Admixture Mapping of Coronary Artery Calcified Plaque in African Americans With Type 2 Diabetes MellitusClinical Perspective
Background—The presence and severity of coronary artery calcified plaque (CAC) differs markedly between individuals of African and European descent, suggesting that admixture mapping may be informative for identifying genetic variants associated with subclinical cardiovascular disease.
Methods and Results—Admixture mapping of CAC was performed in 1040 unrelated African Americans with type 2 diabetes mellitus from the African American-Diabetes Heart Study, Multi-Ethnic Study of Atherosclerosis and Family Heart Study using the Illumina custom ancestry informative marker panel. All cohorts obtained computed tomography scanning of the coronary arteries using identical protocols. For each ancestry informative marker, the probability of inheriting 0, 1, and 2 copies of a European-derived allele was determined. Linkage analysis was performed by testing for association between each ancestry informative marker using these probabilities and CAC, accounting for global ancestry, age, sex, and study. Markers on 1p32.3 in the GLIS1 gene (rs6663966, logarithm of odds [LOD]=3.7), 1q32.1 near CHIT1 (rs7530895, LOD=3.1), 4q21.2 near PRKG2 (rs1212373, LOD=3.0), and 11q25 in the OPCML gene (rs6590705, LOD=3.4) had statistically significant LOD scores, whereas markers on 8q22.2 (rs6994682, LOD=2.7), 9p21.2 (rs439314, LOD=2.7), and 13p32.1 (rs7492028, LOD=2.8) manifested suggestive evidence of linkage. These regions were uniformly characterized by higher levels of European ancestry associating with higher levels or odds of CAC. Findings were replicated in 1350 African Americans without diabetes mellitus and 2497 diabetic European Americans from Multi-Ethnic Study of Atherosclerosis and the Diabetes Heart Study.
Conclusions—Fine mapping these regions will likely identify novel genetic variants that contribute to CAC and clarify racial differences in susceptibility to subclinical cardiovascular disease.
Despite similar or more detrimental cardiovascular disease (CVD) risk factor profiles, African Americans (AAs) have markedly lower levels of coronary artery calcified plaque (CAC) relative to European Americans (EAs).1,2 This observation is consistent in persons with and without type 2 diabetes mellitus (T2DM)3,4 and suggests that CVD risk factors have differential impacts on atherosclerosis based on ethnicity. The underlying cause(s) of ethnic differences in CAC are not well understood and likely reflect the interplay between multiple genetic and nonconventional CVD risk factors.5 Wassel et al6 reported positive association between CAC in AAs and proportion of European ancestry. We performed admixture mapping in 1040 AAs with T2DM to determine genomic regions contributing to ethnic differences in subclinical CVD. T2DM led to higher levels of CAC with the potential for improved power to better discriminate susceptible individuals. Regional admixture mapping (RAM), or mapping by admixture linkage disequilibrium (MALD), is a gene mapping tool used to identify genetic variants associated with differences in prevalence and distribution of a trait between 2 or more ancestral populations.7 The premise of RAM is that if a genetic variant underlies ethnic differences in disease, then it will be easier to map the location of that variant in a recently admixed population relative to panmictic ancestral populations. The proportion of alleles at the marker locus that have ancestry from the high-risk population will be higher in affected individuals than expected, by chance, assuming no other influential evolutionary forces (genetic selection or drift).
Clinical Perspective on p 105
RAM successfully identified disease genes or loci involved in various complex traits and diseases.8–10 Recent methodological developments revealed that combining admixture mapping and genetic association testing can lead to improved power.11
Materials and Methods
The discovery data were pooled from three study cohorts where CAC was quantified by cardiac computed tomography (CT) using identical protocols.12 Statistical analyses included 1040 unrelated individuals; 22 samples were removed because they failed quality control.
