Carbohydrate-dependent interaction between AMY1A/AMY2B copy number variation and APOA5 rs651821 associated with hypertriglyceridemia in Korean middle-aged adults

Abstract

Background

Hypertriglyceridemia arises from complex interactions between genetic and dietary factors. AMY1A/AMY2B copy number variation (CNV) and APOA5 rs651821 have been associated with triglyceride metabolism, and their effects may be modified by carbohydrate intake. This study investigated the combined influence of AMY1A/AMY2B CNV, APOA5 rs651821, and dietary carbohydrate intake on the odds of hypertriglyceridemia in middle-aged Korean adults.

Methods

This cross-sectional study of 595 Korean adults from the Korean Genome and Epidemiology Study (KoGES) 2001–2002 evaluated amylase gene CNV and the APOA5 rs651821 single-nucleotide polymorphism (SNP) in relation to hypertriglyceridemia (triglyceride ≥ 150 mg/dL). A CNV segment encompassing AMY1A/AMY2B (chr1:104,144,265–104,220,453) and the APOA5 rs651821 SNP, which has previously been linked to hypertriglyceridemia, was selected for analysis. Participants were categorized into diploid and duplication CNV groups, as well as into TT and C-carrier SNP groups. Multivariable logistic regression was performed to assess the associations of CNV and SNP status with hypertriglyceridemia, adjusting for covariates. CNV–SNP interactions were examined using multiplicative and additive models to estimate odds ratios, confidence intervals (CI), and attributable proportions. Subgroup analyses, stratified by the median carbohydrate intake, were performed to evaluate the potential three-way interaction effects.

Results

In the fully adjusted model, the independent association of AMY1A/AMY2B duplication with hypertriglyceridemia was attenuated to a marginal trend, whereas APOA5 rs651821 C-carrier showed a significant independent association (OR = 2.00; 95% CI: 1.39–2.88). Compared with diploid × TT, the duplication × C-carrier group showed higher odds of hypertriglyceridemia (OR = 3.11; 95% CI: 1.69–5.72); however, multiplicative and additive two-way interaction metrics were not significant. In carbohydrate-stratified analyses, the duplication × C-carrier group showed the highest odds in the low-carbohydrate group (< median, 73.2% of energy: OR = 4.33; 95% CI: 1.78–10.56), whereas the corresponding association in the high-carbohydrate group (≥ median, 73.2% of energy) did not remain significant after FDR correction. The three-way interaction was nominally significant but did not remain significant after FDR correction.

Conclusions

Although the independent effect of AMY1A/AMY2B CNV was attenuated by clinical covariates, its synergistic interaction with the APOA5 C allele remains a robust predictor of hypertriglyceridemia. AMY1A/AMY2B duplication may increase starch digestion and hepatic de novo lipogenesis, whereas the APOA5 C allele may be associated with reduced triglyceride clearance via impaired lipoprotein lipase activation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40795-026-01357-6.

Introduction

Hypertriglyceridemia, characterized by elevated fasting serum triglyceride (TG) concentrations, contributes to lipotoxicity, oxidative stress, and chronic inflammation, thereby accelerating metabolic dysfunction [1, 2]. Elevated TG levels are frequently detected in individuals with metabolic conditions such as obesity, diabetes, and cardiovascular diseases (CVD) [3, 4]. In individuals with type 2 diabetes, this lipid abnormality has been identified as an independent contributor to the onset and progression of atherosclerosis [5, 6].

Globally, the prevalence of hypertriglyceridemia is estimated to be approximately 32.6% in the adult population, with a marked difference between sexes: approximately 36.9% in men and 23.8% in women [7]. Recent epidemiological data from the Korea National Health and Nutrition Examination Survey (KNHANES, 2007–2022) indicate that hypertriglyceridemia remains a significant public health burden among Korean adults, as reported in the Dyslipidemia Fact Sheet in South Korea, 2024 [8], underscoring the need for continued surveillance and targeted intervention in this population.

Hypertriglyceridemia arises from a combination of genetic and environmental factors. Hepatic overproduction of very-low-density lipoprotein (VLDL) driven by obesity and insulin resistance, together with dietary fat-induced intestinal chylomicron overproduction, collectively elevates circulating TG levels [9]. Among these genetic factors, copy number variations (CNVs) in the amylase genes have recently gained attention owing to their roles in carbohydrate metabolism and lipid homeostasis. In addition to environmental and metabolic factors, genetic variability contributes substantially to an individual’s susceptibility to hypertriglyceridemia. The salivary and pancreatic amylase genes, AMY1A and AMY2B, exhibit CNVs that have been linked to interindividual differences in carbohydrate metabolism and related metabolic phenotypes [10]. Recent studies have suggested that lower AMY1A or AMY2B CNV predisposes individuals to dyslipidemia, including elevated TG levels, possibly through impaired glucose and insulin dynamics [11].

