*2.1. Study Population*

This study was conducted with a local community-based cohort emanating from the Korean population-based cohorts of the Korean Genome and Epidemiology Study [17]. The local community-based cohort included residents living in rural Ansung and urban Ansan since 2001. All subjects provided informed consent at baseline. The cohort was examined by follow-up surveys every two years, and the eighth follow-up survey was performed in 2018. This study used data from the second follow-up survey conducted from 2005 to 2006.

From a total of 7515 subjects, aged 43–74 years, we excluded 2617 subjects with missing data, those with daily energy consumption <500 kcal or >4500 kcal, those with previous history and presence of diabetes, renal disease, thyroid disease, cardiovascular disease, cancer, hysterectomy, and ovariectomy, and those who received medications for those diseases. Finally, 2527 male and 2371 female subjects were included in this study (Figure 1).

**Figure 1.** A flow chart of subject selection.

Dyslipidemia was defined as dyslipidemia diagnosis, related drug use, and abnormal lipid profile (low-density lipoprotein-cholesterol ≥ 160 mg/dL, TG ≥ 200 mg/dL, total cholesterol (TC) ≥ 240 mg/dL, and high-density lipoprotein-cholesterol <40 mg/dL). Blood pressure was the average of three measurements with five minutes interval, taken in the morning after 10 min of rest in sitting position. Coffee intake was assessed using the food-frequency questionnaire. Depending on the amounts of coffee intake per week, the subjects were divided into those who consumed less than one cup of coffee per day (low coffee intake group) and those who consumed more than or equal to one cup of coffee per day (high coffee intake group). A cup was estimated as much as 150 mL. This study was approved by the Institutional Review Board of Ewha Womans University, Seoul, Korea (IRB No. 129-17).

#### *2.2. Genotyping and Analysis of Single Nucleotide Polymorphisms*

Genomic DNA was collected from peripheral blood samples of the subjects and genotyped on Affymetrix Genome-Wide Human SNP Array 5.0, as previously described [18]. Among SNPs in four loci encoding ADORAs, 79 SNPs were included in the platform. The missing call rate (>5%), deviation from Hardy–Weinberg equilibrium (HWE) (*<sup>p</sup>* <sup>&</sup>lt; <sup>1</sup> <sup>×</sup> 106), or minor allele frequency (*<sup>p</sup>* <sup>&</sup>lt; 0.05) was used to eliminate 38 inadequate SNPs in the sample population. Among the remaining 38 SNPs, 30 SNPs were removed due to high levels of pairwise linkage disequilibrium (LD) (Figure 2). Finally, eight of the 79 SNPs were used for further analysis (Table 1).

**Figure 2.** Linkage Disequilibrium (LD) block of genetic variants in the *Adenosine Receptor* (*ADORA*) gene family. The LD blocks are for variants in (**a**) *ADORA1*, (**b**) *ADORA2A*, (**c**) *ADORA2B*, and (**d**) *ADORA3* loci, respectively. Square boxes indicate SNPs used for further analysis.

**Table 1.** The list of selected SNPs in the *Adenosine Receptor* (*ADORA*) gene family.


<sup>1</sup> Alleles are presented as minor/major alleles. SNP, single nucleotide polymorphism; Chr, chromosome; MAF, minor allele frequency; HWE, Hardy–Weinberg equilibrium.