African American-Diabetes Heart Study
The African American-Diabetes Heart Study (AA-DHS) contained AAs with T2DM recruited from 2 Wake Forest School of Medicine studies: the Diabetes Heart Study (DHS) and AA-DHS. DHS is a cross-sectional study of EA and AA families with siblings concordant for T2DM. AA-DHS started after DHS and enrolled only unrelated AAs. AA-DHS objectives were to improve understanding of the striking ethnic differences in CAC observed between AAs and EAs. T2DM was diagnosed after the age of 30 years in the absence of diabetic ketoacidosis. The DHS family of studies is described in Bowden et al.13 Subjects who underwent prior coronary artery bypass surgery were not included in the MALD analysis, as the CAC mass score could be impacted.14 Those with prior myocardial infarction or stroke were included. The final analysis included 635 unrelated AAs obtained by selecting all AA-DHS participants and 1 AA from each of 150 DHS sibpairs (the sibling with the most complete phenotypic data and best kidney function). The study was approved by the Wake Forest School of Medicine Institutional Review Board; all participants provided written informed consent.
Multi-Ethnic Study of Atherosclerosis
The Multi-Ethnic Study of Atherosclerosis (MESA) investigates the prevalence, correlates, and progression of subclinical CVD.15 Recruitment was restricted to diabetic and nondiabetic individuals free of clinical CVD. The cohort included participants with previously and newly diagnosed diabetes mellitus. Data for this analysis were collected at baseline, with exception of HbA1c collected at examination 2. MESA contributed data on 302 unrelated AA participants.
Family Heart Study
A substudy of the Family Heart Study (FamHS) was conducted to obtain cardiac CT scans for CAC.16 University of Alabama in Birmingham recruited AAs, some with T2DM. For this analysis, individuals with coronary artery bypass surgery were excluded. FamHS contributed data on 103 unrelated AA participants.
Replication and validation efforts were conducted using existing genome-wide association (GWAS) data in MESA SHARe and in the EA DHS study. The MESA SHARe data came from 1350 AAs without diabetes and 2497 EAs with diabetes. The AAs with diabetes were included in our discovery sample. Analyses in DHS were conducted on 920 EAs with T2DM.
In summary, analyses were performed on 1040 unrelated AAs with T2DM, 1350 unrelated AAs without diabetes mellitus, and 3417 EAs with T2DM. MESA diabetic EA participants were also unrelated.
Genotyping of Ancestry Informative Marker Selection
DNA was extracted from peripheral blood using PureGene system (Gentra Systems, Minneapolis, MN). The streamlined panel of 1509 ancestry informative markers (AIMs) was typed using the Illumina AA admixture panel covering all 22 autosomes and chromosome X; analyses were restricted to autosomes.
Calcified plaque was measured in the coronary arteries using single and multidetector CT systems incorporating a standardized scanning protocol based on the National Heart Lung and Blood Institute’s MESA.17 Traditionally, the Agatston or Calcium score was reported. However, this scoring system adds noise to CT measurement of calcified plaque, compared with volume-based measures. We use the calcium mass score (SmartScore, General Electric Healthcare, Waukesha, WI) derived from the volume score, but also accounting for the density of calcified plaque on a pixel-by-pixel basis. Additional scoring parameters included a 90 Hounsfield Unit threshold and 2 adjacent pixels to define the maximum calcified lesion size and the program accounted for slice thickness.
Quality control checks similar to those performed in a GWAS were run before performing the main analyses. Twenty-nine single-nucleotide polymorphisms (SNPs) were flagged because Hardy–Weinberg equilibrium testing had P values <10−3, and 2 additional SNPs were flagged because their call rate was <98% (none showed a significant result). Six individuals with call rates <90% were dropped. We dropped 1 individual with a heterozygosity score outside the mean±4 times SE confidence interval. There was no indication of first-degree familial relationships in the analyzed dataset. The estimated kinship coefficient ranged between 0 and 0.13. Lower levels of familial relationship would be harder to detect with the small number of markers. We did not pursue evaluation of familial relationships further.
