Heritability and Genome-Wide Association Study of Plasma Cholesterol in Chinese Adult Twins

Dyslipidemia represents a strong and independent risk factor for cardiovascular disease. Plasma cholesterol, such as total cholesterol (TC), low density lipoprotein cholesterol (LDL-C), and high density lipoprotein cholesterol (HDL-C), is the common indicator of diagnosing dyslipidemia. Here based on 382 Chinese twin pairs, we explored the magnitude of genetic impact on TC, HDL-C, LDL-C variation and further searched for genetic susceptibility loci for them using genome-wide association study (GWAS). The ACE model was the best fit model with additive genetic parameter (A) accounting for 26.6%, common or shared environmental parameter (C) accounting for 47.8%, unique/non-shared environmental parameter (E) accounting for 25.6% for the variance in HDL-C. The AE model was the best fit model for TC (A: 61.4%; E: 38.6%) and LDL-C (A: 65.5%; E: 34.5%). While no SNPs reached the genome-wide significance level (P < 5 × 10−8), 8, 14, 9 SNPs exceeded the suggestive significance level (P < 1 × 10−5) for TC, HDL-C, LDL-C, respectively. The promising genetic regions for TC, HDL-C, LDL-C were on chromosome 11 around rs7107698, chromosome 5 around rs12518218, chromosome 2 around rs10490120, respectively. Gene-based analysis found 1038, 1033 and 1090 genes nominally associated with TC, HDL-C, LDL-C (P < 0.05), especially FAF1, KLKB1 for TC, KLKB1 for HDL-C, and NTRK1, FAF1, SNTB2 for LDL-C, respectively. The number of common related genes among TC, HDL-C and LDL-C was 71, including FAF1, KLKB1, etc. Pathway enrichment analysis discovered known related pathways-zinc transporters, metal ion SLC transporters for TC, cell adhesion molecules CAMs, IL-6 signaling for HDL, FC epsilon RI signaling pathway, NFAT pathway for LDL, respectively. In conclusion, the TC and LDL-C level is moderately heritable and the HDL-C level is lowly heritable in Chinese population. The genomic loci, functional genes and pathways are identified to account for the heritability of plasma cholesterol level. Our findings provide important insights into plasma cholesterol molecular physiology and expect future research to replicate and validate our results.


INTRODUCTION
Dyslipidemia represents a strong and independent risk factor for cardiovascular disease (1), a leading cause of death worldwide (2,3). Besides, many studies indicate that dyslipidemia can increase the risk of preeclampsia, colorectal cancer, diabetic macular edema and polycystic ovary syndrome (4-7). The prevalence of dyslipidemia is high and is increasingly prevalent in China (8,9). Plasma cholesterol, such as total cholesterol (TC), low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), is a common indicator of diagnosing dyslipidemia. Therefore, exploring factors affecting plasma cholesterol homeostasis is a crucial step toward providing early prevention and therapeutic targets for dyslipidemia.
The plasma cholesterol level is mediated by a combination of genetic and environmental factors. At present, the heritability for plasma cholesterol level has been estimated in several studies, with the heritability of TC, HDL-C, LDL-C level ranging from 0 to 89% (10-13), 22 to 93% (13)(14)(15)(16), and 22 to 91% (10,13,16,17), respectively. Additionally, several genome-wide association studies (GWAS) have attempted to find susceptible genetic loci located in the corresponding gene affecting plasma cholesterol level. A GWAS conducted by Middelberg RP et al. (18) for plasma cholesterol found genes LPL, LIPC, CETP associated with HDL-C level and CELSR2, APOB, TOMM40 with LDL-C level. Another two GWAS studies identified SRGAP2, HOXC13 and CD47, DUSP4 that were associated with HDL-C and TC level, respectively (19,20).
However, the known genetic variations only explain a small proportion of the genetic contribution and many potential genetic genes and loci remain to be discovered. Besides, allele frequencies, life style and environmental contributions differ between Chinese and other ethnic populations. Twin samples will have a higher power in genetic study, especially in human complex diseases (21). Here based on 382 Chinese twin pairs, we explored the genetic effect on TC, HDL-C, LDL-C variation and further searched for genetic susceptibility loci for these traits using GWAS (22).

