|
| HOME | SUBSCRIPTIONS | CURRENT ISSUE | PAST ISSUES | CARDIOSOURCE | SEARCH | HELP | FEEDBACK |
|
|
|
|||||||||
|
|||||||||||
|
J Am Coll Cardiol, 2007; 49:2053-2060, doi:10.1016/j.jacc.2007.02.039
(Published online 3 May 2007). © 2007 by the American College of Cardiology Foundation |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,*,
* Cardiovascular Department, LDS Hospital, Intermountain Medical Center, University of Utah, Salt Lake City, Utah
Genetic Epidemiology Division, Department of Biomedical Informatics, University of Utah, Salt Lake City, Utah
Cardiology Division, Department of Internal Medicine, University of Utah, Salt Lake City, Utah
Genetic Research, Intermountain Healthcare, Salt Lake City, Utah.
Manuscript received December 1, 2006; revised manuscript received January 29, 2007, accepted February 5, 2007.
* Reprint requests and correspondence: Dr. Benjamin D. Horne, Cardiovascular Department, LDS Hospital, Intermountain Medical Center, 8th Avenue and C Street, Salt Lake City, Utah 84143. (Email: benjamin.horne{at}intermountainmail.org).
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionConclusionsAppendixReferences |
|---|
Background: The CETP gene plays a critical role in lipoprotein metabolism, but the common and well-studied TaqIB variant is inconsistently predictive of CAD.
Methods: From a deoxyribonucleic acid bank of 10,020 individuals, nondiabetic nonsmoking patients (n = 4,811) with angiographically defined, clinically significant CAD (
70% stenosis) or normal coronaries were genotyped for 11 CETP tSNPs. Myocardial infarction (MI) and lipid levels were evaluated as secondary end points.
Results: Analysis of single tSNPs, corrected for multiple comparisons (p < 0.00485), identified allele +1086A to be associated with CAD (p = 0.0034). Suggestive allelic and significant genotypic associations were found for –631AA (odds ratio [OR] = 3.95, p = 0.004 vs. CC) and +2389GA (OR = 1.21, p = 0.003 vs. GG). Haplotype analysis by linkage disequilibrium (LD) group revealed a CAD association for LD group B (p = 0.0025 across T+1086A, C+878T, C+408T) and near significance for LD group A (p = 0.013 across C-631A, MspI, G+2389A). A weak protective trend for TaqIB was eliminated by adjustment for other tSNPs, and haplotype analyses suggested that TaqIB was simply a marker for other tSNPs or haplotypes. No tSNP or haplotype associations with MI were found.
Conclusions: Multiple, less common SNPs and haplotype variants underlie CETP-related CAD risk, for which the common TaqIB variant is simply a poor marker. The occurrence of risk-related variants on separate haplotypes suggests genetic-risk complexity and allelic heterogeneity. (Database Registry of the Intermountain Heart Collaborative Study; http://clinicaltrials.gov/ct/show/NCT00406185?order=1; NCT00406185 [ClinicalTrials.gov] ).
| ||||||||||||||||
The 22-kilobase cholesteryl ester transfer protein (CETP) gene, located on chromosome 16q21, specifies a 493 amino acid protein, expressed in multiple tissues, that localizes in the circulation primarily on larger, lipoprotein-AI–containing HDL-cholesterol (C) particles, where its principal role is to catalyze the exchange of triglycerides (TG) on apoB-containing particles (e.g., LDL-C, very LDL-C) for cholesteryl esters from HDL-C (6). Higher CETP concentrations typically are associated with lower HDL-C (7,8). TaqIB and other variants of CETP have been associated with reduced CETP activity. However, despite the consistent association these variants have with higher HDL-C levels, their clinical phenotype remains uncertain (9). The previous postulate that CETP activity is inversely related to CAD risk (10–15) has been challenged by the recent, unexpected finding that pharmacological inhibition of CETP (16) increases cardiovascular risk (17).
