Introduction

Global mortality due to chronic diseases, such as metabolic syndrome (MetS) and cardiovascular disease (CVD), is increasing in the twenty first century with the aging population (Partridge et al. 2011). An elevated plasma triglyceride (TG) level (above 150 mg/dL) is a well-known risk factor for CVD and MetS in the general population (Hokanson and Austin 1996; Grundy et al. 1999; Woodward et al. 2005). Willett et al. (1983) reported that cigarette smoking significantly increased TG levels in white women. Wu et al. (2001) found a strong association between increased TG levels and smoking and drinking habits in Chinese men. Previous Korean studies reported a positive association between TG levels and cigarette smoking (Park and Kang 2004; Oh et al. 2005). Park and Kang (2004) showed that elevated TG levels were significantly associated with DNA damage in current smokers.

TG levels are affected by multiple genetic and environmental factors (Iselius et al. 1985; Heller et al. 1993). Previous genome-wide association studies (GWAS) identified single nucleotide polymorphisms (SNPs) in several genes such as glucokinase (hexokinase 4) regulator (GCKR, 2p23), transcription factor 7-like 2 (TCF7L2, 10q25.3), and aldehyde dehydrogenase 2 family (ALDH2, 12q24.2) as well as the apolipoprotein cluster (APOC) and apolipoprotein E (APOE), which are both located in chromosomal region 19q13.2, that are associated with TG levels in Asian populations (Ng et al. 2008; Kathiresan et al. 2009; Teslovich et al. 2010). Two variants, rs12970134 and rs17782313, located near the melanocortin 4 receptor gene (MC4R, 18q22) were significantly associated with obesity-related traits such as TG level in Indians (Janipalli et al. 2012). Two recent studies replicated the association of several genetic variants (APOB, GCKR, LPL, APOE, etc.) with TG levels in Korean populations (Park et al. 2011; Kim et al. 2011).

Previous population-based studies have reported that several genetic variants interacted with cigarette smoking for increased TG levels. Waterworth et al. (2000) showed that alleles 3238G (rs5128) and -482T (rs2854117) in apolipoprotein C-III (APOC3, 11q23.1) interacted with smoking habit to increase TG levels in middle-aged men (P = 0.042 and P = 0.009, respectively). The common variant, H + H + genotype of the HindIII (rs320) SNP located in the lipoprotein lipase gene (LPL, 8p22) increased serum TG levels among smoking men in a Spanish cohort (Ariza et al. 2010). Tan et al. (2012) identified several genetic variants (LPL, ALDH2, APOA5, etc.) and environmental factors (i.e. smoking and drinking habits) for serum TG levels in a healthy Chinese male GWAS. For example, one SNP, the GG genotype of rs671 (ALDH2), was associated with increased serum TG levels in current drinkers (P = 3.3 × 10−5). Furthermore, none of the lipid-associated genetic variants, such as PCSK9, APOB, TRIB1, FADS1, LDLR, and APOC1, that have been identified in European populations, showed interactions with drinking or smoking for TG levels in Koreans (Park et al. 2011). However, the number of gene-environment interaction studies on TG levels in Koreans is limited, since most genetic variants were selected based on European populations.

In the current study, we evaluated the effects of genetic and environmental factors on plasma TG levels in 664 elderly Koreans (mostly over 60 years old). We also investigated gene and cigarette smoking interactions on plasma TG levels.

Materials and Methods

Study subjects and data collection

A population-based prospective cohort study, the Hallym Aging Study (HAS) recruited elderly individuals with a mean age of 68 years mostly living in Chuncheon city, Republic of Korea. Chuncheon city is divided into 1,408 areas based on the 2,000 population census, and 200 of these areas were randomly selected. The HAS was designed to investigate the quality of life and healthcare in this elderly population. The initial study population included 1,510 individuals (451 were aged 45 to 64 years, and 1,059 were over 65 years of age). A second panel survey was conducted in 2004, which included 918 adults (437 men and 534 women). A study flow diagram is shown in Fig. 1.

