Identification of eight genetic variants as novel determinants of dyslipidemia in Japanese by exome-wide association studies

We have performed exome-wide association studies to identify single nucleotide polymorphisms that influence serum concentrations of triglycerides, high density lipoprotein (HDL)–cholesterol, or low density lipoprotein (LDL)–cholesterol or confer susceptibility to hypertriglyceridemia, hypo–HDL-cholesterolemia, or hyper–LDL-cholesterolemia in Japanese. Exome-wide association studies for serum triglycerides (13,414 subjects), HDL-cholesterol (14,119 subjects), LDL-cholesterol (13,577 subjects), hypertriglyceridemia (4742 cases, 8672 controls), hypo–HDL-cholesterolemia (2646 cases, 11,473 controls), and hyper–LDL-cholesterolemia (4489 cases, 9088 controls) were performed with HumanExome-12 DNA Analysis BeadChip or Infinium Exome-24 BeadChip arrays. Twenty-four, 69, or 32 loci were significantly (P < 1.21 × 10−6) associated with serum triglycerides, HDL-cholesterol, or LDL-cholesterol, respectively, with 13, 16, or 9 of these loci having previously been associated with triglyceride-, HDL-cholesterol–, or LDL-cholesterol–related traits, respectively. Two single nucleotide polymorphisms (rs10790162, rs7350481) were significantly related to both serum triglycerides and hypertriglyceridemia; three polymorphisms (rs146515657, rs147317864, rs12229654) were significantly related to both serum HDL-cholesterol and hypo–HDL-cholesterolemia; and six polymorphisms (rs2853969, rs7771335, rs2071653, rs2269704, rs2269703, rs2269702) were significantly related to both serum LDL-cholesterol and hyper–LDL-cholesterolemia. Among polymorphisms identified in the present study, two polymorphisms (rs146515657, rs147317864) may be novel determinants of hypo–HDL-cholesterolemia, and six polymorphisms (rs2853969, rs7771335, rs2071653, rs2269704, rs2269703, rs2269702) may be new determinants of hyper–LDL-cholesterolemia. In addition, 12, 61, 23, or 3 polymorphisms may be new determinants of the serum triglyceride, HDL-cholesterol, or LDL-cholesterol concentrations or of hyper–LDL-cholesterolemia, respectively.


INTRODUCTION
Dyslipidemia, including hypertriglyceridemia, hypohigh density lipoprotein (HDL)-cholesterolemia, and hyper-low density lipoprotein (LDL)-cholesterolemia, is a multifactorial disorder that results from an interaction between genetic background and environmental factors [1,2]. Given that dyslipidemia is an important risk factor for coronary artery disease, ischemic stroke [3], and colorectal cancer [4], its personalized prevention is an important public health goal.
Genome-wide association studies (GWASs) [5][6][7][8] and gene-centric meta-analysis [9] have implicated various genes and loci as determinants of blood lipid levels or in predisposition to dyslipidemia in European-ancestry populations. Genetic variants associated with lipid profiles have been extensively investigated, with one recent study identifying 157 such loci, including 62 variants not previously reported [10]. Recent GWASs [11,12] or studies based on exome [13] or whole-genome [14] sequencing in Europeanancestry populations also identified low-frequency or rare variants related to circulating lipid levels. Although several polymorphisms have been shown to be related to blood lipid profiles in the Japanese population [15], genetic variantsincluding low-frequency or rare variants-that influence circulating lipid levels or contribute to genetic susceptibility to dyslipidemia in Japanese remain to be identified definitively.
We have now performed exome-wide association studies (EWASs) to identify single nucleotide polymorphisms (SNPs)-in particular, low-frequency or rare coding variants with moderate to large effect sizes-that influence the serum concentrations of triglycerides, HDL-cholesterol, or LDL-cholesterol or confer susceptibility to dyslipidemia in Japanese. We used Illumina arrays that provide coverage of functional SNPs in entire exons including such variants.