Given the critical role of CETP in reverse cholesterol transport and hence potentially in CAD risk, this study comprehensively evaluated CETP genetic variation via SNPs and SNP haplotypes in a large, angiographically phenotyped patient cohort. The goal was to better define CETP genotype-phenotype associations and optimally capture both common and uncommon genetic variation (18,19) with a recently described set of tagging (t) SNPs (20).
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsAppendixReferences |
|---|
Assessment of CAD.
The presence of CAD, the primary study end point, was determined by standard coronary angiographic procedures. The patient’s angiogram was graded by a cardiologist (who was unaware of genetic test results). Patients were categorized as being free of CAD (i.e., free of CAD or with minimal, <10% stenosis), having moderate CAD (i.e., most severe lesion 10% to 69% stenosis), or having significant CAD (i.e., 1 lesion of 70% stenosis). Moderate CAD was classified as an indeterminate phenotype and patients in that group were excluded from statistical analyses. Because it is the most clinically relevant phenotype and the decision to treat is based on the presence of flow-limiting lesions (i.e., 70% stenosis), CAD presence was defined as those with significant CAD, and non-CAD control subjects were those patients free of CAD.
Secondary end points. The secondary end point of MI was defined as any MI in the patient’s prior history or on hospital admission. The MI events were determined from Intermountain Healthcare’s electronic record repository and were ruled-in by electrocardiography and/or biomarker measurements. An MI was defined by biomarkers as a creatine kinase-myocardial band (CK-MB) >6 mg/dl and a CK-MB index >3% in the appropriate clinical setting. For MI association analyses, the subgroup of patients with moderate CAD was included.
Lipid measurements were also evaluated as secondary end points. Total cholesterol (TC) and TG were measured from fasting blood samples drawn during hospital stay but before angiography with dry-slide technology on a VITROS 950 analyzer (Ortho Clinical Diagnostics, Raritan, New Jersey). The HDL-C was measured after sample treatment with VITROS HDL-Cholesterol Magnetic Reagent. The LDL-C was calculated from those values.
Clinical information.
Demographic and health history data were obtained from physicians and hospital records and stored in a research database. These data included age, gender, smoking status, hypertension, hyperlipidemia, diabetes, and family history of early CAD or MI. Smoking was considered present for active smokers or those with a >10 pack-year history. Prevalent diabetes mellitus, hyperlipidemia, and hypertension were physician-reported from clinical and laboratory findings or were based on current use of relevant medications. For laboratory findings, diabetes was defined as a fasting blood sugar 126 mg/dl, hypertension as systolic blood pressure 140 mm Hg or diastolic blood pressure 90 mm Hg, and hyperlipidemia as fasting TC 200 mg/dl or LDL-C 130 mg/dl. Patient-reported family history was positive if a first-order relative had suffered cardiovascular death, MI, or coronary revascularization before age 65 years.
Genotyping. Previously, sequence variation in the CETP gene was determined among 50 unrelated individuals in the promoter, 16 exons, exon-intron boundaries, and 3' untranslated region (20). From 32 SNPs, 11 tSNPs were selected from 7 linkage disequilibrium (LD) groups (20) with an analysis method that evaluated inferred haplotypes from 339 unrelated individuals (19). Those 11 tSNPs were included in this study. Haplotypes were determined from these tSNPs as discussed in the next section.
Genotyping of rs1800776 (C-631A) and rs11076175 (MspI) was performed with Simple Probe chemistry and melt curve analysis on the Light Typer (Roche Diagnostics, Indianapolis, Indiana). The other tSNPs were genotyped with 5' exonuclease (Taqman) chemistry on the ABI Prism 7000 (Applied Biosystems, Foster City, California). Assay validation used direct sequencing via ABI Big Dye v 3.1 terminator chemistry.
Study hypothesis and design. The primary study hypothesis was that 1 or more of the 11 tSNPs or a tSNP-defined haplotype would predict an altered risk of angiographic CAD. Secondary end points of interest were MI and lipid levels (e.g., HDL-C). Study patients were separated a priori into three mutually exclusive groups, with the primary study population being nondiabetic, nonsmoking patients. To reduce unwanted variation from non-cholesterol genetic and environmental factors, diabetic patients and smokers were excluded from the primary analysis and evaluated as secondary study populations as: 1) any smoker, and 2) nonsmoking diabetic patients (mutually exclusive). Analysis was restricted to those patients with complete genotypic data for all 11 tSNPs.