Twin Samples Collection and Phenotypic Measurement
The twin samples were collected from Qingdao Twin Registry and the details can be found in the literatures (23,24). Twins who were pregnancy and lactation, as well as incomplete cotwin pairs were excluded. Twin pairs that taking cholesterollowering drugs also were excluded. The study included 382 twin pairs for heritability analysis and 139 dizygotic (DZ) twin pairs for GWAS with a mean age of 51.6 ± 7.7. All twin samples undertook the blood sampling and a physical examination after a 10-12 h overnight fast and completed a questionnaire. The zygosity of same sex and blood type was identified by using Abbreviations: DZ, dizygotic; GWAS, genome-wide association study; MZ, monozygotic; TC, total cholesterol; VEGAS2, Versatile Gene-based Association Study-2. 16 multiple short tandem sequence repeat DNA markers (25,26). According to the standard procedure using an automatic biochemical analyzer (Hitachi 7600; Hitachi, Tokyo, Japan), we tested participants' fasting blood cholesterol sample in the Qingdao Diabetes Hospital. Friedewald equation was used to calculate the LDL-C level: LDL-C (mmol/L) = TC-HDL-C-(TG/2.2) (mmol/L) (27).
The Regional Ethics Committee of the Qingdao CDC Institutional Review Boards approved this study and the ethical principles followed the Helsinki Declaration. All subjects signed the written informed consent.
We used the IMPUTE2 (28) software to impute un-typed SNPs using the LD information from 1,000 Genomes Project Phase 3 reference panel (29) (CHS N = 171 and CHB N = 142). We used R 2 > 0.6, MAF > 0.05 and HWE > 1 × 10 −4 to filter the imputed data, and the 7,405,822 SNPs were used to explore the association with plasma cholesterol level.

Statistical Analysis
Heritability Data preparation and statistical description were performed with SPSS version 22.0. We used the structural equation models (SEM) to evaluate the genetic variance components with Mx software 1 . The Blom's formula was used to guarantee the normal distribution or approximate normal distribution because of the deviation of the distribution of all indicators. Pearson's productmoment correlation coefficient was used to calculate intraclass phenotypic correlations. The correlation in MZ twins was significantly higher than DZ twins, reflecting the importance of genetic effects in plasma cholesterol levels.
The total phenotypic variance could be decomposed to additive genetic variance (A), common or shared environmental variance (C), and unique/non-shared environmental variance (E). The full ACE model was firstly fitted, the likelihood-ratio χ 2 -test was applied to test whether the contributions of A or C to the model had statistical significance by comparing the full model (ACE) and their nested models (CE and AE). The Akaike information criterion (AIC), which is equal to the goodness-offit χ 2 -value minus twice the degrees of freedom, was used to indicate the parsimony of each model and a lower AIC indicated a better fit. Heritability (h 2 ) was calculated in the best-fitting model based on the ratio of additive genetic variation to total phenotypic variation, with adjusting for the effect of age, sex and education level. We calculated the power of twin pairs for additive genetic influences by Mx software, and the calculation results showed that the power of our heritability analysis was above 90%.

SNPs-based analysis
We used the genome-wide efficient mixed-model association (GEMMA) to test the association between plasma cholesterol level and SNP genotypes after adjusting for the following covariates: sex, age, and education level (30). We used Quantilequantile (Q-Q) and Manhattan plots to illustrate overall significance level (P < 5 × 10 −8 ) and suggestive level (P < 1 × 10 −5 ) autosomal and chromosomal X SNPs (31). In addition, we analyzed the enrichment of cell-type enhancers for the typed GWAS results of the regulatory domains outside the coding regions by using online HaploReg v4.1 software 2 (32,33). SNPs with P < 1 × 10 −5 were selected as query SNPs, and an uncorrected P < 0.05 for enrichments of cell-type enhancers were reported.

Gene-based analysis
The gene-based analysis integrated all SNPs within a gene to increase the signal or strength of association, which correcting for linkage disequilibrium (LD) and gene size. The gene-based test was implemented in Versatile Gene-based Association Study-2 (VEGAS2) 3 which uses 1000 Genomes data to simulate the correlations of SNPs across the autosomes and chromosome X. (34,35). We used SNPs from "1000G East ASIAN Population". The genome-wide significant gene for the association was defined as P < 2.63 × 10 −6 (0.05/19,001) as 19,001 genes being evaluated.

Pathway enrichment analysis
We used PASCAL to compute pathway-scored (36)(37)(38). In this approach, genetic markers SNPs were first mapped to genes, and the association scores of all genes in the pathway were computed. We then combined the genes scores of the same pathways to calculate the pathway scores. We used chi-squared or empirical score to evaluate pathway enrichment of high-scoring (possibly fused) genes, avoiding any standard binary enrichment tests with inherent P-value thresholds. Pathways and their corresponding gene annotation were obtained from KEGG, Reactome, and Biocarta (as defined in MSigDB) 4 .

