Multiple Genetic Determinants of Variation of Plasma Lipoproteins in Alberta Hutterites
Abstract We hypothesized that variation of nine candidate genes in lipoprotein metabolism would be associated with variation in fasting plasma lipoprotein variables in 718 Alberta Hutterites, a genetic isolate. We measured plasma lipids, lipoproteins, and apolipoproteins and analyzed DNA for genotypes of apolipoprotein (apo) B (APOB), paraoxonase (PON), lipoprotein lipase (LPL), VLDL receptor (VLDLR), apo CIII (APOC3), LDL receptor–related protein (LRP), hepatic lipase (HL), LDL receptor (LDLR), and apo E (APOE). Using a multivariate analysis, we found that (1) genotypes of APOB, PON, LPL, LDLR, and APOE were significantly associated with variation of plasma apo B–related traits; (2) genotypes of PON, LPL, and APOC3 were significantly associated with variation in plasma triglycerides; and (3) genotypes of VLDLR, APOC3, LDLR, and APOE were significantly associated with variation in plasma apo AI and HDL cholesterol. Regression analysis showed that between 3.2% and 7.8% of the total variation in plasma lipoproteins was accounted for by variation in the candidate genes tested. The observations demonstrate a modest but significant genetic component of variation in plasma lipoprotein levels that is due to the candidate genes studied in this normolipemic human genetic isolate.
- Received February 19, 1995.
- Accepted March 13, 1995.
Complex quantitative traits, such as plasma lipoprotein profile, are considered to be influenced by both genetic and nongenetic factors.1 Approaches to analysis of the genetic determinants of interindividual variation in plasma lipoproteins must take into account nonmendelian inheritance, genetic heterogeneity, incomplete penetrance, and interactions among gene products and between gene products and environmental factors.2 There are at least 20 gene products that participate in lipoprotein metabolism.2 Many population association studies in which DNA markers for lipoprotein metabolism are used have been reported, with a high frequency of contradictory results.2 The problems with traditional association studies in outbred populations have been well documented.2 To minimize the confounding features of ethnic and genetic heterogeneity, we studied the North American Hutterite Brethren, a genetically isolated population. The Hutterites are a useful population for the study of genetic determinants of plasma lipoproteins because their gene pool is small, their indices of relatedness are high, and they share many environmental factors.3 4 5 We have previously identified significant associations between plasma lipoproteins and genotypes of apolipoprotein (apo) E (APOE),3 lipoprotein lipase (LPL),3 and paraoxonase (PON)4 in the Hutterites.
We wanted to determine the contributions of several candidate genes to plasma lipoproteins in the Hutterites. On the basis of established or putative roles for the candidate gene products in lipoprotein metabolism,2 we selected nine genes for the current study: apo B (APOB), PON, LPL, VLDL receptor (VLDLR), apo CIII (APOC3), LDL receptor–related protein (LRP), hepatic lipase (HL), LDL receptor (LDLR), and APOE. We also genotyped clotting factor VII (F7), whose product was not expected to be associated with variation in lipoproteins. We used a multivariate statistical model to minimize the possibility of spurious associations that might have been identified had we made multiple pairwise comparisons.
The Hutterite Brethren are an Anabaptist sect who have an agrarian lifestyle and live on communal farms called colonies.3 4 5 Colonies are surrogates for extended families; women marry between colonies, but men remain within a colony.6 The Hutterites have had a high intrinsic growth rate, and their population remains closed to migration.6 A high degree of consanguinity relative to the founders has accumulated over about 12 generations, with the average inbreeding coefficient of the current generation being .05.5
The prevalence of atherosclerosis risk factors appears to be comparable to that in other populations.7 Subjects from 21 colonies of two Alberta Hutterite sects took part in the Canadian Heart Health Survey screening for coronary heart disease risk factors.8 9 Physical examination included determination of body mass index (BMI), defined as weight/(height2) (in kg/[m2]). Plasma samples from 846 Hutterites were obtained with informed consent. Exclusion criteria included an inadequate blood sample for all biochemical or genetic determinations. The study was approved by ethical review panels of the Universities of Alberta and Toronto.
