Article ID: CJ-24-0491
Background: There are limited data on the use of whole-exome sequencing (WES) to diagnose severe hypertriglyceridemia. Our aim was to identify candidate genes linked to triglyceride levels via a genome-wide association study (GWAS) and to recruit participants with severe hypertriglyceridemia for WES to assess allelic variants in the candidate genes.
Methods and Results: A GWAS was conducted involving 120,140 participants to identify lead loci associated with blood triglyceride levels. Following the identification of these lead loci, WES was performed on DNA samples from 29 participants with hypertriglyceridemia whose triglyceride levels exceeded 800 mg/dL to assess variations in the corresponding genes. In the GWAS of 120,140 participants, the apolipoprotein A5 (APOA5) locus on chromosome 11 showed the strongest association with blood triglyceride levels (lead single nucleotide polymorphism [SNP] rs2075291; P=3.07×10−108), along with 5 independent SNPs (most significant P=7.84×10−167). Other key loci included BUD13 homolog (BUD13; P=2.73×10−62), glucokinase regulator (GCKR; P=2.63×10−24), and lipoprotein lipase (LPL; P=1.50×10−11). WES in 29 hypertriglyceridemia patients identified additional genes, including ALDH1A2, APOC1, LPL, RGS7, and SIK3, showing significant allele frequency variations and potential roles in lipid metabolism.
Conclusions: Our study confirms the role of known genetic loci in triglyceride metabolism and hypertriglyceridemia while uncovering novel loci, offering new perspectives on lipid regulation and potential avenues for therapeutic advancements.
Hypertriglyceridemia was typically defined by plasma triglyceride levels exceeding 150 mg/dL, and affects approximately 10% of the global population diagnosed with dyslipidemia.1,2 Elevated plasma triglyceride levels, or hypertriglyceridemia, are categorized into 3 levels: moderate (200–400 mg/dL), severe (400–800 mg/dL), and very severe (greater than 800 mg/dL).3 Moderately elevated triglyceride levels are commonly linked to lifestyle factors such as diet, obesity, and a lack of regular exercise, escalating the risk of cardiovascular diseases.4 The more severe stages of hypertriglyceridemia are linked to increased cardiovascular risk. These risks are due, in part, to alterations in blood properties, including viscosity and inflammation, which are crucial in understanding the pathophysiology of cardiovascular diseases.5,6 Studies suggest that these changes may affect cardiovascular health, underscoring the importance of managing elevated triglyceride levels to mitigate the risk of cardiovascular disease.7
Moderate hypertriglyceridemia is usually attributed to common genetic variations combined with lifestyle factors, and is less often associated with rare genetic variants. In contrast, severe and very severe hypertriglyceridemia are more likely to be linked to rare genetic mutations and may represent monogenic forms of the disease, particularly when mutations in specific genes such as lipoprotein lipase (LPL),8,9 apolipoprotein C2 (APOC2),10 and apolipoprotein A5 (APOA5) are present.11,12 In these conditions, the high baseline triglyceride levels are due to metabolic defects.
Carrasquilla et al. have reviewed extensive genetic research on hypertriglyceridemia, observing that most studies have used genome-wide association studies (GWAS) or whole-exome sequencing (WES) to identify over 300 genetic loci with significance for triglyceride levels.13 However, these studies primarily focused on individuals of European genetic ancestry. Although WES has revealed numerous genetic loci associated with hypertriglyceridemia levels, comparative genetic research in Asian populations, particularly among Taiwanese cohorts, remains limited. To address this gap, the aim of the present study was to use WES to establish molecular diagnoses for Taiwanese participants clinically diagnosed with severe hypertriglyceridemia. The primary objectives of the study were to: (1) identify candidate genes associated with triglyceride levels through GWAS using data from the Taiwan Biobank; and (2) recruit 30 participants with a clinical history of severe hypertriglyceridemia (defined as triglyceride levels exceeding 800 mg/dL) and perform WES to assess the presence of allelic variants in the candidate genes identified by the GWAS analysis.
Participants At baseline, 189,102 participants were recruited for the study. Genotyping in the Taiwan Biobank was conducted in 2 phases. Initially, participants were genotyped using the TWB1 array, followed by 120,143 later participants who were genotyped using the TWB2 array. All GWAS were performed on blood samples. The TWB1 array, developed by Affymetrix, involved experimental and analytical work performed by the National Center for Genome Medicine and included approximately 653,000 single nucleotide polymorphisms (SNPs). Building on the data from TWBv1.0 and insights from whole-genome sequencing, the TWB2 array was specifically customized for the Taiwanese population, using the Axiom Genome-Wide Array Plate system to include around 750,000 SNPs. Due to differences between the genotyping arrays, combining data from the 2 cohorts is not recommended. Consequently, this study focused on the TWB2 cohort, which provided a larger sample size and more comprehensive genomic data. After excluding 3 individuals lacking blood triglyceride data, 120,140 participants were included in the final analysis.
