosis and HCC⁽ Reference Gatselis, Ntaios and Makaritsis ^{96 )}. In addition, as a mechanism of adaptation to increased energy intake, KC can regulate fatty acid oxidation, increase hepatic lipid storage and insulin resistance⁽ Reference Takeda, O’Dea and Doedens ^{97 )}. These events occur in response to the secretion of inflammatory cytokines, thus suggesting a beneficial role for M2 KC in metabolic disorders⁽ Reference Han, Jung and Morel ^{98 ,} Reference Liu, Chen and Wang ^{99 )}. Stimulating M2 macrophages induces M1 macrophage apoptosis that preserves against NAFLD progression⁽¹⁰⁰⁾. Moreover, macrophage-mediated inflammation is associated with TLR activation. TLR signalling is related to hepatic damage, NAFLD, fibrosis and HCC⁽ Reference Ekihiro and Brenner ^{101 –} Reference Iimuro and Fujimoto ^{103 )}. It has been shown that the TLR involved in NAFLD beginning and progression are TLR2, TLR4 and TLR9⁽ Reference Miura, Kodama and Inokuchi ^{104 ,} Reference Rivera, Gaskin and Allman ^{105 )}. Stimulation of TLR4 on KC activates NF-κB, mitogen-activated protein kinases (MAPK), extracellular signal-regulated kinase 1 (ERK1), p38, c-Jun N-terminal kinase (JNK) and interferon regulatory factor 3 (IRF3), and induces the production of inflammatory cytokines and type I IFN, hence promoting tissue damage, increase in leucocyte infiltration, and secretion of profibrogenic cytokines. In effect, it has been demonstrated that inactivation of TLR4 induces the attenuation of steatosis and NASH in several experimental models⁽ Reference Csak, Velayudham and Hritz ^{106 ,} Reference Spruss, Kanuri and Wagnerberger ^{107 )}. In addition, it was shown that TLR9^–/– mice are protected from NASH and that this protection is consistent with a decrease in IL-1b production by KC⁽ Reference Miura, Kodama and Inokuchi ^{104 )}. Pathogen-associated molecular pattern molecules (PAMP) and damage-associated molecular pattern molecules (DAMP) can as well be recognised by NLR⁽ Reference Martinon, Mayor and Tschopp ^{108 ,} Reference Pedra, Cassel and Sutterwala ^{109 )}, whose activation induces the formation of the inflammasome⁽ Reference Dixon, Flask and Papouchado ^{110 )}. Commonly, activation of TLR and NLR leads to the production of proinflammatory cytokines and the recruitment of immune cells in the liver, as well as in other tissues, such as adipose tissue, muscle, hypothalamus, pancreatic islets and blood vessels. The resultant chronic inflammation promotes insulin resistance and energy imbalance and contributes to fatty liver disease, type 2 diabetes mellitus (T2DM) and atherosclerosis, all of which are metabolic complications of obesity⁽ Reference Csak, Ganz and Pespisa ^{111 )}. Of importance, inflammasome activation has been related to NAFLD development. Among various inflammasome complexes, the NLRP3 inflammasome is the best characterised and has been linked with this and other diseases⁽ Reference Eun-Kyeong, Jin Kyung and Dong-Min ^{112 )}. Altogether, these data support the notion that KC play a critical role in maintaining immune homeostasis of the liver. Thus, the balance of the products involved in inflammatory signalling pathways is crucial to determine NASH attenuation or progression.

Dendritic cells

Liver DC are a heterogeneous population of hepatic antigen-presenting cells whose main function is to induce T cell-mediated immunity. Hepatic DC populations are described as phenotypically immature and tolerogenic⁽ Reference Doherty ^{113 )}, although in an environment of chronic inflammation, DC are transformed to potent inducers of immune responses, becoming efficient antigen-presenting cells and a source of proinflammatory cytokines⁽ Reference Rahman and Aloman ^{114 –} Reference Thomson and Knolle ^{117 )}.