The RAM model that we applied is described in Redden et al.18 Briefly, for the ith individual, let Yi denote the phenotype of interest, let P1i and P2i represent the ancestry of proportions of this individual’s parents, and finally let I (Gij = k) denote dummy variables, indicating whether this individual has inherited exactly k copies of a particular allele from an ancestral population at the jth marker. The RAM model can be written as:(1)
where f( ) is the appropriate link function, and Ai is the individual ancestry (average of P1i and P2i the ancestry proportion of the 2 parents of the ith individual). Note that P1i and P2i do not have to be observed to fit this model. First, existing software will only provide an estimate of Ai. Second, the product of parental ancestries can be estimated based on the individual genotype (Redden et al18 contains details on a maximum likelihood estimate of [P1iP2i] that can be obtained without parental genotypes). Third, when the 2 parents have similar ancestry proportion, because it is likely the case under assortative mating where ancestry is closely linked with socioeconomic status, we have P1iP2i ≈ Ai2. We used ADMIXMAP19 to compute these estimates. ADMIXMAP used a combination of Bayesian and classical approaches to fit a multilevel model for the distribution of individual ancestry proportion in the population and the stochastic variation of ancestry on hybrid chromosomes.
In practice, I (Gij = k), where k=0, 1, or 2 and will not be available; however, the probability of inheriting exactly 0, 1, or 2 alleles from a specific ancestral population can be estimated. These probabilities are identity by descent probabilities; therefore, Equation 1 allows for testing for linkage controlling for individual and parental ancestries.20,21 This model can be fitted easily using standard statistical packages, facilitating inclusion of covariates and interaction effects. Details regarding the computation of these probabilities based on the observed genotypes and the estimated individual ancestry proportions are in the appendix. We use logarithm of odds (LOD) scores to present evidence against the null hypothesis to reinforce the idea that we are performing linkage analysis.
Analyses were run using Log(CAC+1) as a continuous trait and CAC dichotomized (individuals with CAC >10 treated as cases and CAC <10 as controls). This is justified based on the assumption that factors governing presence of CAC may differ from those influencing amount of CAC once calcification is initiated.22 Age, sex, and study were included as covariates in Equation 1. Online-only Data Supplement Table I suggests the distribution of Hb1Ac, body mass index, smoking status, and the use of lipid-lowering medication were statistically different among the three studies. However, the model adjusted for age and sex, and the study allowed us to capture these differences.
Significant Effect and Correction for Multiple Testing
We excluded chromosome X from our analyses; therefore, the MALD analyses were run on 1426 SNPs. A strict Bonferroni correction would place the significance threshold at 3.5×10−5 for a 2-sided test, an excessively conservative threshold. We chose to prioritize AIMs that reached a LOD score of 2.5, corresponding to an alpha level of 7×10−4. This approach can be viewed as somewhat conservative for identifying suggestive evidence of linkage. Following Reich and Patterson’s recommendation,23 we also repeated the analyses using the software ANCESTRYMAP.24 The results were similar to our unadjusted analyses and are not discussed further.
To confirm the preliminary results, additional mapping was performed using a sparse set of markers located on 3 chromosomal regions: 1p32.3 containing the largest score (LOD=3.7); 9p21.2 with a LOD of 2.7 and strong prior evidence of involvement in coronary artery disease; and 11p15.4 with suggestive evidence of linkage (LOD=2.5), the lowest LOD score considered suggestive. SNPs for fine mapping were selected from HapMap rel27 with preference given to AIMs with a δ-value of ≥0.5. We used an r2 threshold of 0.3 to minimize the linkage disequilibrium among the 53 selected SNPs. Pairwise linkage disequilibrium between markers was evaluated using Haploview.25 Genotyping was performed using the MassARRAY SNP Genotyping System (Sequenom Inc., San Diego, CA).26 SNP genotyping was >94.4% efficient, and 71 blind duplicates were 99.8% concordant. Quality control checks similar to those described herein identified 2 SNPs with Hardy–Weinberg equilibrium test P values <10−3. These analyses were conducted in the discovery sample. Results obtained with these SNPs are shown in online-only Data Supplement Table Va–Vc.
