Atherosclerosis and Lipoproteins |
From the Division of Epidemiology (J.M.P., D.K.A., L.D.A.), School of Public Health, University of Minnesota, Minneapolis, Minn; Section of Preventive Medicine and Epidemiology (R.H.M.), Boston University School of Medicine, Boston, Mass; Department of Psychiatry (H.C.), University of Utah, Salt Lake City, Utah; Department of Public Health Sciences (S.S.R.), Wake Forest University School of Medicine, Winston-Salem, NC; Division of Biostatistics (M.A.P.), Washington University Medical School, Saint Louis, Mo; and Department of Epidemiology (G.H.), University of North Carolina, Chapel Hill, NC.
Reprint requests to Donna K. Arnett, PhD, Division of Epidemiology, School of Public Health, University of Minnesota, 1300 S 2nd St, Suite 300, Minneapolis, MN 55454. E-mail arnett{at}epi.umn.edu
| Abstract |
|---|
|
|
|---|
Key Words: HDL cholesterol genome scan anonymous marker linkage quantitative trait loci genetic epidemiology
| Introduction |
|---|
|
|
|---|
At least 9 twin studies311 and 14 family studies5,9,1223 have estimated the heritability of serum HDL-C. Heritability estimates ranged from 24% to 83%, with most studies in the 40% to 60% range. Variability in heritability estimates across studies could be due to a number of factors, including differences in study sample sizes, statistical methods, ages of the study participants, covariate adjustment, reliability of the HDL-C measurement, and whether only twins or families were studied. Despite these differences, most studies found HDL-C to be a highly heritable characteristic, warranting further research into the genetic sources of phenotypic variation.
The demonstration that a heritable trait has a multimodal distribution pattern lends further support to a significant genetic component to the variation in the trait. Segregation analysis has been used to determine the probable mode of inheritance of a disease or trait, as well as to estimate the residual polygenic heritability. At least 9 studies2432 have applied segregation analysis to the detection of major gene loci that influence variation in HDL-C. After adjustment for environmental covariates, major gene effects were detected in some studies, but the estimated allele frequency for low HDL-C was very small (range 0.003 to 0.086). These low-frequency alleles would explain little of the total variability in HDL-C. However, environmental modifiers of the major gene may obscure its detection unless gene-environment interactions are incorporated into the model. Mahaney et al32 developed a model that included genotype-specific sex and covariate effects and age-by-sex interactions and found a major gene effect for high HDL-C with a high-frequency allele (0.84). Consistent with these findings, all studies indicate a significant residual polygenic heritability (0.30 to 0.60), similar to results from twin and family studies.
These heritability and segregation results prompted the initiation of anonymous marker genome scans to identify genomic regions that exhibit linkage to genetic variation in HDL-C. Currently, there are 5 published reports of genome-wide scans for identification of quantitative trait loci (QTLs) that contribute to variation in either total HDL-C or HDL-C subfractions. Two of these studies were conducted in large Mexican-American families and found significant linkage of HDL-C33 and various HDL-C subclass phenotypes34 to new genomic regions. Using a 10-cM map, investigators in the San Antonio Family Diabetes Study mapped total HDL-C (logarithm of the odds [LOD] 3.6) to chromosome 9p (between markers D9S925 and D9S1121).33 Investigators in the San Antonio Family Heart Study used a 15-cM map and found significant linkage of the HDL2a subfraction (LOD=4.87) to chromosome 8q (near marker D8S1128) and to a location near the hepatic lipase gene on chromosome 15 (LOD=3.26). Several other LOD scores >1.9, considered "suggestive" of linkage, warrant further investigation. Using a denser map (
6 cM), Newman et al35 found suggestive linkage of HDL-C to location 45.3 cM on chromosome 5p in 522 adult individuals from a genetically isolated population in Canada (P=0.02). In addition, Kort et al36 found a QTL on chromosome 11q23 linked to the low HDL-C phenotype in large Utah pedigrees (LOD=3.5). Imperatore et al37 reported suggestive linkage of a location near marker D3S2053 on chromosome 3q (LOD=2.64). A recently published genome scan of several lipid phenotypes in the Rochester Family Heart Study found no evidence for linkage of HDL-C to any autosomal location.38 The authors did report suggestive linkage of apolipoprotein (apo) A-II to both chromosome 4 (LOD=2.35 at 169 cM) and chromosome 5 (LOD=2.13 at 79 cM). The goal of the present study was to conduct genomic scans with a dense marker map and a larger sample size than the previously reported studies as part of the National Heart, Lung, and Blood Institute (NHLBI) Family Heart Study (FHS).
