Human and molecular genetics shed lights on fatty liver disease and diabetes conundrum

Abstract The causal role of abdominal overweight/obesity, insulin resistance and type 2 diabetes (T2D) on the risk of fatty liver disease (FLD) has robustly been proven. A consensus of experts has recently proposed the novel definition of ‘metabolic dysfunction‐associated fatty liver disease, MAFLD’ instead of ‘nonalcoholic fatty liver disease, NAFLD’, emphasizing the central role of dysmetabolism in the disease pathogenesis. Conversely, a direct and independent contribution of FLD per se on risk of developing T2D is still a controversial topic. When dealing with FLD as a potential risk factor for T2D, it is straightforward to think of hepatic insulin resistance as the most relevant underlying mechanism. Emerging evidence supports genetic determinants of FLD (eg PNPLA3, TM6SF2, MBOAT7, GCKR, HSD17B13) as determinants of insulin resistance and T2D. However, recent studies highlighted that the key molecular mechanism of dysmetabolism is not fat accumulation per se but the degree of hepatic fibrosis (excess liver fat content—lipotoxicity), leading to reduced insulin clearance, insulin resistance and T2D. A consequence of these findings is that drugs that will ameliorate liver fat accumulation and fibrosis in principle may also exert a beneficial effect on insulin resistance and risk of T2D in individuals with FLD. Finally, initial findings show that these genetic factors might be directly implicated in modulating pancreatic beta‐cell function, although future studies are needed to fully understand this relationship.


| INTRODUC TI ON
Fatty liver disease (FLD) is defined by excessive hepatic fat accumulation mainly due to metabolic derangement and excess in alcohol intake. 1 Abdominal overweight/obesity, insulin resistance and type 2 diabetes (T2D) are among the strongest acquired risk factors for the development of FLD and its progression to advanced fibrosis, cirrhosis and hepatocellular carcinoma. [2][3][4] The causal role of abdominal overweight/obesity, insulin resistance and T2D on risk of FLD development and progression has robustly been proven. 5 The opposite, namely a direct and independent contribution of FLD per se on risk of developing T2D, is still a controversial topic.
However, it is becoming clear that the link between FLD and T2D is more complex than previously thought. Human genetic variations primarily increasing liver fat content do not have a direct effect on insulin resistance. 6 Indeed, recent evidence suggests that quality of fat, rather than quantity, is more important in causing the increase in insulin resistance. 6,7 Furthermore, the effect of gender in the development of FLD should not be dismissed. 8 A growing body of evidence suggests that gender and its related biological components represent strong determinants of FLD development and progression. 9 In agreement, also derangement in glucose metabolism has a sexual dysmorphism. [10][11][12][13][14][15][16] Among the unknown questions, there is also if genetic determinants of FLD interact specifically with sex. Increasing clinical evidence now suggests that FLD may precede and/or promote the development of T2D and other cardiometabolic diseases. 17 Thus, FLD appears to be a good biomarker for predicting risk of incident T2D, irrespective of established risk factors and may be also used to stratify the risk of cardiometabolic diseases and personalize prevention. When dealing with FLD as a new risk factor for T2D, it is straightforward to think of liver fat content contributing directly to hepatic insulin resistance and diabetes as the most likely mechanism. 18 However, as will be discussed in greater detail, emerging data are now challenging this notion.
Very recently, a consensus of experts has proposed to replace the 'nonalcoholic fatty liver disease, NAFLD' with a more appropriate term, namely 'metabolic dysfunction-associated fatty liver disease, MAFLD'. 19,20 This novel term emphasizes that derangement in hepatic lipid and glucose handling, namely metabolic dysfunction, is the key player in the pathogenesis of chronic liver disease.
In particular, they propose a set of novel affirmative criteria for diagnosing MAFLD (mainly based on the presence of overweight/ obesity, T2D or other metabolic syndrome traits), irrespective of other concomitant liver diseases. However, this term has not been unanimously accepted 21 and therefore, in this review we will use the term FLD.
In this review article, we will focus on the contribution of human genetics to the multifaceted and bidirectional relationship between FLD and T2D, highlighting the potential clinical use of FLD for a better risk stratification of T2D and its related chronic vascular complications (mainly cardiovascular and chronic kidney disease).

