Genetic Variants at the APOE Locus Predict Cardiometabolic Traits and Metabolic Syndrome: A Taiwan Biobank Study

Several apolipoprotein genes are located at the APOE locus on chromosome 19q13.32. This study explored the genetic determinants of cardiometabolic traits and metabolic syndrome at the APOE locus in a Taiwanese population. A total of 81,387 Taiwan Biobank (TWB) participants were enrolled to undergo genotype–phenotype analysis using data from the Axiom Genome-Wide CHB arrays. Regional association analysis with conditional analysis revealed lead single-nucleotide variations (SNVs) at the APOE locus: APOE rs7412 and rs429358 for total, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) cholesterol levels; CLPTM1 rs3786505 and rs11672748 for LDL and HDL cholesterol levels; and APOC1 rs438811 and APOE-APOC1 rs439401 for serum triglyceride levels. Genotype–phenotype association analysis revealed a significant association of rs429358 and rs438811 with metabolic syndrome and of rs7412, rs438811, and rs439401 with serum albumin levels (p < 0.0015). Stepwise regression analysis indicated that CLPTM1 variants were independently associated with LDL and HDL cholesterol levels (p = 3.10 × 10−15 for rs3786505 and p = 1.48 × 10−15 for rs11672748, respectively). APOE rs429358 and APOC1 rs438811 were also independently associated with metabolic syndrome (p = 2.29 × 10−14) and serum albumin levels (p = 3.80 × 10−6), respectively. In conclusion, in addition to APOE variants, CLPTM1 is a novel candidate locus for LDL and HDL cholesterol levels at the APOE gene region in Taiwan. Our data also indicated that APOE and APOC1 variants were independently associated with metabolic syndrome and serum albumin levels, respectively. These results revealed the crucial role of genetic variants at the APOE locus in predicting cardiometabolic traits and metabolic syndrome.


Introduction
The APOE gene is located within the APOE-C1-C4-C2 gene cluster on chromosome 19q13.32, which encodes four amphipathic apolipoproteins and two hepatic control regions that regulate the hepatic expression of these genes [1,2]. Several other genes, such as TOMM40 and CLPTM1, that are not directly associated with lipid metabolism are also located close to APOE. This genomic locus is characterized by strong linkage disequilibrium (LD) between different polymorphic genotypes in this region, which tend to be coinherited faithfully. Thus, genome-wide association studies (GWASs) have indicated that most of The TWB is a population-based research consortium. The study population is comprised of 107,494 TWB Han Chinese participants who had no history of cancer and were recruited between 2008 and 2020 from centers across Taiwan. We excluded 26,107 participants because of (1) fasting for <6 h (2862 participants), (2) no imputation data (12,289 participants), and (3) quality control (QC) for the GWAS with identity by descent PI_HAT > 0.187 (10,956 participants). Figure 1 depicts the flowchart of participant enrollment. We also excluded participants with a history of gout, hypertension, hyperlipidemia, and diabetes mellitus as indicated by their serum uric acid level, blood pressure status, and lipid and glucose metabolism parameters, respectively. Definitions of hypertension, diabetes mellitus, obesity, hyperlipidemia, and current smoking are provided in Supplementary Method 1. This study was approved by the Research Ethics Committee of Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation (approval number: 05-X04-007), and the Ethics and Governance Council of the Taiwan Biobank (approval numbers: TWBR10507-02 and TWBR10611-03). Each participant signed an informed consent form before participating in the study.
ple genes being situated at chromosome 19q13.32, we investigated the genetic determinants of cardiometabolic traits and metabolic syndrome by using regional association analysis in TWB participants.

TWB Participants
The TWB is a population-based research consortium. The study population is comprised of 107,494 TWB Han Chinese participants who had no history of cancer and were recruited between 2008 and 2020 from centers across Taiwan. We excluded 26,107 participants because of (1) fasting for <6 h (2862 participants), (2) no imputation data (12,289 participants), and (3) quality control (QC) for the GWAS with identity by descent PI_HAT > 0.187 (10,956 participants). Figure 1 depicts the flowchart of participant enrollment. We also excluded participants with a history of gout, hypertension, hyperlipidemia, and diabetes mellitus as indicated by their serum uric acid level, blood pressure status, and lipid and glucose metabolism parameters, respectively. Definitions of hypertension, diabetes mellitus, obesity, hyperlipidemia, and current smoking are provided in Supplementary Method 1. This study was approved by the Research Ethics Committee of Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation (approval number: 05-X04-007), and the Ethics and Governance Council of the Taiwan Biobank (approval numbers: TWBR10507-02 and TWBR10611-03). Each participant signed an informed consent form before participating in the study.