| Methods |
|---|
|
|
|---|
Phenotyping
Participants (n=5348) were asked to fast the night before attending a 3.5-hour clinic examination that included measurements of anthropometry, blood pressure, and pulmonary function, as well as a carotid artery ultrasound and an ECG. Blood was drawn in all participants for standard laboratory tests (lipids, hemostatic factors, insulin, glucose, blood chemistries, and hematology) and for long-term storage. Interviews and questionnaires assessing lifestyle; medical, personal, and reproductive histories; medication use; food frequency intake; physical exercise; and psychosocial characteristics were conducted. An additional 627 remote site examinations were conducted for participants unable to visit the clinic. Although not all of the clinic components could be completed during remote site visits, blood was also collected from individuals who visited the remote sites. Blood samples from both clinic visits and remote site examinations (total N=5975) were stored for additional phenotyping and genotyping at a later date.
The laboratory at Fairview-University Medical Center at the University of Minnesota performed all blood assays on samples from 5975 participants. HDL-C was measured by COBAS FARA (Roche Diagnostic Systems) after precipitation of the other lipoprotein fractions by dextran sulfate.40 The laboratory variability of the HDL-C assay was calculated in 299 blind replicates, yielding a very high reliability coefficient of 0.98. Body mass index (BMI) was calculated as weight (in kilograms) divided by height (in meters) squared.
Genotyping: Mammalian Genotyping Service Sample
A total of 1027 white individuals from 101 families were genotyped for the entire genome by the NHLBI Mammalian Genotyping Service (MGS) in Marshfield, Wis.41 Highly polymorphic autosomal markers (n=383) from the CHLC Screening Set 10 were used, with an average distance of
9.5 cM between markers. These 101 families were chosen because they were the largest FHS pedigrees containing at least 1 individual above the age-sex specific 80th percentile of the Individual Risk Score for CHD (a score derived for each individual using sex-specific proportional hazards models to predict the age of onset of CHD, based on carotid artery intima-media thickness, lipids, BMI, blood pressure, hypertension, diabetes, and CHD family risk score). In general, these pedigrees include marker data on a proband, his/her spouse, his/her siblings, and his/her children. The marker heterozygosity for this sample was 76%. Fluorescent dyes were incorporated into the FHS DNA samples during amplification by a standardized polymerase chain reaction protocol. The DNA fragments were then separated by electrophoresis on a 6% acrylamide gel with 100 lanes. A Scanning Fluorescence Detector (SCAFUD) was used to read the alleles, and software automatically scored each individual genotype. Detailed information on the genotyping methods, instruments, and software used can be found at the Marshfield Laboratory World Wide Web site.42
Marker inconsistencies were evaluated in 2 steps with the genetics programs ASPEX,43 PEDSYS,44 and MAPMAKER/SIBS.45 Reported pedigree relationships were validated with the marker data. Although these steps ensure that the genotypic data will be consistent with the reported relationships as a whole, there may be isolated markers that remain inconsistent with the reported family structure. To resolve these inconsistencies, either a single subject in a pedigree or the entire family would be set to "missing" for that genetic marker. After quality control processing at the FHS Coordinating Center, the average missing data rate across all MGS markers was 2.8±2.2%, and 91.4% of these markers were complete for at least 95% of participants. At least 95% of the anonymous markers were complete for 93.4% of all subjects (n=959). Because of the limited number of genotypes scored by MGS, no phantom replicates were provided; however, internal quality control procedures were followed at the MGS Laboratory.
Statistical Analysis
To increase the relative contribution of genetic components of variance to the total variance of HDL-C, we have removed the effects of important environmental predictors of HDL-C.46 Adjustment for these covariates was achieved by generating a nested series of sex-specific residual phenotypes. These residual phenotypes were created by regressing covariates on HDL-C and retaining those covariates when they explained a significant portion of the total phenotypic variance. An HDL-C residual phenotype was created separately for males and females and included the covariates age, age2, FHS field center, and BMI. These covariates explained
11% of the trait variation in men and 13% in women. This sex-specific residual phenotype was approximately normally distributed.