This study was approved by the Institutional Review Board of MacKay Memorial Hospital, Taipei, Taiwan (22MMHIS412e).
Phenotype Participants were recruited across 23 collection sites throughout Taiwan, where blood and urine samples were obtained, and questionnaire surveys were conducted. In this study, blood triglyceride levels (mg/dL) were used as a continuous variable for GWAS analysis. Sex, age, education, and the first 10 principal components (PC1–10) were included as covariates to adjust for potential confounding factors. To assess the significance of differences in baseline characteristics, participants with triglyceride levels greater than 800 mg/dL were classified as “cases,” whereas the remaining participants were considered “controls.” Demographic characteristics, including sex, age, education, marital status, smoking history, body mass index, waist-to-hip ratio, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, fasting glucose, and HbA1c, were compared between the 2 groups.
Quality of GWAS The quality control process for the GWAS was conducted in 2 parts, focusing on both sample and marker filtering. Initially, 120,161 individuals were included, and samples with a genotyping call rate >98% were retained (n=120,160). After performing sex checks (n=119,936 retained), samples with heterozygosity outside 3 standard deviations were excluded, leaving 119,135 individuals. After removing participants with a relatedness coefficient (kinship) >0.0884, 101,762 individuals were included in the analysis. For marker quality control, the initial set of 16,201,508 SNPs was filtered based on an information score >0.8 (n=11,276,986 retained) and inclusion of only biallelic SNPs (n=10,205,224 retained). Markers with a genotyping call rate >98% were retained (n=9,074,987), followed by the exclusion of those with a minor allele frequency <0.01 (n=5,203,311). Finally, markers failing Hardy-Weinberg equilibrium at P<5.7×10–7 were excluded, resulting in 5,195,776 SNPs for the final analysis.
Data Analysis We conducted genome-wide association analysis using PLINK version 1.9 to examine the relationship between triglyceride levels and SNPs. A linear regression model was applied for each SNP, adjusting for sex, age, education, and PC1–10 to account for population stratification. SNPs with P<5×10−8 were considered genome-wide significant. Subsequent post-GWAS analysis was performed using Functional Mapping and Annotation of Genome-Wide Association Studies (FUMA) version 1.5.2. The FUMA platform was used to identify candidate SNPs, lead SNPs, and independent significant SNPs, and to annotate these variants based on their functional significance. In addition, we compared our findings with previously reported traits in the GWAS Catalog to assess consistency with known genetic associations. Annotation of SNPs included functional consequence prediction and gene mapping based on positional and expression data, enhancing the biological interpretation of the identified genetic loci.
WESParticipants This study was approved by the Institutional Research Board of National Taiwan University Hospital (NTUH-REC no. 202012278RINB). All participants were volunteers recruited from the outpatient clinic of National Taiwan University Hospital in Taipei, Taiwan, and all provided informed consent.
The study included 30 participants with hypertriglyceridemia, aged ≥30 years. The recorded maximum triglyceride levels spanned from January 10, 2022, to February 8, 2023. Participants with triglyceride levels exceeding 800 mg/dL were included in the study because this threshold is indicative of severe refractory hypertriglyceridemia. We excluded participants with secondary hyperlipidemia, including those with thyroid disease, nephrotic syndrome, and severe liver or biliary diseases, as well as those from whom informed consent was not obtained.
Measurement of Blood Lipids Fasting blood samples were collected from participants and triglycerides were measured using colorimetry (Beckman Coulter, Indianapolis, IN, USA) in the National Taiwan University Hospital laboratory, which is accredited by College of American Pathologists and ISO15186. Changes in triglyceride levels between baseline (maximum values) and the lowest values were used as a measure of the effectiveness of drug treatment (see Table 1). A summary of the triglyceride (mg/dL) measurement data, including frequency, mean, minimum, and maximum values, for the 30 participants is presented in Supplementary Table 1.