Heritability
The final sample consisted of 382 twin pairs for heritability analysis and 139 DZ twin pairs for GWAS analysis with a mean age of 51.6 ± 7.7. The mean value ± SD of TC, HDL-C, LDL-C level for all subjects was 4.9 ± 1.2 umol/L, 1.5 ± 0.5 umol/L, 2.8 ± 0.9 umol/L, respectively (Supplemental Table 1 Table 2

SNPs-Based Analysis TC level
A sample of 139 DZ twin pairs including 1,365,181 qualified SNPs was included for the present GWAS. The Q-Q plot about TC level illustrated the relationship between the observed and expected GWAS P-values (Supplemental Figure 1). The genomic inflation factor (λ-statistic = 1) revealed no evidence of inflation of the test statistics due to population stratification. And weak association was shown due to the slight deviation in the upper right tail from the null distribution. No SNPs reached the genome-wide significance level (P < 5 × 10 −8 ) as illustrated in Manhattan plot (Supplemental Figure 2). However, there were 8 SNPs exceeding the threshold for suggestive significance level (P < 1 × 10 −5 ) (Supplemental Table 3). The strongest association SNP was rs7107698 (P = 2.29 × 10 −6 ). As the locus zoom plots illustrated, one chromosomal loci 11p15.4 showed nominal association with TC level (Supplemental Figure 3). Four SNPs (P = 2.29 × 10 −6 to 2.45 × 10 −6 ) were located at AMPD3 gene on chromosome 11p15.4. By HaploReg v4.1, three cell-type specific enhancers (uncorrected P < 0.05) of primary neutrophils from peripheral blood (P = 0.04), ovary (P = 0.04) and NHDF-Ad adult dermal fibroblast primary cells (P = 0.01) were identified for TC level (Supplemental Table 4).

LDL-C level
The Q-Q plot about LDL-C level was shown in Supplemental Figure 4. None of the SNPs reached the genome-wide significance level (P < 5 × 10 −8 ) as illustrated in Manhattan plot (Supplemental Figure 5). However, there are 9 SNPs exceeding the threshold for suggestive significance level (P < 1 × 10 −5 ) (Supplemental Table 5).

Gene-Based Analysis
In the gene-based analysis, none genes were found to achieve genome-wide significance level and the number of genes that were nominally associated with TC, HDL-C, LDL-C level was 1,038, 1,033, and 1,090, respectively (P < 0.05). The top 20 genes were ranked by their P values (TC: Supplemental Table 11, HDL-C: Table 3, LDL-C: Supplemental Table 12). LOC101929154, RIMS2, INTS9, KRT72, CCDC149, CUBN genes for HDL-C level and AMPD3, ABR genes for TC level had already been shown in the suggestive

chisq-P, Chi-square p-value. Chi-squared method (gene-score p-value were ranked and transformed to a uniform distribution, these values were then transformed by a chi-square quantile function, and summed). The content discussed in detail were in bold. emp-P, empirical p-value. Empirical sampling method (gene-scores are transformed with chi-square quantile function and summed, then Monte Carlo estimate of the p-values were obtained by sampling random sets of the same size).
level SNPs-based analysis, whereas the other genes were novel. The number of common nominally associated genes (P < 0.05) between TC VS HDL-C level, TC VS LDL-C level, HDL-C VS LDL-C level was 195, 504, 91, respectively. And 71 common genes were found among the three traits (Supplemental Table 13).

Pathway Enrichment Analysis
PASCAL program was used in the pathway enrichment analysis. The number of pathways that were nominally associated with TC, HDL-C, and LDL-C level was 816, 529, and 813, respectively (P < 0.05). The top 20 pathways were ranked by their emp-P-values (TC: Supplemental Table 14; HDL-C: Table 4; LDL-C: Supplemental Table 15). The Q-Q plots were shown in Supplemental Figures 7-9. Pathway analysis revealed several biological processes significantly associated with TC level: zinc transporters, metal ion SLC transporters, ERKS ARE inactivated, MAPK targets nuclear events mediated by map kinases, ERK MAPK targets, transport of glucose and other sugars bile salts and organic acids metal ions and amine compounds, amino acid transport across the plasma membrane, etc.
Several biological pathways significantly related to HDL-C levels emerged: cell adhesion molecules CAMs, IL-6 signaling, tight junction, bile salt and organic anion SLC transporters, transport of glucose and other sugars bile salts and organic acids metal ions and amine compounds, signaling by ILS, IL6 pathway, natural killer cell mediated cytotoxicity, sulfur metabolism, chondroitin sulfate biosynthesis, glycosaminoglycan biosynthesis chondroitin sulfate, leukocyte transendothelial migration, etc.
The pathways associated with LDL-C involved: FC epsilon RI signaling pathway, NFAT pathway, metabolism of amino acids and derivatives, ERBB signaling pathway, IL1R pathway, HIV infection, etc.