The APOA5 gene is essential for TG metabolism as it modulates VLDL secretion and TRL clearance [12, 13]. APOA5 is a 366–amino acid protein predominantly produced and secreted by the liver and is associated with high-density lipoprotein (HDL). It is considered a major determinant of plasma TG concentration in fasting and postprandial states [14–16]. Functionally, APOA5 downregulates VLDL-TG production and enhances LPL-mediated hydrolysis of VLDL-TGs, thereby contributing to the reduction in circulating TG levels [14, 17]. SNPs within APOA5, particularly − 1131T > C (rs651821) and 56 C > G (rs3135506), have been identified as key functional variants strongly associated with TG metabolism and susceptibility to dyslipidemia and CVD [18, 19]. Among the variants most intensively examined, − 1131T > C (rs651821) and 56 C > G (rs3135506) are regarded as functional tag SNPs representing APOA5 polymorphisms [17, 20]. The − 1131 C allele exhibits a higher prevalence in East Asian populations (> 25%) compared with Western populations (9–16%) [19, 21, 22]. The presence of the minor C allele at rs651821 has been correlated with augmented plasma TG levels and an increased risk of CVD, independent of other contributing factors [17, 20, 21]. Although APOA5 effects on TG metabolism are known to involve multi-variant haplotypes, including rs662799 and rs2266788, rs651821 was selected as the primary variant of interest in the present study for two reasons. First, rs651821 has been most consistently and strongly associated with plasma TG levels in Korean and East Asian cohort studies, including population-based GWAS [23, 24]. Second, owing to the strong linkage disequilibrium structure among APOA5 variants in East Asian populations, rs651821 serves as a representative tag SNP that captures the predominant haplotypic variation in this genomic region, thereby reflecting the broader haplotype signal without necessitating full multi-variant haplotype analysis [24].

Dietary carbohydrate intake interacts with AMY1A CNV to influence metabolic phenotypes. AMY1A CNV modulates starch digestion efficiency: individuals with lower copy numbers produce less salivary amylase, resulting in slower starch breakdown, greater glycemic fluctuations, and potentially greater lipid accumulation, whereas higher copy numbers confer more stable postprandial glucose and metabolic responses [25–28].

Nutrient–gene interactions also extend to APOA5, a major determinant of plasma TG metabolism. The rs651821 variant within this gene has been robustly associated with increased TG concentration and greater vulnerability to hypertriglyceridemia in diverse populations [29]. Accumulating evidence suggests that the dietary composition, especially carbohydrate intake, modulates the expression of APOA5-related phenotypes, affecting TG metabolism [25]. Moreover, studies examining nutrient–gene interactions have shown that macronutrient composition, particularly the proportion of dietary carbohydrates or fiber, can modulate lipid responses in APOA5 variant carriers [25, 30]. These findings imply that carbohydrate quality and quantity may interact with genetic predisposition to affect TG metabolism, underscoring the importance of considering both diet and genotype when managing hypertriglyceridemia.

Considering that AMY1A/AMY2B CNV and APOA5 rs651821 are independently associated with TG metabolism and that dietary carbohydrate intake influences these metabolic processes, it is important to investigate their combined effects on hypertriglyceridemia. This study hypothesized that the interaction between AMY1A/AMY2B CNV and APOA5 rs651821 is influenced by dietary carbohydrate intake and contributes synergistically to the risk of hypertriglyceridemia in middle-aged Korean adults. This study aimed to identify gene to gene–diet interactions and provide new insights into precise nutritional strategies for managing TGs. To our knowledge, this is the first study to investigate the three-way interaction among AMY CNV, APOA5 genotype, and dietary carbohydrate intake in relation to hypertriglyceridemia.

Materials and methods

Study design and participants

This study used the community-based cohort of the Korean Genome and Epidemiology Study (KoGES) 2001–2002 [31]. The cohort was established with individuals aged 40 years or older from the general population of the Ansan and Anseong regions and collected extensive data on health status, health behaviors, dietary patterns, and genetic information. Initiated in 2001, follow-up surveys have been conducted biennially. To date, the study has been ongoing for over 20 years, reaching the 10th follow-up survey, representing the longest-running cohort study in Korea. While KoGES was established as a prospective cohort with repeated follow-up assessments, this analysis was conducted cross-sectionally based on baseline data obtained at the time of CNV measurement.

Figure 1 illustrates the selection process for the study population. Of the 10,030 individuals enrolled at baseline, following the application of the exclusion criteria, a total of 595 participants were retained in the study: (1) absence of CNV data (n = 9,030); (2) absence of APOA5 rs651821 genotype data (n = 343); (3) absence of TG levels or biochemical information (n = 21); (4) absence of dietary carbohydrate intake information (n = 17); and (5) missing covariates data, including sex, age, body mass index (BMI), total energy, smoking status, alcohol consumption status, and physical activity (n = 24). To evaluate potential selection bias caused by the reduction in sample size, baseline demographic, clinical, lifestyle, and dietary characteristics were compared between participants included in the analytic sample and those excluded from the baseline KoGES cohort [32]. The study protocol was reviewed and approved by the Institutional Review Board of Ewha Womans University, Korea (institutional review board number ewha-202603-0023-01), and all participants provided written informed consent.

Fig. 1.

Flow diagram of the study design, with participant inclusion and exclusion criteria

Data acquisition and clinical assessment

Anthropometric measurements included height, weight, and waist circumference. BMI was calculated from measured height and weight. Waist circumference was measured at the midpoint between the lowest rib and the iliac crest. Blood pressure was measured using a mercury sphygmomanometer in a sitting position after adequate rest.

Following a 12-hour overnight fast, venous blood samples were collected in the morning and processed using SST and EDTA tubes. Fasting plasma glucose, triglycerides, and HDL cholesterol were measured using automated chemistry analyzers (Hitachi 7600 until August 2002; ADVIA 1650, Siemens, from September 2002 onward).