DC are localised in the portal areas and are divided depending on the expression of specific markers into: plasmacytoid (PDCA-1⁺; pDC), and myeloid or classical (PDCA-1^–; mDC). The latter are also subgrouped into DC103⁺/CD11b^– type 1 (mDC1) and DC103^–/CD11b⁺ type 2 (mDC2) cells⁽ Reference Rahman and Aloman ^{114 ,} Reference Eckert, Klein and Kormek ^{115 )}. Following hepatic injury DC express the TLR4–myeloid differentiation factor 2 (MD-2) complex, produce inflammatory cytokines such as IL-12 and TNF-α, and express co-stimulatory molecules (CD40, CD80 and CD86)⁽ Reference Seki and Brenner ^{76 )}. However, the role of DC in the pathogenesis of liver diseases is still unclear. DC ablation has been demonstrated to either prevent or worsen liver damage and inflammation depending on the experimental setting⁽ Reference Rahman and Aloman ^{114 –} Reference Heymann and Take ^{116 )}. Previous reports have indicated that mDC can expand in the early phases of steatohepatitis, developing an immune-stimulating phenotype able to promote T lymphocyte responses⁽ Reference Ibrahim, Nguyen and Rehman ^{118 ,} Reference Henning, Graffeo and Rehman ^{119 )} related to NASH progression⁽ Reference Sutti, Jindal and Locatelli ^{120 ,} Reference Wolf, Adili and Piotrowitz ^{121 )}. Nevertheless, diphtheria toxin-induced DC depletion aggravates liver inflammation and hepatocyte death in experimental NASH, suggesting that DC may also act as negative regulators of inflammation⁽ Reference Henning, Graffeo and Rehman ^{119 )}.

In this sense, Heier et al. ⁽ Reference Heier, Meier and Julich-Haertel ^{122 )} investigated the role of type 1 myeloid DC (CD103⁺/DC mDC1) in the pathophysiology of steatosis and NASH. They used basic leucine zipper transcription factor, ATF-like-3 (Batf3)-deficient animals, which lack CD103⁺ DC and two dietary models to induce steatosis and NASH based on mice fed with a high-sucrose diet or a choline/methionine-deficient diet, respectively. It was shown that Batf3^–/– mice that were administered both diets exhibited aggravated steatosis and lobular inflammation than similarly treated wild-type animals⁽ Reference Heier, Meier and Julich-Haertel ^{122 )}. Accordingly, CD103⁺/mDC1 deficiency increased liver infiltrating monocyte/macrophages as well as their release of inflammatory mediators. However, hepatic transaminases and fibrosis markers were not significantly affected⁽ Reference Heier, Meier and Julich-Haertel ^{122 )}. Moreover, the administration of bone marrow-derived CD103⁺/DC mDC1 to Batf3^–/–-deficient mice was conducive to the diminishment of monocyte recruitment, hepatic transaminases and liver C-C motif chemokine ligand (CCL) 2 expression, without altering the level of steatosis⁽ Reference Heier, Meier and Julich-Haertel ^{122 )}. High-sucrose diet-fed Batf3^–/– mice provided with bone marrow-derived CD103⁺/DC1 further significantly reduced proinflammatory monocyte influx along with serum TAG, cholesterol and NEFA. The authors thus proposed murine CD103⁺/mDC1 as a protective DC subtype that affects the pro-anti-inflammatory balance and preserves the liver from metabolic damage⁽ Reference Heier, Meier and Julich-Haertel ^{122 )}. It can be concluded that they are crucial participants in the inflammatory process involved in steatohepatitis in mice⁽ Reference Bernsmeier and Albano ^{123 )}. Moreover, human CD141⁺/mDC1, that are considered the equivalent of CD103⁺/mDC1 in mice⁽ Reference Dutertre, Wang and Ginhoux ^{124 )}, are abundant in healthy human livers⁽ Reference Kelly, Fahey and Fletcher ^{125 )}, but interestingly, while pDC and mDC2 expand in diseased livers, CD141⁺/DC are depleted in chronic liver diseases⁽ Reference Kelly, Fahey and Fletcher ^{125 )}, suggesting differential involvement of mDC subsets in liver immunity⁽ Reference Bernsmeier and Albano ^{123 ,} Reference Kelly, Fahey and Fletcher ^{125 )}.

In conclusion, it is evident that our understanding of the complex signalling network involved in the regulation of DC functions is incomplete, but overall these results suggest that DC may represent a key factor in the progression of NAFLD.