EWASs for hypertriglyceridemia, hypo-HDLcholesterolemia, or hyper-LDL-cholesterolemia
We performed the EWAS for hypertriglyceridemia with 13,414 subjects (4742 cases, 8672 controls), that for hypo-HDL-cholesterolemia with 14,119 subjects (2646 cases, 11,473 controls), and that for hyper-LDLcholesterolemia with 13,577 subjects (4489 cases, 9088 controls). Characteristics of the subjects are shown in Table 3. In the study of hypertriglyceridemia, age, the frequency of men, body mass index (BMI), and the prevalence of smoking, hypertension, diabetes mellitus, chronic kidney disease, and hyperuricemia as well as the serum concentrations of triglycerides and LDLcholesterol and the ratio of LDL-cholesterol to HDLcholesterol were greater, whereas the serum concentration of HDL-cholesterol was lower, in subjects with hypertriglyceridemia than in controls. In the study of hypo-HDL-cholesterolemia, age, the frequency of men, BMI, and the prevalence of smoking, hypertension, diabetes mellitus, chronic kidney disease, and hyperuricemia as well as the serum concentration of triglycerides and the ratio of LDL-cholesterol to HDL-cholesterol were greater, whereas the serum concentration of HDL-cholesterol was lower, in subjects with hypo-HDL-cholesterolemia than in controls. In the study of hyper-LDL-cholesterolemia, BMI, the prevalence of smoking, the serum concentrations of triglycerides and LDL-cholesterol, and the ratio of LDLcholesterol to HDL-cholesterol were greater, whereas age, the prevalence of chronic kidney disease, and the serum concentration of HDL-cholesterol were lower, in subjects with hyper-LDL-cholesterolemia than in controls. related to both serum triglycerides and hypertriglyceridemia; three polymorphisms (rs146515657, rs147317864, rs12229654) were significantly related to both serum HDL-cholesterol and hypo-HDL-cholesterolemia; and six polymorphisms (rs2853969, rs7771335, rs2071653, rs2269704, rs2269703, rs2269702) were significantly related to both serum LDL-cholesterol and hyper-LDL-cholesterolemia. Among polymorphisms identified in the present study, two polymorphisms (rs146515657, rs147317864) may be novel determinants of hypo-HDL-cholesterolemia, and six polymorphisms (rs2853969, rs7771335, rs2071653, rs2269704, rs2269703, rs2269702) may be new determinants of hyper-LDL-cholesterolemia. In addition, 12, 61, 23, or 3 polymorphisms may be new determinants of the serum triglyceride, HDL-cholesterol, or LDL-cholesterol concentrations or of hyper-LDL-cholesterolemia, respectively.

Linkage disequilibrium and haplotype analysis
Given that eight SNPs (rs7771335, rs2071653, rs2853969, rs2269704, rs2269703, rs495089, rs2269702, rs1233399) associated with hyper-LDL-cholesterolemia were all located at chromosome 6p22.1-p21.3, we examined linkage disequilibrium among these SNPs as well as the relation of their haplotypes to this condition. The eight SNPs were all in strong linkage disequilibrium (Supplementary Table 11). Haplotype analysis revealed that the haplotypes A were significantly (P < 4.31 × 10 -4 ) associated with hyper-LDL-cholesterolemia, with the former haplotype being protective against and the latter representing a risk factor for this condition (Supplementary Table 12).

Relation of identified SNPs to serum triglyceride, HDL-cholesterol, or LDL-cholesterol levels
We examined the relation of genotypes of identified SNPs to serum triglyceride, HDL-cholesterol, or LDLcholesterol levels by one-way analysis of variance. The 46 SNPs identified in the EWAS for serum triglyceride concentration, including the two SNPs also found to be associated with hypertriglyceridemia (rs10790162 of BUD13, rs7350481 at 11q23.3), were all significantly [P < 0.0011 (0.05/46)] associated with serum triglyceride level (Supplementary Table 13). The 104 SNPs identified in the EWAS for serum HDL-cholesterol concentration, including two of the three SNPs found to be associated with hypo-HDL-cholesterolemia (rs146515657 of USP4, rs12229654 at 12q24.1), were all significantly [P < 4.76 × 10 -4 (0.05/105)] associated with the serum HDLcholesterol level, as was the SNP associated with hypo-HDL-cholesterolemia alone (rs147317864 of TRABD2B) (Supplementary Table 14). The 40 SNPs identified in the EWAS for serum LDL-cholesterol concentration, including one SNP also found to be associated with The relation of genotypes of SNPs to the serum concentration of triglycerides was examined by linear regression analysis. a Major allele/minor allele.  Table 15).