Statistical considerations. Differences in CAD diagnosis and MI were evaluated for each tSNP by the chi-square test. Multivariable logistic regression provided adjusted odds ratios (ORs) and 95% confidence intervals (CIs) after forced entry of covariables (age, gender, cardiac risk factors). Adjustment also was made in other models for TC, LDL-C, HDL-C, TG, and self-reported ethnicity. Evaluation of tSNP associations with lipid levels was performed by Student t test (TG level was log-transformed first).
Haplotype-defining tSNP groups were assigned a priori on the basis of prior LD group findings (20). Inference of phased haplotypes and comparison of haplotype frequency differences between patients with and without CAD or MI was performed with SimHap and on the basis of 1,000 simulations. Haplotypes of frequency >0.01 that contained a variant allele were compared with the LD group’s wild-type haplotype. Extended haplotypes with tSNPs from multiple LD groups were then evaluated to find the best CAD risk-discriminating haplotypes. Haplotype analyses did not employ multivariable adjustment, owing to software restrictions. Finally, association analyses were performed at the patient level to obtain ORs for the best haplotypes on the basis of carriage of 0, 1, or 2 copies of the haplotype (ignoring phase uncertainty).
Pair-wise LD was calculated and plotted with Haploview. Other statistical analyses were performed with SPSS (version 13.0; SPSS Inc., Chicago, Illinois). Two-tailed p values were designated as significant for p 
0.00485 on the basis of correction for multiple comparisons of the 11 tSNPs (20) and were considered suggestively associated for p 0.05.
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsAppendixReferences |
|---|
|
1 genotype). Pair-wise LD between the 11 tSNPs is shown in Figure 1. Various tSNP associations with TC, LDL-C, HDL-C, and TG were found (Table 2).
|
|
|
|
Role of TaqIB. Both allelic (Table 3) and genotypic (AA [B2B2] vs. GG [B1B1]: OR = 0.86, p = 0.08) analysis showed a weak association of tSNP TaqIB with CAD. The association was moderately reduced by adjustment for traditional risk factors (p = 0.12) and eliminated by adjustment for tSNPs C–631A, MspI, T+1086A, C+878T, G+2389A, and C+408T (OR = 1.04, p = 0.74). In the LD group D haplotype to which TaqIB belongs, some potential associations were suggested, but the evidence was spread across multiple haplotypes (Table 4). Haplotypes across LD groups A, B, and D were considered (i.e., by adding TaqIB to the 6 tSNPs previously mentioned), but this did not increase the discriminatory power of any haplotype. Instead, the CAD association evidence for the 6-tSNP CAATGC haplotype from LD groups A and B was split between two 7-tSNP haplotypes (data not shown).
Association with MI. Thirty percent of patients had experienced an MI (Table 1). No single tSNP was associated with MI after correction for multiple tests (online ), and no haplotype was associated with MI (online ).
Nonsmoking diabetic patients. Of the 1,592 nonsmoking diabetic patients (not studied in the preceding text), 1,270 were included in this secondary analysis (160 were excluded owing to moderate CAD and 162 owing to missing genotypic data). Gender (p = 0.009), hypertension (p < 0.001), hyperlipidemia (p < 0.001), and family history (p < 0.001) were all different compared with nonsmoking nondiabetic patients. No significant tSNP or haplotype associations were found among diabetic patients (see online ).
Smokers. Of the 1,910 smokers not studied in the primary analysis, 1,460 were included in secondary analysis (177 were excluded owing to moderate CAD and 273 owing to missing genotypes). Smokers were younger (61 vs. 63 years, p = 0.036) and more likely (p < 0.001) to be male and have CAD, hypertension, hyperlipidemia, and family history. Only A+1825C was CAD-associated (see online ), and when pooled with the primary population, an interaction of smoking with A+1825C was suggested (test of interaction: p = 0.001).
