Effects of a Major Gene for Apolipoprotein A-I Concentration Are Thyroid Hormone Dependent in Mexican Americans

ApoA-I is the principal protein in HDL-C, constituting about 70% to 80% of HDL protein mass.¹ It is synthesized in the small intestine and liver. In addition to its role as a structural element of HDL-C, apoA-I is an activator of LCAT, which esterifies cholesterol that is attracted to the surface of nascent HDL. This esterified cholesterol moves to the core of HDL, increasing its size and changing its shape.² The cholesteryl ester is transported to VLDL, IDL, and LDL; taken up by the LDL receptor pathway in the liver; and eliminated from the body. By virtue of its function as a cofactor for LCAT, apoA-I plays a critical role in this process of reverse cholesterol transport, ie, the movement of cholesterol from the tissues to the liver, where it is excreted. Consistent with the importance of apoA-I for reverse cholesterol transport, several studies in humans³⁴ and animal models⁵ have found an inverse association of apoA-I and atherosclerosis.

The structural gene for apoA-I (APOA1) maps to the long arm of chromosome 11, in a gene cluster that also includes the structural loci for apoC-III (APOC3) and apoA-IV (APOA4). APOA1 is expressed in the liver and intestine. Recent studies have shown that APOA1 expression is mediated by a number of factors, including the thyroid hormone T₃.⁶

T₃ is formed primarily by the metabolism, via monodeiodination,⁷ of thyroxine outside the thyroid. More than 99.5% of total T₃ is bound to transport proteins, with the remaining amount forming free T₃. The thyroid hormones T₃ and T₄ play a major role as regulators of cellular metabolism and gene expression in most organ systems and tissues.⁸ They appear to exert their effects by direct stimulation of transcription.⁸

Effects of thyroid hormones on apoA-I have been demonstrated in numerous studies. Plasma apoA-I concentrations in women are increased in hyperthyroidism⁹¹⁰ and decreased in hypothyroidism.¹¹ Administration of thyroid hormone (T₃ or T₄) increases plasma apoA-I concentrations in rats.¹² In contrast to the finding in humans, plasma apoA-I concentrations in hypothyroid rats have generally been found to be unchanged (reviewed in Reference 13).

T₃ has been shown to increase apoA-I mRNA levels in rat liver but not in the intestine.¹⁴¹⁵ Strobl et al¹⁶ found that T₃ increased both APOA1 gene expression and posttranscriptional stability of RNA. Romney et al⁶ have shown that the regulation of APOA1 gene expression by T₃ is mediated in part by a thyroid hormone response element (THRE) located at the 5′ end of APOA1. APOA1 promoter activity is increased in the presence of the THRE site and decreased in its absence. No previous studies in any species have investigated whether the effects of T₃ on plasma apoA-I levels or on APOA1 gene expression are genetically mediated.

Evidence from several family studies indicates that plasma concentration of apoA-I is influenced by genetic factors. However, previous analyses have yielded conflicting results concerning the existence and magnitude of genetic effects on apoA-I levels. In a segregation analysis of a large pedigree ascertained through cases of early myocardial infarction, Hasstedt et al¹⁷ found no evidence for genetic transmission of apoA-I level. Moll et al,¹⁸ on the other hand, obtained evidence for a major gene influencing quantitative variation in apoA-I in a sample of families at high risk for coronary artery disease. Borecki et al,¹⁹ in bivariate segregation analyses of apoA-I, apoA-II, and HDL-C in a 51-member pedigree, concluded that a major gene that influences apoA-II in this family also has pleiotropic effects on apoA-I and HDL-C. It was also postulated that a second major gene may influence apoA-I and HDL-C. In a segregation analysis of randomly ascertained families in Rochester, Minn, Moll et al²⁰ detected a major gene influencing apoA-I levels, but only in a subset of the families. This was the first population-based study to detect a major locus effect on human apoA-I levels. In an analysis of apoA-I levels in baboons measured on two different diets, Blangero et al,²¹ using bivariate segregation analysis methods,²² found evidence for two major genes, in addition to polygenic factors, that affected apoA-I concentrations on the two diets. Interaction between genotype and diet could be detected for both major genes. Prenger et al,²³ in a family study of cardiac catheterization patients, detected a major gene with a dominant allele responsible for elevated levels of apoA-I. In analyses of Donner Laboratory data in Genetic Analysis Workshop 8, both Xu et al²⁴ and Blangero et al²⁵ found evidence for a major gene controlling apoA-I levels. The Blangero group also found that apoA-I levels affected the distribution of cholesterol among HDL subclasses.