Heritability
Our investigation was based on 382 twin pairs to explore the heritability of TC, HDL-C, LDL-C level, and confirmed the genetic variants underlying these trait by GWAS. The correlation in MZ twins was significantly higher than DZ twins, which reflected the existence of genetic variance in TC, HDL-C, LDL-C levels (Supplemental Table 2). The ACE model was the best fit model with A accounting for 26.6%, C accounting for 47.8%, E accounting for 25.6% for the variance in HDL-C level. The best fit model for TC level was AE model with A accounting for 61.4%, E accounting for 38.6% and for LDL-C level was also AE model with A accounting for 65.5%, E accounting for 34.5% ( Table 1). The magnitude of the additive genetic components (A) for TC level obtained here was in line with several previous twin studies (15,16,27,(39)(40)(41)(42)(43)(44)(45)(46)(47). HDL-C level is close to some other twin studies (11,14,27,48,49). LDL-C level is similar to the others (27,42,43,45,47,50).

SNP-Based Analysis TC level.
While no genome-wide significant SNPs were found in GWAS for TC level, one SNP rs7107698 on chromosome 11 was found as promising genetic regions. The AMPD3 gene around the rs7107698 has been linked to the TC level. Adenosine monophosphate deaminase 3 (AMPD3) is an enzyme that catalyzes the hydrolytic deamination of adenosine monophosphate to inosine monophosphate and participates in the pathway of adenylate catabolic (51). The activation of AMP deaminase leads to the increase of uric acid and the production of mitochondrial oxidative stress, which further stimulates the formation of de novo lipogenesis and the activation of ATPcitrate lyase and the synthesis of long-chain saturated fatty acids and eventually affects TC level (52,53). Additionally, three highly correlated SNPs (rs4909928, rs12184411, kgp6520322) were detected within AMPD3 gene, showing suggestive evidence of association with TC level.

HDL-C level.
This cluster of suggestive SNPs (kgp6737496, rs12518218, rs7729225) is linked to LOC101929154, a non-coding RNA. Although we have not fully elucidated the biological function of this ncRNA, mutations in base pairs may affect their regulatory domain to regulate HDL-C levels. Another gene CUBN including suggestive SNPs rs12414709 and rs17345993 has been linked to HDL-C metabolism (54). Cubilin (CUBN) encodes high affinity HDL-C and lipid-poor apoA-I endocytosis receptor, which participate in the renal clearance of filterable apolipoprotein AI/HDL-C (55,56).

LDL-C level
SNPs rs10490120, rs4953640, rs74263479, rs17037869 were included in the FSHR gene. Follicle stimulating hormone receptor (FSHR) may interact with FSH in human hepatic tissue, which can elevate LDL-C level by blocking the expression of LDLR (57,58). Another KCNK9 around the suggestive SNP rs13251143 also has evidence for dyslipidemia. A great deal of studies has shown that KCNK9 gene is associated with obesity, HDL-C, adiponectin levels and aldosterone production (59-61). All the above factors are related to dyslipidemia (62)(63)(64).
As for cell-type enhancers, the relationship between TC level and ovary has been studied. Women with polycystic ovary syndrome had higher TC and LDL-C level, lower HDL-C level (65,66). Many studies have showed that Estrogen Receptor 1 (ESR1) is associated with TC, HDL-C, LDL-C level (67).

Imputation
Genotype imputation substantially increases available SNPs for analysis in GWAS. Substantive SNPs introduced by imputation reached the suggestive level associated with cholesterol. SNPs rs17578959 and rs28845526 were located in the PID1 gene, which served as the regulator of the LDLR-related protein 1 (LRP1) function and controlled the processing of postprandial lipoproteins (68). Solute Carrier Family 13 Member 1 (SLC13A1) gene can regulate the cholesterol level by participating in transport of glucose and other sugars, bile salts and organic acids, metal ions and amine compounds pathways.