All biochemical determinations were made in the J. Alick Little Lipid Research Laboratory at St Michael’s Hospital as described.3 4 8 10 Non-HDL cholesterol was defined as the difference between total cholesterol and HDL cholesterol; non-HDL cholesterol contained LDL and VLDL cholesterol.3 4 8
Sufficient DNA and phenotypic information was obtained for most analyses from 793 Hutterites. Established methods were used to obtain genotypes for APOB codons 3611 and 4154,11 PON codon 192,12 LPL intron 6,13 VLDLR trinucleotide repeat in the 5′-untranslated region (5′-tri),14 APOC3 Sac I site polymorphism in the 3′-untranslated region (3′-UTR),15 LRP tetranucleotide repeat in the 3′-untranslated region (3′-tetra),16 HL codon 202,17 LDLR exon 12 HincII site polymorphism,18 and APOE isoforms.19 As a control genetic variable for association with lipoprotein phenotypes, genotypes for the F7 codon 353 protein polymorphism were determined as described.20
sas (version 6) was used for all statistical comparisons.21 The distribution of each variable was significantly nonnormal in this dataset. Therefore, for parametric statistical analyses, each variable was transformed and subjected to analysis of normality. Logarithmically transformed total cholesterol, LDL cholesterol, non-HDL cholesterol, and HDL cholesterol had distributions that were not significantly different from normal (data not shown). After triglyceride values were transformed as described,22 their distribution was not significantly different from normal. Logarithmic transformation also made the distributions of apo AI and apo B more normal. The transformed variables were used for statistical analyses, but the nontransformed values are presented in the tables.
ANOVA was performed using the general linear models procedure to determine the sources of variation for biochemical traits, with F tests computed from the Type III sums of squares.21 This form of the sums of squares applies to unbalanced study designs and reports the effect of an independent variable after adjustment for all other variables included in the model. Dependent variables were transformed plasma total, LDL, non-HDL, and HDL cholesterol; triglycerides; apo AI; and apo B. Independent variables were age, log BMI, sex, and colony of origin, with the latter variable included to correct for variation that was related to other shared genetic and environmental factors. Also included as independent variables were genotypes of APOB, PON, LPL, VLDLR, APOC3, LRP, HL, LDLR, and APOE. F7 genotype was included as a negative control variable in all multivariate analyses because it had no obvious mechanistic basis for association with plasma lipoproteins.
Regression analysis was performed and partial regression coefficients were used to estimate the percent of phenotypic variation that was due to genotypic variation. This analysis is complex for datasets containing subjects in the age range of 18 to 74 years. The approach used for analysis of data from the Lipid Research Clinics Project22 was used. We included three terms (age, age2, and age3) to adjust for the age-dependent changes in lipoproteins and the natural log of BMI to adjust for the effects of obesity.3 4 22
There were so many significant phenotype-genotype comparisons from the ANOVA (summarized in Table 1⇓) that, for presentation in this article, we elected to focus on the 17 associations that were significant at a corrected nominal level of P<.005. The general linear models procedure for least-squares means was used to determine the level of significance in pairwise comparisons between genotype classes. Least-squares means, also known as population marginal means, are the values for class means after adjustment for all covariates in the model.21 The analysis with regard to biochemical traits was used to determine whether an allele had either a trait-lowering or trait-raising effect and which genotypic class(es) differed significantly from the others.
