|Home | About | Journals | Submit | Contact Us | Français|
Resequencing genes provides the opportunity to assess the full spectrum of variants that influence complex traits. Here we report the first application of resequencing to a large population (n = 3,551) to examine the role of the adipokine ANGPTL4 in lipid metabolism. Nonsynonymous variants in ANGPTL4 were more prevalent in individuals with triglyceride levels in the lowest quartile than in individuals with levels in the highest quartile (P = 0.016). One variant (E40K), present in ~3% of European Americans, was associated with significantly lower plasma levels of triglyceride and higher levels of high-density lipoprotein cholesterol in European Americans from the Atherosclerosis Risk in Communities Study and in Danes from the Copenhagen City Heart Study. The ratio of nonsynonymous to synonymous variants was higher in European Americans than in African Americans (4:1 versus 1.3:1), suggesting population-specific relaxation of purifying selection. Thus, resequencing of ANGPTL4 in a multiethnic population allowed analysis of the phenotypic effects of both rare and common variants while taking advantage of genetic variation arising from ethnic differences in population history.
Adipocytes secrete a variety of proteins that regulate glucose and lipid metabolism1. The metabolic effects of these proteins have been largely deduced from studies in mice; less is known about their importance in humans. As a first step toward elucidating the role of adipokines in lipid metabolism in humans, we examined the effects of sequence variation in ANGPTL4, a gene whose expression is induced in adipose tissue and liver during fasting2. Mice with a genetic deletion of ANGPTL4 have lower plasma triglyceride levels3, whereas hepatic overexpression of ANGPTL4 causes hypertriglyceridemia4, hepatic steatosis5 and a reduction in fat mass6. Although the exact role of ANGPTL4 is not clear, it appears to inhibit lipoprotein lipase4, an enzyme that hydrolyzes the triglycerides in circulating lipoproteins to release fatty acids for uptake by adjacent tissues7. Thus, ANGPTL4 may act in a paracrine manner to regulate the partitioning of fatty acids between sites of storage (adipose tissue) and sites of oxidation (heart, skeletal muscle and liver)8.
To determine how sequence variations in ANGPTL4 influence energy metabolism in humans, we sequenced the seven exons and the intron-exon boundaries of the gene (Supplementary Table 1 online) in 3,551 participants in the Dallas Heart Study, a population-based probability sample of Dallas County residents (1,830 African American, 601 Hispanic, 1,045 European American and 75 other ethnicities) whose lipid and glucose metabolism has been characterized in detail9,10. We identified a total of 93 sequence variations (65 in African Americans, 45 in European Americans and 31 in Hispanics), most of which were rare: more than half (n = 49) were found only in one subject, and 86% (n = 80) had a minor allele frequency below 3% (Supplementary Fig. 1 and Supplementary Table 2 online).
To examine the phenotypic effects of sequence variation in ANGPTL4, we stratified the population by race, sex and trait level and then compared the number of nonsynonymous variants in the top and bottom quartiles of the distribution. We excluded variants found in both quartiles. The first trait we examined was fasting levels of plasma triglyceride. Individuals with factors known to affect triglyceride levels (lipid-lowering drugs, diabetes mellitus and heavy alcohol use) were excluded from the analyses. After these exclusions, the number of individuals with nonsynonymous sequence variants in the bottom quartile (n = 13) was significantly greater than the number in the highest quartile (n = 2; P = 0.016) (Fig. 1); we found all sequence variations in separate individuals except R336C, which was present in three European Americans in the lowest quartile of triglycerides. Although European Americans comprised less than one-third of the sample (1,045 out of 3,551), 10 of the 13 individuals with a nonsynonymous sequence variant found in the low-triglyceride group were European Americans. Four of the sequence variations in the low-triglyceride group (IVS3+1 G > A, K217X, K245fs and S302fs) are predicted to truncate translation or interfere with splicing of the mRNA. In contrast to the nonsynonymous variants, the number of synonymous and noncoding variants in the upper and lower tails of the distribution was identical (n = 15).
We performed similar analyses for body mass index (BMI), high-density lipoprotein cholesterol (HDL-C) levels and fasting plasma insulin levels. We did not find any significant differences in the number of nonsynonymous sequence variants in ANGPTL4 in the lowest and highest quartiles of any of these traits (Supplementary Table 3 online).