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsAppendixReferences |
|---|
Adjustment for traditional risk factors diminished some of the genetic associations. However, because genotype (present since conception) predates the appearance of clinical risk factors, it can be argued that unadjusted genetic risk might be more relevant to pathophysiological considerations (i.e., with genes potentially explaining risk factors rather than the reverse). In contrast, TC, LDL-C, HDL-C, and TG concentrations did not affect tSNP associations, although several tSNP associations with lipid levels were found. Thus, some of the CETP tSNP effects might relate to mechanisms beyond lipid concentrations, including lipoprotein functionality.
Rationale for elucidating genetic associations. Elucidation of the genetic underpinnings of CAD is of major scientific interest because: 1) it can provide crucial insights into disease pathogenesis; 2) it might be applied to individual risk prediction ("individualized medicine"); 3) it can stimulate the discovery of therapeutic interventions (i.e., guiding interventions directed at specific genes, gene products, or pathways); and 4) it underlies the pharmacogenetics of individual patient-drug interactions (3,21).
A logical approach to the discovery of genetic factors for CAD is to focus on genes in biologic pathways implicated in disease pathophysiology (3,22). Topol et al. (22) divided these into four over-arching (and interdependent) pathways: lipid metabolism, arterial inflammation, endothelial integrity, and thrombosis. The CETP gene represents a critical element in the lipid metabolic pathway; hence, understanding its potential contribution to genetic risk is crucial. However, despite the consistent association of CETP variants (including the common and well-studied TaqIB) with HDL-C levels, their impact on CAD end points remains obscure (9,23). Adding to the uncertainty about CETP’s role in CAD risk is the recent unexpected observation that pharmacological inhibition of CETP (i.e., with torcetrapib) (16) increases cardiovascular risk (17), whereas CETP activity previously had been postulated to be inversely related to CAD risk (10–15).
Reasons for past failures. A number of reasons have been proposed for the frequent failure to replicate SNP associations (24,25). One is the testing for associations in adequately large samples so as to minimize both Type I (false positive) and Type II (false negative) errors. A second problem is inadequate characterization and standardization of study populations, including imperfect ascertainment of clinical phenotypes, mixed genetic background of patients (e.g., population stratification), and evaluation of differing outcomes (i.e., CAD vs. MI). Another is a failure to address the problem of multiple comparisons with prospective, restrictive testing with adequate statistical correction. Still another source of complexity is the frequent dissociation of intermediate biomarkers (i.e., lipids/lipoproteins) and clinical outcomes (angiographic CAD or MI) (9,23,26).
In addition, a tested SNP might not be biologically functional but a marker of a functional variant in some but not other populations (18,19). Notably, CETP promoter SNP C-629A (and many other promoter SNPs) is in high LD with TaqIB (20,27). Furthermore, tSNP C+878T is in almost exact LD with exonic SNP I405V, whereas tSNP T+1086A resides on a subset of haplotypes containing those and 6 other SNPs (20). When functional SNPs are unknown, tests of variants in LD groups (i.e., clusters of variants that are linked/co-inherited) might more reliably characterize susceptibility (19,28).
Finally, the basic assumption in previous studies of "common disease, common variants" might not be generally correct. This hypothesis (29) implies that only very common variants (e.g., minor allele frequency [MAF] = 0.20 to 0.50) need be considered (30). However, less frequent, common (but not rare) genetic variants (i.e., MAF = 0.01 to 0.20) also might exert substantial susceptibility (31) for CAD and MI through the mechanisms of allelic (32) and locus heterogeneity. Indeed, multiple less common, non-synonymous variants in PCSK-9, ABCA1, APOA1, and LCAT have recently been reported to explain a major portion of population variance of LDL-C or HDL-C, further challenging the common disease, common variant hypothesis (4,5).