Genomic DNA Extraction and Genotyping
By using a PerkinElmer Chemagic 360 instrument (PerkinElmer, Waltham, MA, USA), genomic DNA was extracted after blood sampling. Single-nucleotide variation (SNV; formerly referred to as "single-nucleotide polymorphism") genotyping was performed using custom TWB chips on the Axiom Genome-Wide Array Plate System (Affymetrix, Santa Clara, CA, USA).

Clinical Phenotypes and Laboratory Examinations
Clinical phenotypes used for the study were body mass index (BMI), waist-hip ratio, waist circumference, and systolic, diastolic, and mean blood pressures. We also collected the following biochemical data: fasting plasma glucose and glycated hemoglobin (HbA1c) levels for glucose metabolism; total, LDL, and HDL cholesterol and triglyceride levels for lipid profile; and aspartate aminotransferase (AST), alanine aminotransferase (ALT), γ-glutamyl transferase (γ-GT), albumin, total bilirubin, serum creatinine, and uric acid levels for liver and renal functional tests. BMI and estimated glomerular filtration rate (eGFR) were calculated as reported previously [34]. White and red blood cell counts, platelet counts, hematocrit, and hemoglobin levels were used for the hematological analysis. Because urine creatinine levels were unavailable, only the spot urine albumin level was used to evaluate albuminuria.

Regional Association Analysis with Conditional Analysis
To determine which variant was the lead SNV around the APOE gene region for lipid profile, we conducted a regional association analysis by including TWB participants enrolled after the exclusion criteria were applied, as reported previously [34,35]. In brief, QC was performed for GWAS using the Axiom Genome-Wide CHB 1 and CHB 2 Array Plates (Affymetrix), each comprising 611,656 and 640,160 SNVs, respectively. Genome-wide genotype imputation was performed using SHAPEIT and IMPUTE2, and the East Asian population from the 1000 Genome Project Phase 3 study was used as the reference panel. With imputation, QC was then performed by filtering SNVs with an IMPUTE2 imputation quality score of >0.3. All samples enrolled for the analysis had a call rate of ≥97%. The indels were removed using VCFtools. For SNV QC, an SNV call rate of <97%, a minor allele frequency of <0.01, and the violation of the Hardy-Weinberg equilibrium (p < 10 −6 ) were the criteria for exclusion from subsequent analyses. After QC, in the APOE region, 325 SNVs at positions between 45.2 and 45.6 Mb on chromosome 19q13.32 were enrolled for illustration and analysis.

Statistical Analysis
By using the Kolmogorov-Smirnov test, continuous variables were tested for normal distribution. All variables were not compatible with normal distribution and were expressed as medians and interquartile ranges. Categorical data were presented as percentages. Lipid profile and urine albumin levels were logarithmically transformed to achieve adherence to a normality assumption before being examined using analysis of variance and regression. A general linear regression was used to analyze the association between the studied phenotypes and investigated genotypes after adjustment for age, sex, BMI, and current smoking status. To evaluate the independent effect of the investigated genotypes on the risk of lifestyle and atherosclerotic risk factors, we used multiple logistic regression analysis. Stepwise multivariable linear regression was then performed to determine the independent correlates of lipid profile and serum albumin levels. Regional association analysis was performed using the PLINK software package. Genome-wide significance was defined as p < 5 × 10 −8 . In genotype-phenotype association analysis, the Bonferroni method was used to correct for multiple comparisons where applicable on the basis of the 33 traits analyzed (0.05/33 = 0.0015). LDmatrix software (https://analysistools.nci.nih.gov/LDlink/?tab=ldmatrix, accessed on 12 January 2021) was used to analyze LD. All statistical analyses were performed using SPSS (version 22; SPSS, Chicago, IL, USA).