Demographic Characteristics of the Study Participants
| GWAS (n=120,140) |
WES (n=29) |
|
|---|---|---|
| Sex | ||
| Male | 39,946 (33) | 24 (83) |
| Female | 80,194 (67) | 5 (17) |
| Age (years) | ||
| <55 | 73,478 (61) | 11 (38) |
| 55–65 | 37,600 (31) | 10 (34) |
| >65 | 9,062 (7) | 8 (28) |
| Overweight or obesity | ||
| No | 94,991 (79) | 12 (41) |
| Yes | 25,066 (20) | 17 (59) |
| Marital status | ||
| Single | 19,133 (15) | 2 (7) |
| Married | 84,981 (70) | 16 (55) |
| Divorced | 10,525 (8) | 2 (7) |
| Widowed | 5,447 (4) | 1 (3) |
| Unknown | 54 (0.04) | 8 (28) |
| Education | ||
| Bachelor’s degree | 72,079 (60) | 9 (31) |
| High school diploma or less | 48,038 (39) | 6 (21) |
| Unknown | 23 (0.01) | 14 (48) |
| Current smoking | ||
| No | 88,747 (73) | 16 (55) |
| Yes | 31,347 (26) | 5 (17) |
| Unknown | 46 (0.04) | 8 (28) |
| Current drinking | ||
| No | 113,008 (94) | 14 (48) |
| Yes | 7,040 (5) | 6 (21) |
| Unknown | 92 (0.08) | 9 (31) |
| Medication | ||
| Crestor | NA | 1 (3) |
| Ezetrol | NA | 1 (3) |
| Lipanthyl | NA | 14 (48) |
| Lipitor | NA | 4 (14) |
| Livalo | NA | 2 (7) |
| Lopid | NA | 2 (7) |
Data are presented as n (%). GWAS, genome-wide association study; NA, not assessment; WES, whole-exome sequencing.
Assessment of Other Covariates Participants characteristics were collected from the chart records of the National Taiwan University Hospital and included sex, age, body mass index, and smoking and alcohol drinking habits. Information regarding the medications used for lipid management was obtained from the medical charts. The medications used to lower triglyceride levels included statins, such as Crestor (rosuvastatin), Lipitor (atorvastatin), and Livalo (pitavastatin); ezetimibe (Ezetrol/ezetimibe); and fibrates, such as Lipanthyl (fenofibrate) and Lopid (gemfibrozil).
WES Analysis For WES analysis, 100 ng DNA per sample was used as input material. Libraries for sequencing were prepared using Agilent SureSelect Human All Exon V7+NCV (Agilent Technologies, Santa Clara, CA, USA), Roche KAPA HyperExome+Mitochondria capture (Roche, Santa Clara, CA, USA), or Twist Exome 2.0 (Twist Bioscience, San Francisco, CA, USA). Each sample was assigned an index code to identify and differentiate sequence data. This step ensured accuracy and efficiency as per the manufacturers’ recommendations. DNA fragment size, controlled by an enzymatic shearing system, ranged from 180 to 280 base pairs (bp), impacting sequencing efficiency and data quality. Overhanging sequences were converted to blunt ends using exonuclease and polymerase activities. The 3′ ends were adenylated, and adapter oligonucleotides were attached. DNA fragments with adapters were then amplified via polymerase chain reaction (PCR). Post-PCR, libraries underwent liquid-phase hybridization with biotin-labeled probes and were captured using magnetic beads coated with streptavidin to isolate target genes. Libraries were further amplified through another PCR round and purified using the AMPure XP system (Beckman Coulter, Beverly, CA, USA). Quality assessment was performed with the Qsep400 System (Bioptic Inc., Taipei, Taiwan), and quantification was done using a Qubit 2.0 Fluorometer (Thermo Scientific, Waltham, MA, USA). Sequencing was performed on an Illumina NovaSeq platform, generating 150-bp paired-end reads, facilitated by Genomics BioSci & Tech Co.
WES was performed on the Galaxy platform, an open-source, web-based platform designed for computational biomedical research,14 including quality control, trimming, alignment, postalignment processing, variant calling, and annotation. Pre-adapter trimming quality metrics for FastQ files from a 30-sample dataset are listed in Supplementary Table 2. High-throughput sequencing technology from Illumina was used to generate raw data. Before analysis, sequence reads in the FastQ format were uploaded to Galaxy. The initial quality assessment of the raw reads was performed using FastQC,15 which offers a detailed analysis of various quality metrics, such as base quality score distributions, sequence quality scores, and GC content. Based on this analysis, necessary quality adjustments were identified and implemented. After quality control, reads were processed using Trimmomatic to trim adapters and remove low-quality bases.16 This step ensures that only high-quality, clean reads are used for alignment, thereby minimizing biases and errors. The quality control metrics for postadapter trimming of FastQ files from a 30-sample dataset are presented in Supplementary Table 3. The processed reads were mapped to the human reference genome (GRCh38) using the Burrows-Wheeler Aligner (BWA),17 known for its efficiency in aligning low-divergent sequences to large reference genomes. After alignment, the resulting BAM files were sorted, and duplicates were marked with Picard Tools (http://broadinstitute.github.io/picard/) to avoid overestimating read depth. This step is crucial for accurate variant calling because it ensures subsequent analyses are based on unique sequencing data. Variants were then identified from the processed BAM files using the Genome Analysis Toolkit (GATK),18 which detects single-nucleotide variants and indels by analyzing the alignments. GATK’s robust framework ensures high-confidence variant detection, essential for downstream genetic association studies and diagnostic applications.
