Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 1.
Published in final edited form as: J Hum Genet. 2020 Aug 5;66(2):111–121. doi: 10.1038/s10038-020-0816-9

Genome-wide association studies in Samoans give insight into the genetic architecture of fasting serum lipid levels

Jenna C Carlson 1,2, Daniel E Weeks 1,2, Nicola L Hawley 3, Guangyun Sun 4, Hong Cheng 4, Take Naseri 5, Muagututi’a Sefuiva Reupena 6, John Tuitele 7, Ranjan Deka 4, Stephen T McGarvey 8,9, Ryan L Minster 1
PMCID: PMC7785639  NIHMSID: NIHMS1619878  PMID: 32759990

Abstract

The current understanding of the genetic architecture of lipids has largely come from genome-wide association studies (GWAS). To date, few GWAS have examined the genetic architecture of lipids in Polynesians, and none have in Samoans, whose unique population history, including many population bottlenecks, may provide insight into the biological foundations of variation in lipid levels. Here we performed a GWAS of four fasting serum lipid levels: total cholesterol (TC), high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides (TG) in a sample of 2849 Samoans, with validation genotyping for associations in a replication cohort comprising 1798 Samoans and American Samoans. We identified multiple genome-wide significant associations (P < 5 × 10−8) previously seen in other populations—APOA1 with TG, CETP with HDL, and APOE with TC and LDL—and several suggestive associations (P < 1 × 10−5), including an association of variants downstream of MGAT1 and RAB21 with HDL. However, we observed different association signals for variants near APOE than what has been previously reported in non-Polynesian populations. The association with several known lipid loci combined with the newly identified associations with variants near MGAT1 and RAB21 suggest that while some of the genetic architecture of lipids is shared between Samoans and other populations, part of the genetic architecture may be Polynesian-specific.

Introduction

The Samoan Islands, comprising both the U.S. Territory of American Samoa (American Samoa) and the Independent State of Samoa (Samoa), have experienced a rise in prevalence of cardiovascular disease and other non-communicable diseases in the last 30 years partly due to economic modernization, rapid urbanization, and lifestyle changes such as increased caloric intake and sedentary behavior [1–3]. A 2010 population-based survey in Samoa, which gathered the “discovery cohort” studied further here, found that many Samoans are at elevated risk of cardiovascular disease based on known risk factors— increased total cholesterol (TC), low-density lipoprotein cholesterol (LDL), and triglycerides (TG), as well as decreased high-density lipoprotein (HDL) [4–6]—with 47% of the 2938 adult Samoans examined in the study having elevated TC (≥5.2 mmol/L), 88% of men and 91% of women having elevated LDL (>2.59 mmol/L), and 43% of women having low HDL (<1.29 mmol/L) [1].

Our current understanding of the genetic component of serum lipid level variation has been largely due to genomewide association studies (GWAS) [7]. The Global Lipids Genetics Consortium found strong evidence for 157 loci associated with one or more of these traits using a sample of 188,577 individuals of European, East Asian, South Asian, and African ancestry [8]. However, few GWAS of serum lipid levels have been conducted in Pacific Islanders [9, 10] and to our knowledge only one has included a small number of Polynesians [11]. Previous studies have estimated the heritability of serum lipid levels in Samoans, ranging from 16% for HDL to 23% for TG, and have identified genetic susceptibility loci via linkage analysis [12], warranting further study of the genetic architecture of serum lipid levels in Samoans. Samoans are a genetically-isolated founder population, with unique evolutionary history, making them particularly useful in genomic studies [13, 14]. Thus, genomic studies of serum lipid levels could reveal novel lipid-altering loci specific to Pacific Islander populations, as well as highlight susceptibility loci shared with global populations.

Here we report the results of a GWAS of fasting TC, LDL, HDL, and TG in up to 2,849 individuals from independent Samoa followed by replication in up to 1798 individuals from independent Samoa and American Samoa, as part of ongoing GWAS of cardiometabolic disease and adiposity-related traits in the Samoan Islands [1]. We identified multiple genome-wide significant associations previously seen in other populations—APOA1 with TG, CETP with HDL, and APOE with TC and LDL—and several suggestive associations, including an association between variants downstream of MGAT1 and RAB21 with HDL.

Methods

Discovery cohort and genotyping

The discovery cohort data are available from dbGaP (accession number: phs000914.v1.p1). The discovery cohort of 2849 individuals is drawn from a population-based sample recruited from Samoa in 2010 (Table 1). The sample selection, data collection methods, and phenotyping, including the laboratory assays for serum lipid and lipoprotein levels, have been previously reported [1, 13]. Briefly, adults aged 24.5 to <65 years old who reported Samoan ancestry (based on having four Samoan grandparents) were recruited from 33 villages from the islands of ‘Upolu and Savai’i of Samoa through a population-based sample strategy [1]. Consenting participants completed interviews targeting lifestyle factors related to cardiometabolic health (health history, socio-economic position, dietary intake, and physical activity) and anthropometric measurements (height, weight, blood pressure, body composition), and gave fasting whole blood samples for biochemical and genetic assays. A description of the prevalence of non-communicable diseases and associated risk factors is provided in Hawley et al. [1]. Serum lipid levels were derived from fasting whole blood samples collected after a minimum 10-h overnight fast. Genotyping was performed using Genome-Wide Human SNP 6.0 arrays (Affymetrix). Extensive quality control was conducted on the basis of a pipeline developed by Laurie et al. [15]. Additional details for sample genotyping and genotype quality control are described in Minster et al. [13]. This study was approved by the institutional review board of Brown University and the Health Research Committee of the Samoa Ministry of Health. All participants gave written informed consent via consent forms in Samoan language. Imputation was not performed in this study because prior experience with this population using extant imputation panels such as the Phase 3 1000 Genomes panel showed that the resulting imputed genotypes did not correlate well with observed genotypes [13]. Moreover, extant imputation panels do not contain any haplotypes from individuals of known Polynesian or Pacific Island descent [16–18]. The extent of genetic similarity between currently available reference panels and these Samoan study samples is unknown. We believe it would be inappropriate to use imputed genotypes based on continental reference panels for these samples as they are drawn from a genetically homogeneous Samoan population. Future reference panels that include Polynesian haplotypes are needed to leverage unobserved genetic information in genetically isolated Polynesian populations like those in this study sample.