Baseline Characteristics
The TWB participants' clinical, demographic, and laboratory data stratified by sex are presented in Table 1. Compared with female participants, male participants had significantly higher BMI, waist circumference, and waist-to-hip ratio; systolic, diastolic, and mean blood pressures; circulating levels of LDL cholesterol, triglyceride, fasting plasma glucose; uric acid and creatinine; liver injury and function parameters of AST, ALT, γ-GT, albumin, and total bilirubin levels; and hematocrit, red blood cell count, and hemoglobin (all p < 0.0001). By contrast, total cholesterol and HDL cholesterol levels, eGFR, albuminuria, and platelet counts (all p < 0.0001) were higher in women than in men.

Genotype-Phenotype Association Analysis for the APOE Locus Lead SNVs
We tested the association of APOE locus lead SNVs with clinical, metabolic, and biochemical phenotypes; hematological parameters; lifestyle habits; and risk factors for atherosclerosis (Supplementary Tables S1-S6). For the lipid profile, after adjustment for age, sex, current smoking status, and BMI, genome-wide significant associations were noted between rs7412, rs429358, and rs439401 and all four lipid traits, between rs438811 and total and LDL cholesterol and triglyceride levels, and between rs11672748 and rs3786505 and HDL and LDL cholesterol levels. Genome-wide significant associations were also noted between rs429358 and rs438811 and the risk of metabolic syndrome. With Bonferroni correction, we also observed a significant association (p < 0.0015) between rs7412, rs438811, and rs439401 and serum albumin levels, between rs429358 and eGFR and ALT levels, between rs438811 and waist-hip ratio and ALT levels, and between rs11672748 and rs3786505 and erythrocyte counts.

Genotype-Phenotype Association Analysis for the APOE Locus Lead SNVs
We tested the association of APOE locus lead SNVs with clinical, metabolic, and biochemical phenotypes; hematological parameters; lifestyle habits; and risk factors for atherosclerosis (Supplementary Tables S1-S6). For the lipid profile, after adjustment for age, sex, current smoking status, and BMI, genome-wide significant associations were noted between rs7412, rs429358, and rs439401 and all four lipid traits, between rs438811 and total and LDL cholesterol and triglyceride levels, and between rs11672748 and rs3786505 and HDL and LDL cholesterol levels. Genome-wide significant associations were also noted between rs429358 and rs438811 and the risk of metabolic syndrome. With Bonferroni correction, we also observed a significant association (p < 0.0015) between rs7412, rs438811, and rs439401 and serum albumin levels, between rs429358 and eGFR and ALT levels, between rs438811 and waist-hip ratio and ALT levels, and between rs11672748 and rs3786505 and erythrocyte counts.

Stepwise Linear Regression Analysis between the APOE Locus-Lead SNVs and Lipid Profile
Stepwise linear regression analyses using age, sex, BMI, current smoking status, and APOE locus lead SNVs revealed that these variants together contributed to 2.39%, 8.79%, 0.41%, and 0.47% of the total variation in serum total, LDL, and HDL cholesterol and triglyceride levels, respectively (Table 3). Moreover, the APOC1 rs438811 variant was independently associated with serum albumin levels (Supplementary Table S7). Stepwise logistic regression analysis demonstrated that APOE rs429358 is independently associated with metabolic syndrome (odds ratio: 1.20 for each C-allele, 95% confidence interval: 1.15-1.26, p = 2.29 × 10 −14 ; Table 4).

Discussion
We investigated the associations between genetic variants around the APOE locus, cardiometabolic traits, and metabolic syndrome in a Taiwanese population. We demonstrated that APOE ε2, ε3, and ε4 variants, defined by the APOE rs7412 and APOE rs429358 genotypes, were the strongest genetic determinants in the APOE locus of total, LDL, and HDL cholesterol levels in Taiwan. Both APOE and CLPTM1 variants were independently associated with LDL and HDL cholesterol levels, and the genetic variants between the APOE and APOC1 gene region, including rs429358 and rs438811, were lead SNVs for se-rum triglyceride levels. Our data also provide evidence for the association of APOE locus variants with metabolic syndrome and serum albumin levels. To the best of our knowledge, this is the first report revealing CLPTM1 as a novel candidate locus for LDL and HDL cholesterol levels independent of the APOE variants in an Asian population. Both CLPTM1 variants had weak LD with the lead SNVs in the APOE and APOC1 genes. APOE is a fascinating multifunctional apolipoprotein that affects cardiovascular and neurological health through common variants [4,36]. Our results suggest that a more complete genotyping in this chromosomal region may help to elucidate more of the crucial roles of the APOE locus variants on cardiometabolic and neurological disorders.