SNPs associated with serum HDL-cholesterol or LDL-cholesterol levels
The ubiquitin specific peptidase 4 gene (USP4) is located at chromosome 3p21.31 (NCBI Gene, https:// www.ncbi.nlm.nih.gov/gene) and is expressed in various tissues and organs (The Human Protein Atlas, http://www. proteinatlas.org). Ubiquitin-specific proteases promote posttranslational protein modification by reversing protein ubiquitination and thereby activates multiple biological processes including the cell cycle, DNA repair, and intracellular signaling [29]. USP4 accelerates the growth, invasion, and metastasis of colorectal cancer [29,30]. We have now shown that rs146515657 [T/C (N650S)] of USP4 was significantly associated with both the serum HDL-cholesterol concentration and hypo-HDLcholesterolemia, with the minor C allele being related to a lower serum level of HDL-cholesterol, although the molecular mechanism underlying this association remains unclear.
The TraB domain containing 2B gene (TRABD2B) is located at chromosome 1p33 (NCBI Gene) and is expressed ubiquitously (The Human Protein Atlas). TRABD2B encodes the metalloprotease TIKI2 that inhibits the Wnt/β-catenin signaling by cleavage of the amino terminus of Wnt protein [31]. The expression of TIKI2 was reduced in osteosarcoma specimens and the increased expression of TIKI2 inhibited the growth of osteosarcoma in vivo, suggesting TIKI2 suppresses the growth of osteosarcoma [32]. We have now shown that rs147317864 [C/T (A262T)] of TRABD2B was significantly associated with both the serum concentration of HDL-cholesterol and hypo-HDL-cholesterolemia, with the minor T allele being related to a reduced level of serum HDLcholesterol, although the functional relevance underlying this association remains unclear.
The myelin oligodendrocyte glycoprotein gene (MOG) is located at chromosome 6p22.1 (NCBI Gene) and is highly expressed in brain (The Human Protein Atlas). The MOG protein is localized to the outer surface of the myelin sheath of neurons in the central nervous system and is a key antigen for autoimmune responses that result in inflammation and demyelination [33,34]. Both B cell responses and antibodies to MOG have also been detected in patients with demyelinating diseases such as multiple sclerosis and acute disseminating encephalomyelitis [34]. We have now shown that rs2071653 (C/T) of MOG was significantly associated with both the serum LDL-cholesterol concentration and hyper-LDL-cholesterolemia, with the minor T allele being related to an increased serum level of LDL-cholesterol, although the underlying molecular mechanism remains unknown.
The protein phosphatase 1 regulatory subunit 18 gene (PPP1R18) is located at chromosome 6p21.33 (NCBI Gene) and is expressed in various tissues (The Human Protein Atlas). Protein phosphatase 1 binds to regulatory subunits that target the enzyme to different intracellular locations to exert its activity toward specific substrates [35]. The PPP1R18 protein is a regulatory subunit that targets protein phosphatase 1 to the F-actin cytoskeleton [36].
We have now shown that rs2269704 (C/T) of PPP1R18 was significantly associated with both the serum LDLcholesterol concentration and hyper-LDL-cholesterolemia, with the minor T allele being related to an increased serum LDL-cholesterol level, although the molecular mechanism underpinning this association is unclear. The nurim gene (NRM) is located at chromosome 6p21.33 (NCBI Gene) and is expressed ubiquitously (The Human Protein Atlas). The NRM protein contains transmembrane domains and resides within the inner nuclear membrane, where it is tightly bound to the nuclear envelope [37]. NRM is expressed in a broad range of cancers, with its expression level being correlated with tumor grade [38]. NRM deficiency was found to alter the shape of the nuclear envelope and to enhance ultraviolet light-induced apoptosis in HeLa cells, implicating NRM in suppression of apoptosis [38]. We have now shown that rs2269703 (G/A) of NRM was significantly associated with both the serum LDL-cholesterol concentration and hyper-LDL-cholesterolemia, with the minor A allele being related to an increased serum LDL-cholesterol level, although the molecular mechanism is unknown.
The mediator of DNA damage checkpoint 1 gene (MDC1) is located at chromosome 6p21.33 (NCBI Gene) and is expressed in various tissues and organs (The Human Protein Atlas). MDC1 is a nuclear protein required for activation of the intra-S phase and G 2 -M phase checkpoints of the cell cycle in response to DNA damage [39]. We have now shown that rs2269702 (A/G) of MDC1 was significantly associated with both the serum LDL-cholesterol concentration and hyper-LDLcholesterolemia, with the minor G allele being related to an increased serum LDL-cholesterol level, although the molecular mechanism remains unclear.