Additional Analyses in MESA
Approximately 1 million SNPs are typed using the Affymetrix Genome-Wide Human SNP Array 6.0. First, we focused on 1350 AAs without diabetes in MESA (AAs with diabetes were already included in our initial analysis). Second, we analyzed the 2497 EA participants with diabetes mellitus. We again conducted association tests between Log (CAC+1) and CAC as binary outcome with a cut point of 10. SNPs were selected located 1000 kb upstream and downstream of the sentinel AIM that had a LOD score >2.5. SNPs were either directly genotyped or imputed using IMPUTE.27 Imputation in the EA sample was performed using 1000 Genomes EUR (BCBI Build 37) as the reference panel, and the combination HapMap I + HapMap II+Utah residents with ancestry from northern and western Europe (CEU)+Yoruba in Ibadan, Nigeria (YRI)+Han Chinese in Beijing, China (CHB)+Japanese in Tokyo, Japan (JPT) (National Center for Biotechnology Information Build 36) served as the reference panel in the AA sample. We used the expected genotype (allelic dosage) for imputed SNPs. The analyses were adjusted for age, sex, and individual admixture proportion. A Bonferroni correction was applied in each region by dividing the nominal alpha level (5%) by the effective number of independent markers.28 Therefore, the significance threshold was lower than the typical threshold used for a whole-genome scan.
Analyses in European Americans From the DHS
We genotyped SNPs located 1000 kb upstream and downstream of each marker that had a LOD score >2.5. Linear mixed effect models were fitted to account for the familial correlation as measured by the realized kinship coefficient matrix. These models were fitted using the Genome-Wide Association analyses with Family data package in R29 adjusting for age, sex, and principal components. The significance threshold was again established using the effective number of tests.28
Demographic and Clinical Characteristics
Demographic and clinical characteristics of 1040 unrelated AAs with T2DM from AA-DHS, MESA, and FamHS have been reported previously,5 summarized in Table 1. Participants had a mean (SD) age of 58.8 (10.2) years, diabetes mellitus duration of 10.4 (8.5) years, HbA1c of 7.9% (2.0), fasting serum glucose of 153.8 (62.68) mg/dL, high-density lipoprotein-cholesterol of 48.1 (13.4) mg/dL, and low-density lipoprotein-cholesterol of 109.8 (35.7) mg/dL. The mean (SD) CAC mass score was 276.0 (628.7) Hounsfield Unit, with a median CAC mass score of 10 Hounsfield Unit. Participants were 54.5% female, 23.7% current smokers, and 35.6% former smokers. Approximately 45% took lipid-lowering medications, 36% insulin, and 42% angiotensin-converting enzyme inhibitors. Demographic and clinical characteristics of each study are provided in online-only Data Supplement Table I. Analyses were adjusted for study to account for variations among the 3 cohorts. We repeated the analyses stratified by study, and results were similar to those observed in study-adjusted analyses (online-only Data Supplement Table II).
Ancestry Proportion Distribution
The allele frequency of each AIM estimated in the Yoruban and Center d’Étude du Polymorphisme Humain populations was supplied as prior probabilities in the estimation process. The average proportion of genome-wide African ancestry in the combined sample was 80% (SD 11.5%). When the sample was split based on a CAC cutpoint of 10 Hounsfield Unit, the average proportion of African ancestry was 79% for individuals with CAC ≥10 and 81% for CAC <10 (P=0.002). The association between CAC and African ancestry proportion remained statistically significant after adjusting for age, sex, and study with an odds ratio of 0.24 (P=0.02) when CAC was analyzed as a binary outcome and a parameter estimate β=−1.8 (P=0.01) when Log(CAC+1) served as the outcome.