APOE Variants and Lipid Profile
APOE is a key regulator of plasma lipid levels, and the Apoe-/-mouse model is the most widely used animal model of atherosclerosis, with markedly elevated plasma cholesterol levels [4,37]. The ApoE3 isoform possesses both the lipid-binding ability and affinity for the LDL receptor to mediate appropriate lipolytic processing and endocytosis of TRL remnant particles and exists as the parent form of three common APOE isoforms with normal plasma lipid levels [11]. ApoE2 and apoE4 differ from apoE3 by the single amino acid substitution with Arg158Cys (rs429358) for apoE2 which is located near the LDL receptor recognition site and Cys112Arg (rs7412) for apoE4, which affected the organization and stability of both the N-terminal and C-terminal domains of APOE. ApoE2 exhibits impaired binding to the LDL receptor and an inability to promote clearance of TRL remnant particles. ApoE4 results in enhancing the binding ability of VLDL particles and impairs lipolytic processing in the circulation, which is associated with a higher VLDL cholesterol/HDL cholesterol ratio as a more proatherogenic lipoprotein cholesterol distribution [11]. In a meta-analysis of more than 800,000 individuals from BioBank Japan, UK Biobank, and FinnGen, a strong association between rs7412 and LDL-C levels was noted, with a very low p value of 3 × 10 −3040 [38]. Our data also reveal that the APOE ε2, ε3, and ε4 variants defined by the APOE rs7412 and APOE rs429358 genotypes were the strongest genetic determinants of LDL cholesterol levels in our Taiwanese cohort, which together contributed to 8.72% of LDL cholesterol levels.

CLPTM1 Polymorphism and Serum Lipid Levels
Clptm1 is a multi-pass transmembrane protein whose biological function has not been fully elucidated. CLPTM1 variants have been suggested to be one of the genetic risk factors of non-syndromic oral clefts [39]. Genetic variants around the CLPTM1 have been associated with episodic memory performance [40]. The associations between CLPTM1 gene variants and expression and Alzheimer disease have been observed in genome-wide and transcriptome-wide association studies [27,40,41]. Several GWASs have revealed CLPTM1 variants associated with various lipid traits and apolipoprotein B levels in European populations [30,31,42,43]. Our data also indicate that CLPTM1 variants were associated with LDL and HDL cholesterol levels independent of APOE variants in a Taiwanese cohort. These results support the crucial role of CLPTM1 in lipid metabolism. Clptm1 has been demonstrated to be a pan-GABAA receptor-associated protein that interacts with multiple subunits and traps GABAA receptors in the endoplasmic reticulum and Golgi apparatus to scale phasic and tonic inhibitory transmission and modulate activity-induced inhibitory homeostasis [29,44,45]. The γ-aminobutyric acid (GABA) is the principal inhibitory neurotransmitter that acts on GABA receptors and modulates cholesterol metabolism in macrophages to increase cholesterol efflux and inhibit the formation of human macrophage-derived foam cells [46]. Through integrative analysis of multiomics data from the Avon Longitudinal Study of Parents and Children and replicated in the TwinsUK study, CLPTM1 is identified to be associated with LDL cholesterol levels by the proposed powerful adaptive gene-based test integrating information from gene expression, methylation, and enhancer-promoter interactions with the signal driven by both low-frequency and common variants [47]. Thus, further study is needed to unravel the molecular mechanism(s) for the associations between CLPTM1 variants and lipid profiles.

APOE-APOC1 Polymorphisms and Serum Triglyceride Levels
Both rs438811 and rs439401 variants, located between APOE and APOC1, have been associated with serum lipid levels in several GWASs [42,[48][49][50]. Apolipoprotein C1 (apoC1) is the smallest of all apolipoproteins and is associated with TRLs and HDL and exchanges between lipoprotein classes. It acts on lipoprotein receptors by inhibiting binding mediated by APOE and modulating the activities of several enzymes [51]. The rs438811 variant is in moderate LD with APOE rs7412 and rs429358 variants and is associated with total, LDL cholesterol, and triglyceride levels, whereas the rs439401 variant is in weak LD with the APOE rs7412 and rs429358 variants and is associated with all four lipid profile variables. For serum triglyceride levels, these two variants were the most crucial genetic determinants in the APOE locus in TWB participants.