Table 1.

Demographic, anthropometric, and blood biochemistry statistics of the genotyped discovery and replication cohorts

2010 Discovery Cohort
 2002–2003 Replication Sample Set
 1994–1995 Replication Sample Set
 2002–2003 Replication Sample Set
 1994–1995 Replication Sample Set
Samoa, n = 2849
 Samoa, n = 490
 Samoa, n = 468
 American Samoa, n = 592
 American Samoa, n = 248
Women
 Men
 Women
 Men
 Women
 Men
 Women
 Men
 Women
 Men
n = 1703 n = 1146 n = 245 n = 245 n = 243 n = 225 n = 337 n = 255 n = 137 n = 111
age (years) 44.8 (11.1) 45.6 (11.1) 44.0 (17.0) 40.7 (16.3) 43.3 (8.6) 43.2 (8.9) 43.0 (16.0) 43.2 (16.6) 43.3 (9.1) 45.5 (10.6)
BMI (kg/m2) 34.8 (6.7) 31.3 (5.9) 33.2 (7.7) 28.8 (5.4) 31.9 (5.7) 28.8 (5.0) 36.5 (8.4) 33.4 (7.6) 37.4 (6.9) 34.9 (6.0)
total cholesterol (mg/dL) 199.3 (36.1) 200.3 (38.7) 202.3 (35.9) 195.5 (40.6) 199.1 (37.2) 197.4 (35.0) 187.1 (38.6) 189.6 (37.8) 197.8 (31.2) 194.1 (31.0)
HDL (mg/dL) 46.5 (10.8) 43.7 (11.2) 47.1 (10.3) 46.3 (11.2) 43.4 (11.0) 42.5 (11.4) 40.9 (8.8) 40.0 (8.8) 34.8 (7.6) 32.8 (8.3)
LDL (mg/dL) 130.0 (32.6) 129.6 (35.3) 130.3 (32.8) 128.6 (37.5) 140.2 (34.8) 133.5 (33.0) 118.5 (33.7) 118.9 (35.1) 136.6 (32.4) 136.0 (37.5)
trigylcerides (mg/dL) 114.9 (80.5) 139.3 (113.1) 110.9 (58.9) 120.1 (91.3) 85.6 (72.8) 102.8 (97.2) 130.8 (78.1) 200.2 (206.5) 118.9 (63.4) 168.2 (109.0)
Open in a new tab

Replication cohort and genotyping

The replication cohort of 1798 individuals contains two sample sets recruited from Samoa and American Samoa (Table 1). Although there is substantial economic variation across the two polities, with American Samoans generally having a higher standard of living, the Samoans from both territories form a single socio-cultural unit with frequent exchange of mates; genetically, they represent a single homogenous population [3, 19]. The first sample set, referred to as the 1990–95 replication sample set, contains 716 unrelated individuals derived from a longitudinal study of adiposity and cardiovascular disease risk factors among adults from American Samoa and Samoa (Table 1). Detailed descriptions of the sampling and recruitment have been reported previously [20–22]. Briefly, participants were recruited from 46 villages and worksites in American Samoa in 1990 and 9 villages in Samoa (Western Samoa, at the time of recruitment) in 1991 and followed up 4 years later in 1994 and 1995, respectively. All participants were, at baseline, free of self-reported history of heart disease, hypertension, or diabetes. This study was approved by the institutional review boards of the Miriam Hospital and the American Samoa Department of Health, and the Health Research Committee of the Samoa Ministry of Health. All participants gave written informed consent via consent forms in Samoan language.

The second sample set, referred to as the 2002–2003 replication sample set, contains 1082 individuals from American Samoa and Samoa and was drawn from an extended family-based genetic linkage analysis of cardiometabolic traits (Table 1) [12, 23–26]. Probands and relatives were unselected for obesity or related phenotypes. The recruitment process and criteria used for inclusion in this study have been described in detail previously [24, 26]. This study was approved by the institutional review boards of Brown University and the American Samoa Department of Health and the Health Research Committee of the Samoa Ministry of Health. All participants gave written informed consent via consent forms in Samoan language. Imputation was not performed in these studies for the same rationale as the discovery cohort above, but also because genome-wide marker data was not available for the samples in these studies.

In both replication sample sets, blood samples were collected in the morning after a minimum of 10 h fasting, from which serum lipid levels were derived using assay methods published previously [12, 20]. Genotyping of variants selected for validation in the replication cohort (described below) was performed using custom-designed TaqMan OpenArray Real-Time PCR assays (Applied Biosystems). SNPs that could not be genotyped using OpenArray assays were genotyped individually using TaqMan SNP Genotyping assays (Applied Biosystems). Eight variants could not be genotyped due to technical difficulties.

Statistical analyses

Prior to association analyses, residuals were generated for all four lipid traits. First, traits were transformed to normality with the Box–Cox power transformation; secondly, model selection was performed using step-wise linear regression with initial model covariates previously associated with serum lipid levels: age, age2, sex, log transformed BMI, fasting glucose, smoking status, farming status (as a measure of physical activity), and interactions between age, age2, and sex. The final TC model adjusted for age, age2, sex, age × sex, and age2 × sex; the final LDL and TG models adjusted for age, age2, sex, and age2 × sex; the final HDL model adjusted for age and sex.

Preliminary associations were performed, and variants were selected for validation without consideration of hypolipidemic medication use, as it was not measured. However, participants did self-report use of heart disease medication. Sensitivity analysis revealed that this self-reported use of medication to treat heart disease was significantly associated with TC and LDL (results not shown); individuals reporting such medication use (n = 17) were excluded from analyses. The prioritization of variants for validation genotyping was updated using these analyses, but only after available resources were fully expended. Unfortunately, not all variants that should have been prioritized for validation genotyping were successfully genotyped. All results presented are those of the corrected analyses, removing the individuals with heart disease medication use.

Additional sensitivity analysis was performed for TG by excluding one outlying observation (i.e., TG >4 standard deviations above mean); results did not change qualitatively, and, since the recorded value was within the range of plausible values for TG, the individual was retained for presented analyses.