Admixture mapping was performed using CAC mass score as a binary trait (online-only Data Supplement Table III), and as a continuous outcome (online-only Data Supplement Table IV). These complementary analyses led to similar conclusions in most cases. When evidence of linkage was detected in a region with both outcomes (quantity and presence), we focused only on the result observed with the continuous outcome because the observed LOD scores tended to be greater. The result with the lower P value is reported in all cases. Results shown for presence (CAC >10 versus CAC <10) were only observed with this outcome because presence and amount of CAC may be governed by different processes.30,31 We considered the additive, dominant, and recessive models, allowing us to effectively reduce the models described in Equation 1 to 1 degree of freedom tests, with potential improvement in power.32 Adjusted P values computed using the maximum of the test statistic observed with the dominant, recessive, and additive models appeared to be 1 order of magnitude lower than their unadjusted counterparts.33 SNPs at 1p32.3 (rs6663966, LOD=3.7), 4p21.2 (rs1712373, LOD=3.0), 8q22.2 (rs6994682, LOD=2.7), 9p21.2 (rs439314, LOD=2.7), and 13q32.1 (rs7492028, LOD=2.8) revealed significant evidence of linkage with Log (CAC + 1). We also detected significant evidence of linkage at 1q32.1 (rs7530895, LOD=3.1), 4q31.1 (rs1872861, LOD=2.7), and 11q25 (rs6590705, LOD=3.3) with the binary outcome. Results for these and other SNPs reaching a LOD score >2.5 are summarized in Table 2. This table also lists the average probability of inheriting exactly 2 copies of the European-derived allele at each marker. Based on the global ancestry proportion estimate, the probability that an AA individual would inherit exactly 2 alleles at random from the European population is 4%. For individuals with CAC mass scores >10, we observed that the mean probability of inheriting exactly 2 European-derived alleles at rs6994682, for example, was ≈6.5%, >60% higher than the expected probability under the null. The observed mean was 4.1% in those with CAC mass scores <10.
RAM Results Stratified by Study
Analyses were repeated stratified by study, instead of accounting for study as in the initial model. Online-only Data Supplement Table II shows the study-specific LOD score and meta-analysis LOD score for each SNP reaching a LOD >2.5 in the analysis adjusted for study. In general, the 2 analysis strategies yielded similar results and would lead to similar inferences in most cases. The AA-DHS–specific LOD scores tended to be higher than those observed in the other studies, except for rs9309717 and rs7933164, which had stronger LOD scores in MESA (AA-DHS had a larger sample size). The extent to which the difference in sample size explains differences in LOD scores is not clear. Ascertainment differences may also contribute.
Fine mapping was performed at 1p32.3 (highest LOD score 3.7), 9p21.2 with established involvement in cardiac disease,34–36 and 11p15.4 (lowest LOD score 2.5). Online-only Data Supplement Table Va to Vc summarizes the admixture mapping results in these regions using an additional set of AIMs typed around each sentinel SNP. The results of conditional analyses adjusted for the sentinel AIM are included. Evidence of support for the linkage peak observed and refinement of the interval on 1p32.3 was observed with four SNPs distal to the sentinel SNP (rs6663966) showing nominal evidence of association (P<0.040; online-only Data Supplement Table Va). Analysis of 19 additional AIMs on 9p21.2 also provided evidence of support for the linkage observed. Twelve SNPs located around the sentinel SNP (rs439314) showing evidence of association (P=0.034−4.5×10−4; online-only Data Supplement Table Vb). Finally, sporadic evidence of support was observed with markers around the 11p15.4 sentinel SNP (online-only Data Supplement Table Vc). In all 3 regions, conditional analyses accounting for the sentinel SNP revealed that each signal was primarily driven by the previously identified SNP with little evidence of additional AIMs contributing.