Next, we determined if any of the sequence variants in ANGPTL4 with a minor allele frequency (MAF) > 1% were associated with trait levels in the three ethnic groups. Only one variant, E40K, which had a MAF of 1.3% in European Americans, was significantly associated with plasma triglyceride levels (Supplementary Table 4 online). The median plasma triglyceride level was 29 mg/dl (27%) lower in carriers than in noncarriers (P = 0.004) in the sample (Table 1). Plasma triglyceride levels cluster with other metabolic risk factors for cardiovascular disease, including adiposity, blood pressure, lipoprotein size distribution, HDL-C levels and insulin sensitivity11. We did not find any significant differences between the carriers and noncarriers in the Dallas Heart Study in mean BMI; blood pressure; fasting levels of glucose, insulin or cholesterol; or hepatic triglyceride content (Table 1). Plasma levels of low-density lipoprotein cholesterol (LDL-C) were significantly higher in carriers than the noncarriers (P = 0.002). The frequencies of the E40K allele in African Americans and Hispanics in the DHS were too low to allow meaningful statistical analysis in these populations (Supplementary Table 5 online).
To confirm these findings, we tested for association between ANGPTL4[E40K] and metabolic phenotypes in two larger population-based studies: the Atherosclerosis Risk in Communities (ARIC) study12 (n = 15,792) and the Copenhagen City Heart Study (CCHS) (n = 10,135)13. In European Americans in the ARIC population, the median plasma triglyceride level was reduced by 16 mg/dl (15%) in the 343 heterozygous carriers (P = 2.2 × 10−10, Table 1). Carriers had significantly lower levels of fasting insulin and LDL-C and significantly higher levels of HDL-C (56 ± 16 versus 51 ± 17 mg/dl, P = 4.0 × 10−7) (Table 1). In the CCHS population, median plasma levels of triglyceride were also significantly lower (141 mg/dl versus 158 mg/dl; P = 1.0 × 10−5) and HDL-C levels significantly higher (63 ± 18 mg/dl versus 60 ± 19 mg/dl, P = 0.0005) in carriers than in non-carriers (Table 1). The E40K heterozygotes also had significantly lower plasma levels of LDL-C. Plasma insulin levels were not available in the CCHS population. As in the Dallas Heart Study, ANGPTL4[E40K] was not associated with BMI, blood pressure or plasma levels of glucose in either ARIC or CCHS.
Thus, the E40K variation was systematically associated with lower plasma triglyceride concentrations in three independent populations. The effects of the variant on other metabolic phenotypes were apparent in the two larger study populations, ARIC and CCHS. Taken together, our data indicate that sequence variation in ANGPTL4 primarily affects plasma levels of triglycerides but also affects other related metabolic parameters, including HDL-C, LDL-C and possibly fasting insulin levels. Furthermore, it is possible that the promoter region and other segments of ANGPTL4 that were not resequenced in this study may contain functionally significant sequence variants.
A notable finding in our population-based study was that loss-of-function alleles in ANGPTL4 were much more common in European Americans than in African Americans. The ratio of nonsynonymous to synonymous variants was substantially greater in European Americans (4:1) than in African Americans (1.3:1) (Fig. 2). To determine if these population differences were a result of natural selection, we analyzed the frequency distribution of the ANGPTL4 sequence variations identified in the Dallas Heart Study using three test statistics (Tajima’s D, Fu and Li’s D and Fu and Li’s Ds)14,15. All three tests showed a significant excess of rare (new) alleles in African Americans, European Americans and Hispanics, consistent with purifying selection acting at the ANGPTL4 locus (Supplementary Table 6 online). Tests for selection have the potential to be confounded by demographic factors such as population expansion, which can also give rise to an excess of rare variants15. However, demographic factors would be expected to affect functional and neutral variants similarly. The disproportionate accumulation of nonsynonymous variants in European Americans is unlikely to result from population expansion and is most consistent with relaxation of purifying selection in this population. We speculate that loss-of-function variants in ANGPTL4 may have been less deleterious to reproductive fitness in Europeans than in Africans. As ANGPTL4 has a role in triglyceride transport and fatty acid metabolism2, changes in selective pressure on ANGPTL4 may reflect shifts in energy use associated with acclimation to colder climates or changes in diet.
Recently, population-based samples have been proposed as an alternative to case-control cohorts for association and resequencing studies16,17. The present study, the first large-scale sequencing survey of unselected individuals in which the resulting genetic data have been coupled to phenotype information, illustrates some of the strengths and limitations of this approach. A major advantage of unselected, well-characterized samples is that sequence variants identified can be tested against multiple phenotypes. The detailed phenotype database available on the Dallas Heart Study participants allowed us to test the spectrum of metabolic consequences associated with genetic variation in ANGPTL4. The use of a multiethnic population for resequencing capitalizes on genetic variation that arises from ethnic differences in population history. Previously, in the Dallas Heart Study we found highly informative sequence variants in PCSK9 and NPC1L1 that were largely confined to African Americans18–20. In contrast, the ANGPTL4 variants identified in this study were more informative in European Americans. Another advantage is that both rare and common sequence variants can be evaluated using multiple approaches.