| FLD and increased risk of diabetes: epidemiological evidence
A body of evidence shows that FLD, as detected by imaging methods, is an early predictor for the development of incident T2D. 3,4 In Table 1, we included the observational studies, published in the last 5 years, investigating the association between FLD and risk of incident T2D. [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40] Collectively, all these studies have consistently documented that FLD was strongly associated with an increased risk of incident T2D, independently of age, sex, adiposity measures and other potential confounding factors ( Table 1). The increased risk of incident T2D ranged approximately from a 50% 30 to 3.5-fold increase 36 in individuals with FLD, becoming even higher in sexstratified analyses. 35 The significant association between FLD and increased risk of incident T2D was also confirmed among FLD individuals with prediabetes. 39 Notably, the increase in the risk of incident T2D was found to be proportional to the severity of liver steatosis assessed by ultrasonography or computed tomography. 23,33,34 For example, in a large prospective cohort study of 18,111 Chinese nondiabetic subjects, Li et al showed that the incidence rates of T2D at 4.6-year follow-up progressively increased with the ultrasonographic severity of FLD at baseline, accounting for 18.1% of incident T2D cases in the moderate-severe FLD group, 10.6% in the mild FLD group and 4.7% in the normal group, respectively (P < .001). In the multivariable Cox regression analysis, the adjusted hazard ratios (HRs) for incident T2D were, respectively, 2.34 (95% CI 1.9-3.0) and 1.88 (95% CI 1.6-2.2) in individuals belonging to the moderate-severe and mild FLD groups, when compared with those in the non-FLD group (P-trend < 0.001). 23 Similarly, in a prospective cohort study of 41,650 Chinese nondiabetic individuals followed for a mean period of 3.6 years, it has been reported that FLD on ultrasonography was independently associated with increased incidence of both T2D (adjusted HR 1.62, 95% CI 1.5-1.8) and prediabetes (adjusted HR 1.12, 95% CI 1.1-1.2).
In particular, compared with subjects without FLD, the HRs for T2D development were significantly greater in those belonging to the se-

TA B L E 1 (Continued)
middle-aged individuals (mostly of Asian ethnicity) has provided further strong evidence for a causal relationship between FLD and risk of T2D. 41 Collectively, all these epidemiological studies support the notion that FLD (defined radiologically or histologically) is strongly as-

| FLD and risk of T2D chronic complications: epidemiological evidence
The global prevalence of FLD diagnosed by ultrasonography and magnetic resonance spectroscopy among individuals with T2D is currently estimated to be approximately 55%, with the highest rates reported from Europe (68%) and West Asia (67%), followed by South Asia (58%), Latin America (57%), East Asia (52%), the United States (52%) and Africa (30%). 42 These rates for the global FLD prevalence are nearly twice those observed in the general population from the same regions. 42,43 Similarly, the global prevalence of histologically proven nonalcoholic steatohepatitis (NASH) and advanced fibrosis among individuals with FLD and T2D is very high, accounting for 37% and 17%, respectively. 42 Additionally, T2D has been adversely related to the onset of FLD long-term complications, such as cirrhosis, hepatocellular carcinoma, liver-related mortality and all-cause mortality. [44][45][46][47][48] In this context, T2D seems to be not only a major driven of FLD global burden but also an important risk factor for liver disease progression. with a longer follow-up period (6.5 years). 52 Accumulating evidence also suggests that FLD is associated with valvular heart disease (mainly aortic-valve sclerosis) and increased risk of cardiac arrhythmias (mainly permanent atrial fibrillation), especially in individuals with T2D. 53,54 This supports the notion that the diagnosis of FLD identifies a subset of subjects at higher risk of cardiovascular disease over time. 55 In the last decade, a growing body of epidemiological evidence also suggests that FLD is significantly associated with an increased prevalence and incidence of microvascular complications of diabetes, especially with chronic kidney disease. 56

| Common genetic variants associated with risk of FLD
In the last decade, several common genetic variants have been reported to confer increased genetic susceptibility to or protection against FLD. 59 Notably, these common genetic variants had a several fold larger effect if compared to common variants of susceptibility in other complex disease traits, including T2D or obesity. A detailed discussion of the association between rare genetic variants of FLD and risk of insulin resistance and diabetes is beyond the scope of this review article. Briefly, rare mutations in apolipoprotein B (APOB) predispose to familial hypobetalipoproteinaemia and progressive liver disease due to impaired triglycerides assembly into very lowdensity lipoproteins and failure to secrete triglycerides from the liver. 60 Consistently with common genetic variations, despite higher liver fat content, the risk of insulin resistance and diabetes seems not to be greatly increased in carriers of APOB variants. [61][62][63][64][65] Moreover, although the coexistence of obesity, visceral adiposity and insulin resistance promotes the development of hepatic fat accumulation in these subjects, familial hypobetalipoproteinaemia represents a condition that per se leads to higher degree of FLD. 66,67 In this section, we will discuss the evidence of an association between common genetic variants of FLD and T2D or insulin resistance.