Analyses in Other Study Samples
Results in additional samples are summarized in online-only Data Supplement Table VIa and VIb, including associated Manhattan plots. These plots show supportive evidence in identified regions, indicating that the association is not likely an uncontrolled artifact. Analyses in the MESA sample with AAs without diabetes identified numerous nearby SNPs with P values ranging between 10−3 and 10−4 in each region. The chromosome 8 region had an SNP whose P value was 2.5×10−5 with additional supporting evidence around it. Results are shown in Figure 1A for Log (CAC + 1) and Figure 1B for presence of CAC.
We conducted association tests using available GWAS data on EAs in MESA and DHS. Figure 2A shows that the strongest results in MESA EAs were on chromosome 11 near the 11p15.4 peak, with several SNPs located near the sentinel marker with P values between 10−4 and 10−5. We also observed associations with Log (CAC + 1) near 13q32.1 with P values between 10−3 and 10−4 and between the 2 sentinel makers on chromosome 1 with P values ranging between 0.01 and 0.001. There was no evidence of association near the 9p21.2 region. Results were similar when presence of CAC was used as the outcome except for the chromosome 1 regions, which had P values between 10−2 and 10−3 (Figure 2B). As shown in Figure 3A and 3B, DHS provided minimal evidence of support for genomic regions from the initial analysis, with P values 10−2 to 10−3. One SNP near 2p25.3 had stronger evidence (P value between 10−3 and 10−4).
We note that these SNPs would not meet genome-wide significance in a GWAS setting under a strict Bonferroni correction threshold. However, considered in this context, they strongly support initial results.
Nearly all published reports demonstrate that AAs have lower amounts of CAC than EAs.1–4 CAC predicts risk of CVD death and myocardial infarction equally among members of all population groups.37,38 Furthermore, AAs manifest lower levels of CAC despite greater exposure to conventional CVD risk factors, supporting underlying inherited or biological mechanisms. The present analyses were directed at identifying genetic regions and variants governing susceptibility to development and progression of CAC in AAs with diabetes.
All AA participants with T2DM in the AA-DHS, MESA, and FamHS cohorts with CAC were included. Autosomal AIMs on the Illumina AA admixture panel were genotyped, and our RAM method applied to compute the probability of inheriting 0, 1, and 2 alleles from the ancestral European population. These probabilities were subsequently used to test for linkage with CAC, analyzed as a continuous (Log [CAC+1]) and a binary trait (cases had CAC >10). We noted the significant negative association between African ancestry and CAC, replicating the sentinel observation of Wassel et al.6 This reverses the longstanding concept that AAs are at higher CVD risk than EAs. In fact, AAs are at lower biological risk for myocardial infarction based on less CAC, and they have significantly lower rates of myocardial infarction than EAs, provided equal access to healthcare.39,40 Lack of healthcare access likely contributes to higher CVD rates in the general AA population. We extend these results by pinpointing specific genomic regions, where excess European ancestry appears to confer higher risk for initiation of CAC, progression, or both.
The RAM analysis suggested regions on chromosomes 1, 2, 4, 8, 11, and 13 may harbor genetic variants conferring higher level or odds of CAC. We also replicated the 9p region associated with CVD in several ethnic groups. Significant novel associations with CAC were observed on chromosomes 1, 4, and 11 with LOD scores between 3.0 and 3.7. Our effort to replicate these findings is limited by the fact that few studies outside of the AA-DHS, DHS, and MESA have large-scale genetic and CT data on AAs with T2DM. We extended our results in AAs without diabetes from MESA. A limitation of our validation approach is that, independently of the power of the test being conducted, the meaning of a negative result is unclear. Null results may indicate the initial finding is false-positive or an interaction exists with diabetes mellitus that cannot be captured adequately in samples without diabetes.