Association between lipid residuals and autosomal genotypes of 659,492 SNPs with minor allele frequency ≥0.05 and Hardy-Weinberg Equilibrium test P value ≥5 × 10−5 was assessed using linear mixed modeling in GenABEL, including previously-derived empirical kinship estimates to adjust for subject relatedness [13, 27]. The association between X chromosome genotypes and the lipid phenotypes were calculated in GenABEL, without adjustment using the empirical kinship estimates. Genomic inflation due to population stratification and cryptic relatedness was assessed by estimating λGC using the lower 90% of the P value distribution [28]. GWAS P values in the discovery cohort (PD) were compared with a threshold for genome-wide significance of PD < 5 × 10−8 and a suggestive association threshold of PD < 1 × 10−5. Statistical power to detect signals at these thresholds was calculated using the Genetic Power Calculator [29].

Gene-set enrichment analysis with MAGENTA was also performed to identify any biological pathways enriched for discovery association signals [30]. Briefly, gene scores were obtained from the most significant P value among SNPs located within each gene using the association results from each lipid GWAS. Genes scores were adjusted for confounding factors including gene size, number of variants, and linkage disequilibrium-related properties by using step-wise multiple linear regression. The 95th percentile of all gene scores was used as the enrichment cutoff for each trait [31]. Gene-set enrichment P values were obtained for highly ranked gene scores. Gene sets were obtained from Gene Ontology (April 2010), pathway information from the Ingenuity (June 2008) and KEGG (June 2010), and biological processes and molecular function from PANTHER (January 2010).

For each of the lipid traits, the INRICH program [32] was used to test for enrichment of known genes (as constructed from Teslovich et al. [33] and Willer et al. [8]) INRICH tests if more known genes are contained in associated intervals than expected by chance, using permutation based on 1 million replicates to generate experimentwide empirical P values. For each lipid trait, we defined the associated intervals as 100 kb intervals centered on the most significant SNP within association peaks with PD < 1 × 10−4.

We selected 21 regions demonstrating at least suggestive association for association validation in the replication cohort. Subsequent analyses—which excluded 17 participants who were taking heart disease medication—revealed additional 10 regions should have selected for validation. These regions were not followed-up because resources were not available for genotyping; six additional regions from the initial 21 failed to genotype and resources were not available to reattempt the genotyping. The variant from each locus with smallest P value across the four lipid scans (defined as 1 Mb windows surrounding the peak SNP) or a proxy SNP in high linkage disequilibrium with the lowest-P value SNP was selected as representative of the locus for replication genotyping.

Statistical association was measured in the 1990–1995 and 2002–2003 replication sample sets independently, and results were combined using meta-analysis (see below). Association analyses for both sample sets were performed using GenABEL [34] in R [35], using the same regression models as in the discovery cohort but additionally adjusting for polity (American Samoa or Samoa); the 2002–2003 sample set was additionally adjusted using expected kinship, as derived from familial pedigree information [36].

Prior to meta-analysis, quality control was performed using EasyQC to check for strand and allele frequency consistency [37]. P value-based meta-analysis using sample sizes as weights was performed using METAL [38] to generate two P values: one for the meta-analysis of the two replication cohorts (PR) and one for the replication cohorts and discovery sample together (PDR). Resulting meta-analysis signals were evaluated based on genome-wide significance and suggestive thresholds (as described above) and by the contribution of the replication sample to the signal. Effect directions for meta-analysis results of peak SNPs were qualitatively compared with those of previously reported lead SNPs.

For ease of reference, any locus identified here with a corresponding signal within 1 mega base pairs (Mb) in a prior lipid study is referred to by the previously prescribed locus name; [8, 10] for loci not previously associated with lipid traits, the symbol of the gene nearest the peak SNP in the locus or the hyphen-separated symbols of the nearest two genes is used as the locus label.

Results

The demographic, anthropometric, and biochemical characteristics of the 2849 participants composing the discovery cohort for this GWAS of serum lipids levels and the 1798 participants composing the replication cohort are presented in Table 1. A detailed description of the discovery cohort and its trends compared with the historical sample sets making up the replication cohort has been previously reported [1]. Briefly, the average age was similar for all cohorts; average BMI was higher among women compared with men and in American Samoa compared with Samoa; average BMI for men and women in Samoa was higher in more recent studies; average lipid levels are largely similar across cohorts with minor exceptions.

We assessed 659,492 unique genome-wide markers for association with 4 traits—TC, HDL, LDL, and TG—in up to 2849 Samoans in the discovery cohort. Relatedness within the discovery cohort was well controlled using the empirical kinship coefficients; λGC ranged between 1.03 and 1.07 for the four lipid traits (Figs. S2, S4, S6, and S8 in S1 Appendix). We observed 38 genome-wide suggestive or significant associations across 31 loci from the four GWAS (Fig. 1; Figs. S1, S3, S5, and S7 in S1 Appendix; Tables S1, S7, S10, and S13 in S1 Appendix).

Fig. 1.

Fig. 1

Open in a new tab

Manhattan plots for GWAS of four lipid traits in the discovery cohort of 2849 Samoans. The dashed and solid lines denote genomewide suggestive and genome-wide significant P value thresholds (P < 1 × 10−5 and P < 5 × 10−8, respectively). Peaks are labeled with the candidate gene or closest gene in the region if they have at least suggestive association in the discovery cohort for at least one trait and demonstrate evidence of replication or have been previously associated

Genome-wide significant association was observed in the discovery cohort between all four traits and markers near APOE: TC and rs4420638 (PD = 2.67 × 10−16, Fig. 2e); HDL and rs4420638 (PD = 9.07 × 10−9, Fig. 2f); LDL and rs1160985 (PD = 2.61 × 10−20; Fig. 2g); and TG and rs4420638 (PD = 7.44 × 10−10, Fig. 2h). In addition, HDL was associated with markers near CETP (rs289708, PD = 1.19 × 10−11), and TG, with APOA1 (rs6589566, PD = 3.98 × 10−18, Fig. 2c). Suggestive associations were observed between lipid levels and markers at an additional 28 loci, including the MGAT1 and RAB21 loci and HDL (Fig. 2a, d), and APOA1 with TC (rs3741298, PD = 1.63 × 10−7, Fig. 2b). We had 80% power at α = 1 × 10−5 and α = 5 × 10−8 to detect SNPs that account for 1.0% and 1.5%, respectively, of the residual variance in a phenotype.