Fig. 1
figure 1

Flow chart of the Hallym Aging Study

Full size image

After informed consent was obtained from each participant, we collected self-reported questionnaires, anthropometric measurements, and clinical traits, including plasma TG levels, from each participant at three-year intervals. Questionnaires were administered by interviewers face-to-face. Anthropometric and clinical traits were measured by a clinical team at Chuncheon Sacred Heart Hospital. The study protocols and procedures were approved by the Institutional Review Board of Hallym University (HIRB-2007-001).

Clinical measurements

BMI was calculated as an individual’s weight in kilograms divided by the square of his/her height in meters. We averaged three measurements of systolic blood pressure (SBP) and diastolic blood pressure (DBP) in one sitting using a standard protocol. We stored blood samples in a −70 °C deep freezer until analysis (DF8510; ilShinBioBase Co., Ltd., Korea). Plasma concentrations of total cholesterol (TC), triglyceride (TG), high density lipoprotein cholesterol (HDL-C), and fasting blood sugar (FBS) were measured by an automatic analyser (Hitachi 7600-210; Hitachi Medical Corp., Hitachi, Japan) using enzymatic methods. Plasma insulin was estimated with Elecsys E170 (Hitachi Medical Corp.) and Versamax (Molecular Devices, Sunnyvale, CA, USA) systems. We estimated the homeostasis model assessment of insulin resistance index (HOMO-IR) using the formula described by Matthews et al. (1985). Other clinical data and procedures have been described elsewhere (Cho et al. 2009).

Genotyping and quality control

We extracted genomic DNA (gDNA) from peripheral blood or buffy coat using the FlexiGene DNA Extraction Kit (Qiagen Inc., Valencia, CA, USA), and DNA concentration was then estimated using a Nanobiometer (MECASYS Co., Ltd., Korea). The gDNA of each sample was separated by 1 % agarose gel electrophoresis to assess DNA quality. gDNA was quantified on a spectrofluorometer (Perkin Elmer Inc., CT, USA) by using the PicoGreen dsDNA quantification kit (Invitrogen Corp., Carlsbad, CA, USA). We excluded samples with low DNA concentrations (<50 ng/μL) before the genotyping test.

We selected three to ten SNPs located within 2 kb upstream and downstream of each of five candidate genes, ADLH2, APOE, GCKR, MC4R, and TCF7L2 for TG levels, and two genes of interest (GATA2 and HNF1A) for cigarette smoking by using the literature database, HuGE Navigator (http://hugenavigator.net/HuGENavigator/home.do) and dbSNP (www.ncbi.nlm.nih.gov/projects/SNP/). We genotyped 715 Koreans for 32 SNPs with the Golden Gate Genotyping Assay (Illumina Inc., San Diego, CA, USA). We excluded three individuals and three SNPs with genotyping call rates <95 % (rs3751150, rs3751151, and rs4420638). We evaluated pair-wise linkage disequilibrium (LD), Hardy–Weinberg equilibrium (HWE), and minor allele frequency (MAF) by using the Haploview software package, v. 4.2 (Barrett et al. 2005).

Statistical analysis

We conducted univariate and multivariate analyses using the STATA software package, v. 11.2 (StataCorp., College Station, TX, USA). We excluded TG levels of outliers (mean + 3SD) and missing values in the baseline survey (N = 664). We transformed TG level to a natural logarithmic scale (logTG) to normalize the distribution of data. We performed a trend test for the median TG level according to smoking status habit using a one-sided Jonckheere-Terpstra test (Pirie 1983). To detect confounding factors, we performed single linear regression (SLR) analyses to evaluate associations of logTG with age, gender, history of hypertension, four lifestyle factors (regular exercise, smoking, drinking, and sleep), and eight clinical variables (BMI, waist circumference, SBP, DBP, TC, HDL-C, albumin, FBS, fasting insulin, and HOMA-IR). We identified a best-fit model using a general linear model (GLM) with stepwise backward elimination to estimate the beta coefficient (β) and 95 % confidence interval (CI) after adjusting for confounding factors.

We conducted a single SNP analysis to identify the associations of 27 SNP markers near or within seven genes with logTG levels under three genetic models (i.e. additive, dominant, and recessive) after adjustment for age and sex (GLM 1); and for the best-fitting GLM 2. To account for multiple testings, we used a significance threshold less than 0.0018 based on Bonferroni correction. We finally evaluated the interactions of 27 SNP markers with smoking habit on plasma TG levels in 664 individuals.