Statistical Analysis Descriptive statistics, which include means and standard deviations, were used to examine the distribution of variables such as age, body mass index, and highest triglyceride levels. Fisher’s exact test was used to analyze the significance of differences in minor allele frequencies of single-nucleotide variants between our 30 participants and Asian population data from the National Center for Biotechnology Information (NCBI) dbSNP database (GRCh38).19 Subsequently, 24 single-nucleotide variants were further examined using the Genotype-Tissue Expression (GTEx) project to explore the relationship between potential single-nucleotide variants implicated in severe hypertriglyceridemia and gene expression.20 The normalized effect size (NES) was a statistical measure indicating the extent to which a genotype, such as a particular single-nucleotide variant, influences gene expression. The P value tested the statistical significance of this association; a smaller P value suggests that the relationship between the genotype and gene expression levels is more likely to be genuine rather than occurring by chance. Our focus was primarily on gene expression levels in tissues including the heart (left ventricle), cells (cultured fibroblasts), whole blood, artery, adipose (visceral), artery (aorta), and adipose (subcutaneous). However, when more than one tissue type was relevant, we selected the tissue with the smaller P value.
Gene Network After gene expression analysis, 5 genes were used to construct a gene network via the ZS-REVELEN website (https://revelen.zsservices.com/).21 ZS-REVELEN integrates various types of biological data, including genetics, proteomics, and potentially transcriptomics. Using the ZS-REVELEN algorithm, potential gene interactions can be predicted and, with a relevance score cut-off of 0.01, known gene interactions, co-expression patterns, and other bioinformatics insights explored. This approach helped us identify significant gene interactions among these 9 genes and, using this gene network interaction, to attempt to elucidate the causes of hypertriglyceridemia.
The characteristics of participants in the GWAS (n=120,140) and WES analysis (n=29) are summarized in Table 1. Briefly, 66.8% of the GWAS cohort was female, most (61.2%) participants in the GWAS cohort were aged <55 years, and 20.9% were overweight or obese. In the GWAS cohort, 70.7% were married, 60.0% held a Bachelor’s degree, 73.9% were non-smokers, and 94.1% did not consume alcohol (Table 1).
In the WES cohort, most (83%) participants were male and 38% were aged <55 years (Table 1). Slightly more than half the participants (59%) were overweight or obese, 55% were married, and 31% had a Bachelor’s degree. Notably, 55% of WES participants were non-smokers and 48% abstained from alcohol (Table 1).
GWASThe strongest associations were observed at the APOA5 locus on chromosome 11 (Table 2; Figure 1; Supplementary Figures 1–3), where the lead SNP was rs2075291 (P=3.07×10−108). Of the 5 independent SNPs at this locus, the most significant reached a P value of 7.84×10−167. Similarly, BUD13 homolog (BUD13) on chromosome 11 exhibited strong associations with blood triglyceride levels, with the lead SNP rs180326 having a P value of 2.73×10−62 and there being a total of 14 independent SNPs. Other notable loci included glucokinase regulator (GCKR) on chromosome 2 (P=2.63×10−24) and LPL on chromosome 8 (P=1.50×10−11), both showing multiple independent signals.
Key Gene Loci Associated With Blood Triglyceride Levels in the Genome-Wide Association Study
| Gene | Chr | Start | End | Lead SNP | Independent SNPs | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| rs ID | Position | Beta | SE | P value | No. SNPs |
rs ID | Minimum P value |
||||