Fig. 2.

Fig. 2

Open in a new tab

Regional association plots for selected loci. Regional association plots generated in LocusZoom [49] showing -log10 (P values) for SNPs in (a) MGAT1 locus and HDL, (b) APOA1 locus and TC, (c) APOA1 locus and TG, (d) RAB21 locus and HDL, and the APOE locus and (e) TC, (f) HDL, (g) LDL, and (h) TG. Points are color coded within each plot according to pairwise linkage disequilibrium (r2) with the labeled SNPs; the saturation of the color of each plotted SNP measures the linkage disequilibrium (r2) with the labeled SNP sharing the same color

Gene-set enrichment analysis with MAGENTA highlighted, at a <5% false-discovery rate (FDR), several lipid homeostasis pathways and gene ontologies for HDL and TG (Tables S8 and S14 in S1 Appendix). Four gene sets were below the FDR for both HDL and TG: HDL particle remodeling, reverse cholesterol transport, cholesterol efflux, and phospholipid efflux. An additional 12 gene sets were implicated for HDL and three gene sets for TG. The HDL particle remodeling and reverse cholesterol transport gene sets had significant enrichment for TC (Table S3 in S1 Appendix), and a single gene set was implicated with LDL, the amylase pathway (Table S11 in S1 Appendix). All four traits had significant enrichment for known TC, HDL, LDL, and TG loci using the INRICH method (Tables S5, S9, S12, and S15 in S1 Appendix).

Validation of peak SNPs was attempted for 21 loci. At loci with multiple associated variants, the most significant variant was chosen as representative of the locus. For some loci, the exclusion of participants using self-reported heart disease medication resulted in a different peak SNP. Thus, for the APOE locus rs1160985 was genotyped instead of rs4420638 (PD = 2.67 × 10−16 for TC, PD = 9.07 × 10−9 for HDL, and PD = 7.44 × 10−10 for TG); for the APOA1 locus rs964184 was genotyped instead of rs3741298 (PD = 1.63 × 10−7 for TC) or rs6589566 (PD = 3.98 × 10−18 for TG); for the MGAT1 locus rs1038143 was genotyped instead of rs249356 (PD = 1.06 × 10−6 for HDL); for the APOB locus rs754523 was genotyped instead of rs1469513 (PD = 2.71 × 10−6 for LDL).

We successfully genotyped the peak SNP, or a proxy SNP, in the replication cohorts for 15 loci. Two loci (APOA1 with TG, PDR = 1.81 × 10−29; APOE with TC, PDR = 4.29 × 10−21, and LDL, PDR = 1.53 × 10−27) demonstrated genome-wide significant associations in the discovery-replication meta-analysis (Table 2 and Tables S1, S7, S10, and S13 in S1 Appendix). An additional four associations demonstrated evidence of replication with consistent directions of effect and suggestive joint PDR values (GCKR with TG, PDR = 5.62 × 10−8; MGAT1 with HDL, PDR = 2.91 × 10−7; APOA1 with TC, PDR = 1.72 × 10 −6; RAB21 with HDL, PDR = 5.92 × 10−7). Three associations had suggestive joint PDR values driven by the discovery associations only (APOB with LDL, PDR = 5.81 × 10−6; LIPC with HDL, PDR = 9.15 × 10−7; CDH4 with HDL, PDR = 8.77 × 10−6); associations at APOB and CDH4 had consistent directions of effect. Among the remaining loci with at least suggestive association in the discovery sample, but not in the discovery-replication meta-analysis, consistent effect directions were also seen for TC and APOB and ZHX2; LDL and ALG10 and CPNE8 (Table 2 and Tables S1, S7, S10, and S13 in S1 Appendix).

Table 2.