Two studies have examined relationships between apoA-I concentrations and candidate loci. In an association study, Kondo et al²⁶ found that in smokers, apoA-I levels are influenced by allelic variation at the CETP locus. In a sibpair study Bu et al²⁷ concluded that a gene or genes influencing apoA-I and apoA-II levels were linked to the APOA2 locus.

In the present study we sought to determine whether a major gene influences plasma concentrations of apoA-I and, in view of the well-established effects of T₃ on apoA-I plasma concentrations, whether the expression of genes influencing apoA-I levels is mediated by T₃. Additionally, we tested the hypothesis that polymorphic variation at the apoA-I structural locus accounts for some of the quantitative variability in apoA-I plasma levels.

Methods

Subjects

The subjects in this study were 376 Mexican American individuals in 24 extended pedigrees ascertained as part of the San Antonio Family Heart Study, a population-based study of the genetics of atherosclerosis, diabetes, and obesity in Mexican Americans living in San Antonio, Tex. The pedigrees ranged in size from 2 to 60 examined individuals (mean=15.7) and included 146 males and 230 females. These 376 individuals resided in 225 households, with the number of households per pedigree ranging from 1 to 11. Families were ascertained without regard to any specific disease, through a randomly chosen 40- to 60-year-old proband. To be eligible, a proband was required to have at least six first-degree relatives (excluding parents) at least 16 years of age and living in San Antonio. Pedigrees include the proband and his or her spouse, and all age-eligible first-, second-, and third-degree relatives of both. Table 1 provides more detailed information regarding the composition of the pedigrees. It shows all possible pairwise relationships within the 24 pedigrees, stratified by household-sharing status.

Data are available on numerous quantitative phenotypes related to heart disease, diabetes, and obesity in these family members. The phenotypic measures include plasma concentrations and size distributions of lipoproteins and apolipoproteins, as well as sex hormones, thyroid hormones, and fasting and 2-hour insulin and glucose. Information has been gathered on diet using a food-frequency questionnaire designed according to procedures developed and validated by Willett et al²⁸ and modified by us using a database that included those food items which in aggregate account for 80% to 85% of the consumption of total calories, total protein, total fat, saturated fat, cholesterol, total carbohydrate, starch, and sucrose in Mexican Americans living in San Antonio.²⁹ Data on other environmental covariates, including smoking behavior, alcohol use, physical activity, and prescription drug use, are also available. Genotypes have been determined for more than 20 candidate loci, and a repository of immortalized lymphocytes provides a source of DNA for each family member.

Samples

Blood samples were obtained from participants of the San Antonio Family Heart Study after an overnight fast. Plasma and serum samples were stored as aliquots at −80°C. All procedures were approved by the Institutional Review Board of the University of Texas Health Science Center at San Antonio, and written informed consent was obtained from all participants.

Assay of ApoA-I

Plasma concentration of apoA-I was measured at Medical Research Laboratories, Cincinnati, Ohio using a nephelometric method.³⁰³¹ The interassay CV for quality control samples was 3.5%. In the current sample of 376 Mexican Americans, the mean apoA-I plasma level was 145.5 mg/dL, with an SD of 26.3 and a range of 86 to 232 mg/dL.

