Although many genetic variants have been associated with NAFLD, almost all the studies reported to date have used a cross-sectional design, and clinical translation has been poor. In this study, we investigated the longitudinal associations between NAFLD and SNPs in the Korean population. Although we did not see any genome-wide significant associations with the development or regression of NAFLD that met the Bonferroni correction criteria, the present study has a role as an exploratory study. Previously, we performed a GWAS with a cross-sectional design and found that the PNPLA3 and SAMM50 genes are significantly associated with NAFLD, even after adjusting for age, sex and BMI, in the Korean population [12]. Considering these results, the presence of NAFLD is significantly associated with SNPs, while longitudinal changes in NAFLD, such as development or regression, may not be associated with genetic variants.
PNPLA3 is a well-known genetic variant that is associated with NAFLD, and the severity of steatohepatitis and fibrosis has been validated in various ethnic groups via GWAS [20, 21]. A recent GWAS showed that four genetic variants (s759359281 in SLC30A10, rs13107325 in SLC39A8, rs58542926 in TM6SF2, and rs738409 in PNPLA3) had variable effects on liver fat measured using MRI and other metabolic traits [22]. The pathogenic mechanism through which the PNPLA3 variant contributes to NAFLD development and progression has been extensively investigated. In brief, PNPLA3 is involved in lipid metabolism and modulates the accumulation of hepatic triglycerides [23, 24]. In an exome-wide study, impaired function of transmembrane 6 superfamily member 2 (TM6SF2) promoted NAFLD by reducing very low-density lipoprotein secretion [25]. In a histologically confirmed case-control study, rs58542926, located in the TM6SF2 locus, was a low-frequency variant with a modest effect on NAFLD, even after conditioning on PNPLA3-rs 738409 and metabolic risk factors [26]. In addition, the membrane bound O-acyltransferase domain-containing 7 (MBOAT7) gene was associated with the risk of NAFLD in the European Caucasian population [27]. A GWAS identified variants in the MBOAT7 and TM6SF2 genes as new risk loci for alcohol-related cirrhosis [28]. Recently, it has been found that the association of rs6834314 with ALT reflects its association with NAFLD and that 17-beta hydroxysteroid dehydrogenase 13 (HSD17B13) plays a role in NAFLD through its enzymatic activity [29]. In addition, combined effects of multiple genetic variants on NAFLD severity were suggested in a multicenter biopsy-based study [30]. However, all of these studies were based on a cross-sectional design; thus, knowledge of the longitudinal effects of genetic variants on the development or regression of NAFLD is limited.
Several studies have been conducted to investigate whether genetic variants influence changes in clinical characteristics. A previous study regarding the effect of the rs738409 PNPLA3 allele on the ability of weight loss to decrease liver fat showed that hepatic fat decreased by 45% in the PNPLA3-148MM group and by 18% in the PNPLA3-148II group, suggesting a role of rs738409 in lifestyle modification [31]. However, this study did not evaluate fatty liver as a variable, unlike our research. In a single-blind randomized controlled trial, the researchers correlated the PNPLA3 rs738409 gene polymorphism with changes in metabolic profile and intrahepatic triglycerides. Although the presence of the G allele in the PNPLA3 rs738409 gene polymorphism was associated with a greater reduction in metabolic parameters, including intrahepatic triglycerides, body weight, the waist-to-hip ratio, blood total cholesterol, and low-density lipoprotein levels, there was no significant difference in the remission of NAFLD among patients with different PNPLA3 rs738409 genotypes [32]. Consistent with the previous study, we did not reach the threshold for significance in this study; only 1 SNP (rs4906353) in the gene that encodes kinesin light chain 1 (KCL1) on chromosome 14 showed an association with the development of NAFLD, with marginal significance; however, this SNP did not pass the Bonferroni correction for genome-wide significance.
One explanation for this result may be related to the pathogenesis of NAFLD development. The development or regression of NAFLD is determined by a combination of environmental factors, such as gut microbiota and dietary components, and multiple genetic factors [33]. Indeed, hepatic steatosis-related genetic variants are associated not only with nonalcoholic steatohepatitis or fibrosis but also abnormal metabolic traits, including serum lipids, glucose and anthropometric measures [34]. In addition, the PNPLA3 genotype has been elucidated as a modifier of NAFLD-associated metabolic systemic diseases such as carotid atherosclerosis [35] and chronic kidney disease [36]. However, the effect of genetics via SNPs may be slow and could be compensated by various factors, and changes in NAFLD may take longer than we expected.
Another possible explanation is the decreased power caused by the smaller sample size. While the threshold of P-value < 0.05 is considered to be statistically significant in conventional observational studies, GWAS results have much smaller p-values. For GWAS, the genome-wide significance threshold has usually been suggested to be a P-value < 5 × 10− 8 [37]. Since a larger sample size in an association study has a higher chance of having a statistical significance, a genetic association that is found to be significant in its initial GWAS is generally replicated in multiple studies. Thus, complex interactions, including the characteristics of the study population, the types of genetic data or arrays used for the analysis, minor allele frequencies of SNPs, and different patterns of linkage disequilibrium, should be taken into account when interpreting the GWAS data [38]. To confirm the longitudinal association of NAFLD with genetic variants, future studies with larger sample sizes are needed.
Although this study did not demonstrate a significant association between SNPs and the development or regression of NAFLD, it is the first study to investigate the longitudinal association between SNPs and the risk of NAFLD in an apparently healthy population that presented for health check-ups. This study has several limitations. First, the population in this study was based on subjects who voluntarily underwent health check-ups at a single center in Korea. It may not be representative of the general population, and there could be regional and economic selection bias compared with the entire Korean population. Second, the validation set was recruited from the same single-center population as the discovery set. Since some subjects from the same population were sampled in both sets, the selection of subjects may affect the results of this study. In the future, a validation study should be performed using a different population set recruited from a different location. Third, although genetic determinants affect changes in NAFLD, the development or regression of NAFLD is largely associated with the worsening or improvement of metabolic traits, which is related to residual confounding factors, such as lifestyle modifications, changes in medical treatment or health status, and could not be evaluated in this study. Since the effect of lifestyle and diet may surpass the effect of SNPs, further study is needed to take into account changes in lifestyle and diet to confirm the actual association of SNPs with the development of NAFLD. Fourth, although the effect of genetics via SNPs may be slow and could be compensated by various factors, and the median follow-up duration in this study may be too short to investigate the change in NAFLD. When genetic variants affect the development of NAFLD, a long-term follow-up period is necessary. Finally, although ultrasonography for the NAFLD diagnosis is an operator-dependent procedure and not sensitive enough to detect minor steatosis, histological changes using liver biopsy could not be evaluated in the present study.