The chromosome 1p32.3 linkage peak was located in GLI-Similar family zinc finger 1 gene. This SNP appeared to be monomorphic in the Center d’Étude du Polymorphisme Humain population, and its minor allele frequency in Yorubans is 26%. The minor allele frequency in this sample is 42%. The low-density lipoprotein receptor-related protein 8 gene (LRP8) located ≈200 kb from this linkage peak and nearby genes remains of interest. LRP8 encodes an apolipoprotein E receptor, a member of the low-density lipoprotein receptor family. The apolipoprotein E receptor is involved in cellular recognition and internalization of lipoproteins. LRP8 is associated with coronary artery disease in European-derived populations. Admixture mapping peaks are generally not as well defined as GWAS signals. Therefore, the reported signals may in fact be tracking the effect of other nearby genes. The second chromosome 1 LOD peak is located at 1q32.1 near the chitotriosidase (CHIT1) gene. The allele frequency at the tested marker is 97% and 33% in the Center d’Étude du Polymorphisme Humain and Yoruban HapMap 3 samples, respectively, relative to 43% in the study sample. A positive correlation between chitotriosidase and atherosclerosis has been reported in European-derived samples,41,42 and serum chitotriosidase activity may predict CVD events.43
Two regions on chromosome 4 had LOD scores >2.7. The first identified with Log (CAC + 1) with a LOD of 3.0 is 4p21.2 near the PRKG2 gene, a region including genes involved in bone metabolism. Relationships between bone disease and arterial calcification are well established; numerous studies report inverse relationships between bone mineral density and atherosclerosis in EAs and AAs.44–46 The 9p21 region was previously associated with coronary artery disease susceptibility in predominantly European-derived populations.36,47,48 AAs with excess European ancestry in this region had higher prevalence and higher levels of CAC. The chromosome 11 and 13 peaks appear to be novel with regard to this phenotype; further replication and validation efforts are needed.
Admixture mapping has limitations. These AIMs are not 100% informative, and the European ancestry at each marker is probabilistically determined. In addition, the power of RAM is maximized when the average ancestry proportion from the 2 ancestral populations is close to (0.50, 0.50); these proportions were estimated to be approximately (0.80, 0.20) in this sample. Nonetheless, admixture mapping remains a valuable and proven gene-mapping technique. In fact, combining admixture mapping and association tests can lead to a non-negligible gain in power.11
In addition to fine mapping the genome-wide significant linkage peaks, attention should be paid to underlying candidate genes involved in vitamin D metabolism, calcium handling, and bone health. We demonstrated inverse relationships between CAC and bone mineral density in the AA-DHS.49 Thus, susceptibility to osteoporosis and development of CAC appear to be linked in EAs and AAs. CYP2R1, an enzyme important in activation of vitamin D, is located under the chromosome 11 MALD peak. BMP3 and PRKG2, involved in multiple aspects of cartilage and bone development, are located under the chromosome 4 MALD peak. Additional genes under these MALD peaks with involvement in calcium, bone, and arterial metabolism include chromosomes 1 (ADORA1, BTG2, FMOD, and PRELP) and 8 (RNF19A). It is likely that population-specific susceptibility to CAC reflects, in part, the widely appreciated racial differences in calcium and vitamin D metabolism.50
In conclusion, genomic regions on chromosomes 1, 2, 4, 8, 9, 11, and 13 identified by admixture mapping appear to contribute to ethnic differences in susceptibility to CAC between AAs and EAs. Fine mapping under these peaks is likely to detect causative genes, potentially leading to improved understanding of the biological causes of this phenomenon and development of novel antiatherosclerotic therapies.
The investigators acknowledge the cooperation of the DHS and AA-DHS participants and study recruiter Cassandra Bethea.