| Patatin-like phospholipase domaincontaining 3
To date, the patatin-like phospholipase domain-containing 3 (PNPLA3) rs738409 encoding for an isoleucine to methionine substitution at position 148 (I148M) of the protein is the most robust genetic determinant of FLD. 68,69 This genetic variant is associated with insulin resistance or T2D mainly in individuals with obesity but not in those with normal weight. 68,70-76 A possible reason for this association is that obesity uncovers the effect of the PNPLA3 variant, increasing its effect size. 73,77 Additionally, quality of intrahepatic lipids, rather than quantity, may exert a major impact on the development of insulin resistance and glucose intolerance. 6,7,[78][79][80][81][82] In particular, in metabolically related FLD, but not in PNPLA3-related FLD, the liver was found to be predominantly enriched with saturated triglycerides and with markers of de novo ceramides synthesis. 6

| Transmembrane 6 superfamily member 2
A body of evidence shows that the rs58542926 in transmembrane 6 superfamily member 2 (TM6SF2) (E167K) is a robust genetic determinant of FLD, 87-89 inducing a reduction in APOB100 containing lipoprotein lipidation and secretion. 90,91 Furthermore, studies have also investigated the relationship between FLD, insulin sensitivity and T2D among individuals carrying the TM6SF2 E167K. As for the PNPLA3 I148M, lines of evidence have described the TM6SF2 E167K as a potential risk variant for T2D development, 92,93 mainly linked to increased hepatic and adipose insulin resistance and impaired pancreatic beta-cell function. 94 On the other hand, TM6SF2 E167K has been reported to be associated with preserved insulin sensitivity, estimated by HOMA-IR and adipose insulin resistance or measured by hyperinsulinaemic euglycaemic clamp. 84,95

| Membrane bound O-acyltransferase domaincontaining protein 7
The Similarly, no association was found between the MBOAT7 rs641738 and HOMA-estimated insulin resistance among Asian adult individuals. 84 However, in a multiethnic cohort of 860 obese youths, Umano et al showed that MBOAT7 rs626283 (ie a genetic variant in strong linkage disequilibrium with the MBOAT7 rs641738) was associated with both hyperisulinaemia and impaired insulin sensitivity in European individuals, but not in Hispanics and African Americans. 102

| Glucokinase regulator
The rs1260326 in glucokinase regulator (GCKR) (P446L) reduces GCKR ability to inhibit glucokinase, resulting in constitutive activation of glucose uptake and increased hepatic de novo lipogenesis. 103 This results in the occurrence of FLD with lower insulin resistance and decreased risk of T2D as shown in several ethnic groups, mostly European and Asian populations. [104][105][106][107][108][109][110][111][112][113][114][115][116] Notably, as for other genetic variants, a GCKR-related protection against development of T2D was not observed in African American individuals, 113,114,116 supporting that the impact of GCKR variant on T2D risk and its related clinical traits might differ depending on ethnicity. Moreover, the association of the GCKR variant with fasting glucose, insulin levels and insulin sensitivity seems to be less pronounced in children or adolescents compared to adults, suggesting that the GCKR-induced hypoglycaemic effect might become more evident with increasing age. 117,118 Unexpectedly, the rs1260326 or rs780094 (an intronic variant in high linkage disequilibrium) in GCKR gene variants have been associated with increased 2-hour postload plasma glucose levels. 106,114,119 Finally, inconsistent results have been reported regarding the association between GCKR polymorphisms and pancreatic beta-cell function, as estimated by the HOMA-B index. 106,110,114

| Causal relationships between FLD, insulin resistance and diabetes: Mendelian randomization studies
In the last few years, an increasing number of studies have applied a Mendelian randomization approach to establish a possible causal relationship between FLD and its related metabolic traits, that is insulin resistance and T2D. 89,124 Interestingly, it has been shown that the presence of genetically determined fatty liver (by using a genetic risk score including PNPLA3, TM6SF2, GCKR and MBOAT7 variants) was causally associated with greater insulin resistance, as estimated by HOMA-IR, in individuals at risk of progressive liver disease (ie those with suspected NASH or severe obesity), but not in the general population. 89 However, it should be noted that as reported by Stender et al these genetic variants strongly interact with obesity 125 and, therefore, it is not surprising that the deleterious metabolic effect of these genetic variants was observed principally among those at higher risk for FLD.
Moreover, this study also suggested that FLD per se does not directly cause insulin resistance, but the risk is mainly mediated by the degree of liver fibrosis, in other words by the duration and severity of liver disease (Figure 1). 89 Within this context, hyperinsulinaemia might be secondary to intrahepatic accumulation of specific lipotoxic species in addition to fibrosis-induced defect in hepatic insulin clearance. 59 Similarly, given the well-recognized association between cirrhosis and increased risk of incident T2D, 126  I148M was associated with chronic hyperglycaemia and increased visceral adiposity, but not with insulin resistance. Interestingly, the authors proposed that PNPLA3-induced reduction in glucose tolerance was largely mediated by pancreatic chronic inflammation, leading to impaired pancreatic insulin and glucagon secretion. 124 Taken all this together, it would appear that genetically determined liver steatosis does not carry the same diabetogenic risk associated with the metabolically determined liver steatosis. Moreover, quality of intrahepatic lipids, rather than quantity, decides whether the accumulation of fat in the liver will result in changes in glucose metabolism rather than only a deleterious effect for the hepatocyte.
Based on this evidence, it is likely that the use of drugs that will ameliorate liver steatosis and fibrosis in principle should also exert a beneficial effect on insulin resistance and risk of T2D as-