Myeloid-derived suppressor cells

MDSC are a heterogeneous population of immune cells defined by their ability to suppress T-cell activation through the production of IL-10, transforming growth factor-β (TGF-β) and arginase⁽ Reference Gabrilovich and Nagaraj ^{126 )}. Various cells in the liver, particularly HSC⁽ Reference Höchst, Schildberg and Sauerborn ^{127 –} Reference Hsieh, Chou and Yang ^{129 )}, as well as liver-derived soluble factors, including hepatocyte growth factor and acute-phase proteins⁽ Reference Yen, Yen and Hsu ^{130 )}, can induce the differentiation of MDSC from myeloid cells. Although MDSC are also present in the healthy liver⁽ Reference Chen, Akbar and Abe ^{131 )} they are expanded during chronic liver disease⁽ Reference Pallett, Gill and Quaglia ^{132 )} and have been linked to HCC progression⁽ Reference Schneider, Teufel and Yevsa ^{133 )}.

Results from patients with liver disease and experimental animal models have been obtained, supporting the notion that MDSC are involved in the pathogenesis of liver inflammation and HCC⁽ Reference Hammerich and Tacke ^{134 )}. In addition, it has been reported that the depletion of hepatic MDSC increases the expression of fibrosis markers⁽ Reference Roseboom, de Rooij and Painter ^{12 )}, thus exerting a protective role. Moreover, it has been shown that MDSC can counteract proinflammatory cells in NAFLD⁽ Reference Yao, Abe and Kawasaki ^{135 )}.

In human and mouse models of liver cancer, MDSC are induced in the tumour environment and suppress anti-tumoral immune responses⁽ Reference Hammerich and Tacke ^{134 )}. In this sense, the CCL2–C-C motif chemokine receptor 2 (CCR) axis plays a pivotal role in the migration of MDSC in cancer, since the impairment of CCL2–CCR2 signalling inhibits tumour growth⁽ Reference Huang, Lei and Zhao ^{136 –} Reference Hale, Itani and Buchta ^{138 )}. Yao et al. ⁽ Reference Yao, Abe and Kawasaki ^{135 )} have found that the expression of CCL2 is up-regulated in the livers of NAFLD mice, and that CCL2 stimulates the migration of MDSC in vitro. Thus, the CCL2–CCR2 pathway might contribute to MDSC accumulation in the steatotic liver, therefore promoting HCC pathogenesis.

Innate and adaptive immune lymphoid cells

Hepatic innate lymphocyte populations (NK cells, NKT cells, mucosal associated invariant T cells and γδ T cells) are potent cytokine producers that influence both innate and adaptive immune responses in the liver. Some studies indicate that NK cells can be activated in response to elevated levels of several cytokines, such as IL-12, interferon-c and IL-1 during NASH⁽ Reference Ganz and Szabo ^{139 ,} Reference Tian, Chen and Gao ^{140 )}. It was reported that administration of a high-fructose diet to mice deficient in NK cells prevents the development of steatosis, showing that NK cells may be implicated in the evolution to NASH⁽ Reference Bhattacharjee Kumar, Arindkar and Das ^{141 )}. In addition, NK cells play a critical role in inhibiting the development of fibrosis through direct killing of early and senescent activated HSC⁽ Reference Krizhanovsky, Yon and Dickins ^{142 ,} Reference Radaeva, Wang and Radaev ^{143 )}. Furthermore, the influx of NKT cells into the liver during NASH is mediated by enhanced expression of the chemokine (C-X-C motif) ligand (CXCL) 16 on SEC and macrophages, which binds C-X-C chemokine receptor (CXCR) type 6 on the surface of NKT cells⁽ Reference Wehr, Baeck and Heymann ^{144 )}. In this respect, up-regulation of CXCL16 is accompanied by the production of IL-4 and IFN-γ by NKT cells, which worsens inflammation via macrophage activation⁽ Reference Wehr, Baeck and Heymann ^{144 )}. Activation of the Hedgehog pathway has also been implicated in NKT cell recruitment in NASH livers of methionine/choline-deficient diet-fed mice. Moreover, activation of the Hedgehog pathway drives HSC activation and fibrogenesis⁽ Reference Syn, Agboola and Swiderska ^{145 )}. Then, NKT cells seem to be involved in the development of NASH.