Quantitative-trait linkage analyses were performed with version 2.1 of the program GENEHUNTER.47,48 All analyses were completed with a variance-components model to examine evidence for linkage of a QTL for total HDL-C with an autosomal genome-wide map of 383 anonymous markers. GENEHUNTER estimates the amount of variance in a quantitative trait that can be attributed to a QTL at that marker position. Maximum likelihood values for the mean trait value, additive and dominant variance components for the QTL, additive and dominant variance components due to other, unlinked loci, and an environmental variance were calculated. We tested the hypothesis that the genetic variance due to the QTL equals zero by comparing the log likelihoods of a model in which the observed variances between relatives was determined as a function of the identity-by-descent relationships at a given marker locus with a model in which the QTL variance components were constrained to zero. The difference in these 2 log likelihoods is similar to the classic LOD score in linkage analysis. The multipoint identity-by-descent method used the proportion of alleles shared identical by descent at genotyped loci to estimate identity-by-descent sharing at 5 evenly spaced points between adjacent markers along a chromosome for each nuclear family. Allele frequencies were derived from the entire (n=1027) sample genotyped by MGS.
Because of computational limitations, GENEHUNTER was unable to use all data from every family genotyped for the MGS marker set. As a result, we were only able to perform analyses on families with 24 max bits or fewer (max bits are defined as twice the number of nonfounders [NF] minus the number of founders [F], or 2NF-F). To meet these constraints, we modified our pedigrees in 2 ways. First, we divided the 4 largest families into 8 new families. Three additional families were "trimmed" by removing 9 genotyped individuals from the analysis based on their relative contribution to the power to detect linkage. These individuals were chosen from the largest sibships within a pedigree and were either missing covariate information or were closest to the sex-specific raw HDL-C mean value. Table 1 shows the pedigree size distribution after modification. All pedigrees included at least 5 individuals, with a mean pedigree size of 9.7. The 9 individuals deleted represented <1% of our total genotyped sample.
|
| Results |
|---|
|
|
|---|
|
Residual heritability of HDL-C was estimated at 53%. Multipoint linkage analyses were performed across all 22 autosomes for sex-specific and age-, age2-, FHS field center, and BMI-adjusted total HDL-C. As shown in the Figure, only 1 LOD score approached genome-wide significance,49,50 occurring at location 39.9 cM on chromosome 5p (LOD=3.64, P2=0.0002) with a 1-LOD support interval of 33.6 to 53.7 cM. This LOD score peak is in the 9-cM region between markers D5S2845 and D5S1470. Both markers are trinucleotide repeats and have high heterozygosities of 0.68 and 0.83. A significant portion of the genetic variance (63%) was partitioned to a dominant QTL. When only additive genetic variances were modeled, the maximum LOD score was attenuated to 2.95. Location 27.5 cM on chromosome 13p (LOD=2.36, P2=0.004) produced the second-highest LOD score peak in our data. The 1-LOD support interval for this possible QTL was 20.4 to 32.8 cM and was between markers D13S1493 and GATA86H01. However, when only the additive genetic variance was modeled, there was no evidence for linkage at this location (LOD=0.16). Additional QTLs met the nominal significance level (approximate LOD score >1.30, P2
0.05) on chromosomes 1, 4, 6, and 8 (Table 3).
|
|
| Discussion |
|---|
|
|
|---|
The Rochester Family Heart Study researchers reported another suggestive linkage of apoAII to location 169 cM on chromosome 4q (LOD=2.35). We also found evidence of a QTL affecting HDL-C only 11 cM telomeric to this location at 180.8 cM (LOD=1.33, P2=0.046). We found no overlap of our linkage results with either report from the San Antonio Investigators33,34 or with those by Imperatore et al37 or Kort et al.36
HDL-C is a complex trait that likely has important polygenic and environmental determinants. The HDL-C literature includes multiple studies indicating a strong heritable component, even after adjustment for important environmental predictors. As new candidates are described, investigators rush to genotype their study families DNA to replicate the reported findings, often with little success. The heterogeneity of study populations and often the small sample sizes of many linkage studies have led to a confused literature, with important linkage findings in one study not replicated in many others. In the case of HDL-C, previously reported findings were from very different populations in North America, including Mexican-Americans in Texas,33,34 the genetically isolated Hutterites in Canada,35 Pima Indians in Arizona,37 and whites in Minnesota.38 Although the sample sizes were generally large (range 415 to 1484), the family structures varied from sibships37 to extended pedigrees of varying complexities33,34,38 to a very large inbred kindred.35 Also, the covariate adjustment strategies ranged from minimal in the Pima Indians36 and Rochester Family Heart Study37 to complete in the San Antonio Family Heart Study.34 Our covariate adjustment most closely resembled that of the Rochester Family Heart Study investigators.