Significant and suggestive loci and replication genotyping

Total cholesterol
Locus SNP Chr BP EA OA Dir PD PR PDR Known gene Traits SAM EAS SAS EUR AMR AFR
APOB rs754523 2 21311691 G A + + + 6.25E-06 0.178 1.20E-05 APOB C,H,L,T 0.247 0.265 0.140 0.309 0.306 0.201
PDE4D rs7711093 5 59593138 G A + + − 3.01E-06 0.747 5.37E-04 0.506 0.643 0.553 0.856 0.775 0.852
LUCAT1 rs10072084 5 90539203 C T + .. 9.48E-06 0.541 0.559 0.422 0.232 0.429 0.822
FILIP1 rs2951921 6 76165524 T C + .. 9.04E-07 0.073 0.022 0.063 0.015 0.030 0.293
ZHX2 rs7841763 8 123971081 T C + + + 4.82E-06 0.631 1.03E-04 0.043 0.023 0.146 0.102 0.058 0.169
APOA1 rs964184a 11 116648917 C G + + + 5.37E-05 0.009 1.72E-06 APOA1 C,H,L,T 0.560 0.760 0.771 0.838 0.723 0.779
SIRT2 rs10405150 19 39387919 C T + .. 6.34E-06 0.056 0.147 0.144 0.081 0.117 0.774
ZNF283 rs16976816 19 44339377 G A + .. 9.78E-06 0.977 0.970 0.994 0.987 0.976 0.864
APOE rs1160985a 19 45403412 C T + + + 2.13E-13 3.36E-09 4.29E-21 APOE C,H,L,T 0.724 0.659 0.590 0.554 0.447 0.378
HDL
Locus SNP Chr BP EA OA Dir PD PR PDR Known gene Traits SAM EAS SAS EUR AMR AFR
STON1-GTF2A1L rs6739536 2 48831901 A G −.. 1.58E-06 0.762 0.676 0.837 0.918 0.842 0.609
MGAT1 rs1038143a 5 180213878 T C − − − 3.72E-06 0.016 2.91E-07 0.309 0.110 0.137 0.148 0.081 0.013
AKAP7 rs3777486 6 131584648 G A − + − 3.09E-06 0.808 4.37E-04 0.976 0.909 0.898 0.858 0.793 0.949
CSMD1 rs1626142 8 4345284 T C − .. 7.67E-06 0.612 0.400 0.435 0.298 0.392 0.654
RAB21 rs328733 12 72197574 T C − − − 2.57E-06 0.036 5.92E-07 0.788 0.626 0.700 0.869 0.738 0.611
ZNF10 rs2292029 12 133734113 A G − .. 4.05E-06 0.179 0.112 0.254 0.245 0.156 0.008
HS6ST3 rs16953620 13 97508453 A G − .. 8.48E-06 0.968 0.950 0.949 0.848 0.911 0.836
LIPC rs10438284 15 58629424 G A − + − 4.00E-07 0.133 9.15E-07 LIPC C,H,T 0.286 0.326 0.173 0.272 0.199 0.047
CETP rs289708 16 57038162 T C − .. 1.19E-11 CETP C,H,L,T 0.905 0.815 0.800 0.860 0.823 0.597
LIPG rs16950739 18 47138509 T C − .. 1.07E-07 LIPG C,H 0.019 0.058 0.190 0.057 0.140 0.004
APOE rs1160985a 19 45403412 C T − + − 0.003 0.342 0.004 APOE C,H,L,T 0.724 0.659 0.590 0.554 0.447 0.378
CDH4 rs817687 20 59753355 A G − − − 2.31E-06 0.237 8.77E-06 0.986 0.924 0.945 0.967 0.855 0.610
LDL
Locus SNP Chr BP EA OA Dir PD PR PDR Known gene Traits SAM EAS SAS EUR AMR AFR
APOB rs754523a 2 21311691 G A + + + 3.25E-06 0.158 5.81E-06 APOB C,H,L,T 0.247 0.265 0.140 0.309 0.306 0.201
KALRN rs6789134 3 123942339 G A + + − 3.22E-06 0.925 1.96E-04 0.078 0.161 0.120 0.046 0.118 0.212
ZHX2 rs7841763 8 123971081 T C + + − 1.80E-06 0.557 3.78E-05 0.043 0.023 0.146 0.102 0.058 0.169
SH2D4B rs10509415 10 82473065 A C + .. 7.96E-06 0.706 0.510 0.735 0.758 0.808 0.708
ALG10 rs3912355 12 34079616 C T + + + 2.12E-06 0.544 3.97E-05 0.855 0.872 0.651 0.605 0.732 0.863
ALG10B rs10880642 12 38554152 A G + .. 5.56E-06 0.802 0.715 0.555 0.481 0.532 0.299
CPNE8 rs11169807 12 39244161 C T + + + 4.77E-06 0.418 4.13E-05 0.794 0.644 0.536 0.505 0.408 0.898
LINC02408 rs17104016 12 67969929 A T + .. 9.29E-06 0.574 0.734 0.924 0.853 0.906 0.838
LINC00922 rs254371 16 65943650 T C + .. 9.04E-06 0.555 0.698 0.650 0.599 0.693 0.884
ZNF283 rs16976816 19 44339377 G A + .. 1.78E-06 0.977 0.970 0.994 0.987 0.976 0.864
APOE rs1160985 19 45403412 C T + + + 2.61E-20 5.07E-09 1.53E-27 APOE C,H,L,T 0.724 0.659 0.590 0.554 0.447 0.378
Triglycerides
Locus SNP Chr BP EA OA Dir PD PR PDR Known gene Traits SAM EAS SAS EUR AMR AFR
GCKR rs780094 2 27741237 T C + + + 9.84E-07 0.01 5.62E-08 GCKR C,T 0.334 0.476 0.198 0.411 0.360 0.132
CD200 rs2399416 3 112059213 A G + + − 5.12E-06 0.668 9.29E-04 0.021 0.148 0.140 0.393 0.272 0.101
SPIN1 rs7861888 9 90886340 A G + .. 4.24E-06 0.706 0.719 0.897 0.921 0.842 0.989
APOA1 rs964184a 11 116648917 C G + + + 2.37E-17 8.97E-14 1.81E-29 APOA1 C,H,L,T 0.440 0.240 0.229 0.162 0.277 0.221
KIRREL3 rs3018434 11 126805881 G A + .. 4.16E-06 0.916 0.803 0.920 0.866 0.857 0.974
APOE rs1160985a 19 45403412 T C + − + 0.312 0.836 0.507 APOE C,H,L,T 0.276 0.341 0.410 0.446 0.553 0.623
Open in a new tab

PDdiscovery cohort GWAS p value

PR replication cohort p value

PDRjoint discovery and replication cohorts p value

Known gene Loci observed in Teslovich et al. [33] or Willer et al. [8]

Traits Traits locus associated with in Teslovich et al. [33] or Willer et al. [8]

C TC

H HDL

L LDL

T TG

EA effect allele, OA other allele (reference allele), Dir direction of the effect in each of the four samples (+ indicates the effect allele is increasing the trait value on the raw scale), SAM Samoan effect allele frequency (EAF), EAS East Asian EAF, SAS South Asian EAF, EUR European EAF, AMR Admixed American EAF, AFR African EAF

Significant P values < 5 × 10−8 are in bold, while suggestive P values between 1 × 10−5 and 5 × 10−8 are in italics

a

The SNP genotyped in the replication population was not the peak SNP at this locus for this trait

We compared the directions of effect to those previously reported in Willer et al. [8] and Teslovich et al. [33] (as made available by the Center for Statistical Genetics at the University of Michigan: http://csg.sph.umich.edu/willer/public/lipids2013/ and http://csg.sph.umich.edu/willer/public/lipids2010/ ) for APOB, GCKR, APOA1, LIPC, LIPG, and APOE (i.e., genome-wide suggestive loci that have been previously associated with lipid traits). For all variants from established lipid loci, the locus name and associated traits from Teslovich et al. and Willer et al. are given in Table 2. We observed a consistent direction of effect for the representative SNP for all associations (Table 2).