On the other hand, populations of adaptive lymphocytes (CD4⁺ and CD8⁺ T cells and B cells) are found in healthy livers⁽ Reference Norris, Collins and Doherty ^{146 ,} Reference Pruvot, Navarro and Janin ^{147 )}. Depending on the cytokine environment, T CD4⁺ cells can assume a proinflammatory phenotype (Th1) or an anti-inflammatory phenotype (Th2)⁽ Reference Romagnani ^{148 )}. The balance between Th1 and Th2 T cells is important to maintain immune system homeostasis. There is a third type of CD4⁺ cells called Th17, which play a pro-inflammatory role⁽ Reference Tang, Bian and Zhao ^{149 )}. The implication of the adaptive immunity in stimulating adipose tissue inflammation has been well studied in obesity⁽ Reference Sell, Habich and Eckel ^{150 )}. A similar mechanism is implicated in the beginning of inflammation in NASH, where different studies have demonstrated that both macrophages and lymphocytes are the most common inflammatory infiltrates in these livers⁽ Reference Brunt ^{151 )}. For instance, individuals with NASH have augmented frequencies of IFN-γ-producing memory CD4⁺ and CD8⁺ T cells⁽ Reference Inzaugarat, Ferreyra Solari and Billordo ^{152 )}. Dysfunctional chemotaxis is one of the molecular mechanisms driving T cell infiltration into the liver. In this sense, it has been shown that peripheral CD4⁺ T cells from both obese mice and NASH patients migrate more promptly toward the chemokine CXCL12 when compared with T cells from healthy animals or donors⁽ Reference Boujedidi, Robert and Bignon ^{153 )}. Likewise, Th17 cell infiltration is found in NASH livers and it is well known that IL-17 secretion exacerbates hepatic steatosis and inflammation⁽ Reference Tang, Bian and Zhao ^{149 )}. Moreover, IL-17 signalling in HSC up-regulates the expression of profibrotic genes while absence of IL-17 in a murine model of liver fibrosis diminishes the levels of proinflamamtory cytokines and cell death⁽ Reference Tan, Qian and Jiang ^{154 ,} Reference Meng, Wang and Aoyama ^{155 )}. In addition, different studies have reported that the progression to NASH is marked by an increment of the Th17:regulatory T cells (Tregs) ratio in peripheral blood and liver⁽ Reference Rau, Schilling and Meertens ^{156 )}.

B cells play an important role in NASH pathogenesis. It has been shown that 40–60 % of patients with NASH have circulating antibodies against lipid peroxidation-derived antigens⁽ Reference Albano, Mottaran and Vidali ^{157 )}. In addition, the elevated titres of these antibodies are in parallel with an increased risk of developing liver fibrosis⁽ Reference Nobili, Parola and Alisi ^{158 )}. Furthermore, Winer et al. ⁽ Reference Winer, Winer and Shen ^{159 )} have demonstrated that B cells rapidly increase in serum and adipose tissue of mice fed a HFD.

B cell activating factor (BAFF) is a well-known cell survival and maturation factor for B cells. An increase in serum level of BAFF in human NASH has been demonstrated, and the serum BAFF level correlates with B cell content in the liver⁽ Reference Kawasaki, Abe and Tada ^{160 )}. Moreover, BAFF receptor-deficient mice showed improved obesity and insulin resistance induced by HFD but, in addition, displayed enhanced hepatic steatosis that indicates a protective role of BAFF in hepatic steatosis⁽ Reference Kawasaki, Abe and Tada ^{160 ,} Reference Kim and Do ^{161 )}. Nevertheless, to better understand the B cell’s contribution in sustaining hepatic inflammation and promoting fibrosis in NASH, further studies will be required.