Association between APOE Gene Region Variants and Metabolic Syndrome
Metabolic syndrome involves a complex of interrelated cardiometabolic derangements linked to insulin resistance and chronic inflammation and is characterized by clustering of cardiovascular risk factors, including hypertension, dyslipidemia, central obesity, and glucose intolerance [52]. Metabolic syndrome is also associated with higher risks of type 2 diabetes mellitus, nonalcoholic fatty liver disease, atherosclerotic cardiovascular disease, ischemic stroke, and mortality [48,[53][54][55][56]. A high prevalence of metabolic syndrome has been observed in the Asia-Pacific region, including Taiwan [57]. Strong genetic components interacting with environmental factors have been associated with the risk of metabolic syndrome. Elucidating genetic risk factors for metabolic syndrome remains a challenging task in the prevention and management of the syndrome complex. Studies for the association between genetic variants on the APOE locus and metabolic syndrome are not consistent with some GWASs; some genetic association studies have demonstrated APOE or APOC1 variants associated with the risk of metabolic syndrome [58][59][60][61], whereas others have reported no evidence of this association [62][63][64]. Our results reveal an independent association between APOE rs429358 genotypes and metabolic syndrome. One study suggested an association between APOE rs429358 genotypes and metabolic syndrome, which became nonsignificant after Bonferroni correction [64]. Thus, additional studies are required to elucidate the role of APOE variants, especially rs429358, in the risk of metabolic syndrome.

Association between APOE Gene Region Variants and Serum Albumin Levels
Albumin is a multifaceted protein synthesized in hepatocytes involving anti-inflammatory, antioxidant, anticoagulant, and antiplatelet aggregation activities, and serum albumin is the most abundant human plasma protein [65]. In our previous study on TWB participants, we found pleiotropic effects of common and rare GCKR exonic mutations on cardiometabolic traits, including serum albumin levels [34]. In a cross-population atlas of genetic associations for 220 human phenotypes, using 179,000 BioBank Japan participants and a meta-analysis of studies involving a total of 628,000 UK Biobank and FinnGen participants, Sakaue et al. [38] also revealed the APOE/APOC1 region as a novel locus for serum albumin levels with APOC1 rs438811 as the candidate lead SNV. This is consistent with our finding that the rs438811 genotype was significantly associated with serum albumin levels independent of GCKR variants.

Limitations
This study has several limitations. First, we used genome-wide genotyping arrays with imputation data; however, whole-exome sequencing or even whole-genome sequencing is more powerful in detecting whether other functional variants are more important genetic determinants for cardiometabolic traits and metabolic syndrome in the APOE locus. Second, ethnic genetic heterogeneity in the genetic association studies suggests that our data may not be applicable to other ethnic populations. Third, not all genetic associations reached genome-wide significance in our analysis, and studies with larger sample sizes and metaanalyses including transethnic population studies may help validate our results.

Conclusions
Our data reveal that APOE ε2, ε3, and ε4 variants were the strongest genetic determinants of total, LDL, and HDL cholesterol levels in the APOE locus and CLPTM1 is a candidate locus for LDL-C and HDL-C levels, independent of other gene variants around the APOE gene region. The association between APOE and APOC1 variants and metabolic syndrome and serum albumin levels further supported the crucial role of APOE region variants in cardiometabolic disorders in a Taiwanese population.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/genes13081366/s1, Method 1: Definitions of hypertension, diabetes mellitus, current smoking, microalbuminuria and metabolic syndrome; Table S1: Lead singlenucleotide variants for serum lipid levels profile around the APOE gene region in regional association studies with conditional analysis; Table S2: Association of the APOE rs7412 genotypes with metabolic and hematological phenotypes; Table S3: Association of the APOE rs429358 genotypes with metabolic and hematological phenotypes; Table S4: Association of the APOC1 rs438811 genotypes with metabolic and hematological phenotypes; Table S5: Association of the APOE rs439401 genotypes with metabolic and hematological phenotypes; Table S6: Association of the CLPTM1 rs3786505 genotypes with metabolic and hematological phenotypes; Table S7: Association of the CLPTM1 rs11672748 genotypes with metabolic and hematological phenotypes; Table S8: Stepwise linear regression analysis for serum albumin levels: including GCKR and APOE region variants.  Informed Consent Statement: Each participant signed an informed consent form before participating in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.