| RGS7 | 1 | 240931554 | 241520530 | rs144080505 | 241155782 | 10.61 | 1.76 | 1.75E-09 | 2 | rs144080505; rs116354677 | 1.75E-09 |
| GCKR | 2 | 27719709 | 27746554 | rs1260326 | 27730940 | 8.18 | 0.80 | 2.63E-24 | 5 | rs1260326; rs780096; rs1919128; rs3749147; rs6719175 | 2.63E-24 |
| MLXIPL | 7 | 73007524 | 73038873 | rs3812316 | 73020337 | −9.13 | 1.42 | 1.35E-10 | 6 | rs3812316; rs77048596; rs35493868; rs7805504; rs11974409; rs145099458 |
1.35E-10 |
| LPL | 8 | 19759228 | 19824769 | rs308 | 19817476 | −10.05 | 1.49 | 1.50E-11 | 9 | rs308; rs145391587; rs17410962; rs9644638; rs6983430; rs115849089; rs4128744; rs149764042; rs264 |
3.63E-16 |
| GFRA2 | 8 | 21547915 | 21669869 | rs6587005 | 21614564 | 14.16 | 2.28 | 5.41E-10 | 1 | rs6587005 | 5.41E-10 |
| INSL6 | 9 | 5131979 | 5185668 | rs138749195 | 5140707 | 23.90 | 3.90 | 9.35E-10 | 1 | rs138749195 | 9.35E-10 |
| FADS2 | 11 | 61560452 | 61634826 | rs174562 | 61585144 | −5.15 | 0.82 | 3.98E-10 | 4 | rs174562; rs174584; rs174594; rs509360 | 3.98E-10 |
| BUD13 | 11 | 116618886 | 116643704 | rs180326 | 116624703 | 16.42 | 0.98 | 2.73E-62 | 14 | rs180326; rs964184; rs603446; rs3741298; rs595049; rs10892037; rs1240776; rs1787689; rs180377; rs180373; rs180371; rs61905084; rs180349; rs180327 |
7.15E-104 |
| APOA5 | 11 | 116660083 | 116663136 | rs2075291 | 116661392 | 35.13 | 1.58 | 3.07E-108 | 5 | rs2075291; rs149121905; rs75476300; rs76240380; rs182500870 |
7.84E-167 |
| APOA1 | 11 | 116706467 | 116708666 | rs7116797 | 116707338 | 8.08 | 0.85 | 2.13E-21 | 16 | rs7116797; rs61907566; rs7396061; rs1076485; rs7111854; rs1815786; rs7130297; rs4938343; rs641620; rs74830; rs1263172; rs5104; rs595049; rs10892037; rs645901; rs5128 |
7.23E-34 |
| SIK3 | 11 | 116714118 | 116969153 | rs199986368 | 116902343 | 29.05 | 3.25 | 3.79E-19 | 1 | rs199986368 | 2.83E-28 |
| PAFAH1B2 | 11 | 117014983 | 117047610 | rs117592676 | 117017710 | 24.94 | 2.86 | 3.13E-18 | 3 | rs117592676; rs7111854; rs111530588 | 3.13E-18 |
| ALDH1A2 | 15 | 58245622 | 58790065 | rs1532085 | 58683366 | 4.62 | 0.81 | 1.22E-08 | 1 | rs1532085 | 1.22E-08 |
| APOC1 | 19 | 45417504 | 45422606 | rs12721046 | 45421254 | 7.80 | 1.33 | 4.56E-09 | 3 | rs12721046; rs56131196; rs75627662 | 1.76E-21 |
ALDH1A2, aldehyde dehydrogenase 1 family member A2; APOA1, apolipoprotein A1; APOA5, apolipoprotein A5; APOC1, apolipoprotein C1; Beta, effect size; BUD13, BUD13 homolog; Chr, chromosome; FADS2, fatty acid desaturase 2; GCKR, glucokinase regulator; GFRA2, GDNF family receptor alpha 2; INSL6, insulin like 6; LPL, lipoprotein lipase; MLXIPL, MLX interacting protein like; PAFAH1B2, platelet activating factor acetylhydrolase 1b catalytic subunit 2; RGS7, regulator of G protein signaling 7; SIK3, SIK family kinase 3; SNP, single-nucleotide polymorphism.
Manhattan plot of genome-wide association study on blood triglyceride levels in the Taiwanese population.
WES
Our WES analysis of 29 participants with severe hypertriglyceridemia identified several genetic variants with allele frequencies >50% (Table 3), which were subsequently compared to allele frequencies in the broader Asian population. Significant differences were observed at several loci. For example, GCKR (lead SNP rs1260326, P=2.63×10−24) and APOA5 (lead SNP rs2075291, P=3.07×10−108) showed distinct allele distributions between the patient cohort and the Asian population. The MLX interacting protein like (MLXIPL) and LPL loci also exhibited notable allele frequency variations, suggesting that these genetic variants may play a unique role in triglyceride metabolism in this specific population. It should be noted that loci were excluded if no data were available from the NCBI dbSNP database or if the sample size did not meet the 1 : 4 ratio (<120). Detailed information is provided in Supplementary Table 4.