The effect allele frequencies (EAFs) in the two samples —discovery and replication—were largely similar for each of the 15 successfully genotyped SNPs (Tables S1, S7, S10, and S13 Appendix). However, many loci had markedly different EAFs between Samoans and other 1000 Genomes populations (Table 2). For example, compared with 1000 Genomes populations, there were higher EAFs in Samoans for rs964184 near APOA1 (G allele frequency: 0.440 in Samoans vs. <0.277 in 1000 Genomes populations), rs1160985 near APOE (C allele frequency: 0.724 in Samoans vs. <0.659 in 1000 Genomes populations), and rs1038143 near MGAT1 (A allele frequency: 0.309 in Samoans vs. <0.148 in 1000 Genomes populations).

Discussion

In this study, we examined four measures of fasting lipid levels—TC, HDL, LDL, and TG—for associations with 659,492 SNPs from a genome-wide array in a discovery cohort of 2849 Samoans, with follow-up genotyping of significant and suggestive findings in a replication cohort comprising 1798 Samoans from Samoa and American Samoa. Thirty-one loci had at least suggestive evidence of association with one or more lipid traits in the discovery cohort, of which eight have been reported to be associated with lipid levels previously: APOB, GCKR, MGAT1, APOA1, LIPC, CETP, LIPG, and APOE [8, 10, 33]. Enrichment analyses highlighted known lipid metabolism gene sets and previously associated lipid loci.

We observed a difference in the architecture of the statistical association signals between the four lipid traits and variants near APOE (Fig. 2e–h). The peak SNP for TC and LDL was rs1160985, an intronic variant in TOMM40 upstream of APOE; whereas the peak SNP for HDL and TG was rs4420638, an intergenic variant downstream of APOC1 and APOE. rs1160985 demonstrated evidence of replication for TC and LDL but not for HDL and TG, consistent with the discovery findings. Sanna et al. report two independent signals for APOE, led by missense variants rs7412 and rs429358 [39]. Neither rs7412 nor rs429358 was genotyped in these samples so APOE isoform could not be determined. However, rs4420638, the sentinel SNP in our analyses of TC, HDL, and TG and in Sanna et al., is in high LD with rs429358. The sentinel SNP from our analysis of LDL was rs1160985, which is not in LD with rs4420638 in Samoans (r2 = 0.093). It appears that this variant may be flagging an additional independent signal than the two reported in Sanna et al., as it is in low LD with rs7412 in all 1000 Genomes population (max r2 = 0.23 in Southern Han Chinese; LDPop accessed 5 May 2020 [40]). While this could support a shared genetic architecture for TC and LDL and for HDL and TG in Samoans, this study was not positioned to adequately capture the association signal present at this locus. Further fine-mapping of this locus, including sequencing or imputation of the 19q13.2 region, is needed to dissect the genetic architecture of APOE and lipid levels in Polynesians and would give further insight into how much genetic architecture is shared between Samoans and better-studied populations.

We did not observe suggestive or genome-wide significant association with several loci which have figured prominently in multiple lipid GWAS (e.g., LPL, LDLR, CILP2, FADS1/2/3, ANGPTL3, SORT1, PPP1R3B, MIX-IPL, HNF4A, PCSK9, GALNT2, HMGCR) either because we lacked sufficient power to detect their effects, the effects are negligible in Samoans, or the allele frequencies of associated variants are different enough in Samoans to hinder detection. However, it is important to note that this study was not designed to evaluate the effect of known lipids loci in Samoans, nor were previously associated loci examined specifically.

We detected and replicated a suggestive association between HDL and a variant on 5q35.3 (Fig. 2a). While the peak SNP lies within an intron of BTNL8, the variant selected for follow-up genotyping is intergenic, downstream of MGAT1. Previous association between the MGAT1 locus and HDL have been reported, including a significant association with a variant in MGAT1, rs634501, in a study of Han Chinese and a suggestive association with variants near MGAT1 in a GWAS in the Micronesian population of Kosrae [10, 41]. Furthermore, in Europeans, variation near MGAT1 has been associated with BMI, serum fatty acid levels and composition, and glucose response [42–44]. The encoded MGAT enzyme plays a major role in the absorption of dietary fat in the intestine [45]. As this variant has only demonstrated evidence of association in non-European population, it may be tagging another variant or haplotype that is rare or absent in Europeans but is present in Pacific Island or South Asian populations and is itself associated with HDL.

We also detected and replicated a novel suggestive association between HDL and a variant downstream of RAB21 (Fig. 2d). Unlike the variant downstream of MGAT1, we observed similar allele frequencies between Samoans and 1000 Genomes populations (i.e., <20% difference between Samoans and another 1000 Genomes population) in the HDL-associated variant downstream of RAB21. This region, 12q21, was previously seen in linkage analysis with both univariate and bivariate scans of TC and LDL [12]. The individuals included in this linkage analysis are also included in our 2002–2003 replication sample set, however, they do not appear to be driving the association signal near RAB21 (Table S7 in S1 Appendix). Variation near RAB21 has been previously associated with obesity [46]. RAB21 belongs to the family of monomeric GTPases involved in control of cellular membrane trafficking and is involved in the targeted trafficking of integrins and the regulation of cell adhesion and migration [47, 48].

This study is limited in drawing conclusions about the genetic architecture of lipids in Samoans, as replication genotyping was unavailable for many loci and due to the lack of genome-wide imputation. Future studies, evaluating the evidence of association between associations seen here in separate cohorts as well fine-mapping loci with genotype imputation (given the availability of a relevant reference panel), are necessary to fully evaluate the genetic architecture of lipids in Samoans.

This is the first GWAS of lipid phenotypes in Samoans, and we observed association with many known lipid loci, which was further supported by the gene-set enrichment analysis highlighting lipid metabolism gene sets. However, the difference in association results near APOE, coupled with evidence of Pacific-Islander-specific associations with MGAT1 and RAB21 suggest that some, but not all, of the genetic architecture of lipids is shared between Samoans and other populations. Given this evidence of a partially distinct genetic architecture of lipids in Samoans, further investigation and fine-mapping of lipid loci, especially that across multiple ethnicities, is warranted.