Fasting insulin was also measured from fasting blood samples and was used to estimate insulin resistance. Insulin resistance was assessed using the homeostasis model assessment of insulin resistance (HOMA-IR), calculated as fasting insulin (µIU/mL) × fasting glucose (mg/dL) / 405 [33]. Information on medication use and physician-diagnosed diseases, including lipid-lowering medication, oral antidiabetic medication, insulin treatment, antihypertensive medication, thyroid medication, diabetes, thyroid disease, and kidney disease, was obtained from baseline questionnaire data. Antidiabetic medication use was defined as current use of oral antidiabetic medication or insulin treatment [31].

Alcohol consumption was classified as non-drinker, former drinker, or current drinker. Smoking status was categorized as never, former, or current smoker. Physical activity was assessed as metabolic equivalent of task (MET-h/week), weighted by activity intensity level.

Classification of hypertriglyceridemia

Hypertriglyceridemia was classified into the hypertriglyceridemia group when TG levels were 150 mg/dL or higher, and the non-hypertriglyceridemia group when TG levels were less than 150 mg/dL [34].

Carbohydrate intake assessment

Carbohydrate intake was collected using a semi-quantitative food frequency questionnaire [35]. Daily nutrient intake was estimated using reported consumption frequency and portion sizes for 103 food items. Intake frequency was assessed using nine categories (seldom, once a month, 2–3 times a month, 1–2 times a week, 3–4 times a week, 5–6 times a week, once a day, twice a day, and 3 times a day). Portion size was classified as small, medium, or large, with the average intake of the Korean population used as the reference.

Genotyping and CNV determination

CNV was detected using a comparative genomic hybridization 720 K array chip containing 720,356 probes [36]. Raw data in .CEL file format, including probe signal intensities, were processed using PennCNV (version 1.0.3) to extract the signal intensities, perform CNV calling, and apply GC correction. Preprocessing steps, including raw data acquisition, signal intensity extraction, CNV calling, and GC adjustment, were conducted by the National Biobank of Korea at the Korea Disease Control and Prevention Agency.

For CNV calling, probes with the same sequence ID were grouped into unique CNV segments. Quality control was performed by removing segments with fewer than five probes, segments with a positive-to-negative probe ratio < 90%, segments with CNV values of zero across all samples, and segments located on sex chromosomes. Outliers were excluded using the interquartile range method, and the UCSC LiftOver tool was employed to convert the coordinates from hg18 to hg19 [37].

A total of 16,051 autosomal CNV cases passed quality control. Segments within the diploid range (log2 [1.32/2] to log2 [2.64/2]) were considered normal and were set to zero. CNVs were classified by log2 ratios as follows: < log2 (0.87/2), deep deletion; between log2 (0.87/2) and log2 (1.32/2), heterozygous loss; between log2 (1.32/2) and log2 (2.64/2), diploid; between log2 (2.64/2) and log2 (3.36/2), gain; and > log2 (3.36/2), amplification [38]. The CNV burden was calculated as the total number of autosomal CNV segments with nonzero values per sample, with each non-diploid segment counted as one discrete CNV event.

Selection of CNV and SNP

To identify CNV loci in the AMY region, multivariable linear regression analyses were performed to assess associations between genomic segments within the AMY locus and serum TG levels, adjusting for age, sex, BMI, smoking and alcohol consumption status, MET, total energy intake, lipid-lowering medication, antidiabetic medication use, antihypertensive medication, thyroid medication, thyroid disease, kidney disease, diabetes diagnosis, HOMA-IR, and dietary fat. The segment at chr1:104,144,265–104,220,453 demonstrated a statistically significant association with TG levels and was selected for further analysis. This segment encompasses AMY1A and AMY2B and represents a duplicated state (Database of Genomic Variants nsv950526). Information on the selected CNV loci is shown in Fig. 2. SNP genotyping was performed using the Korean Biobank Array (Korean Chip, Seoul, Korea). The APOA5 rs651821 variant was selected based on prior evidence of a strong association with hypertriglyceridemia, a key component of MetS, in middle-aged Korean adults [39, 40].

Fig. 2.

Significant copy number variation segment at the AMY locus linked to triglycerides. Each point represents a CNV segment tested for association with TG levels. The red circle indicates the AMY1A/AMY2B duplication (nsv950526); the gray shading marks the region of consistent association signals. CNV, copy number variation; TG, triglyceride

Statistical analyses

To assess the representativeness of the analytic sample, baseline characteristics were additionally compared between participants included in the analysis and those excluded from the baseline KoGES cohort. Standardized mean differences (SMDs) were calculated to evaluate the magnitude of imbalance between groups, with values around 0.1 interpreted as indicating small imbalance [32]. Differences in demographic and clinical profiles were examined across groups classified according to the hypertriglyceridemia status. Means with standard deviations were reported for continuous variables and compared using the Student’s t-test, whereas categorical variables were described as frequencies and percentages and assessed using the chi-square test.

For CNV analysis, participants with a heterozygous loss (n = 3) were excluded. The remaining participants were classified into diploid and duplication groups, the latter of which included both gain and amplification variants. The association between the CNV status and hypertriglyceridemia was subsequently examined. For SNP analysis, participants were classified into the C-carrier group, defined as carriers of the risk allele C, and the TT group, defined as non-carriers. Associations between the SNP genotypes and hypertriglyceridemia were assessed. To ensure consistency in sample size across analyses, three individuals with a heterozygous loss of CNVs were excluded prior to analysis.

Logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations of individual CNVs, APOA5 rs651821, and their combined genotype groups with hypertriglyceridemia. The multivariable models were adjusted for age, sex, BMI, smoking status, alcohol consumption, physical activity, total energy intake, lipid-lowering medication, antidiabetic medication use, antihypertensive medication, thyroid medication, thyroid disease, kidney disease, diabetes diagnosis, HOMA-IR, and dietary fat intake.