Results

The baseline characteristics of 664 subjects along with the TG levels are shown in Table 1. TG levels did not differ between men and women (166.2 vs. 156.4 mg/dL, P > 0.05). Variations in the mean TG level according to smoking status in each group are shown in Table 2. Plasma TG levels did not differ by smoking status in any group (P > 0.05). Among 17 variables, we observed statistically significant associations for ten variables with P values less than 0.05 in SLR analyses (data not shown). The best-fitting model included six variables, age, sex, SBP, TC, HDL, and FBS, by using stepwise backward elimination (6 × 10−25 < P < 0.05).

Table 1 Baseline characteristics of study subjects in the Hallym Aging Study, 2007
Full size table
Table 2 Variation in the plasma triglyceride levels according to smoking status by gender group, Hallym Aging Study, 2007
Full size table

All of SNP markers that were near or within seven genes (i.e. ADLH2, APOE, GCKR, MC4R, TCF7L2, GATA2, and HNF1A) showed HWE P values greater than 0.05 and MAFs >1 %. 24 SNPs were described after considering LD (Table 3). We identified significant associations for seven SNPs with plasma TG levels of 664 individuals after adjustment for age and sex. Three variants, rs2713604 (GATA2), rs2713603 (GATA2), and rs3751152 (HNF1A), were not significant after adjusting for age, sex, SBP, TC, HDL, and FBS. For the four remaining SNPs, the CT and TT genotypes of rs7412 (APOE) showed the strongest association with increased TG levels under the dominant model (β = 0.21; 95 % CI, 0.11–0.30; P = 1.9 × 10−5). In addition, we found that two novel genetic variants, the AA genotype of rs2860288 (GATA2) and the GG genotype of rs2393791 (HNF1A), increased plasma TG levels (β = 0.23; 95 % CI, 0.05–0.41 and β = 0.09; 95 % CI, 0.00–0.17, respectively). Two additional SNPs, rs671 (ALDH2) and rs7412 (APOE), remained significant after Bonferroni correction in GLM 2 (P < 0.0018). These variants lead to non-synonymous substitutions that modify the amino acid composition of the gene products (Table 3).

Table 3 Association of 24 SNPs with plasma triglyceride levels in 664 subjects
Full size table

Among the 24 SNPs, we identified statistically significant interactions of four genetic variants, GCKR (rs1260326), GATA2 (rs2713604 and rs2713603), and HNF1A (rs2464196) with smoking on plasma TG levels (Fig. 2 and Table 4). As shown in Table 4, while the mean TG levels of non-smokers who were C allele carriers of rs1260326 (GCKR) (G+) and of current smokers not carrying a C allele (E+) were 161.9 mg/dL and 172.7 mg/dL, respectively, smokers carrying the CC genotype (G+ E+) had significantly lower TG levels (124.6 mg/dL) (β = -0.33, P G×E = 0.017). Plasma TG levels were highest in current smokers with the CC genotype of rs2713604 (GATA2) (195.6 mg/L). The ancestral G and C alleles of rs2713603 and rs2464196 were associated with increased plasma TG levels in smokers (180 mg/dL, P G×E = 0.046 and 171 mg/dL, P G×E = 0.028, respectively). However, two missense variants in rs1260326 (Leu446Pro) and rs2464196 (Ser487Asn) showed no significant association with mean logTG after adjustment for age and sex (P = 0.536 and P = 0.373, respectively; Fig. 3).

Fig. 2
figure 2

The effect of interactions between four SNPs and smoking on triglyceride levels in Koreans. The interaction models shown in A–D are for rs1260326 (GCKR), rs2713604 (GATA2), rs2713603 (GATA2), and rs2464196 (HNF1A) with smoking status, respectively (P, P value for the interaction)

Full size image
Table 4 Interactions between genetic polymorphisms and smoking status for plasma triglyceride levels
Full size table
Fig. 3
figure 3

Variations in the mean log-transformed triglyceride (logTG) levels according to each genotype. The mean logTG levels shown in A–D are for rs1260326 (GCKR), rs2713604 (GATA2), rs2713603 (GATA2), rs2464196 (HNF1A) after adjustment for age and sex, respectively (P, adjusted P value)

Full size image

Discussion

High TG levels can be caused by multiple risk factors such as obesity, family history, a lack of exercise, blood pressure, smoking, and genetic variants (Iselius et al. 1985; Heller et al. 1993). TG levels significantly predicted the risk of CVD and type 2 diabetes mellitus (T2DM) (Woodward et al. 2005; Pearson et al. 2003; Tirosh et al. 2008). Many prospective cohort studies have reported a positive association between elevated TG levels and cigarette smoking (Willett et al. 1983; Grundy et al. 1999; Wu et al. 2001; Oh et al. 2005). In the current study, plasma TG levels were not associated with smoking in 664 elderly Koreans.