Hepatic non-haematopoietic cells

In addition to hepatic immune cell populations, non-haematopoietic cells in the liver play key roles in local and systemic innate immunity and inflammation. HSC reside in the space between hepatocytes and the liver SEC, known as the space of Disse⁽ Reference Weiskirchen and Tacke ^{162 )}. In physiological conditions, HSC store vitamin A and lipids⁽ Reference Weiskirchen and Tacke ^{162 ,} Reference Thapa, Chinnadurai and Velazquez ^{163 )}. However, in response to a liver injury, such as in NAFLD, HSC become activated, transdifferentiate to myofibroblasts and begin to synthesise large amounts of extracellular matrix, leading to fibrosis⁽ Reference Ramadori and Saile ^{164 –} Reference Dranoff, Kruglov and Robson ^{166 )}. HSC activation is a result of the release of proliferative and fibrogenic cytokines, in particular platelet-derived growth factor (PDGF) and TGF-β from different cells, such as hepatocytes, SEC, macrophages, NK cells and lymphocytes⁽ Reference Elpek ^{167 )}. Guy et al. ⁽ Reference Guy, Suzuki and Zdanowicz ^{168 )} have found that ballooned hepatocytes generate Sonic Hedgehog, a ligand of the Hedgehog-signalling pathway that induces HSC activation and contributes to NASH progression in mice. TLR4 is another factor implicated in the activation of HSC⁽ Reference Seki and Brenner ^{76 ,} Reference Seki, De Minicis and Osterreicher ^{169 )}. Hepatocytes express TLR4 and TLR2, and it has been shown that TLR2 is up-regulated during inflammatory conditions⁽ Reference Liu, Gallo and Green ^{170 ,} Reference Matsumura, Degawa and Takii ^{171 )}. In addition, a role for biliary epithelial cells in portal and septal fibrosis has been demonstrated⁽ Reference Chiba, Sasaki and Kitamura ^{172 )}. Murine biliary cells express CD14, myeloid differentiation factor 2 (MD-2), TLR2, TLR3, TLR4 and TLR5⁽ Reference Harada, Ohira and Isse ^{173 )}, and are involved in antigen presentation to T cells⁽ Reference Lleo and Invernizzi ^{174 )}, thus playing an additional role in hepatic immune function. In addition, SEC constitutively express TLR4 and CD14⁽ Reference Uhrig, Banafsche and Kremer ^{175 )}, and are capable of presenting antigens to classical T cells⁽ Reference Thomson and Knolle ^{117 ,} Reference Crispe ^{176 )}.

DNA methylation

DNA methylation is one of the main epigenetic modifications that contribute to altered gene expression in NAFLD. As mentioned above, PPARγ has been found to reduce the expression of proinflammatory cytokines and contribute to insulin resistance in patients with NAFLD⁽ Reference Sun, Fan and Qiao ^{50 –} Reference Sookoian, Rosselli and Gemma ^{52 )}. It has been shown that the promoter region of PPARγ becomes hypermethylated as fibrosis augments in NASH liver biopsies from patients. Thus, it functions like a negative regulator of HSC activation and liver fibrogenesis⁽ Reference Mann, Chu and Maxwell ^{180 ,} Reference Zeybel, Hardy and Wong ^{181 )}. Increasing evidence demonstrated that DNA methylation and insulin resistance in NAFLD patients are critical players for the transition from steatosis to fibrotic NASH⁽ Reference Tian, Wong and Chan ^{59 )}.

It is well known that NAFLD is linked to obesity and the metabolic syndrome⁽ Reference Younossi, Stepanova and Afendy ^{182 )}. Adipose tissue enlargement related to the metabolic syndrome is associated with immune system activation and chronic inflammation, which is also involved in the development of insulin resistance⁽ Reference Sell, Habich and Eckel ^{150 )}. In addition, altered methylation levels on the UBASH3B (ubiquitin-associated and SH3 domain-containing protein B) and TRIM3 (tripartite motif-containing 3) genes in obesity have been reported, genes which are both involved in regulating T cell and macrophage proliferation and function⁽ Reference Wang, Zhu and Snieder ^{183 )}. Another study has succeeded in linking altered DNA methylation to systemic inflammation in obesity. It was shown that TNF-α promoter was hypomethylated in leucocytes from whole blood in obese individuals⁽ Reference Hermsdorff, Mansego and Campión ^{184 )}. Moreover, another study identified global DNA hypermethylation in B cells from obese and T2DM individuals⁽ Reference Simar, Versteyhe and Donkin ^{185 )}, suggesting altered B cell function in metabolic diseases. In addition, the up-regulation of the DNA DNMT3B in macrophages has been linked to elevated levels of SFA, leading to M1 polarisation and adipose tissue inflammation⁽ Reference Yang, Wang and Liu ^{186 )}. Furthermore, oxidative stress has certainly been linked to inflammation, and mitochondria are the principal sources and targets of reactive oxygen species. As mentioned above, elevated levels of MT-ND6 methylation in mitochondria have been found in NASH patients⁽ Reference Pirola, Gianotti and Burgueño ^{54 )}. When NAFLD progresses to HCC, inflammation and oxidative stress also play a critical role in this process. The suppressor of cytokine signalling (SOCS) 1 is a negative regulator of cytokine signalling, epigenetically regulated and, indeed, a suppressor of inflammation. SOCS1 promoter hypermethylation is one of the best-known epigenetic changes in macrophages and HCC⁽ Reference Herath, Leggett and MacDonald ^{187 –} Reference Amodio, Bellizzi and Leotta ^{189 )}. It was demonstrated that loss of SOCS1 expression inside tumour cells through promoter hypermethylation is markedly associated with overproduction of inflammatory cytokines such as TNF-α and IL-6⁽ Reference Cheng, Huang and Ma ^{190 )}. Moreover, increased promoter methylation of the Janus tyrosine kinase (JAK)/signal transducer and activator of transcription (STAT) inhibitor SOCS2 also in the tumours of glycine N-methyltransferase (GNMT) knockout mice has been shown, leading to epigenetic activation of the inflammatory pathway⁽ Reference Martinez-Chantar, Vazquez-Chantada and Ariz ^{191 )}. Is evident that the remodelling of the immune system in metabolic disease and HCC is intimately linked to epigenetic modifications and methylation of CpG dinucleotides is the most well-studied epigenetic mechanism that plays crucial roles in cell fate decisions and oncogenic transformation till now.