Present rationale. This study was designed to address several weaknesses of previous SNP association studies. The study was large, providing power to discover modest risk associations and reduce error. Sources of statistical noise (e.g., smoking) were excluded. Multiple comparisons were addressed by appropriate statistical correction. Relevant clinical end points (angiographic CAD, clinical MI) rather than biomarkers (lipids) were assessed and tested separately. The SNPs were selected from a comprehensive study (20) so as to maximally account for genetic variation in CETP.
Together, testing of single tSNPs and of haplotypes can be complementary. If single tSNPs are themselves disease-associated, they could represent the most powerful disease markers (33). In contrast, if they are not and/or if several functional SNPs exist in close proximity and interact, then haplotypes could represent superior markers (28). This study tested for both, with tSNPs as an appealing set of markers to address the first argument and haplotypes of tSNPs to address the latter. Finally, our study size allowed us to address possible associations of less common as well as the more common CETP polymorphisms with CAD and MI risk.
Implications of study findings. Similar to several other lipid-metabolic genes (4,5), multiple less common CETP variants (MAF = 0.01 to 0.20) were associated herein with CAD risk (e.g., C-631A and T+1086A). This finding, taken together with the previous examples, indicates that the "common variant common disease" hypothesis does not always hold and should not be assumed. Because SNPs with lower MAF necessarily require greater power to detect associations than very common SNPs, candidate gene studies should be much larger than previously envisioned (i.e., powered to adequately evaluate MAFs of 0.01 to 0.20). Although they require replication and extension, these findings revise the understanding of CETP with respect to CAD pathogenesis, therapeutic insights, individual risk predictions, and, eventually, individualized pharmacogenetics.
TaqIB and the present study. The CETP TaqIB has been evaluated extensively with varied results (e.g., see references 7,8,10–12,34,35) and a modest overall association in meta-analysis (15). In the present study, an apparent TaqIB association with CAD (genotypic OR = 0.86, p = 0.08) of similar magnitude to prior reports (15) was eliminated by adjustment for other tSNPs. The confounding of TaqIB resulted primarily because the minor alleles of two tSNPs associated with CAD (C-631A, G+2389A) were not on the haplotypes containing the protective TaqIB variant allele and because the variant alleles of the tSNPs associated with lower CAD risk (T+1086A, C+878T) were coincident with the TaqIB variant. This suggests that previous TaqIB associations likely arose as a result of multiple other CETP variants that are in LD with TaqIB. Thus, TaqIB is a less precise marker of CAD risk than the combination of multiple other less common CETP SNPs and should be replaced by these markers in future studies.
Study limitations. Not all CETP SNPs might be "tagged" by the 11 tSNPs, although they were selected to optimally account for genetic variation in coding and functional regions. Observational studies are subject to potential selection bias, and registry patients were at higher CAD risk than the general population. However, important potential confounders were dealt with by design (smoking and diabetes) or by multivariable adjustment (for other risk factors). The study consisted primarily of Caucasians, and the results might not apply equally to other ancestral groups. The functional effect of tSNP alleles was not directly assessed, but associations with lipid levels might be an indirect marker of functional activity (also, LD group B tSNPs are in LD with I405V and TaqIB [20], which decrease CETP mass [7,34]; MspI in LD group A might increase mass [7,34]).
|
Conclusions |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsAppendixReferences |
|---|
|
Appendix |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsAppendixReferences |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsAppendixReferences |
|---|
A polymorphism and a novel promoter variant are independently associated with CETP concentration Hum Mol Genet 2003;12:111-123.This article has been cited by other articles:
|
|
 |
|
  A. Thompson, E. Di Angelantonio, N. Sarwar, S. Erqou, D. Saleheen, R. P. F. Dullaart, B. Keavney, Z. Ye, and J. Danesh Association of Cholesteryl Ester Transfer Protein Genotypes With CETP Mass and Activity, Lipid Levels, and Coronary Risk JAMA, June 18, 2008; 299(23): 2777 - 2788. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | SUBSCRIPTIONS | CURRENT ISSUE | PAST ISSUES | CARDIOSOURCE | SEARCH | HELP | FEEDBACK |