To evaluate CNV–SNP interactions, a multiplicative interaction (MI) model was applied. Multivariable logistic regression models, including an interaction term, were then fitted to assess MI effects after adjusting for potential confounders. Additive interactions were evaluated by calculating the attributable proportion (AP) and the corresponding 95% CIs, which represented the proportion of hypertriglyceridemia risk attributable to CNV–SNP interactions [41]. Subgroup analyses were conducted according to carbohydrate intake to explore potential 3-way interaction effects [42]. For the primary subgroup analysis, carbohydrate intake was expressed as the percentage of total energy intake and dichotomized at the median value in the analytic sample [43]. Logistic regression analyses were then performed within each carbohydrate intake subgroup. As a sensitivity analysis, both AMY CNV and carbohydrate intake were additionally modeled as standardized continuous variables, and the results are presented in Supplementary Table 4. Analyses were performed in Python (version 3.11.11). Nominal p < 0.05 were reported, and FDR-adjusted p-values were used to evaluate statistical significance in analyses involving multiple related tests.

Results

Baseline characteristics of the participants

The demographic and clinical characteristics of the 595 participants grouped according to hypertriglyceridemia status are summarized in Table 1. The hypertriglyceridemia group had a higher percentage of male participants than the non-hypertriglyceridemia group. BMI, waist circumference, blood pressure (both systolic and diastolic), fasting glucose, and TG concentrations were elevated in individuals with hypertriglyceridemia, whereas HDL cholesterol levels were reduced. Individuals with hypertriglyceridemia exhibited a greater prevalence of metabolic phenotypes, including abdominal obesity, elevated blood pressure, increased fasting glucose, and decreased HDL cholesterol, than those without hypertriglyceridemia. Regarding lifestyle habits, the proportion of nonsmokers was significantly higher in the non-hypertriglyceridemia group than in the hypertriglyceridemia group (64.4% vs. 50.6%, p = 0.0035). In terms of alcohol consumption status, current drinkers were significantly more prevalent than non-drinkers or former drinkers in the hypertriglyceridemia group, whereas non-drinkers were more prevalent than former or current drinkers in the non-hypertriglyceridemia group. No statistically significant differences were observed between groups in terms of total energy intake, carbohydrate intake, or MET levels. To assess the representativeness of the analytic sample, baseline characteristics were compared between the 595 included participants and 9,435 excluded participants from the baseline KoGES cohort (Supplementary Table S1). Overall, the included participants were generally comparable to the excluded participants across demographic, clinical, lifestyle, and dietary characteristics. Most SMD values were below or close to 0.1, suggesting that the magnitude of imbalance was generally small. Although statistically significant differences were observed for sex, diastolic blood pressure, elevated blood pressure, and low HDL cholesterol, the corresponding SMD values indicated small differences in magnitude.

Table 1.

Demographic and clinical characteristics of study participants based on hypertriglyceridemia

Variable	Hypertriglyceride (n = 247)	Non-hypertriglyceride (n = 348)	p-value
Age (year)	53.03 ± 8.40	52.21 ± 8.98	0.2593
Male (n, %)	152 (61.5)	155 (44.5)	6.2 × 10− 5
BMI (kg/m2)	25.19 ± 2.95	23.78 ± 2.89	1.0 × 10− 8
Waist circumference (cm)	85.31 ± 8.01	79.78 ± 8.01	7.3 × 10− 16
Systolic blood pressure (mmHg)	127.53 ± 19.17	119.45 ± 17.98	2.0 × 10− 7
Diastolic blood pressure (mmHg)	84.34 ± 11.51	79.59 ± 10.87	4.1 × 10− 7
Fasting blood glucose (mg/dL)	92.84 ± 25.51	83.91 ± 14.94	1.2 × 10− 6
Triglycerides (mg/dL)	243.47 ± 124.51	102.05 ± 25.32	3.4 × 10− 46
HDL cholesterol (mg/dL)	41.32 ± 7.96	48.22 ± 9.70	5.0 × 10− 20
Abdominal obesity (n, %)	85 (34.4)	71 (20.4)	0.0002
Elevated blood pressure (n, %)	154 (62.3)	143 (41.1)	5.0 × 10− 7
Elevated fasting glucose (n, %)	51 (20.6)	20 (5.7)	6.8 × 10− 8
Low HDL cholesterol (n, %)	150 (60.7)	145 (41.7)	6.8 × 10− 6
Total energy (kcal/day)	2025.41 ± 648.21	1954.96 ± 631.24	0.1852
Carbohydrate (g/day)	364.72 ± 111.84	351.79 ± 109.69	0.1607
Smoking (n, %)			0.0035
Non-smoker	125 (50.6)	224 (64.4)
Former smoker	55 (22.3)	55 (15.8)
Current smoker	67 (27.1)	69 (19.8)
Alcohol consumption (n, %)			0.0040
Non-drinker	95 (38.5)	176 (50.6)
Former drinker	15 (6.1)	27 (7.8)
Current drinker	137 (55.5)	145 (41.7)
MET (hours/week)	167.51 ± 102.45	178.23 ± 104.43	0.2141