Histone modifications

Modification of histones is another key epigenetic change that can alter the immune response⁽ Reference Schoenborn, Dorschner and Sekimata ^{192 ,} Reference Di Spirito and Shen ^{193 )}. In obese individuals, HDAC4 has been reported to be decreased in peripheral blood myeloid cells and to be also inversely correlated with the expression of the proinflammatory CCL5⁽ Reference Abu-Farha, Tiss and Abubaker ^{194 )}. If its expression is restituted, the activation of TNF-α-induced NF-κB is inhibited, indicating its regulatory role in inflammation. Increased histone H3 hyperacetylation in the TNF-α and cyclo-oxygenase 2 promoter regions has been shown in monocytes from type 1 diabetes mellitus and T2DM subjects⁽ Reference Miao, Gonzalo and Lanting ^{195 )}, while histone H3 lysine 4 (H3K4) methylation has been related to monocyte dysfunction by stabilising NF-κBp65 at the promoter sites of inflammatory genes⁽ Reference Li, Reddy and Miao ^{196 )}. The liver is a metabolically active organ that is usually affected by inflammation in obese individuals. In this regard, it has been demonstrated that when hepatocytes are cultured with NEFA the up-regulation of the Brahma-related gene (Brg1) and Brahma protein are induced, which promote inflammation by increasing histone acetylation near the promoter regions of the IL-1, IL-6 and monocyte chemoattractant protein-1 (Mcp1) genes. Thus, Brg1 is involved in NASH pathogenesis, behaving as an epigenetic coordinator of proinflammatory gene transcription⁽ Reference Tian, Xu and Fang ^{197 )}. Moreover, TNF-α and chemokine CCL2 are considered important inflammatory mediators in the development of NAFLD. By a chromatin immunoprecipitation assay, an increase in H3K9 and H3K18 acetylation at TNF-α and CCL2 in obese mice was demonstrated ⁽ Reference Mikula, Majewska and Ledwon ^{198 )}. These results indicate that elevated levels of TNF-α and CCL2 expression in fatty liver are associated with the alterations in histone H3 acetylation⁽ Reference Mikula, Majewska and Ledwon ^{198 )}. Likewise, SIRT1, a class III HDAC, which is involved in the regulation of glucose homeostasis, antihyperlipidaemic activity, insulin sensitivity, oxidative stress, anti-inflammatory activity, and antiaging activity⁽ Reference Colak, Yesil and Mutlu ^{199 )}, has been shown to be down-regulated in the adipose tissue of obese individuals. This results in augmented macrophage recruitment through increased chemoattractant and cytokine production⁽ Reference Gillum, Kotas and Erion ^{200 )}. As mentioned above, this HDAC also inhibits NF-κB activity to diminish the inflammatory response and modulates other cytokines implicated in lipid metabolism⁽ Reference Tian, Wong and Chan ^{59 ,} Reference Herranz and Serrano ^{201 )}. Therefore, liver-specific deletion of SIRT1 has been shown to enhance fatty liver disease and obesity-induced inflammation, while SIRT1 overexpression exhibited a protective effect against steatohepatitis and insulin resistance⁽ Reference Purushotham, Schug and Xu ^{61 ,} Reference Escande, Chini and Nin ^{202 )}. In addition, another study performed in non-human primates has demonstrated that a maternal HFD increases the fetal acetylation of histone H3K14 and decreased SIRT1 expression in their fetal livers⁽ Reference Suter, Chen and Burdine ^{203 )}. Moreover, a significant reduction in the expression of SIRT1 in NAFLD animal models has been reported, and natural SIRT1 activator has shown protective effects on metabolic diseases⁽ Reference Colak, Ozturk and Senates ^{204 )}. Furthermore, the mitochondrial SIRT3 has been shown to be required for maintenance of integrity in response to oxidative stress⁽ Reference Kim, Patel and Muldoon-Jacobs ^{205 )}. Mice deficient in SIRT3 displayed NASH⁽ Reference Hirschey, Shimazu and Jing ^{62 )}, therefore indicating its role in regulating inflammation. Otherwise, several studies have demonstrated that aberrant chromatin modifications link metabolic perturbation to cellular transformation in the liver. In this sense, the expression of SIRT1 was found to be down-regulated in HCC, as it happens in NAFLD⁽ Reference Wang, Sengupta and Li ^{206 )}. Indeed, overexpression of SIRT1 in a HFD model exposed to hepatic carcinogen protected mice from HCC⁽ Reference Herranz, Muñoz-Martin and Canamero ^{207 )}, indicating a tumour-suppressive role for SIRT1.