Assay of T₃

Serum concentration of T₃ was measured by radioimmunoassay using a commercially available kit (Diagnostic Product Corporation). Automated counting, standard curve determination, and sample estimation were performed on an LKB-Wallac 1282 Compu-Gamma counter. Quality control software developed by us was used to identify assay aberrations and inconsistencies. Data files were generated automatically using specialized software developed to interface between our pedigree database system (PEDSYS³² ) and the gamma counter. The mean T₃ level in the sample of 376 individuals was 1.84 nmol/L, with an SD of 0.39 and a range of 0.86 to 3.14. The intra-assay CV was 5.9%, and the interassay CV was 8.7%.

APOA1 Genotyping

Lymphocyte DNA was extracted from blood samples as described elsewhere.³³ All genetic markers were typed by PCR. PCR amplification reactions (20 mL total volume) used 0.5 mg of lymphocyte DNA, 1.0 pmol/mL of each oligonucleotide primer, 0.025 U/mL Taq polymerase (Perkin-Elmer Cetus), and buffer and nucleotide components described by the supplier. Three restriction fragment length polymorphisms were typed in the APOA1-APOC3-APOA4 region on chromosome 11: an Msp I polymorphism in the 5′ promoter region of APOA1 (heterozygosity (H)=0.44),³⁴ an Msp I polymorphism in intron 3 of APOA1 (H=0.30),³⁵ and an Sst I polymorphism in the 3′ flanking region of APOC3 (H=0.28).³⁶ The sequences of amplification primers and PCR temperatures for the APOA1 promoter polymorphism and the APOC3 Sst I polymorphism are given in their respective references. The Msp I polymorphism in APOA1 intron 3 was typed by PCR using forward primer 5′-GGGATCCAGGCTCGGCATTTCTGGCAG-3′ and reverse primer 5′-GCTGCAGGTGGAGGTCACGCTGTCCC-3′. PCR was achieved by using 30 cycles of denaturation (95°C for 1.0 minute), annealing (59°C for 1.5 minutes), and elongation (72°C for 2.0 minutes). The PCR products were digested with Msp I (4 U, 37°C for 4 hours) and electrophoresed on 2.0% agarose gels.

To increase the informativeness of these loci for linkage analysis, we evaluated the three-locus haplotype for each typed individual. Because of the high gametic disequilibria among these markers, only five of the possible eight haplotypes were observed. Out of the 376 individuals sampled, 355 could be haplotyped unambiguously. Haplotype assignment for the remaining 21 individuals was handled probabilistically by conditioning on the observed marginal genotypes in the linkage analysis. The estimated heterozygosity of the APOA1/APOC3 haplotyped system was 0.61±0.01 in these pedigrees.

Statistical Genetic Methods

Segregation Analysis

Major gene effects on apoA-I were detected and characterized by complex segregation analysis³⁷ using the Pedigree Analysis Package (PAP),³⁸ which we modified to allow the effect of an environmental covariate on a quantitative trait to depend on genotype at a major locus.

In complex segregation analysis, various hypothetical models for the transmission of the quantitative trait are compared. Likelihood methods are used to choose the best among these models. Each model is characterized by penetrance parameters that describe the effects of genetic and nongenetic factors on the trait, and for each model, parameter values are estimated that maximize the likelihood of the pedigree data using a quasi-Newton optimization algorithm.³⁹ Each model is compared with the most general by computing −2 times the difference in the log_e likelihoods of the two models. This difference is asymptotically distributed as χ², with degrees of freedom equal to the difference in the numbers of parameters estimated in the two models.

We evaluated a general transmission model and four restricted models: sporadic, polygenic, mendelian, and environmental. In all models, covariate effects on apoA-I levels were estimated simultaneously with other parameters. All models included random environmental and common-household effects, and all except the sporadic included a polygenic component. The mendelian, environmental, and general models included an effect of a major factor contributing to heterogeneity in apoA-I levels. In the mendelian model, this factor is a single genetic locus, and in the environmental, it is a major environmental factor. In the general model, characteristics of the major factor are estimated (that is, the factor may be either genetic or nongenetic). Three transmission probabilities, τ_AA, τ_Aa, and τ_aa, are estimated, representing the probability that factor A is transmitted by individuals of type AA, Aa, and aa, respectively. In the mendelian model, these transmission probabilities are fixed at their mendelian expectations of 1.0, 0.5, and 0.0. In the environmental model, the three transmission probabilities are held equal, yielding random transmission between generations.