Sources of Funding
This study was supported, in part, by the General Clinical Research Center of the Wake Forest University School of Medicine grant M01 RR07122; National Institute of Diabetes and Digestive and Kidney Diseases grant RO1 DK071891 (Dr Freedman); and National Heart, Lung, and Blood Institute grant R01 HL67348 (Dr Bowden). MESA and the MESA SHARe project were conducted and supported by the National Heart, Lung, and Blood Institute in collaboration with MESA investigators. Support was provided by grants and contracts N01 HC-95159, N01-HC-95160, N01-HC-95161, N01-HC-95162, N01-HC-95163, N01-HC-95164, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168, N01-HC-95169, and RR-024156. Funding for SHARe genotyping was provided by National Heart, Lung, and Blood Institute Contract N02-HL-6-4278. Genotyping was performed at the Broad Institute of Harvard and Massachusetts Institute of Technology (Boston, MA) and at Affymetrix (Santa Clara, CA) using the Affymetrix Genome-Wide Human SNP Array 6.0. FamHS was supported by contracts NO1-HC-25104, NO1-HC-25105, NO1-HC-25106, NO1-HC-25107, NO1-HC-25108, NO1-HC-25109, and 5R01HL088215 (Dr Province) from the National Heart, Lung, and Blood Institute. This manuscript has been reviewed by FamHS investigators for scientific content and consistency of data interpretation with previous FamHS publications, and significant comments have been incorporated before submission for publication.
Guest Editor for this article was Christopher H. Newton-Cheh, MD, MPH.
The online-only Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.111.964114/-/DC1.
- Received February 28, 2012.
- Accepted November 9, 2012.
- © 2013 American Heart Association, Inc.
- Bild DE,
- Detrano R,
- Peterson D,
- Guerci A,
- Liu K,
- Shahar E,
- et al
- Wassel CL,
- Pankow JS,
- Peralta CA,
- Choudhry S,
- Seldin MF,
- Arnett DK
- Bowden DW,
- Cox AJ,
- Freedman BI,
- Hugenschimdt CE,
- Wagenknecht LE,
- Herrington D,
- et al
- Bild DE,
- Bluemke DA,
- Burke GL,
- Detrano R,
- Diez Roux AV,
- Folsom AR,
- et al
- Carr JJ,
- Nelson JC,
- Wong ND,
- McNitt-Gray M,
- Arad Y,
- Jacobs DR Jr.,
- et al
- Reich D,
- Patterson N
- Barrett JC,
- Fry B,
- Maller J,
- Daly MJ
- Chen MH,
- Yang Q
- Kronmal RA,
- McClelland RL,
- Detrano R,
- Shea S,
- Lima JA,
- Cushman M,
- et al
- Samani NJ,
- Raitakari OT,
- Sipilä K,
- Tobin MD,
- Schunkert H,
- Juonala M,
- et al
- Greenland P,
- Bonow RO,
- Brundage BH,
- Budoff MJ,
- Eisenberg MJ,
- Grundy SM,
- et al
- Carr JJ
- Young BA,
- Maynard C,
- Boyko EJ
- Boot RG,
- van Achterberg TA,
- van Aken BE,
- Renkema GH,
- Jacobs MJ,
- Aerts JM,
- et al
- Artieda M,
- Cenarro A,
- Gañán A,
- Jericó I,
- Gonzalvo C,
- Casado JM,
- et al
African Americans have markedly lower levels of coronary artery calcified plaque (CAC) than European Americans, despite the presence of more severe conventional cardiovascular disease risk factors. These findings suggest that inherited or biological factors underlie susceptibility to CAC, a measure of subclinical coronary artery disease. In an effort to detect genes underlying susceptibility to CAC, we performed regional admixture mapping in 1040 unrelated African Americans with type 2 diabetes mellitus who had computed tomography-derived measures of CAC. Regional admixture mapping is a gene-mapping technique useful for detecting disease genes in admixed populations. Regional admixture mapping is particularly powerful when the disease prevalence or distribution of the trait of interest varies substantially between the ancestral populations. Regional admixture mapping revealed 11 genomic regions located on chromosomes 1, 2, 4, 8, 9, 11, and 13 that were significantly or suggestively linked with the presence or severity of CAC in African Americans. Strikingly, all 11 regions displayed higher than expected European ancestry in individuals whose CAC score was 10 or higher. These results underscore that susceptibility to coronary artery disease in the admixed African American population relates to European ancestry. Fine mapping under these linkage peaks is likely to detect the genes that regulate the related processes of atherosclerotic coronary artery disease and vascular calcification.