Moreover, mice with genetic ablation of the enzyme that converts methionine to SAM (methionine adenosyl transferase 1A; MAT1A) displayed lower hepatic SAM levels, higher lipogenesis and oxidative stress; thus, they were predisposed to NAFLD-associated HCC⁽ Reference Lu, Alvarez and Huang ^{208 ,} Reference Martinez-Chantar, Corrales and Martinez-Cruz ^{209 )}. On the other hand, mice defective in glycine N-methyltransferase, the SAM catabolic enzyme, displayed higher hepatic SAM levels but also developed NAFLD and HCC⁽ Reference Martinez-Chantar, Vazquez-Chantada and Ariz ^{191 ,} Reference Liao, Liu and Lee ^{210 )}. Thus, either an excess or a defect in SAM liver levels may result in aberrant methylation of histones, deriving from epigenetic modulation of critical metabolic and carcinogenic pathways⁽ Reference Lu and Mato ^{211 )}.

Non-coding RNA

Short non-coding miRNA and long non-coding RNA (lncRNA) act in modulating translation and transcription of target genes, thus regulating a variety of biological functions⁽ Reference Wang, Yao and Lin ^{212 )}. miRNA play an important role in lipid metabolism and inflammation and some of these miRNA have been shown to be epigenetically regulated in NAFLD⁽ Reference Finch, Marquardt and Yeoh ^{213 –} Reference Panera, Gnani and Crudele ^{215 )}. In this sense, the enhancer of zeste homolog 2 (EZH2) is known as a histone–lysine N-methyltransferase enzyme that intervenes in histone methylation and transcriptional repression⁽ Reference Viré, Brenner and Deplus ^{216 )}, catalysing the addition of methyl groups to histone H3 at lysine 27⁽ Reference Cao, Wang and Wang ^{217 )}. Moreover, a study performed in rats demonstrated that EZH2 inhibition induced up-regulation of TNF-α and TGF-β inflammatory genes as well as of miRNA-200b and miRNA-155, subsequently inducing hepatic steatosis in NAFLD⁽ Reference Vella, Gnani and Crudele ^{218 )}. Another study developed in patients with NAFLD showed differential expression of seven miRNA (miRNA-132, miRNA-150, miRNA-433, miRNA-28-3p, miRNA-511, miRNA-517a and miRNA-671) in the visceral adipose tissue, which target genes comprising the obestatin gene, insulin receptor pathway genes and inflammation-related genes⁽ Reference Estep, Armistead and Hossain ^{219 )}.

Even though liver biopsy is the ‘gold standard’ for diagnosis of NAFLD, it is very invasive. Thus, serum miRNA panelling is currently being considered as a non-invasive technique that may be used as a biomarker for diagnosis and monitoring of NAFLD. A study showed that the levels of circulating miRNA-34a, miRNA-122, miRNA-181a, miRNA-192 and miRNA-200b were significantly correlated with inflammatory activity, fibrosis stage and liver enzyme levels in mice⁽ Reference Cermelli, Ruggieri and Marrero ^{220 ,} Reference Tryndyak, Latendresse and Montgomery ^{221 )}.