To make the inference that a major gene is influencing the phenotype in question, all competing models except for the mendelian model must be judged to be significantly worse fitting than the general model. The hierarchical nature of these tests minimizes the potential for false inference of major loci. Therefore, modern segregation analysis is a relatively conservative technique.

Covariate effects were simultaneously estimated within the segregation analysis. Because of the potentially large number of such effects, we first performed a preliminary screening procedure, in which all possible covariate effects were simultaneously estimated within the context of a polygenic model. This model was used to allow for the nonindependence among biological relatives so as not to inflate the type 1 error rate. Covariate effects that passed a significance test at α=0.10 were included in all subsequent genetic analyses.

Covariates considered were sex; sex-specific age (and age² ); serum T₃ concentration; dietary intakes of alcohol, saturated fat, and cholesterol, as determined using a food-frequency questionnaire; total physical activity (assessed by interview); and the following dichotomous variables: diabetic status (according to WHO criteria), diabetic or lipid-lowering medications, exogenous sex hormones (for females), menopausal status (for females), and current smoking status.

Because of the prior evidence that T₃ directly influences APOA1 gene action, the effects of T₃ were allowed to differ among major locus genotypes. We have modified the standard mixed model⁴⁰⁴¹ to allow for genotype×environment interaction. In the mendelian model with genotype-specific covariate effects, the quantitative phenotype of an individual with a given major locus genotype is

where μ_jk is the mean value for genotype jk (jk=AA, Aa, or aa), x_i is the covariate effect, β_jk is the genotype-specific regression coefficient of p_i on x_i for genotype jk, g_i is the polygenic component, h_i is a shared-household effect, and e_i is the random environmental component. Equation 1 is easily extended to an arbitrary number of covariate effects whose regression coefficients, in the absence of genotype×environment interaction, involve the constraint β_AA=β_Aa=β_aa. The phenotypic variance is a function of the covariate and is given bywhere σ²_m(x) is the variance component due to the major gene, σ²_g to polygenes, σ²_h to shared-household effects, and σ²_e to random environment. The hypothesis of no interaction between major genotype and covariate is tested by testing the hypothesis that the genotype-specific regressions are equal (β_AA=β_Aa=β_aa), or alternatively, the hypothesis that the variance component associated with the major gene is not a function of the covariate [σ²_m(x)=σ²_m].⁴⁰

Linkage Analysis

To test the hypothesis that the major gene is linked to a specific candidate gene, we performed linkage analysis using PAP³⁸ with modifications.²¹⁴⁰ In this approach, the allele frequencies, genotype-specific means, covariate effects, and other parameters obtained from segregation analysis define the penetrance function. The linkage parameter to be estimated is the recombination fraction θ between the major gene and the candidate gene. Tests of linkage were based on a comparison of the log₁₀ likelihood of a model in which θ is estimated with one in which θ is constrained to equal 0.5 (no linkage). The difference in log₁₀ likelihoods between these two models is called the lod score. The lod function shows the change in lod score as a function of θ. Linkage is excluded for values of the lod function <−0.834, which corresponds to a value of P=.05, assuming a χ² distribution for the test statistic lod·2·log_e(10). Similarly, since we are dealing with a prespecified hypothesis utilizing a candidate gene (ie, not a random polymorphic marker), significant support for linkage requires a value of P≤.05.

Results

Descriptive Statistics

Table 2 shows the sex-specific raw means and SEs for all of the variables considered in the analysis. Females exhibit significantly higher plasma apoA-I concentrations than do males. T₃ levels are not significantly different between the sexes. In this sample, there are significant differences in environmental lifestyle variables (physical activity, alcohol, diet, and smoking) between the sexes. Males have higher levels of physical activity, higher alcohol intake, higher dietary saturated fat and cholesterol, and a higher proportion of smokers than do females.