General considerations
In previous GWASs of blood lipid traits in East Asian populations [24], a minor allele frequency (MAF) and effect size of identified SNPs were 10% to 33% and -0.088 to -0.050 mg/dL for triglycerides, 12% to 15% and -0.035 to 0.043 mg/dL for HDL-cholesterol, and 26% and 2.203 mg/dL for LDL-cholesterol, respectively. In trans-ancestry GWASs for lipid profiles [10], the MAF and effect size of identified SNPs ranged from 9% to 49% and from -0.033 to 0.037 mg/dL for triglycerides, from 9% to 50% and from -0.051 to 0.034 mg/dL for HDLcholesterol, and from 4% to 48% and from -0.051 to 0.103 mg/dL for LDL-cholesterol. In more recent GWASs that included low-frequency or rare variants, the MAF and effect size of identified SNPs ranged from 1.76% to 3.25% and from -30% to 21% for triglycerides, from 0.20% to 2.01% and from -3 to 17 mg/dL for HDL-cholesterol, and from 0.05% to 3.43% and from -40 to 71 mg/dL for LDLcholesterol [11]; as well as from 1% to 47% and from -0.170 to 0.128 mmol/L for triglycerides, from 5% to 20% and from -0.141 to 0.044 mmol/L for HDL-cholesterol, and from 1% to 21.6% and from -0.049 to 0.648 mmol/L for LDL-cholesterol, respectively [12]. Studies based on exome and whole-genome sequencing identified SNPs with a MAF and effect size from 0.06% to 28.2% and from -98.0 to 51.5 mg/dL, respectively, for LDL-cholesterol [13]; as well as those with a MAF of 0.007% to 4.6% and effect size of -65.3 to 12.0% for triglycerides and -0.087 to 0.40 mmol/L for HDL-cholesterol [14].
In our study, among 46 SNPs associated with the serum triglyceride concentration, one SNP was a rare variant (MAF, 0.3%) with a large effect size (difference in serum triglyceride level among genotypes, 28

Study limitations
There are several limitations to the present study. (1) Our results were not replicated and will therefore require validation in independent subject panels or in other ethnic groups. (2) Subjects who had treatment for other diseases such as diabetes mellitus were included in the study. It was possible that such treatment affected lipid profiles www.impactjournals.com/oncotarget of the subjects. (3) SNPs identified in our study might be in linkage disequilibrium with other polymorphisms in the nearby genes that are actually responsible for the observed associations. (4) Three SNPs associated with hyper-LDL-cholesterolemia were not significantly related to the serum LDL-cholesterol concentration, which may be attributable to the effects of medical treatment. (5) The functional relevance of the observed associations remains to be determined.