Comparative Whole-Exome Sequencing Analysis to Identify Distinct Genetic Variants in Severe Hypertriglyceridemia: Taiwanese Cohort vs. Asian Population Allele Frequencies
| Chr | Gene | Start | End | Whole-exome sequencing |
Asian population | P for difference |
|||
|---|---|---|---|---|---|---|---|---|---|
| Position | Allele frequency |
Sample size |
Reference allele |
Alternative allele |
|||||
| 1 | RGS7 | 240931554 | 241520530 | 241355798 | T:29 | 168 | T=0.006 | A=0.994 | 7.85×10−34 |
| 1 | RGS7 | 240931554 | 241520530 | 241503944 | G:19 | 632 | G=0.413 | T=0.587 | 0.01 |
| 8 | LPL | 19759228 | 19824769 | 19820404 | G:16 | 192 | G=0.682 | A=0.318 | 0.20 |
| 8 | LPL | 19759228 | 19824769 | 19823468 | C:21 | 184 | C=0.332 | G=0.000, T=0.668 | 1.31×10−4 |
| 11 | SIK3 | 116714118 | 116969153 | 116760991 | T:25 | 216 | T=0.236 | A=0.000, C=0.764 | 5.27×10−11 |
| 11 | SIK3 | 116714118 | 116969153 | 116779090 | G:16 | 166 | G=0.976 | A=0.024, T=0.000 | 5.00×10−10 |
| 11 | SIK3 | 116714118 | 116969153 | 116786845 | C:23 | 522 | C=0.425 | T=0.575 | 1.51×10−4 |
| 11 | SIK3 | 116714118 | 116969153 | 116786951 | C:16 | 172 | C=0.971 | T=0.029 | 1.31×10−9 |
| 11 | SIK3 | 116714118 | 116969153 | 116789970 | G:22 | 6,470 | G=0.2198 | A=0.7802 | 1.13×10−9 |
| 11 | SIK3 | 116714118 | 116969153 | 116790676 | C:16 | 6,802 | C=0.9359 | A=0.0641, T=0.0000 | 9.28×10−9 |
| 11 | SIK3 | 116714118 | 116969153 | 116791110 | T:25 | 206 | T=0.248 | C=0.752 | 2.19×10−9 |
| 11 | SIK3 | 116714118 | 116969153 | 116821618 | C:23 | 254 | C=0.315 | T=0.685 | 1.63×10−6 |
| 11 | SIK3 | 116714118 | 116969153 | 116822638 | G:18 | 150 | G=0.880 | A=0.120, T=0.000 | 0.001 |
| 11 | SIK3 | 116714118 | 116969153 | 116832924 | G:26 | 168 | G=0.280 | C=0.720 | 3.54×10−10 |
| 11 | SIK3 | 116714118 | 116969153 | 116836867 | A:15 | 150 | A=0.320 | G=0.680 | 0.05 |
| 11 | SIK3 | 116714118 | 116969153 | 116836968 | A:26 | 574 | A=0.329 | C=0.000, G=0.671 | 8.75×10−10 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58542542 | G:18 | 3,880 | G=0.8015 | T=0.1985, A=0.0000 | 0.01 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58545711 | T:29 | 180 | T=0.000 | C=1.000 | 2.20×10−16 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58545758 | A:18 | 3,770 | A=0.5700 | G=0.4300, T=0.0000 | 0.58 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58545811 | A:27 | 6,330 | A=0.1485 | C=0.0000, G=0.8515 | 1.14×10−31 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58560880 | C:29 | 6,906 | C=0.0659 | A=0.9341, T=0.0000 | 2.18×10−86 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58560910 | A:29 | 3,886 | A=0.0697 | C=0.0000, G=0.9303 | 1.37×10−78 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58610570 | A:21 | 700 | A=0.447 | C=0.553 | 0.003 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58610884 | A:19 | 190 | A=0.632 | G=0.368, T=0.000 | 0.80 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58612038 | C:17 | 180 | C=0.656 | T=0.344 | 0.46 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58646237 | T:19 | 180 | T=0.622 | C=0.378 | 0.73 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58692965 | T:29 | 178 | T=0.163 | G=0.837, C=0.000 | 1.32×10−20 |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58788004 | A:16 | 214 | A=0.734 | G=0.266 | 0.04 |
| 19 | APOC1 | 45417504 | 45422606 | 45419065 | G:19 | 150 | G=0.693 | A=0.000, T=0.307 | 0.68 |
| 19 | APOC1 | 45417504 | 45422606 | 45420238 | T:16 | 438 | T=0.660 | C=0.340 | 0.23 |
| 19 | APOC1 | 45417504 | 45422606 | 45420395 | A:28 | 6,968 | A=0.2448 | G=0.7552 | 2.93×10−19 |
Originally, 53 loci were included in this table. However, loci were excluded if no data were available from the National Center for Biotechnology Information (NCBI) dbSNP database or if the sample size did not meet the 1 : 4 ratio (<120). For detailed information, see Supplementary Table 4. Abbreviations as in Table 2.
Expression Quantitative Trait Loci
Following the identification of significant loci linked to severe hypertriglyceridemia (Table 4), expression quantitative trait loci (eQTL) analysis was conducted on specific SNPs. Several SNPs were associated with gene expression in various tissues. For example, regulator of G protein signaling 7 (RGS7; rs4659599) showed a negative NES of −0.60 with a highly significant P value of 1.5×10−5 in cultured fibroblasts. Similarly, multiple SNPs within SIK family kinase 3 (SIK3) showed strong associations with gene expression in subcutaneous adipose tissue, including rs3825041 (NES=−0.77, P=1.5×10−11) and rs2072560 (NES=−0.84, P=1.7×10−13). Conversely, several SNPs, such as rs75198898, rs3741297, and rs2075291, did not yield corresponding eQTL results, indicating a lack of available data for those variants.