Continuous variables were compared using independent t-tests and are expressed as means ± standard deviations. Categorical variables were compared using chi-square tests and are expressed as numbers and percentages. Abdominal obesity was defined as a waist circumference ≥ 90 cm in men or ≥ 85 cm in women. Elevated blood pressure was defined as systolic blood pressure ≥ 130 mmHg, diastolic blood pressure ≥ 85 mmHg, or current use of antihypertensive medication/a diagnosis of hypertension. Elevated fasting glucose was defined as fasting plasma glucose ≥ 100 mg/dL or current use of antidiabetic medication/a diagnosis of diabetes. Low-HDL cholesterol was defined as HDL cholesterol < 40 mg/dL in men or < 50 mg/dL in women

BMI  body mass index, HDL  high-density lipoprotein, MET  metabolic equivalent of task

Association of AMY CNV with hypertriglyceridemia

To evaluate segment-level probe support for the selected AMY1A/AMY2B CNV region, probe coverage was summarized using the available probe-level corrected ratio files. The target region chr1:104144265–104,220,453 was consistently covered by 24 probes across 1,000 available samples, with a mean probe count of 24.0 ± 0.0 and a mean corrected ratio of 0.0215 ± 0.0555 (Supplementary Table S2).

Table 2 summarizes the results of logistic regression analysis assessing the effect of CNV status in the AMY1A and AMY2B genes on the risk of developing hypertriglyceridemia. After adjustment for age, sex, BMI, smoking status, alcohol consumption, physical activity, total energy intake, medication use, metabolic comorbidities, HOMA-IR, and dietary fat intake, the association between AMY1A/AMY2B duplication and hypertriglyceridemia was attenuated and was no longer statistically significant. Participants with AMY1A/AMY2B duplication showed a marginally higher odds of hypertriglyceridemia than those with diploid copy number, but the association did not reach statistical significance. The adjusted model showed a McFadden pseudo R² of 0.1141, AIC of 746.588, and ROC-AUC of 0.7183.

Table 2.

Association of AMY1A/AMY2B nsv950526 with hypertriglyceridemia based on adjusted odds ratios

nsv950526	Hypertriglyceridemia no. (%)	Non-hypertriglyceridemia no. (%)	OR (95% CI)	p-value
Diploid	186 (40.1)	278 (59.9)	1.00 (Ref)
Duplication	61 (47.7)	67 (52.3)	1.50 (0.98–2.29)	0.0631

CNV status was defined using the segment mean log2 ratio L = log2 (CN/2): diploid − 0.60 ≤ L ≤ 0.40, 1.32 ≤ CN 2.64.; gain 0.40 < L ≤ 0.75, 2.64 < CN ≤ 3.36; amplification L > 0.75, CN > 3.36. Duplication includes gain and amplification. Adjusted for age, sex, smoking, alcohol consumption, MET, BMI, total energy, lipid-lowering medication, antidiabetic medication, antihypertensive medication, thyroid medication, thyroid disease, kidney disease, and diabetes diagnosis, HOMA-IR, and dietary fat intake

Model fit indices for the adjusted logistic regression model were as follows: McFadden pseudo R² = 0.1141, AIC = 746.588, and ROC-AUC = 0.7183. Because only one CNV term was tested in this model, the FDR-adjusted p-value was identical to the nominal p-value

AIC  Akaike information criterion, BMI  body mass index, CI  confidence interval, CNV copy number variation, FDR  false discovery rate, HOMA-IR homeostasis model assessment of insulin resistance, MET  metabolic equivalent task, OR  odds ratio, ROC-AUC  area under the receiver operating characteristic curve

Association of APOA5 SNP with hypertriglyceridemia

Table 3 presents the results of logistic regression analysis assessing the association between APOA5 rs651821 genotypes (TT, TC, and CC) and the risk of hypertriglyceridemia. Logistic regression analysis adjusted for covariates showed significant differences in the risk of developing hypertriglyceridemia according to the genotype. Carriers of the C allele exhibited a significant 2.00-fold increased risk of developing hypertriglyceridemia compared with individuals with the TT genotype (OR 2.00, 95% CI: 1.39–2.88, p = 0.0002).

Table 3.

Association of APOA5 rs651821 with hypertriglyceridemia based on adjusted odds ratios

rs651821	Hypertriglyceridemia no. (%)	Non-hypertriglyceridemia no. (%)	OR (95% CI)	p-value
TT	101 (34.5)	192 (65.5)	1.00 (Ref)
C-carrier	146 (48.8)	153 (51.2)	2.00 (1.39–2.88)	0.0002

C-carrier denotes C-allele carriers (TC/CC); the reference genotype is TT. Adjusted for age, sex, BMI, smoking, alcohol, MET, and total energy, lipid-lowering medication, antidiabetic medication, antihypertensive medication, thyroid medication, thyroid disease, kidney disease, diabetes diagnosis, insulin resistance and dietary fat intake.

BMI  body mass index, CI  confidence interval, MET  metabolic equivalent task, OR  odds ratios

Interaction between AMY1A/AMY2B CNV and APOA5 SNP on the risk of hypertriglyceridemia

The combined associations between AMY1A/AMY2B nsv950526 and APOA5 rs651821 SNP on hypertriglyceridemia risk were evaluated (Fig. 3). Individuals with a diploid copy number and the C allele had a 1.86-fold higher odds of hypertriglyceridemia compared with those with a diploid copy number and the TT genotype (OR 1.86, 95% CI: 1.23–2.81, FDR = 0.0053). The duplication × TT group was not significantly associated with hypertriglyceridemia, whereas individuals with a duplication copy number and the C allele had a 3.11-fold increased risk of hypertriglyceridemia compared with individuals with a diploid copy number and the TT genotype (OR 3.11, 95% CI: 1.69–5.72, FDR = 0.0008). The interaction between CNV and SNP on the risk of hypertriglyceridemia was further evaluated using MI and AP (Supplementary Table S3). Although the combined genotype analysis showed elevated odds in the duplication × C-carrier group, neither MI nor AP was statistically significant.