Determinants of Variation in Biochemical Variables
The summary of the eight ANOVAs is shown in Table 1⇑. At a nominal probability value of .05, F values among the analyses of eight biochemical traits were entered as dependent quantitative variables, with the 11 genetic markers being used as independent qualitative variables in each analysis. At a nominal probability value of .005, chosen to correct for the eight separate ANOVAs, 17 of these F values were significant. In general, the genetic variables that were most significantly associated with variation in plasma apo B–related traits were the genotypes for APOB codon 3611, PON, LPL, LDLR, and APOE. In contrast, the genetic variable that was most significantly associated with triglycerides and HDL cholesterol was the genotype for APOC3. No biochemical trait was significantly associated with genotypic variation in APOB codon 4154 or HL. LRP genotype was significantly associated with variation in total plasma cholesterol but not with any other lipoprotein traits. As expected, genetic variation of F7 was not associated with variation in any biochemical trait.
The details of the ANOVAs are shown in Table 2⇓. Significant associations were found between variation of the logarithm of total plasma cholesterol and age, sex, logarithm of BMI, colony of origin, LDLR genotype, APOB codon 3611 genotype, PON genotype, LRP genotype, and APOE genotype. Of these independent genotype variables, those for LDLR, APOB codon 3611, PON, and APOE were each also significantly associated with variation in the natural logarithms of the related plasma traits LDL cholesterol, non-HDL cholesterol, HDL cholesterol, and apo B. LRP genotype was not significantly associated with any of these related variables. Though not significantly associated with total plasma cholesterol (P=.12), LPL genotype was significantly associated with variation in plasma LDL cholesterol, non-HDL cholesterol, and apo B. Also, VLDLR genotype was significantly associated with variation in plasma apo B and non-HDL cholesterol.
Significant associations were found between variation of transformed plasma triglycerides and age, sex, logarithm of BMI, colony of origin, APOC3 genotype, PON genotype, and APOB codon 3611 genotype. APOE and LDLR genotypes were not significantly associated with variation of transformed plasma triglycerides.
Significant associations were found between variation of both the logarithms of plasma HDL and apo AI and age, sex, logarithm of BMI, colony of origin, APOC3 genotype, VLDLR genotype, APOE genotype, and LDLR genotype. PON genotype was significantly associated with variation in the logarithm of plasma HDL cholesterol but not with the logarithm of plasma apo AI. APOB codon 3611 genotype was significantly associated with variation in the logarithm of plasma apo AI but not with the logarithm of plasma HDL cholesterol.
The ratio of LDL cholesterol to HDL cholesterol was used to create a dependent variable that has been shown to be an index of atherosclerosis risk.22 Significant associations were found between LDL cholesterol:HDL cholesterol and genotypes for APOE, PON, LPL, APOC3, and VLDLR.
Allele and Genotype Frequencies
Allele frequencies for most markers tested were similar to those reported in other Caucasian populations (data not shown). However, the minor allele frequencies in the Hutterites compared with other Caucasians were somewhat increased for the APOB codon 3611, VLDLR, LRP, and F7 genotypes (Table 3⇓). Also, the minor allele frequencies in the Hutterites were somewhat decreased for the APOE gene compared with other Caucasians.23 There was no significant deviation of observed genotype frequencies from those predicted by the Hardy-Weinberg law in this sample of Hutterites for all genotypes included in this analysis (data not shown).
Regression analysis was performed to estimate the percentage of phenotypic variation that was determined by genetic variation. These results are shown in Table 4⇓. The model included age, log BMI, colony of origin, and genotypes and was able to account for 39%, 37%, 48%, 51%, 43%, and 35% of the variation in plasma total cholesterol, LDL cholesterol, non-HDL cholesterol, apo B, triglycerides, and HDL cholesterol, respectively (Figure⇓). In this model, 7.8%, 7.6%, 7.3%, 7.2%, 3.2%, and 5.1% of the total variation in plasma total cholesterol, LDL cholesterol, non-HDL cholesterol, apo B, triglycerides, and HDL cholesterol, respectively, were accounted for by variation in the candidate genes tested. Although these percentages were small, the partial regression coefficients were all significant with a nominal P<.05. In general, the same genotypic variables that were significantly associated with plasma biochemical variables in the regression analysis were also found to be significantly associated in the ANOVA.