Supplementary Material

Supplementary Material

Acknowledgements

We would like to thank the Samoan participants of the study, local village authorities, and the many Samoan and other field workers over the years. We acknowledge the Samoan Ministry of Health, the Samoa Bureau of Statistics, and the American Samoan Department of Health for their support of this research. We give particular thanks to two research assistants, Melania Selu and Vai-moana Lupematasila, who contributed to the 2010 recruitment and continue to assist us in our work in Samoa. This work was funded by the National Institute of Health grants R01HL093093 (STM), R01HL133040 (RLM), R01AG09375 (STM), R01HL52611 (MI Kamboh), R01DK59642 (STM), and R01DK55406 (RD). Genotyping was performed in the Core Genotyping Laboratory at the University of Cincinnati, funded by National Institutes of Health grant P30ES006096 (SM Ho).

Footnotes

Supplementary information The online version of this article (https://doi.org/10.1038/s10038-020-0816-9) contains supplementary material, which is available to authorized users.

Data availability

The discovery cohort data are available from dbGaP (accession number: phs000914.v1.p1). The data from the replication cohort (recruited in 1990–1995 and 2002–2003) are not available as participants were not consented for data sharing.

Conflict of interest

The authors declare that they have no conflict of interest.

References

  • 1.Hawley NL, Minster RL, Weeks DE, Viali S, Reupena MS, Sun G, et al. Prevalence of adiposity and associated cardiometabolic risk factors in the samoan genome-wide association study. Am J Hum Biol. 2014;26:491–501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Keighley E, McGarvey ST, Quested C, Mccuddin C, Viali S, Maga UA. Nutrition and health in modernizing Samoans: temporal trends and adaptive perspectives In: Ohtsuka R, Ulijaszek SJ, editors. Health Change in the Asia-Pacific Region: Biocultural and Epidemiological Approaches. Cambridge, NY: Cambridge University Press; 2007. p. 147–91. [Google Scholar]
  • 3.McGarvey ST. Cardiovascular disease (CVD) risk factors in Samoa and American Samoa, 1990–95. Pac Health Dialog 2001;8:157–62. [PubMed] [Google Scholar]
  • 4.Kannel WB, Castelli WP, Gordon T, McNamara PM. Serum cholesterol, lipoproteins, and the risk of coronary heart disease. Framingham Study Ann Intern Med 1971;74:1–12. [DOI] [PubMed] [Google Scholar]
  • 5.Gordon T, Castelli WP, Hjortland MC, Kannel WB, Dawber TR. High density lipoprotein as a protective factor against coronary heart disease. The Framingham Study. Am J Med 1977;62:707–14. [DOI] [PubMed] [Google Scholar]
  • 6.Roeters van Lennep J, Westerveld HT, Erkelens DW, van der Wall EE. Risk factors for coronary heart disease: implications of gender. Cardiovasc Res. 2002;53:538–49. [DOI] [PubMed] [Google Scholar]
  • 7.Lange LA, Willer CJ, Rich SS. Recent developments in genome and exome-wide analyses of plasma lipids. Curr Opin Lipido. 2015;26:96–102. [DOI] [PubMed] [Google Scholar]
  • 8.Willer CJ, Schmidt EM, Sengupta S, Peloso GM, Gustafsson S, Kanoni S, et al. Discovery and refinement of loci associated with lipid levels. Nat Genet. 2013;45:1274–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Burkhardt R, Kenny EE, Lowe JK, Birkeland A, Josowitz R, Noel M, et al. Common SNPs in HMGCR in micronesians and whites associated with LDL-cholesterol levels affect alternative splicing of exon13. Arterioscler Thromb Vasc Biol 2008;28:2078–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Lowe JK, Maller JB, Pe’er I, Neale BM, Salit J, Kenny EE, et al. Genome-wide association studies in an isolated founder population from the Pacific Island of Kosrae. PLoS Genet 2009;5: e1000365. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Dumitrescu L, Carty CL, Taylor K, Schumacher FR, Hindorff LA, Ambite JL, et al. Genetic determinants of lipid traits in diverse populations from the population architecture using genomics and epidemiology (PAGE) study. PLoS Genet 2011;7:e1002138. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Åberg K, Dai F, Sun G, Keighley E, Indugula SR, Bausserman L, et al. A genome-wide linkage scan identifies multiple chromosomal regions influencing serum lipid levels in the population on the Samoan islands. J Lipid Res. 2008;49:2169–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Minster RL, Hawley NL, Su C- T, Sun G, Kershaw EE, Cheng H, et al. A thrifty variant in CREBRF strongly influences body mass index in Samoans. Nat Genet 2016;48:1049–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kristiansson K, Naukkarinen J, Peltonen L. Isolated populations and complex disease gene identification. Genome Biol 2008;9:109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Laurie CC, Doheny KF, Mirel DB, Pugh EW, Bierut LJ, Bhangale T, et al. Quality control and quality assurance in genotypic data for genome-wide association studies. Genet Epidemiol. 2010; 34:591–602. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Howie B, Marchini J, Stephens M. Genotype imputation with thousands of genomes. G3 (Bethesda). 2011;1:457–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.McCarthy S, Das S, Kretzschmar W, Delaneau O, Wood AR, Teumer A, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat Genet. 2016;48:1279–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Das S, Abecasis GR, Browning BL. Genotype imputation from large reference panels. Annu Rev Genomics Hum Genet. 2018;19:73–96. [DOI] [PubMed] [Google Scholar]
  • 19.Deka R, Mc Garvey ST, Ferrell RE, Kamboh MI, Yu LM, Aston CE, et al. Genetic characterization of American and Western Samoans. Hum Biol 1994;66:805–22. [PubMed] [Google Scholar]
  • 20.McGarvey ST, Levinson PD, Bausser-Man L, Galanis DJ, Hornick CA. Population change in adult obesity and blood lipids in American Samoa from 1976–1978 to 1990. Am J Hum Biol 1993;5:17–30. [DOI] [PubMed] [Google Scholar]
  • 21.Chin-Hong PV, McGarvey ST. Lifestyle incongruity and adult blood pressure in Western Samoa. Psychosom Med 1996;58: 131–7. [DOI] [PubMed] [Google Scholar]