Segregation Analysis

Table 3 summarizes the results of the segregation analyses, with statistical tests for each restricted model compared with the general model. Results are shown for the situation in which the effects of T₃ on apoA-I level are constant across genotypes (no MG×T₃ interaction) and the case in which T₃ effects are genotype-specific (MG×T₃ interaction). When no MG×T₃ interaction is allowed, no clear picture of the inheritance of apoA-I levels emerges. The sporadic model is strongly rejected and the polygenic model is nearly rejected, thus implicating a major factor. However, neither the environmental nor the mendelian model can be rejected in comparison to the general model. Thus, while a major gene effect cannot be ruled out, neither can an effect of a major environmental factor.

In contrast, when MG×T₃ interaction is allowed, our statistical inference becomes unequivocal. The only model not rejected in comparison to the general model is the mendelian model. These results suggest that a single genetic locus involving the differential effects of thyroid hormone is of major importance for apoA-I variation in this population.

The best (most parsimonious) model for the determination of apoA-I concentration is a mendelian model in which a, the allele determining high concentrations of apoA-I, has a frequency of 0.22±0.07. The three genotypes are assumed to be in Hardy-Weinberg proportions, so that the frequencies of the three genotypes AA, Aa, and aa are 0.61±0.11, 0.34±0.08, and 0.05±0.03, respectively. (SEs for genotypic frequencies were obtained using a Taylor series approximation.) At the average T₃ concentration of 1.84 nmol/L, the mean apoA-I concentrations of the three genotypes are 129.7±4.3, 149.4±7.4, and 179.5±10.9 mg/dL.

Fig 1 shows that the three genotypes differ significantly in the relationship between apoA-I and T₃ levels (χ²₁=4.25, P=.039), as evaluated by a χ² statistic with 1 df, which tests the constraint β_AA,Aa=β_aa. The AA and Aa genotypes exhibit an increase in apoA-I concentration with increasing T₃ (β_AA,Aa=14.2±4.0), while the rarer aa homozygote shows a strong negative relationship with T₃ (β_aa=−40.5±18.0).

The variance in apoA-I concentration can be partitioned into a component attributable to the major gene, a polygenic component (due to the additive effects of other genes that are individually undetectable), a shared-household component, and a random environmental component. There was a marginally significant influence of shared household on apoA-I variation (P=.05). On average, shared household environment accounts for 11% of the total phenotypic variation. There was also evidence for significant residual genetic effects (P=.031), which accounted for another 23% of the variation. Similarly, on average, the major gene accounts for 30% of apoA-I variation. However, given the significant interaction between the major gene and T₃ serum levels, the major genic variance is a function of T₃, as shown in Fig 2. The proportion of variance in apoA-I concentration due to this major gene (ie, the major genic heritability) decreases from 46% at a T₃ level of 1.25 nmol/L (representing the 5th percentile of the T₃ distribution) to 18% at a T₃ level of 2.5 nmol/L (the 95th percentile). The upper line shows the total heritability of apoA-I including both major genic and residual genetic variation.

Covariates that significantly influenced apoA-I concentration included sex, age, exogenous sex hormones, dietary saturated fat, and dietary cholesterol. For example, apoA-I concentration in females is 10.37±3.29 mg/dL higher than in males. There is a significant age effect in females, with an increase of 0.44±0.10 mg/dL per year. Exogenous sex hormones (coded as a dichotomous variable) increase apoA-I concentration by 12.39±4.14 mg/dL in females. ApoA-I concentration decreases with increasing consumption of dietary saturated fat (a decrease of 0.12±0.07 mg/dL per gram of saturated fat) and increases with increased dietary cholesterol (an increase of 0.012±0.005 mg/dL per gram of cholesterol).