Study subjects
A total of 14,337 subjects (8354 individuals with dyslipidemia, 5983 controls) was recruited as described previously [40].
Venous blood was collected in the early morning after the subjects had fasted overnight. Blood samples were centrifuged at 1600 × g for 15 min at 4°C, and serum was separated for subsequent analysis. Serum concentrations of triglycerides, HDL-cholesterol, and LDL-cholesterol were measured at the clinical laboratory of each hospital. The 4742 subjects with hypertriglyceridemia and 8672 controls had serum triglyceride concentrations of ≥ 1.69 mmol/L (range, 1.69 to 20.14 mmol/L) and < 1. Individuals with single-gene disorders such as familial hypercholesterolemia or with endocrinologic or metabolic diseases that cause dyslipidemia were excluded from the study. Those taking medications that may cause secondary dyslipidemia were also excluded. Autopsy cases were excluded from controls.
The study protocol complied with the Declaration of Helsinki and was approved by the Committees on the Ethics of Human Research of Mie University Graduate School of Medicine, Tokyo Metropolitan Institute of Gerontology, Hirosaki University Graduate School of Medicine, and participating hospitals. Written informed consent was obtained from each participant or families of the deceased subjects.

EWASs
Methods for collection and extraction of genomic DNA samples were described previously [40]. EWASs for the serum concentrations of triglycerides (13, . Detailed information of the exome arrays and methods of quality control were described previously [40]. Totals of 41,371, 41,225, and 41,347 SNPs passed quality control in the EWASs for hypertriglyceridemia, hypo-HDL-cholesterolemia, and hyper-LDL-cholesterolemia, respectively, and were included in the analysis.

Statistical analysis
The relation of SNP genotypes to the serum concentrations of triglycerides, HDL-cholesterol, or LDL-cholesterol in the EWASs was examined with linear regression analysis. For analysis of characteristics of the study subjects, quantitative and categorical data were compared between cases and controls with the unpaired Student's t test and Fisher's exact test, respectively. Allele frequencies were estimated by the gene counting method, and departure from Hardy-Weinberg equilibrium was identified with Fisher's exact test. The relation of allele frequencies of SNPs to hypertriglyceridemia, hypo-HDL-cholesterolemia, or hyper-LDL-cholesterolemia in the EWASs was examined with Fisher's exact test. To compensate for multiple comparisons of genotypes or allele frequencies with lipid concentrations or dyslipidemia, we applied Bonferroni's correction for statistical significance of association. Given that 41,225 to 41,371 SNPs were analyzed, a P value of < 1.21 × 10 -6 [0.05/(41,225 to 41,371] was considered statistically significant for the EWASs. Quantile-quantile plots for P values of genotypes or allele frequencies in the EWASs are shown in Supplementary Figures 3 and 4, respectively. The inflation factor (λ) was 1.05 for serum triglycerides, 0.97 for serum HDL-cholesterol, 1.06 for serum LDL-cholesterol, 1.20 for hypertriglyceridemia, 1.29 for hypo-HDLcholesterolemia, and 1.20 for hyper-LDL-cholesterolemia. Multivariable logistic regression analysis was performed with hypertriglyceridemia, hypo-HDL-cholesterolemia, or hyper-LDL-cholesterolemia as a dependent variable and independent variables including age, sex (0, woman; 1, man), and genotype of each SNP. A detailed method www.impactjournals.com/oncotarget of analysis was described previously [40]. Relations of genotypes of identified SNPs to serum concentrations of triglycerides, HDL-cholesterol, or LDL-cholesterol were examined by one-way analysis of variance. Bonferroni's correction was also applied to other statistical analysis as indicated. Statistical tests were performed with JMP Genomics version 6.0 software (SAS Institute, Cary, NC).

CONCLUSIONS
The SNPs rs146515657 of USP4 and rs147317864 of TRABD2B may be novel determinants of hypo-HDLcholesterolemia whereas rs2853969 at 6p21.3, rs7771335 at 6p22.1, rs2071653 of MOG, rs2269704 of PPP1R18, rs2269703 of NRM, and rs2269702 of MDC1 may be new determinants of hyper-LDL-cholesterolemia. In addition, 12, 61, 23, or 3 SNPs may be new determinants of the serum triglyceride, HDL-cholesterol, or LDL-cholesterol concentrations or of hyper-LDL-cholesterolemia, respectively. Determination of genotypes for these SNPs may prove informative for assessment of the genetic risk for dyslipidemia in Japanese.