Even the most consistent candidate-gene/HDL-Cphenotype linkages are thought to play relatively small roles in variation of HDL-C at the population level. The well-established apoA-I, lipoprotein lipase, apoC-II, lecithin:cholesterol acyltransferase, and cholesteryl ester transfer protein candidates are thought to explain <2% of the total phenotypic variation in the United States.51 Simulations performed on the FHS family structures indicate that variance-components methods would be capable of detecting QTLs accounting for >10% of the phenotypic variance, so it is not surprising that our LOD score peaks do not overlap with known candidate genes.
As already mentioned, the metabolism of HDL-C is a complex process with multiple structural and enzymatic proteins exerting major influences on the ability of HDL to remove cholesterol from the bloodstream. As such, many candidate genes that significantly affect HDL-C metabolism have been reported in the literature. None of the LOD score peaks listed in Table 3 corresponds to any of these candidate genes. It is to be expected that complex traits, such as HDL-C, are not primarily determined by 1 or even just a few genes. At the same time, it is unlikely that there is 1 unknown gene that solely regulates the observed genetic variability in HDL-C. A review of the literature points to no obvious candidate genes in the general location of the QTL detected on chromosome 5.
In short, these results indicate linkage of a new genomic region to genetic variability in HDL-C. A broad peak with a maximum LOD score of 3.64 (P2=0.0002) is localized to a region near 40 cM on the short arm of chromosome 5. This location is very close to a QTL identified in a sample of Alberta Hutterites. As far as we know, there are no other published reports of this possible QTL. Taken together, the results of all of these genome scans indicate many loci influencing HDL-C. Subtle differences in study samples and marker sets may contribute to the identification of different QTLs and to what appears to be a lack of replication across studies. Anonymous marker genome scans such as this one are largely hypothesis-generating exercises. Although the QTLs identified in different studies may not be the same, the use of genome-wide methods such as these may make it possible to identify multiple-component QTLs that contribute to the polygenic HDL-C trait. As more genome scans are published, the overlapping regions and positional candidate genes should become the focus of more intensive efforts to identify novel mechanisms that influence variability in HDL-C.
| Acknowledgments |
|---|
Received July 16, 2001;
revision received August 3, 2001;
| References |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
D. K. Arnett, A. E. Baird, R. A. Barkley, C. T. Basson, E. Boerwinkle, S. K. Ganesh, D. M. Herrington, Y. Hong, C. Jaquish, D. A. McDermott, et al. Relevance of Genetics and Genomics for Prevention and Treatment of Cardiovascular Disease: A Scientific Statement From the American Heart Association Council on Epidemiology and Prevention, the Stroke Council, and the Functional Genomics and Translational Biology Interdisciplinary Working Group Circulation, June 5, 2007; 115(22): 2878 - 2901. [Abstract] [Full Text] [PDF] |
||||
![]() |
T. A. Lakka, T. Rankinen, T. Rice, A. S. Leon, D. C. Rao, J. S. Skinner, and C. Bouchard Quantitative trait locus on chromosome 20q13 for plasma levels of C-reactive protein in healthy whites: the HERITAGE Family Study Physiol Genomics, October 11, 2006; 27(2): 103 - 107. [Abstract] [Full Text] [PDF] |
||||
![]() |
Y. Yu, D. F. Wyszynski, D. M. Waterworth, S. D. Wilton, P. J. Barter, Y. A. Kesaniemi, R. W. Mahley, R. McPherson, G. Waeber, T. P. Bersot, et al. Multiple QTLs influencing triglyceride and HDL and total cholesterol levels identified in families with atherogenic dyslipidemia J. Lipid Res., October 1, 2005; 46(10): 2202 - 2213. [Abstract] [Full Text] [PDF] |