  • 22.Galanis DJ, McGarvey ST, Quested C, Sio B, Afele-Fa’amuli SA. Dietary intake of modernizing Samoans: implications for risk of cardiovascular disease. J Am Diet Assoc. 1999;99:184–90. [DOI] [PubMed] [Google Scholar]
  • 23.Åberg K, Dai F, Sun G, Keighley ED, Indugula SR, Roberts ST, et al. Susceptibility Loci for Adiposity Phenotypes on 8p, 9p, and 16q in American Samoa and Samoa. Obesity. 2009;17:518–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Dai F, Keighley ED, Sun G, Indugula SR, Roberts ST, Åberg K, et al. Genome-wide scan for adiposity-related phenotypes in adults from American Samoa. Int J Obes. 2007;31:1832–42. [DOI] [PubMed] [Google Scholar]
  • 25.Åberg K, Dai F, Viali S, Tuitele J, Sun G, Indugula SR, et al. Suggestive linkage detected for blood pressure related traits on 2q and 22q in the population on the Samoan islands. BMC Med Genet. 2009;10:107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Dai F, Sun G, Åberg K, Keighley ED, Indugula SR, Roberts ST, et al. A whole genome linkage scan identifies multiple chromosomal regions influencing adiposity-related traits among Samoans. Ann Hum Genet. 2008;72:780–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Chen W-M, Abecasis GR. Family-based association tests for genomewide association scans. Am J Hum Genet. 2007; 81:913–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Devlin B, Roeder K. Genomic control for association studies. Biometrics. 1999;55:997–1004. [DOI] [PubMed] [Google Scholar]
  • 29.Purcell S, Cherny SS, Sham PC. Genetic power calculator: design of linkage and association genetic mapping studies of complex traits. Bioinformatics 2003;19:149–50. [DOI] [PubMed] [Google Scholar]
  • 30.Carithers LJ, Ardlie K, Barcus M, Branton PA, Britton A, Buia SA, et al. A novel approach to high-quality postmortem tissue procurement: The GTEx Project. Biopreserv Biobank 2015; 13:311–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Segrè AV, Groop L, Mootha VK, Daly MJ, Altshuler D. Common inherited variation in mitochondrial genes is not enriched for associations with type 2 diabetes or related glycemic traits. PLoS Genet. 2010;6:e1001058. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Lee PH, O’Dushlaine C, Thomas B, Purcell SM. INRICH: interval-based enrichment analysis for genome-wide association studies. Bioinformatics. 2012;28:1797–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Teslovich TM, Musunuru K, Smith AV, Edmondson AC, Stylianou IM, Koseki M, et al. Biological, clinical and population relevance of 95 loci for blood lipids. Nature 2010;466:707–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Aulchenko YS, Ripke S, Isaacs A, van Duijn CM. GenABEL: an R library for genome-wide association analysis. Bioinformatics 2007;23:1294–6. [DOI] [PubMed] [Google Scholar]
  • 35.R Core Team. R: A language and environment for statistical computing. Vienna, Austria, Austria: R Foundation for Statistical Computing; 2017. [Google Scholar]
  • 36.Therneau TM, Sinnwell J kinship2: Pedigree functions. R package version 1.6.4; 2014. [Google Scholar]
  • 37.Winkler TW, Day FR, Croteau-Chonka DC, Wood AR, Locke AE, Magi R, et al. Quality control and conduct of genome-wide association meta-analyses. Nat Protoc. 2014;9:1192–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics. 2010;26:2190–1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Sanna S, Li B, Mulas A, Sidore C, Kang HM, Jackson AU, et al. Fine mapping of five loci associated with low-density lipoprotein cholesterol detects variants that double the explained heritability. PLoS Genet. 2011;7:e1002198. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Machiela MJ, Chanock SJ. LDlink: a web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants. Bioinformatics. 2015;31:3555–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Wu J, Yin R- X, Zhou Y- G, Zhang Q- H, Wu J- Z, Chen W- X. Association between the MGAT1 rs634501 polymorphism and serum lipid traits in the Chinese Han and Maonan ethnic groups. Int J Clin Exp Pathol. 2018;11:5923–37. [PMC free article] [PubMed] [Google Scholar]
  • 42.Johansson A, Marroni F, Hayward C, Franklin CS, Kirichenko AV, Jonasson I, et al. Linkage and genome-wide association analysis of obesity-related phenotypes: association of weight with the MGAT1 gene. Obesity 2010;18:803–8. [DOI] [PubMed] [Google Scholar]
  • 43.Jacobsson JA, Rask-Andersen M, Risérus U, Moschonis G, Koumpitski A, Chrousos GP, et al. Genetic variants near the MGAT1 gene are associated with body weight, BMI and fatty acid metabolism among adults and children. Int J Obes. 2012;36:119–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Del-Aguila JL, Beitelshees AL, Cooper-DeHoff RM, Chapman AB, Gums JG, Bailey K, et al. Genome-wide association analyses suggest NELL1 influences adverse metabolic response to HCTZ in African Americans. Pharmacogenomics J. 2014;14:35–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Yen C- LE, Stone SJ, Cases S, Zhou P, Farese RV. Identification of a gene encoding MGAT1, a monoacylglycerol acyltransferase. Proc Natl Acad Sci. 2002;99:8512–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Turcot V, Lu Y, Highland HM, Schurmann C, Justice AE, Fine RS, et al. Protein-altering variants associated with body mass index implicate pathways that control energy intake and expenditure in obesity. Nat Genet 2018;50:26–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Pellinen T, Ivaska J Integrin traffic. J Cell Sci. 2006;119(Pt 18):3723–31. [DOI] [PubMed] [Google Scholar]
  • 48.Pellinen T, Tuomi S, Arjonen A, Wolf M, Edgren H, Meyer H, et al. Integrin trafficking regulated by Rab21 is necessary for cytokinesis. Dev Cell 2008;15:371–85. [DOI] [PubMed] [Google Scholar]
  • 49.Pruim RJ, Welch RP, Sanna S, Teslovich TM, Chines PS, Gliedt TP, et al. LocusZoom: regional visualization of genome-wide association scan results. Bioinformatics. 2010;26:2336–7. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material
NIHMS1619878-supplement-Supplementary_Material.pdf (1.4MB, pdf)

RESOURCES