Background
The incidence rates of complex diseases such as obesity and type 2 diabetes have been increasing worldwide 12. Obesity is a major risk factor for type 2 diabetes, hypertension, dyslipidemia, and other cardiovascular complications and it has become a global public health problem 3. Several epidemiological studies have shown that obesity and low levels of high-density lipoprotein cholesterol (HDL-C) are major risk factors for coronary heart disease 45 and there is substantial evidence that obesity and HDL-C are strongly influenced by genetic factors. In fact, a number of studies have identified chromosomal regions harboring quantitative trait loci (QTL) influencing obesity 6 or dyslipidemia 7. Surprisingly, little is known about specific loci commonly influencing obesity and HDL-C concentrations, despite the evidence for appreciable correlation between these phenotypes. The present analysis examines whether there exists any chromosomal regions harboring genes that influence the covariation between BMI and HDL-C phenotypes using the Framingham Heart Study data through a bivariate linkage approach.
Subjects and Methods
The Framingham Heart Study
For purposes of the current analysis, Framingham Heart Study participants from the original cohort were combined with participants from the offspring cohort to maximize the number of individuals per pedigree. Information collected as part of the 12th examination (except for height, which was collected at the 14th examination) of the original cohort, which occurred 1970 to 1971, was combined with information collected during the first examination of the offspring cohort, which occurred from 1971 to 1975. Of the more than 10,000 participants enrolled in either the Framingham Heart Study or the Framingham Offspring Study, genotypic information was available on 1702 individuals from 330 extended pedigrees. The pedigrees include from 2 to 29 genotyped individuals and the genotyped sample includes 394 individuals from the original cohort and 1308 individuals from the offspring cohort.
Phenotypes
We used BMI and HDL-C data collected as part of the Framingham Heart Study. BMI was calculated as weight (in kilograms)/height squared (in meters). HDL-C (mg/dl) was measured by automated enzymatic methods. BMI and HDL-C values were log transformed to minimize the problem of non-normality.
Variance components linkage analysis
A multipoint variance components linkage analysis was used to test linkage between marker loci and a given phenotype, which was based on specifying the expected genetic covariances between pairs of relatives as a function of their identity by descent (IBD) at a marker linked to a QTL 8. It allows for locus-specific effects, residual genetic effects, covariate effects, and random environmental effects. Since the trait-specific linkage analysis (i.e., univariate) cannot exploit the additional information embedded in the correlation pattern between two quantitative traits, a bivariate multipoint linkage analysis was used to exploit the additional information contained in the correlation pattern between two quantitative traits 910.
Univariate genetic linkage analysis
In a simple additive model in which n QTLs and an unknown number of residual polygenes influence a given trait, the covariance matrix (Ω) for a pedigree is given by
![]()
where Πi is a matrix whose elements (πijl) provide the expected proportion of genes that individuals j and l share IBD at a QTL (qi) that is linked to a genetic marker locus, σ2q is the additive genetic variance due to the major locus, Φ is the kinship matrix, σ2g is the genetic variance due to additive genetic effects, I is an identity matrix, and σ2e is the variance due to random environmental effects.
Bivariate genetic linkage analysis
The extension of the variance components linkage approach to the bivariate situation facilitates the testing of linkage of two correlated phenotypes to a single genetic region simultaneously. In the bivariate analysis, trait-specific means, variance components relating to major gene effects (σ2q), residual additive genetic effects (σ2g), random environmental effects (σ2e), covariate effects (age and sex terms), as well as the three associated correlations ρq (correlation caused by a major gene), ρg (correlation caused by residual additive genetic effects), and ρe (correlation caused by random environmental effects) are estimated simultaneously using maximum-likelihood estimates. Using this bivariate model, we tested the null hypothesis that σ2q = 0 for both traits by comparing the log likelihood of this restricted model to that of a model in which σ2q was estimated for the traits. In addition, tests for the hypotheses of complete pleiotropy (i.e., the same major gene in the chromosomal region of interest affects both traits) and coincident linkage (i.e., no shared major gene effects in the chromosomal region of interest on the two traits) can be performed 9. To test complete pleiotropy or coincident linkage, likelihood for the linkage model in which ρq was estimated was compared with the model in which ρq was constrained to -1 (complete pleiotropy [a negative value for ρq was chosen since the overall polygenic correlation between the examined traits was found to be negative]) or to 0 (coincident linkage [i.e., ρq = 0]), respectively. These analytical techniques were implemented in the computer program SOLAR 8.
Results
Descriptive statistics for the examined phenotypes are reported in Table 1. To minimize the problem of non-normality, BMI and HDL-C values were log transformed. Both phenotypes exhibited high heritabilities (h2 ± SE, ln BMI = 0.45 ± 0.05 and ln HDL-C = 0.52 ± 0.06), after adjusting for age and sex influences.