Gene Expression Profiles of Key Genetic Variants Linked to Severe Hypertriglyceridemia
| Chr | Gene | Start | End | Whole-exome sequencing |
Expression quantitative trait loci | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Position | Allele | SNP | Allele | Gene | NES | P value | Tissue | ||||
| 1 | RGS7 | 240931554 | 241520530 | 241355798 | T:29 | rs4659599 | T/A | RGS7 | −0.60 | 1.5e-05 | Cells: Cultured fibroblasts |
| 11 | SIK3 | 116714118 | 116969153 | 116760991 | T:25 | rs3825041 | T/C | RP11 | −0.77 | 1.5e-11 | Adipose: Subcutaneous |
| 11 | SIK3 | 116714118 | 116969153 | 116779090 | G:16 | rs75198898 | NA | NA | NA | NA | NA |
| 11 | SIK3 | 116714118 | 116969153 | 116786951 | C:16 | rs3741297 | NA | NA | NA | NA | NA |
| 11 | SIK3 | 116714118 | 116969153 | 116789970 | G:22 | rs2266788 | G/A | RP11 | −0.79 | 1.6e-12 | Adipose: Subcutaneous |
| 11 | SIK3 | 116714118 | 116969153 | 116790676 | C:16 | rs2075291 | NA | NA | NA | NA | NA |
| 11 | SIK3 | 116714118 | 116969153 | 116791110 | T:25 | rs2072560 | T/C | RP11 | −0.84 | 1.7e-13 | Adipose: Subcutaneous |
| 11 | SIK3 | 116714118 | 116969153 | 116832924 | G:26 | rs5128 | G/C | RP11 | −0.79 | 1.9e-22 | Muscle: Skeletal |
| 11 | SIK3 | 116714118 | 116969153 | 116836968 | A:26 | rs2070665 | A/G | RP11 | −0.89 | 1.1e-26 | Muscle: Skeletal |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58545711 | T:29 | rs2414592 | NA | NA | NA | NA | NA |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58545811 | A:27 | rs6083 | A/G | MINDY2 | −0.19 | 1.5e-11 | Artery: Tibial |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58560880 | C:29 | rs3829462 | NA | NA | NA | NA | NA |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58560910 | A:29 | rs3829461 | NA | NA | NA | NA | NA |
| 15 | ALDH1A2 | 58245622 | 58790065 | 58692965 | T:29 | rs4774310 | T/G | MINDY2 | −0.20 | 2.9e-12 | Artery: Tibial |
| 19 | APOC1 | 45417504 | 45422606 | 45420395 | A:28 | rs11615 | A/G | CD3EAP | −0.42 | 3.9e-39 | Muscle: Skeletal |
NA, not available; NES, normalized effect size. Other abbreviations as in Table 2.
Gene Network Analysis
The gene network analysis incorporating 5 key genes, namely aldehyde dehydrogenase 1 family member A2 (ALDH1A2), apolipoprotein C1 (APOC1), LPL, RGS7, and SIK3, demonstrated significant interconnections with various other genes implicated in lipid metabolism and regulation (Figure 2). ALDH1A2, involved in retinal dehydrogenase activity, interacted with APOC1, which may play a crucial role in lipoprotein metabolism. In addition, LPL was connected to multiple apolipoproteins, such as APOA1 and apolipoprotein B (APOB), highlighting its central role in lipid hydrolysis. The presence of RGS7 within this network suggests its potential involvement in G-protein signaling pathways related to lipid metabolism, whereas SIK3 was linked to several kinases that regulate metabolic processes.
Gene interaction network analysis.
This study identified key genetic loci associated with severe hypertriglyceridemia, highlighting the roles of APOA5, BUD13, RGS7, GCKR, and LPL in lipid metabolism. Our WES analysis further uncovered additional genes, including ALDH1A2, APOC1, RGS7, and SIK3. WES analysis revealed distinct allele frequency patterns for these genes compared with broader populations, suggesting their unique contribution to triglyceride metabolism in individuals with hypertriglyceridemia. eQTL analysis also demonstrated that several of these genes influence gene expression in specific tissues, such as adipose and muscle, which may contribute to the pathophysiology of lipid disorders. Gene network analysis emphasized the interconnections among these genes and their roles in metabolic regulation, offering insights into their collective impact on hypertriglyceridemia and their potential as therapeutic targets.