Different miRNA are differentially expressed both at the tissue and systemic levels in obese or T2DM individuals. The miRNA-146b-5p is involved in the modulation of inflammatory pathways and has been shown to be decreased in monocytes from obese⁽ Reference Hulsmans, Van Dooren and Mathieu ^{222 )} and T2DM⁽ Reference Balasubramanyam, Aravind and Gokulakrishnan ^{223 )} individuals. In addition, TLR4-induced inhibition of miRNA-107 is impaired in obesity, leading to an increased inflammatory response in macrophages⁽ Reference Foley and O’Neill ^{224 )}. Furthermore, the trafficking of inflammatory immune cells to the adipose tissue in obesity is controlled by the expression of chemokines⁽ Reference Sell, Habich and Eckel ^{150 )}. An increased level of CCL2 in obese individuals which was associated with reduced miRNA-126 and miRNA-193b has been shown⁽ Reference Arner, Mejhert and Kulyté ^{225 )}.

Interestingly, the liver-specific miRNA-122 is underexpressed in NASH patients⁽ Reference Cheung, Puri and Eicken ^{71 )}. As mentioned above, different studies have demonstrated that deletion of miRNA-122 in mice resulted in hepatic steatosis, inflammation and HCC⁽ Reference Hsu, Wang and Kota ^{74 ,} Reference Tsai, Hsu and Hsu ^{226 )}. These pathological manifestations were associated with hyperactivation of oncogenic pathways⁽ Reference Hsu, Wang and Kota ^{74 ,} Reference Tsai, Hsu and Hsu ^{226 )} and hepatic infiltration of inflammatory cells that produced IL-6 and TNF-α⁽ Reference Hsu, Wang and Kota ^{74 )}. Furthermore, there is evidence indicating that miRNA-21, miRNA-103, miRNA-143 and miRNA-378 increase oxidative stress and inflammation in animal models with obesity and steatosis⁽ Reference Hulsmans, de Keyzer and Holvoet ^{227 )}. Using a NASH-related HCC model of rodents fed on a choline-deficient, low-methionine and amino acid-defined diet, authors found a significant increase in miRNA-21 expression at early stages of HCC⁽ Reference Wang, Majumder and Nuovo ^{228 )}. Moreover, unsaturated fatty acids in hepatocytes induce steatosis mediated by mammalian target of rapamycin (mTOR)/NF-κB signalling pathway stimulation and concomitantly activate miRNA-21, that in turn suppresses the expression of the phosphatase and tensin homolog gene (PTEN)⁽ Reference Vinciguerra, Sgroi and Veyrat-Durebex ^{229 )}. In HCC, inhibition of miRNA-21 increased PTEN expression, decreased tumour cell proliferation, migration, and invasion, suggesting that miRNA-21 and the tumour-suppressor PTEN pathway are involved in the NAFLD-related HCC development⁽ Reference Meng, Henson and Wehbe-Janek ^{230 )}. In addition, miRNA-143 was shown to not only regulate obesity-associated pathways but also promote HCC development and progression⁽ Reference Tian, Wong and Chan ^{59 )}. Furthermore, using diet-induced NASH models it was demonstrated that miRNA-155 is overexpressed in these livers and it was associated with the early stages of HCC⁽ Reference Pogribny, Starlard-Davenport and Tryndyak ^{231 ,} Reference Yan, Jia and Chen ^{232 )}. At these early stages, up-regulation of miRNA-155 was related to the down-regulation of their target genes CCAAT/enhancer binding protein β (CEBPB) and SOCS1 that have been shown to suppress inflammation and HCC, respectively⁽ Reference Yan, Jia and Chen ^{232 ,} Reference Worm, Stenvang and Petri ^{233 )}. In addition, NF-κB activates miRNA-155 expression, thus linking NF-κB signalling to NAFLD-related HCC via miRNA deregulation⁽ Reference Wang, Majumder and Nuovo ^{228 )}.

On the other hand, the contribution of lncRNA to obesity-related inflammation is not completely understood. lncRNA E33 has been reported to be up-regulated in macrophages from diabetic and diet-induced diabetic mice⁽ Reference Reddy, Chen and Park ^{234 )}. lncRNA E33 overexpression in RAW cells was associated with the up-regulation of several proinflammatory genes while lncRNA E33 silencing reversed these effects⁽ Reference Reddy, Chen and Park ^{234 )}. The expression of non-coding RNA, in particular miRNA, constitutes an important epigenetic mechanism that coordinates downstream features of the immune system, including immune cell differentiation, function and recruitment.