<p>Table 1</p>
Descriptive statistics
Variable
N
Mean
SD
Kurtosis
Skewness
Age
1702
38.4
14.6
-
-
BMI
1686
25.3
4.2
2.5
0.9
HDL-C
1294
50.7
14.2
0.9
0.7
ln BMI
1686
3.2
0.2
0.6
0.3
ln HDL-C
1294
3.9
0.3
0.2
-0.2
As shown in Figure 1, univariate multipoint linkage analysis yielded significant evidence for linkage with ln BMI (LOD = 3.9) and suggestive evidence with ln HDL-C (LOD = 2.7, 150 cM) on chromosome 6. The chromosomal region at 158 cM near markers D6S1009 and GATA184A08 on chromosome 6 showed evidence for linkage to both BMI and HDL-C, although the trait-specific linkage curves peaked at two adjacent locations covering about an 8cM region (Figure 1). Both genetic (-0.26) and environmental (-0.17) correlations between BMI and HDL-C were low. Subsequently, we extended our analytical approach to the bivariate situation to exploit the additional information underlying the pattern of covariation between ln BMI and ln HDL-C phenotypes.
<p>Figure 1</p>
Chromosomal regions linked to ln BMI (A) and ln HDL-C (B) in genomic scans
Chromosomal regions linked to ln BMI (A) and ln HDL-C (B) in genomic scans
![]()
Table 2 presents the results from multipoint bivariate linkage analyses of ln BMI and ln HDL-C. A bivariate LOD score of 6.24 was obtained for the region involving the ordered markers D6S1009-GATA184A08 for the trait pair of ln BMI-ln HDL-C (Figure 2). The corresponding p-value is 1.8 × 10-7 (λ = 28.75). The bivariate LOD score can be converted to univariate LOD equivalent (LODeq) of 5.46 for this chromosomal region, which adds further significance to our finding. The correlation (ρq ± SE) between trait pair ln BMI-ln HDL-C due to QTL effects is -0.54 ± 0.19, which is statistically significant (p = 0.0072). It should be noted that the test for complete pleiotropy (i.e., locus-specific correlation between ln BMI and ln HDL is not significantly different from -1) was rejected (p = 0.0166). However, coincident linkage (i.e., locus-specific correlation between ln BMI and ln HDL is equal to 0) for ln BMI and ln HDL was strongly rejected (p = 0.0072).
<p>Table 2</p>
Results from bivariate linkage analyses of the trait pair ln BMI-ln HDL-C phenotypes
Trait Pair
cMA
ρqB
Bivariate LOD
λC
P
D
LODeqE
ln BMI-ln HDL
152
-0.54
6.24
28.75
1.8 × 10-7
5.46
ADistance from pter; Bρq is correlation due to QTL effects; Clikelihood-ratio statistic λ. DAsymptotic P value, under the assumption that λ is distributed as a 1/4X22 : 1/2X21 : 1/4X20 mixture. ELodeq is the univariate lod score corresponding to the reported bivariate lod score
<p>Figure 2</p>
Univariate and bivariate linkage profiles for ln BMI and ln HDL-C phenotypes
Univariate and bivariate linkage profiles for ln BMI and ln HDL-C phenotypes
![]()
Discussion
The univariate linkage analyses implicated ~8cM chromosomal region, involving the sequential markers D6S1009-GATA184A08, on chromosome 6q influencing both ln BMI and ln HDL-C. The genetic correlation between ln BMI and ln HDL was higher compared with environmental correlation and showed an inverse relationship as expected. This pattern of correlation is consistent with previous observations in which low overall genetic correlations were observed between obesity and lipid measures in a Mexican-American population 11. Since both ln BMI and ln HDL-C have exhibited significant linkage to the same region on chromosome 6q, we specifically examined the mapping pattern of the bivariate trait. The bivariate analysis has resulted in co-localization of these correlated traits at a chromosomal region near marker D6S1009. The tests for complete pleiotropy and coincident linkage indicated that the putative major gene influences both BMI and HDL-C, albeit incompletely. As remarked by Almasy et al. 9, such a situation of incomplete pleiotropy would be expected if either epistatic or gene × environmental interactions influence the putative major gene's effect on one of the two correlated traits, but not both. Also, it is possible that multiple variants in a putative major gene may differentially influence the correlated phenotypes.
It appears that the trait pairs with low to moderate overall genetic correlations (BMI vs. HDL-C) appear to contain much more information to improve power to co-localize precisely the correlated phenotypes to a genetic region. It is important to note that this same region (or genetic locations close to this region) on chromosome 6q has been reported to affect various components of the Insulin Resistance Syndrome including obesity, insulin, lipid, and blood pressure measures [e.g., 121314]. Recently, Atwood et al. 15 reported evidence for linkage of the same region on chromosome 6q23-25 (D6S1009-GATA184A08-D6S2436-D6S305) with six separate measurements of BMI with corresponding maximum LOD scores of 4.64, 2.29, 2.41, 1.40, 0.99, and 3.08, respectively, collected across 28 years of the Framingham Heart Study. This study provides additional evidence in support of the implicated region on chromosome 6 harboring the QTL affecting BMI variation in Framingham data.
Conclusions
Using a bivariate linkage approach, we have found strong evidence for a QTL, which appears to influence both BMI and HDL-C, in Framingham data. Our study suggests that the putative locus lies in the chromosomal region near marker D6S1009 on chromosome 6.