Fig. 3.

Combined association of AMY CNV and APOA5 rs651821 with hypertriglyceridemia risk. Odds ratios (ORs) with 95% confidence intervals (CIs) for CNV–APOA5 rs651821 interaction on hypertriglyceridemia risk; reference category is diploid × TT. The C allele was modeled dominantly. CNV categories: diploid (−0.60 ≤ L ≤ 0.40), gain (0.40 < L ≤ 0.75), amplification (L > 0.75), where L = log2(CN/2). Adjusted for age, sex, BMI, smoking, alcohol, MET, and total energy, lipid-lowering medication, antidiabetic medication, antihypertensive medication, thyroid medication, thyroid disease, kidney disease, diabetes diagnosis, insulin resistance and dietary fat intake. *FDR <0.05. BMI, body mass index; MET, metabolic equivalent of task; CNV, copy number variation

Subgroup analysis

Figure 4 presents the results of the subgroup analyses according to the carbohydrate intake level, adjusted for the same covariates as in Supplementary Table 3. In the low-carbohydrate group (< median, 73.2% of energy), the diploid × C-carrier group showed higher odds of hypertriglyceridemia compared with the diploid × TT reference group (OR = 2.10; 95% CI: 1.12–3.94; FDR = 0.0421). The duplication × TT group was not significantly associated with hypertriglyceridemia, whereas the duplication × C-carrier group showed the highest odds of hypertriglyceridemia in this subgroup (OR = 4.33; 95% CI: 1.78–10.56; FDR = 0.0076).

Fig. 4.

Forest plot of subgroup analysis stratified by median carbohydrate intake. Subgroup analysis of the 3-way interaction among AMY1A/AMY2B CNV, APOA5 rs651821 genotype, and carbohydrate intake. ORs with 95% CIs shown; reference category is diploid × TT. CNV categories and covariates as in Figure 3. BMI, body mass index; MET, metabolic equivalent of task; CNV, copy number variation. Carbohydrate intake was categorized as low (< 73.2% of total energy) or high (≥ 73.2% of total energy) based on the median value

In the high-carbohydrate group (≥ median, 73.2% of energy), the diploid × C-carrier group was significantly associated with higher odds of hypertriglyceridemia (OR = 2.10; 95% CI: 1.13–3.88; FDR = 0.0421). The duplication × TT group was not significantly associated with hypertriglyceridemia. The duplication × C-carrier group showed an elevated OR, but the association did not remain significant after FDR correction. The 3-way interaction among AMY1A/AMY2B CNV, APOA5 rs651821, and carbohydrate intake was nominally significant but did not remain significant after FDR correction. Therefore, these carbohydrate-stratified findings should be interpreted cautiously as exploratory results.

As a sensitivity analysis, AMY CNV and carbohydrate intake were additionally modeled as continuous variables (Supplementary Table S4). In this model, neither continuous AMY CNV nor carbohydrate intake was significantly associated with hypertriglyceridemia, and the continuous AMY CNV × APOA5 C-carrier × carbohydrate intake interaction was also not significant. APOA5 rs651821 C-carrier status remained significantly associated with hypertriglyceridemia.

Discussion

This study examined the joint associations of AMY1A/AMY2B CNV, APOA5 rs651821, and dietary carbohydrate intake with hypertriglyceridemia in Korean middle-aged adults. After additional adjustment for medication use, metabolic comorbidities, HOMA-IR, and dietary fat intake, AMY1A/AMY2B duplication showed a marginal but non-significant association with hypertriglyceridemia. In contrast, APOA5 rs651821 C-carrier status remained significantly associated with higher odds of hypertriglyceridemia. The combined genotype analysis showed elevated odds among individuals carrying both AMY duplication and the APOA5 C allele, but formal multiplicative and additive CNV–SNP interaction metrics were not statistically significant. Carbohydrate-stratified analyses suggested possible diet-dependent patterns; however, these findings should be interpreted cautiously as exploratory because the three-way interaction did not remain significant after FDR correction.

Although the association between AMY1A/AMY2B duplication and hypertriglyceridemia was attenuated after additional adjustment and did not reach statistical significance, the direction of the association may be interpreted in light of previous studies linking amylase gene copy number to starch digestion and postprandial metabolic responses. Variations in the copy number of the amylase genes AMY1A and AMY2B play a major role in determining starch digestive capacity and early postprandial glucose responses [44]. Individuals with a higher number of AMY gene copies exhibit greater salivary and pancreatic amylase activity, leading to more rapid starch hydrolysis and accelerated glucose absorption [28, 45]. In dietary environments where carbohydrates contribute substantially to the total energy intake, such as in this Korean cohort, enhanced enzymatic digestion may be associated with increased hepatic glucose influx and stimulation of lipogenic pathways by activating sterol regulatory element–binding protein–1c and carbohydrate-responsive element–binding protein, ultimately promoting de novo lipogenesis and VLDL-TG production [11, 46, 47]. Given this mechanistic background, the attenuation observed in our study suggests that the metabolic consequences of high amylase gene copy number might be heavily mediated by insulin resistance or masked by the effects of pharmacological treatments in this middle-aged cohort.