Least-squares means for each significant phenotype-genotype association from Table 1⇑ were assessed for pairwise differences that were significant at a nominal P<.005. For systems with three genotypes, a pairwise significant difference between the least-squares mean of one homozygous class and the least-squares means of the other two classes was interpreted as being consistent with a recessive effect of the allele on the biochemical phenotype. For systems with three genotypes, pairwise significant differences between both homozygous classes and between the heterozygous class and both homozygous classes were interpreted as being consistent with a codominant effect of the alleles on the biochemical phenotype. These interpretations depended on the assumption that the numbers of subjects in each genotypic class were sufficient to detect significant differences.
For the APOB codon 3611 system, subjects homozygous for the Q/Q genotype had a significantly higher least-squares mean non-HDL cholesterol than subjects with the R/Q (P=.003) or R/R genotype (P=.05), but pairwise comparison of subjects with the R/Q genotype and those with the R/R genotype did not indicate a significant difference for this trait. This trend was consistent for pairwise comparisons of other apo B–related traits, such as plasma total cholesterol, LDL cholesterol, triglycerides, and apo B itself (data not shown). These observations are consistent with a recessive trait-raising effect of the APOB codon 3611Q allele.
For the PON codon 192 system, subjects homozygous for the Q/Q genotype had significantly lower least-squares mean LDL cholesterol, non-HDL cholesterol, apo B, and LDL cholesterol:HDL cholesterol ratio than the Q/R genotype (P=.0009, .0001, .0001, and .0001, respectively). The pairwise comparisons of least-squares means of these traits between subjects with the Q/R and R/R genotype did not indicate significant differences. These findings are consistent with our previous findings that subjects homozygous for PON codon 192 Q/Q had lower mean levels of plasma apo B–related traits and a less atherogenic lipoprotein profile than subjects with the other genotypes and that this allele had a recessive effect on phenotype.4
For the LPL intron 6 genotype system, subjects homozygous for the absence of the Pvu II site had significantly lower least-squares mean plasma apo B and LDL cholesterol:HDL cholesterol than heterozygous +/− subjects (P=.001 and .002, respectively) or the subjects homozygous for the presence of the Pvu II site (P=.008 and .017, respectively). This trend was consistent for pairwise comparisons of other apo B–related traits, such as plasma total cholesterol, LDL cholesterol, and apo B (data not shown). This is consistent with our previous finding that subjects homozygous for absence of the Pvu II site in LPL intron 6 had lower plasma apo B–related traits and that this allele had a recessive effect on phenotype.3
For the VLDLR 5′-trinucleotide repeat marker system, we observed that subjects homozygous for A3/A3 had higher least-squares mean plasma apo B than subjects with the A1/A1, A1/A2, A1/A3, A1/A4, or A2/A2 genotype (P=.001, .031, .026, .006, and .008, respectively). A similar trend was seen for mean plasma non-HDL cholesterol (data not shown). The observations are consistent with a tendency towards higher mean apo B levels among subjects homozygous for an A3 allele and lower mean apo B levels among subjects with a single A1 allele.
For the APOC3 3′-UTR Sac I site polymorphism marker system, heterozygotes for the presence of the site had lower least-squares mean HDL cholesterol and higher least-squares mean triglycerides than homozygotes for the absence of the site (P=.0006 and .003, respectively). The observations are consistent with triglyceride-raising and HDL cholesterol–lowering effects for the allele marked with the presence of the Sac I site.