The increased flexibility afforded by an unselected sample is achieved at the expense of efficiency, as much of the information and statistical power is provided by individuals in the tails of the distribution. Of 23 European Americans heterozygous for ANGPTL4[E40K] in the Dallas Heart Study, 11 had plasma triglyceride levels in the bottom quartile of the distribution, whereas only one was in the top quartile (P = 0.008, Fig. 3). In the ARIC study, 52 of the carriers (49 heterozygotes and 3 homozygotes) had plasma triglyceride levels below the 5th percentile of the population distribution, whereas only six were above the 95th percentile (P = 1.1 × 10−9, Fig. 2). Thus, comparison of the extremes of the population distribution constitutes a powerful and efficient analytical strategy to capture the effects of both common and rare sequence variants on complex traits.
The coding regions of the ANGPTL4 gene were sequenced in those participants of the Dallas Heart Study (DHS) (n = 3,551) from whom fasting venous blood samples were obtained (Supplementary Note online). The DHS is a population-based probability sample of Dallas County (52% African American, self-identified as ‘black’; 29% European American, self-identified as ‘white’; 17% Hispanic and 2% other ethnicities) in which ethnicity was self assigned according to US census categories9. The study was approved by the institutional review board of University of Texas Southwestern Medical Center, and all subjects provided written informed consent before participation.
The genetic associations observed in the DHS were validated in the ARIC and CCHS studies (Supplementary Note). The ARIC study is a prospective study of atherosclerosis initiated in 1987 (ref. 12) in four communities in the USA (Jackson, Mississippi; Minneapolis, Minnesota; Forsyth County, North Carolina and Washington County, Maryland). A randomly selected cohort of approximately 4,000 persons, ages 45–64 years, was selected from each community12. The protocol for the study was approved by the institutional review boards of all centers, and all participants provided written informed consent that included consent for genetic studies. The Copenhagen City Heart Study is a prospective study of ischemic heart disease initiated in 1976. At the third examination (1991–1994), 10,135 individuals participated, and 9,255 gave blood for DNA analysis13. A total of 9,247 individuals were genotyped in the present study. The study was approved by local ethical committees, and all subjects provided written informed consent.
The exons and flanking introns of ANGPTL4 were sequenced in both directions in 3,551 participants in the Dallas Heart Study as described21. The oligonucleotide primers used for sequencing are shown in Supplementary Table 1. All sequence variants identified were verified by manual inspection of the chromatograms, and missense changes were confirmed by an independent resequencing reaction.
Fluorogenic 5′-nucleotidase assays for the ANGPTL4 alleles encoding E40K or the wild-type protein were developed with the use of the TaqMan assay system (Applied Biosystems). The assays were performed on a 7900HT Fast Real-Time PCR instrument with probes and reagents purchased from Applied Biosystems.
The prevalence of nonsynonymous variants in the upper and lower quartiles of the Dallas Heart Study was compared using Fisher’s exact test. Individuals who had diabetes, used lipid-lowering drugs or consumed more than 30 g alcohol per day for a man or 20 g per day for a woman were excluded from the analyses. Risk factor levels between carriers of each ANGPTL4 variant and noncarriers were compared by analysis of variance (ANOVA). For comparison of plasma lipid levels, we included age, sex and BMI as covariates in the model. Plasma levels of triglyceride and insulin were log transformed before analysis. Selective neutrality for the sequence variants identified was tested by Tajima’s method14 and Fu and Li’s methods15.
The authors thank the Joint Genome Institute’s production sequencing group, T. Hyatt, K. Moller Hansen, M. Refstrup, W.S. Schackwitz, J. Martin and A. Ustaszewska for excellent technical assistance and J. Schageman, C. Lee and K. Lawson for statistical analyses. We are indebted to the staff and participants of the Dallas Heart Study, the ARIC study and the CCHS for their important contributions. We thank J. Goldstein and M. Brown for discussions. This work was supported by grants from the Donald W. Reynolds Foundation, the US National Institutes of Health, the Danish Medical Research Council, The Danish Heart Foundation and the Research Fund at Rigshospitalet (Copenhagen University Hospital). Research conducted at the E.O. Lawrence Berkeley National Laboratory and the Joint Genome Institute was performed under the Berkeley Program for Genomic Applications, funded by the US National Heart, Lung, and Blood Institute (HL066681) and Department of Energy Contract DE-AC02-05CH11231 (University of California).