Much of the prior literature has focused on the potential detrimental metabolic effects of low AMY1 copy numbers. A reduced copy number has been associated with diminished salivary amylase activity [28, 45], altered postprandial glucose handling [48, 49], and metabolomic patterns indicating reduced carbohydrate utilization [50]. However, clinical lipid parameters often show no clear association with AMY1 CNV [51, 52], and meta-analyses have indicated inconsistent results [53]. These discrepancies demonstrate that AMY CNV is not inherently detrimental or protective; instead, its metabolic consequences depend on the dietary composition, enzyme activity, and metabolic context [44, 54].

Participants carrying the APOA5 rs651821 C allele exhibited a markedly increased prevalence of hypertriglyceridemia compared with those homozygous for the TT genotype (48.8% vs. 34.5%). After adjustment for confounders, C allele carriers showed a 2.0-fold increased odds of hypertriglyceridemia (OR = 2.00, 95% CI: 1.39–2.88; p = 0.0002) (Table 3). This strong association aligns with the findings from large-scale East Asian studies, in which rs651821 has consistently emerged as the most significant TG-related locus [18, 55]. The APOA4–APOA5–ZNF259–BUD13 cluster plays a fundamental role in TG metabolism, and functional evidence has confirmed that the protective T allele enhances APOA5 expression via stronger GATA4 binding, whereas the C allele reduces transcription and consequently elevates TG concentrations [56].

Ethnic variation in APOA5 effects has been well demonstrated. In Jewish populations, rs651821-linked haplotypes produced divergent TG phenotypes depending on ancestry, with significant associations in the Ashkenazi and Yemenite groups, but not in Sephardic individuals [19]. These population-specific effects highlight the role of the genetic background in modulating APOA5 function. Consistent with prior evidence, our Korean cohort showed a substantially higher prevalence of hypertriglyceridemia among C allele carriers, supporting the relevance of rs651821 to TG regulation in this population.

This may in turn promote hepatic lipogenic pathways — including activation of sterol regulatory element–binding protein-1c and carbohydrate-responsive element–binding protein — potentially stimulating de novo lipogenesis and VLDL-TG secretion, consistent with established mechanistic models, although direct measurement of amylase activity, insulin dynamics, and DNL markers was not performed in this cohort [46, 57, 58]. The APOA5 rs651821 C allele is thought to reduce APOA5 transcription through weakened GATA4 binding [18], which may result in lower ApoA-V concentrations and diminished LPL-mediated TG clearance [18, 59], consistent with established mechanistic models. Under high-carbohydrate dietary conditions, the combination of potentially increased hepatic TG synthesis and reduced TG clearance may act synergistically to elevate plasma TG levels [60]; however, this mechanistic model is inferred from existing experimental evidence and requires confirmation through direct functional assessment in future studies. The integrated effects of AMY1A/AMY2B CNV and APOA5 rs651821 genotypes can be explained by a unified biochemical pathway linking carbohydrate digestion, hepatic TG synthesis, and triglyceride clearance [19, 55]. Higher AMY copy number increases salivary and pancreatic amylase activity, leading to rapid starch hydrolysis and greater postprandial glucose availability [28]. Conversely, the APOA5 rs651821 C allele may reduce APOA5 transcription and ApoA-V-mediated triglyceride clearance, thereby contributing to elevated plasma triglyceride levels [18]. This hypothesis-driven model, based on prior experimental evidence, may explain why the hypertriglyceridemia risk was the highest among individuals carrying both AMY duplication and the APOA5 C allele in the Korean population. The biological mechanisms proposed here represent hypothesis-driven interpretations based on prior literature and are not directly demonstrated in the current study.

In the carbohydrate-stratified analyses, the pattern of combined genotype associations differed by median carbohydrate intake. In the low-carbohydrate group, the duplication × C-carrier group showed the highest odds of hypertriglyceridemia compared with the diploid × TT reference group (OR = 4.33; 95% CI: 1.78–10.56; FDR = 0.0076), whereas the duplication × TT group was not significantly associated with hypertriglyceridemia. In the high-carbohydrate group, the duplication × C-carrier group showed an elevated OR, but the association did not remain significant after FDR correction. The three-way interaction among AMY1A/AMY2B CNV, APOA5 rs651821, and carbohydrate intake was nominally significant but did not remain significant after FDR correction. These subgroup analyses should be interpreted cautiously given the reduced sample sizes within each stratum, which limit statistical power; the wide confidence intervals observed suggest instability of some estimates.

This study had several notable strengths. First, it is one of the few population-based investigations to simultaneously examine CNVs in amylase genes (AMY1A/AMY2B) and functional variations in APOA5 within a unified gene–gene–diet framework. By integrating genetic variations related to carbohydrate digestion with a well-established regulator of triglyceride clearance, the present study provides a biologically coherent model that links carbohydrate intake to plasma TG regulation.

Second, the study was conducted in a Korean population characterized by a traditionally high-carbohydrate intake, offering a nutritionally relevant context in which the metabolic consequences of amylase gene dosage are likely to be the most pronounced. This dietary background enhances the interpretability of the observed gene–diet interactions and strengthens the plausibility of the proposed mechanisms.

Third, the availability of detailed dietary data enabled stratified and interaction analyses that went beyond single-gene associations, allowing us to demonstrate that the effects of AMY CNV and APOA5 rs651821 are not static but are substantially modified by macronutrient composition. Taken together, these strengths support the robustness of the observed associations and underscore the value of incorporating the dietary context into genetic studies of lipid metabolism.