| Effect of FLD genetics on T2D chronic complications
To date, emerging evidence supports the existence of a significant relationship between some genetic determinants of FLD and susceptibility to diabetic nephropathy, although the topic needs to be further explored. 134 Notably, the PNPLA3 I148M has been associated with lower estimated glomerular filtration rate and increased risk of chronic kidney disease among European individuals with T2D. 135,136 Interestingly, the significant association between the PNPLA3 I148M variant and increased risk of kidney dysfunction was independent of established renal risk factors and severity of FLD, suggesting that the PNPLA3 I148M might be directly involved in the pathophysiology of diabetic nephropathy. In line with this hypothesis, PNPLA3 expression was found to be high in the renal cortex, mainly in podocytes. 136 Conversely, the steatogenic allele in GCKR locus seems to protect against the development of chronic kidney disease among T2D individuals, 137,138 consistently with the GCKR-related hypoglycaemic effect observed in nondiabetic individuals.
Some evidence also suggests that PNPLA3 and TM6SF2 gene variants may protect against cardiovascular risk, whereas variants in GCKR are associated with increased risk of cardiovascular disease, perhaps mediated by a decrease in the atherogenic dyslipidemia in both PNPLA3 and TM6SF2 carriers and an increase in the atherogenic dyslipidemia in GCKR carriers. 139 However, further research is needed to clarify whether 'genetic-related FLD' and 'metabolic-related FLD' exert differential effects on risk of major adverse cardiovascular events. 49,140 F I G U R E 1 Causal relationship between genetically determined fatty liver disease, insulin resistance and diabetes. A Mendelian randomization study published by Dongiovanni et al 89 showed that: 1) genetically determined fatty liver disease (FLD) is causally associated with insulin resistance in individuals at risk of progressive liver disease (eg those with suspected NASH or severe obesity); 2) impairment of insulin sensitivity is mediated by increased hepatic fibrosis (excess liver fat content-lipotoxicity). Similarly, a Mendelian randomization study by Liu et al 124 confirmed that genetically determined FLD causes the development of type 2 diabetes (T2D), although the underlying molecular mechanism(s) has yet to be entirely elucidated. In accord with the well-recognized link between cirrhosis and increased T2D onset, 126 the association between genetically determined FLD and enhanced risk of incident T2D might again be largely mediated by increased hepatic fibrosis

| CON CLUS I ON S AND FUTURE PER S PEC TIVE S
New insights by molecular human genetics robustly support that FLD is causally associated with dysmetabolism and T2D. 89,124 Recent studies highlighted that the key molecular mechanism of dysmetabolism is not fat accumulation per se but the degree of hepatic fibrosis (excess liver fat content-lipotoxicity), leading to reduced insulin clearance, insulin resistance and T2D. 59 Notably, initial findings show that these genetic factors might be directly implicated in modulating pancreatic beta-cell function, 124 although future studies are needed to fully understand this relationship.
In this context, it is worth noting that a consensus of experts has recently proposed novel criteria for diagnosing MAFLD (mainly based on the presence of overweight/obesity, T2D or other metabolic syndrome traits), irrespective of other concomitant liver diseases. 19,20 We believe that this novel definition is the first attempt to define the complexity of FLD and its heterogeneous clinical phenotypes, paving the way for a more fit design of clinical trials that will lead to precision medicine. Finally, it is also reasonable to speculate that the quantitative assessment of liver fat content by novel unconventional methods and the discovery of specific biomarkers of hepatic lipotoxicity will provide a better opportunity to improve the overall risk prediction of incident T2D in all individuals with FLD.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
All authors conceived and designed the review, were involved in drafting and revising the manuscript and approved the final version prior to submission.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analysed in this study.