Previous studies have linked severe hypertriglyceridemia to rare mutations in LPL,8,9APOC2,10 and APOA5,11,12 and our findings reinforce the significance of these loci, especially APOA5, in lipid metabolism. WES revealed allele frequency differences in loci like GCKR, LPL, and MLXIPL, indicating greater genetic complexity beyond monogenic mutations. Furthermore, eQTL analysis and gene network results point to broader regulatory mechanisms in triglyceride metabolism. Notably, several lead SNPs and genes identified in this study have also been associated with triglyceride levels in previous studies (Supplementary Table 5), highlighting both common and novel genetic contributors. For example, the GCKR gene, linked to the lead SNP rs1260326, has previously been associated with various lipid traits, such as triglyceride levels, cholesterol, and polyunsaturated fatty acids, as well as metabolic syndrome and dyslipidemia. Similarly, the MLXIPL gene, associated with rs3812316, has been linked to triglyceride levels, waist-to-hip ratio (adjusted for body mass index), and other lipid-related traits, including phosphatidylcholine and triacylglycerol. The LPL gene (rs308) has also been extensively reported to be associated with triglyceride and HDL cholesterol levels. Although these genes and SNPs have been previously associated with traits relevant to lipid metabolism, our findings further confirm and expand on the genetic loci involved in triglyceride metabolism, particularly in the context of hypertriglyceridemia. In addition, genes such as RGS7, GDNF family receptor alpha 2 (GFRA2), insulin like 6 (INSL6), SIK3, and platelet activating factor acetylhydrolase 1b catalytic subunit 2 (PAFAH1B2), associated with lead SNPs in our study, have not been widely reported in the GWAS catalog for lipid traits, offering potential new insights into genetic contributors to hypertriglyceridemia. Other independent SNPs and genes from our GWAS, compared with previous studies, can be found in the Supplementary Files (Excel).
In our study, we performed WES for 29 participants with triglyceride levels exceeding 800 mg/dL. WES can detect all exonic regions of the genome, whereas GWAS only identify known loci. Rare variants, which may have significant effects in specific populations, are often undetectable by GWAS. WES provides more detailed data, revealing different types of gene mutations, such as missense mutations, deletions, or insertions, and thereby offering deeper insights into gene function and impact. In addition, WES provides more information for personalized medicine by identifying individual genetic risk factors, leading to more precise methods for disease prevention, diagnosis, and treatment.
However, our WES study for participants with hypertriglyceridemia has several limitations. First, the relatively small sample size may limit the generalizability of our findings and our ability to detect associations with lower-frequency genetic variants. Although we used Fisher’s exact test to analyze minor allele frequencies, we acknowledge that this method does not account for clinical variables. To address this, we secured individual-level data from the Taiwan Biobank, allowing us to perform regression analyses that consider such factors, thereby enhancing the validity of our association analysis. These results require replication and validation in larger, more diverse populations to confirm their relevance to triglyceride levels specifically. In the present study, the focus on genetic factors may have overlooked the potential contributions of environmental, lifestyle, and epigenetic factors, which also significantly affect triglyceride levels and overall lipid metabolism. Future studies should integrate these factors to provide a more comprehensive understanding of the determinants of hypertriglyceridemia. Although our study identified several genetic variants and candidate genes associated with hypertriglyceridemia, we did not establish a predictive model for molecular diagnosis based on these findings. Future studies should aim to develop predictive models using these identified genes to enhance molecular diagnosis for patients with hypertriglyceridemia.
In conclusion, this study identified novel genetic loci and highlighted distinct allele frequency patterns contributing to triglyceride metabolism in severe hypertriglyceridemia, providing insights into potential therapeutic targets for lipid disorders. The findings emphasize the roles of tissue-specific gene expression in the pathophysiology of lipid metabolism.
During the preparation of this manuscript, the authors used ChatGPT to assist with grammar checking and sentence revision.
This research was funded by the Ministry of Science and Technology of Taiwan (Grant no. MOST 109-2314-B-002-203).
The authors declare no conflicts of interest.
H.-Y.F. contributed to the study design, data analysis, interpretation of results, and manuscript writing. M.-C.T. contributed to quality analysis, statistical analysis, and interpretation of results. C.-J.L. assisted with data analysis, contributed to the creation of figures, and performed analyses using Genotype-Tissue Expression (GTEx) to assess gene expression levels across various tissues. C.-L.Y. supplied the software to draw the figures, helped create the figures, contributed to the interpretation of results, discussed mechanisms, and used GTEx for analysis of gene expression in potential tissues. P.-J.L. and H.-Y.H. supported the analysis of GWAS and facilitated the download of the Taiwan Biobank reference genome. H.-C.H., T.-C.S., H.-J.L., and Y.-F.L. provided support for data analysis, interpretation of results, and discussion of mechanisms. T.-P.L. played a role in research design, data quality analysis, and statistical analysis. K.-L.C. was responsible for sample collection, project supervision, study design, and manuscript revision.
This study was approved by the Institutional Research Board of National Taiwan University Hospital (No. 202012278RINB) and by the Institutional Review Board of MacKay Memorial Hospital (No. 22MMHIS412e). All procedures followed the ethical standards of the responsible committees on human experimentation and the Declaration of Helsinki.
The data are managed by the laboratory of K.-L.C. Access to the restricted-use dataset requires approval from the laboratory’s Principal Investigator (K.-L.C.).
Please find supplementary file(s);
https://doi.org/10.1253/circj.CJ-24-0491