For the LDLR exon 12 marker system, there was a consistent gene-dosage effect seen for least-squares mean LDL cholesterol, non-HDL cholesterol, and apo B. Subjects homozygous for the absence of the HincII site had significantly higher mean LDL cholesterol, non-HDL cholesterol, and apo B compared with heterozygous +/− subjects (P=.0003, .002, and .007, respectively) or homozygotes for the presence of the HincII site (P=.0001, .0003, and .0001, respectively). Furthermore, heterozygous +/− subjects had higher mean LDL cholesterol, non-HDL cholesterol, and apo B compared with homozygotes for the presence of the HincII site (P=.05, .09, and .03). This trend was also seen in pairwise comparisons of plasma total cholesterol (data not shown). These observations are consistent with a codominant effect of these LDLR alleles. The absence of the HincII site was associated with higher levels of plasma apo B–related traits, and the presence of that site was associated with lower levels of plasma apo B–related traits. Also, subjects homozygous for the absence of the HincII site had higher mean plasma HDL cholesterol than heterozygous +/− subjects or homozygotes for the presence of the HincII site (P=.02 and .001, respectively), but the pairwise comparison of +/− genotype subjects with +/+ genotype subjects was not significant for this trait. This is consistent with a recessive HDL cholesterol–raising effect of the allele marked by the absence of the HincII site.
Finally, for the APOE genotype system, there was a consistent gene-dosage effect seen for least-squares mean LDL cholesterol, non-HDL cholesterol, and LDL cholesterol:HDL cholesterol. Subjects with the 4/3 genotype had significantly higher mean LDL cholesterol, non-HDL cholesterol, and LDL cholesterol:HDL cholesterol compared with subjects with the 3/3 genotype (P=.007, .004, and .013, respectively) or those with the 3/2 genotype (P=.0003, .001, and .0001, respectively). Furthermore, subjects with the 3/3 genotype had higher mean LDL cholesterol, non-HDL cholesterol, and LDL cholesterol:HDL cholesterol compared with subjects with the 3/2 genotype (P=.003, .0001, and .0001). This trend was consistent for pairwise comparisons of plasma total cholesterol and apo B (data not shown). These observations are consistent with a codominant raising effect of the E4 allele on plasma apo B–related traits compared with the E3 allele and are also consistent with a codominant lowering effect of the E2 allele on plasma apo B–related traits compared with the E3 allele. These observations are in agreement with previous observations in this study sample.3
We have identified associations between variation in candidate genes in lipoprotein metabolism and the plasma lipoprotein profile in a genetic isolate. The significant associations were consistent, and they were confirmed by two separate statistical analyses. Genotypes of APOB, PON, LPL, LDLR, and APOE were significantly associated with variation of plasma apo B–related traits, namely apo B and total, LDL, and non-HDL cholesterol. Genotypes of APOC3 were most significantly associated with variation in plasma triglycerides. Genotypes of APOC3 and LDLR were significantly associated with variation in plasma apo AI and HDL cholesterol. In addition, variation of LDL cholesterol:HDL cholesterol was most significantly associated with variation in the genes for PON, LPL, and APOE. Regression analyses showed that between 3.2% and 7.8% of the total variation in plasma lipoproteins was accounted for by variation in the candidate genes tested. These findings suggest that candidate genes in lipoprotein metabolism are important genetic determinants of plasma lipoproteins.
Genotypes of PON, LPL, and APOE in analyses in which individual genetic variables were used have been shown to be significantly associated with variation of plasma apo B–related traits in the Hutterites.3 4 In the present multivariate analysis, variation of these genes was also significantly associated with variation in the plasma apo B–related traits. This study also shows that variation of the LDLR and APOB genes is also significantly associated with variation in the plasma apo B–related traits in the Hutterites. These associations are mechanistically plausible, given that apo B is both the structural protein for the LDL particle and the ligand for receptor-mediated endocytosis of LDL by the LDL receptor. Specifically, the APOB codon 3611 polymorphism is associated with the Ag(h/i) polymorphism, although this substitution does not appear to alter affinity of the particle for the LDL receptor and has not had consistent associations with plasma lipid levels in outbred populations.24
The total contribution of the studied genetic determinants to variation in plasma levels of apo B–related traits, which ranged from 7.2% to 7.8%, was comparable to the contribution to these traits of age, sex, BMI, and colony of origin. The contribution of colony of origin to variation in biochemical traits may have been related to additional genetic and/or environmental factors that were unique to specific colonies. In some instances in this sample, the contributions of individual genes to total phenotypic variation, such as the contribution of LDLR and LPL genotypes to total plasma cholesterol, were comparable (2.96% and 1.11%, respectively) to the percent contribution of sex and BMI.