This study showed six covariates that increased plasma TG levels in the best-fitting GLM (i.e. age, sex, SBP, TC, HDL, and FBS). We confirmed the associations between two gene variants (ALDH2 and APOE) and plasma TG levels after adjustment for the GLM in this population (Teslovich et al. 2010; Tan et al. 2012), while the GCKR, TCF7L2, and MC4R polymorphisms were not significantly associated with TG levels. A missense variant, the AA genotype of rs671 (Glu504Lys) located in the ALDH2 gene, was reported to be related to alcohol metabolism and associated with low TG levels in Chinese populations (Tan et al. 2012; Wang et al. 2013). The T allele of rs7412 (Arg176Cys, in APOE) has been reported to be associated with increased TG levels after adjustment for age, sex, waist circumference, blood glucose, blood pressure, and smoking and drinking status in an additive effect model as well as the risk of Alzheimer’s disease (Ariza et al. 2010; Lescai et al. 2011). Interestingly, two novel genetic variants, the rare G and A alleles of rs2860288 (GATA2) and rs2393791 (HNF1A), respectively, showed a positive association with plasma TG levels in the recessive model. The GATA2 gene (3q21.3) encodes a transcription factor, which contains six exons and plays a role in haematopoiesis and endothelial dysfunction (Connelly et al. 2006). Mutations in GATA2 polymorphisms are associated with a risk of familial early-onset coronary artery disease, sporadic monocytopenia, mycobacterial infection syndrome, and lung cancer (Connelly et al. 2006; Hsu et al. 2011; Kumar et al. 2012). In addition, this gene polymorphism was found to be strongly associated with smoking habit in a healthy US population (Pan et al. 2010). The HNF1A gene was shown to play a critical role in pancreatic β-cell function (Ryffel 2001) and is strongly associated with a risk of T2DM in smokers (Ley et al. 2011). Cauchi et al. (2008) showed obesity-modified associations between genetic variants (e.g. GCKR and HFN1A) and T2DM in European populations. Particularly, they demonstrated the impact of rs1169288 variants (HNF1A) on T2DM susceptibility in non-obese subjects.

Previous population-based studies showed that TG levels were modified by the interactions of several genetic variants (e.g. LPL and the APOE-APOC clusters) with cigarette smoking in general populations (Waterworth et al. 2000; Ariza et al. 2010). However, none of the genetic variants showed statistically significant interactions with cigarette smoking on serum TG level in Asian populations (Tan et al. 2012; Park et al. 2011). In this study, we observed novel interactions of rs1260326 (GCKR), rs2713604 (GATA2), rs2713603 (GATA2), and rs2464196 (HNF1A) with current smoking for plasma TG levels (P < 0.05). A limitation of this study is that the limited number of study subjects could increase the possibility of false negative results in the identification of gene-environment interactions. In addition, we were not able to determine the specific biological mechanisms underlying the statistically significant interactions between the genes and cigarette smoking on increased plasma TG levels.

In conclusion, common susceptibility variants and their interaction with smoking habits affect the inter-individual variability of plasma TG levels. The novel interactions between genetic variants (i.e. rs1260326:GCKR, rs2713604:GATA2, rs2713603:GATA2, and rs2464196:HNF1A) and cigarette smoking reported in the current study need to be validated in large-scale cohort studies. Particularly, verification of the interactions of some genetic variants with cigarette smoking on increased TG levels may be necessary to demonstrate the functional role of those variants in complex human diseases. Future studies focusing on the complex interactions between multiple genetic and environmental factors on TG levels could contribute to personalized preventive and therapeutic approaches to reduce CVD and MetS.