Despite the strengths, this study had several limitations that should be considered when interpreting the findings. First, its cross-sectional design precludes causal inferences. Although the exploratory interaction findings are supported by established biological mechanisms, longitudinal or interventional studies are required to confirm the temporal and causal relationships among AMY CNV, APOA5 genotype, dietary carbohydrate intake, and triglyceride metabolism.

Second, dietary intake was assessed using self-reported measures that are inherently subject to recall bias and measurement error. While such misclassification is likely to be non-differential and would tend to attenuate the observed associations, residual errors in estimating carbohydrate intake may have influenced the magnitude of the detected gene–diet interactions.

Third, direct measurements of amylase enzyme activity, postprandial insulin response, hepatic de novo lipogenesis markers, and lipoprotein lipase activity were not available in this cohort. The mechanistic pathways proposed to link AMY CNV and APOA5 rs651821 to TG metabolism are therefore inferred from established experimental and clinical evidence rather than directly demonstrated in this study. Future studies incorporating these functional biomarkers will be necessary to confirm the proposed biological mechanisms and to establish the clinical relevance of the observed gene–diet interactions in the context of mild-to-moderate hypertriglyceridemia.

Fourth, our genetic analysis was limited to a single APOA5 variant (rs651821). Although rs651821 serves as a well-established functional tag SNP with strong linkage disequilibrium with other APOA5 variants in East Asian populations, formal haplotype analyses incorporating rs662799, rs2266788, and other variants within the APOA5 locus may provide additional mechanistic resolution. Future studies should consider multi-variant haplotype approaches to more comprehensively characterize the contribution of APOA5 to hypertriglyceridemia risk, particularly in the context of gene–diet interactions. Fifth, variations in AMY CNV–TG associations across studies may partially reflect differences in dietary starch exposure. AMY1A copy numbers vary widely across populations as an evolutionary response to long-term starch availability [47, 61]. In modern environments with a high availability of rapidly digestible carbohydrates, individuals with AMY duplications may experience exaggerated hepatic glucose influx and lipogenesis, particularly when coupled with APOA5-mediated defects in TG clearance [58, 62]. This convergence of increased TG synthesis and reduced TG removal may contribute to a metabolic shift toward hypertriglyceridemia.

Finally, the analytic sample was substantially reduced from the original baseline KoGES cohort because the present analysis required complete data on AMY1/AMY2B CNV calls, APOA5 genotype, triglyceride measurements, dietary intake, and covariates. Although the included and excluded participants were generally comparable in observed baseline characteristics, with most SMD values below or close to 0.1, selection bias due to unmeasured characteristics cannot be completely excluded. Therefore, the generalizability of the findings to the entire KoGES baseline population should be interpreted with caution.

Conclusion

In conclusion, this study examined the joint associations of AMY1A/AMY2B CNV, APOA5 rs651821, and dietary carbohydrate intake with hypertriglyceridemia in middle-aged Korean adults. APOA5 rs651821 C-carrier status was consistently associated with higher odds of hypertriglyceridemia, whereas the independent association of AMY1A/AMY2B duplication was attenuated after stringent adjustment for clinical covariates. Notably, individuals carrying both the duplication and the C allele exhibited a substantially increased risk (OR = 3.11). These findings support a biologically coherent model in which AMY1A/AMY2B duplication may be associated with increased carbohydrate digestion and hepatic de novo lipogenesis, whereas the APOA5 C allele may be associated with reduced TG clearance via impaired lipoprotein lipase activation. Although the three-way interaction involving carbohydrate intake did not remain significant after FDR correction, stratified analyses suggested potential diet-dependent modulations, particularly in the low-carbohydrate group. These findings underscore the importance of considering both genomic variation and dietary composition to better understand individual differences in TG metabolism. Future longitudinal and interventional studies incorporating functional biomarkers are warranted to confirm these causal relationships and evaluate the efficacy of personalized dietary strategies based on AMY and APOA5 genotypes.

Abbreviations

AP

Attributable proportion

APOA5

Apolipoprotein A5

BMI

Body mass index

CI

Confidence interval

CNV

Copy number variation

CVD

Cardiovascular disease

HDL

High-density lipoprotein

HOMA-IR

Homeostasis model assessment of insulin resistance

KNHANES

Korea National Health and Nutrition Examination Survey

KoGES

Korean Genome and Epidemiology Study

LPL

Lipoprotein lipase

MET

Metabolic equivalent of task

MetS

Metabolic syndrome

MI

Multiplicative interaction

OR

Odds ratio

SMD

Standardized mean differences

SNP

Single-nucleotide polymorphism

TG

Triglyceride

TRL

Triglyceride-rich lipoprotein

VLDL

Very-low-density lipoprotein

Authors’ contributions

Conceptualization: Shin D; Investigation: Kim M, Ko S-H, Kim S, Shin D; Supervision: Shin D; Writing - original draft: Kim M, Ko S-H, Shin D; Writing - review & editing: Kim M, Ko S-H, Kim S, Shin D; Funding acquisition: Shin D.

Funding

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. RS-2024-00340086).

Data availability

Data used in the manuscript cannot be made publicly available due to ethical restrictions. However, access to the Korean Genome and Epidemiology Study (KoGES) data can be requested by submitting an application to the KoGES administration (https://coda.nih.go.kr).

Declarations

Ethics approval and consent to participate

The study protocol was reviewed and approved by the Institutional Review Board of Ewha Womans University, Republic of Korea (institutional review board number ewha-202603-0023-01), and all participants provided written informed consent.

All participants provided written informed consent prior to participation in the KoGES.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.