Introduction

Adiponutrin, also known as patatin-like phospholipase domain-containing protein 3 (PNPLA3), is expressed in many tissues, including the liver, where it is primarily localized in lipid droplets [1]. PNPLA3 exhibits hydrolase activity on monoacyl-, diacyl- and triacylglycerols/triglycerides (TGs) and lysophosphatidic acid acyltransferase activity, thus participating in the hepatic metabolism of TGs and phospholipids (PLs), while some authors suggest that it may be secondarily associated with glucose homeostasis [2–4].

The single nucleotide polymorphism (SNP) rs738409 in the PNPLA3 gene has been strongly associated with the development and progression of nonalcoholic fatty liver disease (NAFLD) [5, 6]; indeed, this is the most validated SNP linked to the full spectrum of NAFLD [7]. rs738409 (c444C>G) leads to the substitution of methionine for isoleucine in the active center of the protein; this alteration affects access, substrate, and, consequently, enzymatic activity [8], leading to reduced hydrolytic removal of fatty acids and increased accumulation of TGs in hepatocytes and retinol retention in hepatic stellate cells, thus contributing to hepatic steatosis and fibrosis [8, 9]. The risk of nonalcoholic steatohepatitis (NASH) in patients with this variant is maximized when coexists with high adiposity, illustrating the additive effects of genetic and environmental contributors to this disease [10].

While NAFLD is closely associated with insulin resistance, dyslipidemia, and hyperglycemia, the association of rs738409 to lipid and glucose metabolism remains unclear. The main aim of this study was the investigation of the association of rs738409 with serum lipids in patients with dyslipidemia. The secondary aim was the association of SNP with glycosylated hemoglobin (HbA_1c).

Materials and methods

This was a post-hoc analysis of a study in patients with dyslipidemia recruited in an outpatient basis in the dyslipidemia clinic, between 2009 and 2012. All the participants provided informed consent and the study was approved by the Bioethics Committee of the Medical School of Aristotle University of Thessaloniki. The inclusion criteria were: 1) dyslipidemia defined as morning fasting total cholesterol (TC) > 240 mg/dL and/or low-density lipoprotein cholesterol (LDL-C) > 160 mg/dL and/or TGs > 200 mg/dL in at least two measurements with an interval > 1 week; 2) lack of any hypolipidemic therapy or discontinuation of any hypolipidemic agent for at least 3 months.

The patients’ personal and family medical history, anthropometric characteristics [gender, age, and body mass index (BMI)], cigarette smoking, and alcohol consumption were recorded. Morning blood samples were collected after a 12-h fast. Lipid tests were performed with standard methods using a biochemical analyzer (Hitachi 912, Roche, Ramsey, MN, USA). LDL-C was calculated with the Friedewald equation [11]. HbA_1c measurement was determined by high performance liquid chromatography (HA-8121, Menarini, Firenze, Italy). Genomic DNA was extracted from whole-blood samples by means of the Ron’s Blood DNA minikit (BIORON GmbH, Ludwigshafen, Germany) according to the manufacturer’s instruction. Genotyping for rs738409 was accomplished using a previously reported polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method [12].

Results

One hundred seventy-five patients with dyslipidemia were included [97 women, aged 59 (13), BMI 29 (4), type 2 diabetes mellitus (T2DM) 43%, hypertension 69.7%] (Table 1). Genotypic for rs738409 was successfully performed in 165 patients. Genotype distribution did not deviate from Hardy-Weinberg equilibrium (P = 0.810). Gender, age, and BMI did not differ significantly between genotypes (CC vs. CG vs. GG), contrary to smoking, T2DM, and coronary artery disease. Circulating lipids [TC, TGs, LDL-C, high-density lipoprotein cholesterol (HDL-C)] or HbA_1c were not associated by the polymorphism following adjustment for gender, BMI, and T2DM diagnosis, using either the additive or the dominant model of inheritance of the minor allele (Table 2).