Gene-Based Analysis TC level
Fas Associated Factor 1 (FAF1) can inhibit the activation of NF-kB by interfering with the assembly of IêB kinase (IKK) complex and impeding the nuclear translocation of NF-kB RelA in stimulation-dependent manner. NF-kB signaling pathway is closely related to inflammation (69)(70)(71). Epidemiology investigations have revealed that inflammation is associated with a risk of dyslipidemia (72,73). Lamina C et al. found that Kallikrein B1 (KLKB1) was associated with apolipoprotein A-IV concentrations in a genome-wide association meta-analysis. Apolipoprotein A-IV acting as a major component of HDL-C and chylomicron particles participates in reverse cholesterol transport (74).

HDL-C level
Kallikrein B1 (KLKB1) has been studied in a genomewide association meta-analysis on apolipoprotein A-IV concentrations. Apolipoprotein A-IV acting as a major component of HDL-C and chylomicron particles participates in reverse cholesterol transport (74).

LDL-C level
Neurotrophic Receptor Tyrosine Kinase 1 (NTRK1) encodes the neurotrophic tyrosine kinase receptor (NTKR) family members, which are related to classical mitogen-activated protein kinases (MAPK) signaling pathways (75). Further, MAP signaling pathways participated in lipid metabolism (76). Syntrophin Beta 2 (SNTB2) suggestes a role in ERK and SR-BI level, and sphingomyelin metabolism in obesity and have been proved to affect the activity of ABCA1 by stabilizing ABCA1 protein, which affect the catabolism of LDL-C level (77)(78)(79). Fas Associated Factor 1 (FAF1) also plays a key role in regulating the LDL-C level (69-71).

Pathway Enrichment Analysis TC level
Pathway analysis revealed several biological processes significantly associated with TC level: zinc transporters, metal ion SLC transporters, immune system, regulation of transport. Other novel pathways especially top 20 may also be interesting potential candidates for future research and validation.
(1) Zinc serves as a catalytic or structural cofactor for many different proteins and participates in the formation and function of enzyme. One study in young Finns (6-18 years) reported that serum Zinc was positively correlated with TC level (80). (2) Six SLC gene families encode proteins which mediate transport of metals. Transition metal ion such as vanadium, iron, can influence the serum lipid and lipoprotein profiles (81,82). (3) ERKS ARE inactivated, MAPK targets nuclear events mediated by map kinases and ERK MAPK targets participate in the immune system pathway. Immune system is also associated with cholesterol level. Acute phase conditions and immune disorders promote the decrease of HDL-C and change the composition and size of HDL-C (83). (4) Transport of glucose and other sugars bile salts and organic acids metal ions and amine compounds and amino acid transport across the plasma membrane participate in regulation of transport. Several genes (GAS6, GPLD1, PRKCE, WNK1, ESR1, RAC2, FAF1, FFAR1, RNASEL, GCK, SLC30A8, RARRES2, TLR5, TLR1, ACSL3, ABCA7, Th17) enriching in pathway of regulation of transport have been clarified the association with cholesterol level.

HDL level
Several biological pathways significantly related to HDL-C levels include: cell adhesion molecules CAMs, tight junction, chondroitin sulfate biosynthesis, sulfur metabolism, immune system, regulation of transport.
(1) An increased CAMs expression may be a mechanism that decreases plasma HDL-C level (84). (2)

LDL level
The pathways associated with LDL-C were: immune system, metabolism of amino acids and derivatives, ERBB signaling pathway, HIV infection.
(1) FC epsilon RI signaling pathway, NFAT pathway, IL1R pathway constitute the immune system to regulate cholesterol levels. (2) Metabolism of amino acids and derivatives participates in the formation and function of enzyme to regulate cholesterol level. (3) ERBB4 activates sterol regulatory element binding protein-2 (SREBP-2) to enhance LDL-C uptake and cholesterol biosynthesis (88). (4) A cross-sectional epidemiological study in China showed that mean LDL-C was lower in HIV-positive than HIV-negative subjects (89).

Strengths and Limitations
There are two advantages in this study. First, our results were based on the twin data of TC, HDL-C, LDL-C level. Associated individuals, such as twin pairs, will confer increased power in genetic association analysis due to their genetic association (21). And the power of our heritability analysis was above 90%, which suggested that the heritability findings were credible. Second, compared with most GWAS results from European populations, we sampled GWAS from the Qingdao twin population. Our GWAS results of TC, HDL-C, LDL-C level in East Asian population provided a basis for future molecular biology investigations.
This study has two limitations that are noteworthy. First, our study had a relatively small sample size due to the challenges of recruiting and identifying qualified twin pairs. So it should be noted that the power of our analysis might not be sufficient to detect association between SNPs, pathways and cholesterol levels, and our results needed to be furtherly confirmed. Second, our study provided a lot of suggestive results, but no statistically significant results. However, many variants and genes had been confirmed, other novel variants and genes may also be interesting potential candidates for future research and validation.