From the ANOVA we observed that genetic variation in LRP was associated with variation in total plasma cholesterol (P=.034) but with no other phenotype. We also observed that genetic variation of VLDLR was associated with variation in non-HDL cholesterol, apo B, HDL cholesterol, and apo AI (P=.047, .043, .0058, and .033, respectively). The identification of such associations suggests that genetic variation of these two receptors may determine plasma lipoprotein levels. Both LRP and VLDLR, members of the LDLR gene family that were both cloned by use of strategies that exploited their structural homology with LDLR, have been shown to function in the catabolism of chylomicron remnants and VLDL, and these may be the mechanisms that underlie the genetic association with the biochemical determinations that depend on metabolism of these particles.
Genotypes of LPL and PON have been shown to be associated with variation in plasma triglycerides in the Hutterites.3 4 The significant association of the 3′-UTR Sac I site polymorphism of APOC3 with variation in triglycerides and HDL cholesterol is completely consistent with numerous other observations that variants of this gene are associated with variation in these plasma lipoprotein traits.2 25 The basis for this association remains unsolved, but it could be due to linkage disequilibrium between this site polymorphism and functional variants in the promoter region of APOC3 that in turn may have a variable functional impact under different metabolic situations such as diabetes mellitus.25
We previously showed that genetic variation in APOE was significantly associated with variation in plasma apo AI and HDL cholesterol in the Hutterites and that this was mainly due to higher concentrations of these variables among relatively small numbers of subjects with the E2 allele.3 The significant association between genetic variation of APOC3 and plasma apo AI and HDL cholesterol may be due to the fact that plasma triglycerides and HDL cholesterol are not independent variables and likely share common determinants of their plasma concentrations.2 The significant association of genotypes of VLDLR and LDLR with HDL cholesterol and apo AI at a nominal P<.05 in the ANOVA suggests that genetic variation of these receptors might affect plasma HDL cholesterol and apo AI metabolism. There is evidence from in vivo turnover studies that apo AI and HDL metabolism are abnormal in subjects with familial hypercholesterolemia due to genetic defects in LDLR.26 However, because our subjects homozygous for the absence of the HincII site in LDLR exon 12 had higher mean levels for both plasma LDL and HDL cholesterol, the mechanism that underlies the lower plasma HDL cholesterol in subjects with familial hypercholesterolemia cannot explain this association in the Hutterites.
There are at least two explanations for the numerous observed significant associations between variation in candidate genes and lipoprotein variation. First, the genetic variants that encode structural changes in the protein might affect some aspect of function in vivo, and this in turn might affect lipoprotein metabolism and plasma lipoprotein concentrations. Alternatively, in this Hutterite sample, any allelic variant, whether in the coding region or not, may be in linkage disequilibrium with another functional mutation within the candidate gene or even at another gene that is physically or genetically linked to the candidate gene. In this case, the variant serves as a marker for the actual causative mutation. Thus, these associations might be difficult to detect in the general outbred population, where among unrelated subjects many different functional mutations and markers may exist at any locus.