Interestingly, upon comparison of genotype distributions following stratification of patients according to T2DM, we detected no diabetic patients homozygous for the G allele, whereas, in the non-diabetic group, 12.4% had the GG genotype (P = 0.005); no significant association was found between rs738409 and serum lipids/lipoproteins or HbA_1c in the subgroup analysis. When stratifying according to obesity (BMI cut-off 30 kg/m²), no significant difference was observed in genotype distribution. However, subgroup analysis produced a significant association between rs738409 and TC (P = 0.014) and LDL-C (P = 0.046) in the non-obese, with the lowest levels observed in carriers of the CG genotype (Table 3). A post-hoc, multiple pairwise comparison showed significant difference in TC between carriers of the CC and CG genotypes (P = 0.012) but no specific difference in LDL-C following the Bonferroni correction. No such difference was detected in the obese subgroup. The design of this study cannot allow safe conclusions on these different associations observed between non-obese and obese. Although residual confounding may have affected these results (e.g., differences between obese and non-obese in dietary and exercise habits, hepatic and renal function, percentage of NAFLD, etc.), further studies are needed to validate or not these differences and to investigate their pathophysiological and clinical implications.

Discussion

In this study, no association was found between rs738409 SNP and circulating lipids/lipoproteins or HbA_1c in patients with dyslipidemia. In subgroup analysis, TC levels were significantly higher in non-obese, but not in obese, dyslipidemic patients with CC, compared to CG carriers.

A meta-analysis confirming the association between rs738409 and NAFLD did not find any association between the polymorphism and circulating lipids/lipoproteins [13]. Some authors have reported that obese individuals carrying the G allele did not show any difference in serum lipids or glucose compared to non-carriers [14, 15], similar to our findings; however, others showed that individuals carrying the G allele presented with either lower [16] or higher [17] TG levels. Still, a more recent report indicated that homozygotes for GG had elevated TG, but not other lipids/lipoproteins, glucose, or HbA_1c [18]. Finally, in contrast to our results, in a study with obese NAFLD patients, it was shown that those with the CG or GG genotype presented with a 9.6-fold higher risk of glucose disorders [19]. In that study, a higher rate of CG carriers was observed in the group of dyslipidemic patients [19]. Importantly, a recent meta-analysis, which included 63 studies, showed significant associations of rs738409 polymorphism of PNPLA3 with TG and TC, but not with LDL-C and HDL-C. The rs738409 polymorphism was associated with lower TG and TC levels [20]. However, we should highlight that: 1) the overall magnitude of the associations of this SNP with TG and TC was small; 2) the associations of this SNP with TG and TC were not significant in all subpopulations, possibly implying population differences; and 3) most of the studies that were included in this meta-analysis had not separately shown associations between rs738409 polymorphism of PNPLA3 and lipid levels [20], similarly to our study. While we have no convincing explanation for the discrepancies between studies, they may be, at least partly, attributed to underlying population and methodological differences (e.g., ethnicity, dietary habits, obesity, percentage of NAFLD patients).

The main strength of this study was that it may be the first to evaluate the rs738409 SNP in patients with dyslipidemia; in previous studies, dyslipidemia had been evaluated in the setting of patients with NAFLD and/or T2DM. Its limitations include: 1) the relatively small sample size and the lack of an a priori power calculation; however, the sample was sufficient to provide significant difference in TC in the subgroup of the non-obese; 2) the cross-sectional design, which does not allow establishing causal relationships; 3) the lack of a validation set, in spite of the largely negative results; 4) the lack of data on NAFLD; 5) the lack of data on metabolic syndrome, kidney function, and concomitant medications; 6) the lack of recorded plasma glucose levels, since glycemic parameters were not the primary aim of this study.

In conclusion, this study showed no association between rs738409 SNP of PNPLA3 gene and serum lipids/lipoproteins or HbA_1c in patients with dyslipidemia. An association between this SNP and TC in non-obese patients needs validation by other independent studies. This study largely confirms that the rs738409 SNP is not associated with lipid and glycemic profile, contrary to its reported high association with the development and severity of NAFLD.