In the general model against which this mendelian model was tested, the estimated parameter values were close to those estimated for the mendelian model. The estimates of the transmission probabilities were τ_AA =0.93±0.05, τ_Aa=0.62±0.14, and τ_aa =0.00±0.44. The similarity of the general and mendelian models further strengthens evidence for a major gene effect on apoA-I concentration.

Linkage Analysis

To determine whether the major gene that influences apoA-I concentrations is the APOA1 structural gene, we performed analyses using haplotype data for three polymorphic markers in the APOA1-APOC3-APOA4 region. Fig 3 shows the results of the linkage analysis. The maximum likelihood estimate of the recombination fraction (θ) was 0.50±0.22, indicating that there is no evidence for linkage of the major apoA-I gene with the APOA1 structural locus. Fig 3 also shows that the APOA1 locus and surrounding area can actually be excluded as the site for the putative major gene.

Discussion

Strong evidence for a major locus with two alleles (A and a) determining plasma apoA-I concentration was obtained when interaction with T₃ was allowed. This finding suggests that a single locus involving the differential effects of thyroid hormone is of major importance for apoA-I variation in this population and that thyroid hormone is a major determinant of the genetic variation in this quantitative lipoprotein phenotype. The linkage results indicate that the putative major locus is not the APOA1 structural locus. The test for linkage is very specific. It excludes the APOA1 gene from being the characterized major locus; it does not exclude the APOA1 gene from playing any part in quantitative apoA-I variability. Since linkage to the APOA1 structural locus can be excluded, and given the presence of significant major gene×thyroid hormone interaction, we suggest that a variant in one of the thyroid hormone receptor genes may be the locus responsible for variation in plasma apoA-I levels. This hypothesis is currently being tested.

Our results indicate that thyroid hormone is an important determinant of apoA-I variability in normal individuals. In other studies, we have determined that the effect of T₃ on lipoprotein phenotypes is primarily genetic in origin. Genetic variation in T₃ also accounts for a portion of the genetic variance in other phenotypes related to reverse cholesterol transport.⁴² T₃ concentration accounts for ≈16% of the genetic variance in HDL-C; 21% of the genetic variance in apoA-II, another protein component of HDL-C; and 37% of the variance in LpA-I, the portion of HDL-C that contains apoA-I but no apoA-II.

Although genetic studies of normal T₃ variation are few, there is evidence that T₃ concentration is itself heritable. Meikle et al⁴³ presented data on twin pairs that suggest a heritability ranging from 0.42 to 0.98. Comuzzie et al⁴² conducted a quantitative genetic analysis of T₃ levels in the San Antonio Family Heart Study using a maximum likelihood variance decomposition method that allowed for simultaneous estimation of covariate effects. They estimated the heritability of T₃ to be 0.46±0.12, indicating that 46% of the phenotypic variation in T₃ is attributable to additive genetic effects. The estimated heritability of FT₃ was lower, 0.17±0.09. There appear to have been no other studies of the heritability of T₃ or FT₃. Similarly, there are no studies of major gene effects on T₃ levels. Thus, it remains to be determined whether the T₃ effects on apoA-I are mediated by genetic variation in T₃.

In summary, we have examined the genetic determinants of quantitative variability in plasma apoA-I levels in a normal Mexican American population. Our data indicate that ≈53% of the phenotypic variance in apoA-I plasma levels is due to genetic factors, with 30% being due to the effects of a single major gene. The effect of this gene is mediated by serum levels of the thyroid hormone T₃, with lower levels of T₃ leading to increased variation in apoA-I levels among major locus genotypes and higher levels of T₃ leading to diminished differences in apoA-I levels among genotypes. This interaction between the expression of the major gene and circulating thyroid hormone is plausible given the known effects of T₃ on APOA1 gene expression. Additionally, using linkage analysis, we have excluded the APOA1 structural locus as the site of this mutation. Future studies will focus on the localization of this important quantitative trait locus.

Selected Abbreviations and Acronyms

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Footnotes