The results indicate that new candidate genes that have not been used extensively in studies in the past, such as PON, VLDLR, and LRP, are associated with variation in plasma lipoproteins. It is possible that the PON genotype, which has been demonstrated to have an impact in vivo on plasma paraoxonase activity, is the marker for the actual functional basis for the phenotype-genotype association. Further functional studies, such as cell-binding characteristics of lipoproteins that have been incubated with the two isoforms of paraoxonase and studies of lipoprotein turnover in vivo among subjects classified by PON genotype, are required to determine the mechanism that underlies the phenotype-genotype association. In the cases of the VLDLR and LRP genotypes, however, it is not likely that the flanking sequence markers that were used have any functional impact. Thus, if these genes did affect lipoprotein metabolism through DNA changes that affected either receptor structure or expression, the association with the markers used could only be explained by linkage disequilibrium with the functional changes. DNA sequencing would be required to identify the actual changes within these genes encoding receptors that affect function. A similar argument could be made for sequencing the LDLR gene from subjects with different LDLR genotypes, because the LDLR marker we used did not result in a change in the exon 12 coding sequence.
Among the candidate genes that were found not to be associated with variation in any plasma lipoprotein trait were the codon 4154 polymorphism of APOB, the codon 202 polymorphism of HL, and the codon 353 polymorphism of F7. It is always possible that there was insufficient statistical power to detect an association. The lack of a significant phenotype-genotype association for the F7 codon 353 genotype was not surprising, because this genotype was included a priori as a negative control for association with lipoproteins. The lack of a significant phenotype-genotype association when the APOB codon 4154 genotype was used was also not surprising, given that functional studies have failed to show that this protein polymorphism has an effect. Even if the APOB codon 4154 genotype had been in linkage disequilibrium with a functional change in APOB, our multivariate analysis included the APOB codon 3611 genotype, a marker in linkage disequilibrium with APOB codon 4154 genotype in the Hutterites (data not shown). Thus, any contribution of APOB codon 4154 genotype variation to phenotypic variation would not have been independent of APOB codon 3611 genotype variation. Finally, although the HL codon 202 DNA change was itself silent at the protein level, it was an informative diallelic marker; thus the lack of association between any biochemical phenotype and this genotype strongly suggests that genetic variation of HL is not associated with variation in plasma lipoproteins in the Hutterites.
In summary, variation in candidate genes was associated with variation in plasma lipoprotein traits in this Hutterite sample. Candidate gene products may have subtle effects on lipoprotein metabolism that were detected in our relatively large, homogeneous study sample. This sample might be useful in future studies of other candidate genes or anonymous loci to identify associations by use of newer analytical approaches.27 Future analyses in this sample would include genotyping with random markers as part of a general genomic screening to identify new loci that are associated with an extent of variation in the quantitative traits that is similar to or greater than that for the candidate genes reported here.
This work was supported by grants from the Medical Research Council of Canada, the National Health Research and Development Program of Canada, and the Heart and Stroke Foundations of Ontario and Canada. Dr Hegele is a McDonald Scholar of the Heart and Stroke Foundation of Canada. We would like to thank Stanley Chan (APOB codon 3611 genotypes), Kevin Higgins (APOB codon 4154 genotypes), Greg Ip (APOE and LPL genotypes), Ulana Kawun (F7 genotypes), Dennis Lam (LDLR and VLDLR genotypes), Edwin Lee (PON genotypes), Patricia Ram (HL genotypes), Stefan Sadikian (LRP genotypes), and Tammy Znajda (APOC3 genotypes) for their technical assistance. Teresa Lippingwell and Liling Chan archived the phenotypic and genotypic data. Dr Adele Csima, Department of Biostatistics, University of Toronto, provided expert advice regarding our statistical analyses.
Mehrabian M, Lusis AJ. Genetic markers for studies of atherosclerosis and related risk factors. In: Lusis AJ, Rotter JI, Sparkes RS, eds. Molecular Genetics of Coronary Artery Disease: Candidate Genes and Processes in Atherosclerosis. Monograph Hum Genet. Basel, Switzerland: Karger; 1992:363-418.
Hegele RA, Evans AJ, Tu L, Ip G, Brunt JH, Connelly PW. A gene-gender interaction affecting plasma lipoproteins in a genetic isolate. Arterioscler Thromb. 1994;14:671-678.
Hegele RA, Brunt JH, Connelly PW. A polymorphism in the paraoxonase gene associated with variation in plasma lipoproteins in a genetic isolate. Arterioscler Thromb Vasc Biol. 1995;15:89-95.