CONCLUSIONS
In brief, we have verified genetic impact on TC, HDL-C, LDL-C variation through twin modeling. The promising genetic regions for TC, HDL-C, LDL-C level were on chromosome 11 around rs7107698, chromosome 5 around rs12518218, chromosome 2 around rs10490120, respectively. There are 8, 14, 9 SNPs exceeding the threshold for suggestive significance level for TC, HDL-C, LDL-C level, respectively. In the gene-based analysis, the number of genes that was nominally associated with TC, HDL-C, LDL-C level was 1,038, 1,033, and 1,090, respectively. Although our findings need to be replicated and validated, the data reported here could represent a useful reference for GWAS results of TC, HDL-C, LDL-C level in East Asian population and provide a basis for future molecular biology investigations.

DATA AVAILABILITY
The SNPs datasets of this study have been deposited in the European Variation Archive (EVA) (Accession No. PRJEB23749).

AUTHOR CONTRIBUTIONS
HL and DZ designed the study. HD and CX collected samples and phenotypes. HL and CZ assisted in sample data and sequencing data management. HL and WW analyzed the sequencing data and interpreted the analysis results. HL and DZ drafted the manuscript, XT and HD participated in the discussion, and WW, CX, and CZ revised it. All the authors read the manuscript and agreed to publish. All the authors agreed to be responsible for all aspects of the work. . The x-axis shows the -log10 of expected P-values of association from chi-square distribution and the y-axis shows the -log10 of P-values from the observed chi-square distribution. The black dots represent the observed data, and the red line is the expectation under the null hypothesis of no association.
Supplemental Figure 8 | Quantile-quantile plot for PASCAL pathway results of HDL-C level. The x-axis shows the -log10 of expected P-values of association from chi-square distribution and the y-axis shows the -log10 of P-values from the observed chi-square distribution. The black dots represent the observed data, and the red line is the expectation under the null hypothesis of no association.
Supplemental Figure 9 | Quantile-quantile plot for PASCAL pathway results of LDL-C level. The x-axis shows the -log10 of expected P-values of association from chi-square distribution and the y-axis shows the -log10 of P-values from the observed chi-square distribution. The black dots represent the observed data, and the red line is the expectation under the null hypothesis of no association.
Supplemental Table 1 | Descriptive statistics for subjects in all sample and GWAS sample.
Supplemental Table 2 | Descriptive statistics for subjects in all sample and GWAS sample phenotypic correlation coefficients (95% CI) with covariates' effects in MZ and DZ twin pairs.
Supplemental Table 3 | The summary of SNPs with P < 1 × 10 −5 for association with TC in typed GWAS data.
Supplemental Table 4 | Query SNP enhancer summary for TC level in typed GWAS data.
Supplemental Table 5 | The summary of SNPs with P < 1 × 10 −5 for association with LDL-C in typed GWAS data.
Supplemental Table 6 | Query SNP enhancer summary for LDL-C level in typed GWAS data. Table 7 | The summary of SNPs with P < 1 × 10 −5 for association with TC in GWAS imputation.

Supplemental
Supplemental Table 8 | The summary of SNPs with P < 1 × 10 −5 for association with HDL-C in GWAS imputation.
Supplemental Table 9 | The summary of SNPs with P < 1 × 10 −5 for association with LDL-C in GWAS imputation.
Supplemental Table 10 | The comparison between our imputation results of genome-wide association study and other East Asian studies.
Supplemental Table 11 | The top 20 genes from VEGAS2 gene-based analysis showing the strongest association with TC level in typed GWAS data.
Supplemental Table 12 | The top 20 genes from VEGAS2 gene-based analysis showing the strongest association with LDL-C level in typed GWAS data. Table 13 | Common genes among TC, HDL-C and LDL-C level (P < 0.05).

Supplemental
Supplemental Table 14 | The top 20 pathway results-KEGG, Reactome, and Biocarta (emp-P < 0.05) using PASCAL program for TC level in typed GWAS data.
Supplemental Table 15 | The top 20 pathway results-KEGG, Reactome, and Biocarta (emp-P < 0.05) using PASCAL program for LDL-C level in typed GWAS data.