||||
![]() |
Q. Yang, C.-Q. Lai, L. Parnell, L. A. Cupples, X. Adiconis, Y. Zhu, P. W. F. Wilson, D. E. Housman, A. M. Shearman, R. B. D'Agostino, et al. Genome-wide linkage analyses and candidate gene fine mapping for HDL3 cholesterol: the Framingham Study J. Lipid Res., July 1, 2005; 46(7): 1416 - 1425. [Abstract] [Full Text] [PDF] |
||||
![]() |
S. T. Turner, M. Fornage, C. R. Jack Jr, T. H. Mosley, S. L. R. Kardia, E. Boerwinkle, and M. de Andrade Genomic Susceptibility Loci for Brain Atrophy in Hypertensive Sibships From the GENOA Study Hypertension, April 1, 2005; 45(4): 793 - 798. [Abstract] [Full Text] [PDF] |
||||
![]() |
D. E. Arking, G. Atzmon, A. Arking, N. Barzilai, and H. C. Dietz Association Between a Functional Variant of the KLOTHO Gene and High-Density Lipoprotein Cholesterol, Blood Pressure, Stroke, and Longevity Circ. Res., March 4, 2005; 96(4): 412 - 418. [Abstract] [Full Text] [PDF] |
||||
![]() |
X. Wang and B. Paigen Genetics of Variation in HDL Cholesterol in Humans and Mice Circ. Res., January 7, 2005; 96(1): 27 - 42. [Abstract] [Full Text] [PDF] |
||||
![]() |
Y. Bosse, Y. C. Chagnon, J.-P. Despres, T. Rice, D. C. Rao, C. Bouchard, L. Perusse, and M.-C. Vohl Compendium of genome-wide scans of lipid-related phenotypes: adding a new genome-wide search of apolipoprotein levels J. Lipid Res., December 1, 2004; 45(12): 2174 - 2184. [Abstract] [Full Text] [PDF] |
||||
![]() |
G. E. Sonnenberg, G. R. Krakower, L. J. Martin, M. Olivier, A. E. Kwitek, A. G. Comuzzie, J. Blangero, and A. H. Kissebah Genetic determinants of obesity-related lipid traits J. Lipid Res., April 1, 2004; 45(4): 610 - 615. [Abstract] [Full Text] [PDF] |
||||
![]() |
Y. Bosse, Y. C. Chagnon, J-P. Despres, T. Rice, D. C. Rao, C. Bouchard, L. Perusse, and M-C. Vohl Genome-wide linkage scan reveals multiple susceptibility loci influencing lipid and lipoprotein levels in the Quebec Family Study J. Lipid Res., March 1, 2004; 45(3): 419 - 426. [Abstract] [Full Text] [PDF] |
||||
![]() |
X. Wang, I. Le Roy, E. Nicodeme, R. Li, R. Wagner, C. Petros, G. A. Churchill, S. Harris, A. Darvasi, J. Kirilovsky, et al. Using Advanced Intercross Lines for High-Resolution Mapping of HDL Cholesterol Quantitative Trait Loci Genome Res., July 1, 2003; 13(7): 1654 - 1664. [Abstract] [Full Text] [PDF] |
||||
![]() |
M.C. Mahaney, L. Almasy, D.L. Rainwater, J.L. VandeBerg, S.A. Cole, J.E. Hixson, J. Blangero, and J.W. MacCluer A Quantitative Trait Locus on Chromosome 16q Influences Variation in Plasma HDL-C Levels in Mexican Americans Arterioscler. Thromb. Vasc. Biol., February 1, 2003; 23(2): 339 - 345. [Abstract] [Full Text] [PDF] |
||||
![]() |
C. S. Fox, J. F. Polak, I. Chazaro, A. Cupples, P. A. Wolf, R. A. D'Agostino, and C. J. O'Donnell Genetic and Environmental Contributions to Atherosclerosis Phenotypes in Men and Women: Heritability of Carotid Intima-Media Thickness in the Framingham Heart Study Stroke, February 1, 2003; 34(2): 397 - 401. [Abstract] [Full Text] [PDF] |
||||
![]() |
X. Wang and B. Paigen Quantitative Trait Loci and Candidate Genes Regulating HDL Cholesterol: A Murine Chromosome Map Arterioscler. Thromb. Vasc. Biol., September 1, 2002; 22(9): 1390 - 1401. [Abstract] [Full Text] [PDF] |
||||
![]() |
H. Coon, M. F. Leppert, J. H. Eckfeldt, A. Oberman, R. H. Myers, J. M. Peacock, M. A. Province, P. N. Hopkins, and G. Heiss Genome-Wide Linkage Analysis of Lipids in the Hypertension Genetic Epidemiology Network (HyperGEN) Blood Pressure Study Arterioscler. Thromb. Vasc. Biol., December 1, 2001; 21(12): 1969 - 1976. [Abstract] [Full Text] [PDF] |
||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||