Hostetler JA. Hutterite Society. Baltimore, Md: Johns Hopkins University Press; 1974:223.
Connelly PW, MacLean DR, Horlick L, O’Connor B, Petrasovits A, Little JA. Plasma lipids and lipoproteins and the prevalence of risk for coronary heart disease in Canadian adults. Can Med Assoc J. 1992;146:1977-1987.
MacLean DR, Petrasovits A, Nargundkar M, Connelly PW, MacLeod E, Edwards A, Hessel P. Canadian heart health surveys: a profile of cardiovascular risk—survey methods and data analysis. Can Med Assoc J. 1992;146:1971-1976.
Jenkins DJA, Wolever TMS, Rao AV, Hegele RA, Mitchell SJ, Ransom TPP, Boctor DL, Spadafora PJ, Jenkins AL, Mehling C, Relle LK, Connelly PW, Story JA, Furumoto EJ, Corey P, Wursch P. Effect on blood lipids of very high intakes of fiber in diets low in saturated fat and cholesterol. N Engl J Med. 1993;329:21-26.
Johnson JP, Nishina PM, Naggert JK. PCR assay for a polymorphic site in the LPL gene. Nucleic Acids Res. 1991;18:7469.
Jokinen E, Sakrai J, Yamamoto T, Hobbs HH. CGG triple repeat polymorphism in VLDL receptor (VLDL-R) gene. Hum Mol Genet. 1994;3:521.
Tas S. Strong association of a single nucleotide substitution in the 3′-untranslated region of the apolipoprotein CIII gene with common hypertriglyceridemia in Arabs. Clin Chem. 1989;35:256-259.
Zuliani G, Hobbs HH. Tetranucleotide length polymorphism 5′ of the α2-macroglobulin receptor (A2MR)/LDL receptor-related protein (LRP) gene. Hum Mol Genet. 1994;3:215.
Leitersdorf E, Hobbs HH. Human LDL receptor HincII polymorphism detected by gene amplification. Nucleic Acids Res. 1988;16:7215.
Hixson JE, Vernier DT. Restriction isotyping of human apolipoprotein E by gene amplification and cleavage with HhaI. J Lipid Res. 1990;31:545-548.
Humphries SE, Lane A, Green FR, Cooper J, Miller GJ. Factor VII coagulant activity and antigen levels in healthy men are determined by interaction between factor VII genotype and plasma triglyceride concentration. Arterioscler Thromb. 1994;14:193-198.
SAS/STAT Guide for Personal Computers, Version 6. Cary, NC: SAS Institute; 1987.
Green PP, Namboodiri KK, Hannan P, Martin J, Owen ARG, Chase GA, Kaplan EB, Williams L, Elston RC. The collaborative Lipid Research Clinics Program family study, III: transformation and covariate adjustments of lipid and lipoprotein levels. Am J Epidemiol. 1984;119:959-974.
Davignon J, Gregg RE, Sing CF. Apolipoprotein E polymorphism and atherosclerosis. Arteriosclerosis. 1988;8:1-21.
Humphries SE. Lifestyle, genetic factors and the risk of heart attack: the apolipoprotein B gene as an example. Biochem Soc Trans. 1988;21:569-582.
Dammerman M, Sandkuijl LA, Halaas JL, Chung W, Breslow JL. An apolipoprotein CIII haplotype protective against hypertriglyceridemia is specified by promoter and 3′-untranslated region polymorphisms. Proc Natl Acad Sci U S A. 1993;90:4562-4566.
Schaefer JR, Rader DJ, Ikewaki K, Fairwell T, Zech LA, Kindt MR, Davignon J, Gregg RE, Brewer HB Jr. In vivo metabolism of apolipoprotein A-I in a patient with homozygous familial hypercholesterolemia. Arterioscler Thromb. 1992;12:843-848.