Gene expression variation between African Americans and whites is associated with coronary artery calcification: the multiethnic study of atherosclerosis

Abstract

Coronary artery calcium (CAC) is a strong indicator of total atherosclerosis burden. Epidemiological data have shown substantial differences in CAC prevalence and severity between African Americans and whites. However, little is known about the molecular mechanisms underlying initiation and progression of CAC. Microarray gene expression profiling of peripheral blood leucocytes was performed from 119 healthy women aged 50 yr or above in the Multi-Ethnic Study of Atherosclerosis cohort; 48 women had CAC score >100 and carotid intima-media thickness (IMT) >1 mm, while 71 had CAC <10 and IMT <0.65 mm. When 17 African Americans were compared with 41 whites in the low-CAC group, 409 differentially expressed genes (false discovery rate <5%)were identified. In addition, 316 differentially expressed genes were identified between the high- and low-CAC groups. A substantial overlap between these two gene lists was observed (148 genes, P < 10⁻⁶). Furthermore, genes expressed lower in African Americans also tend to express lower in individuals with low CAC (correlation 0.69, P = 0.002). Ontology analysis of the 409 race-associated genes revealed significant enrichment in mobilization of calcium and immune/inflammatory response (P < 10⁻⁹). Of note, 25 of 30 calcium mobilization genes were involved in immune/inflammatory response (P < 10⁻¹⁰). Our data suggest a connection between immune response and vascular calcification and the result provides a potential mechanistic explanation for the lower prevalence and severity of CAC in African Americans compared with whites.

coronary artery calcium (CAC) is highly correlated with total atherosclerosis burden (33). Epidemiological studies have shown that CAC can improve prediction of cardiovascular disease (CVD) events independently of traditional coronary risk factors among various race/ethnic groups (8, 13, 17, 31). Several studies have shown racial/ethnic differences in CAC prevalence and severity that cannot be fully explained by conventional coronary risk factors (3, 10, 24, 28). In all studies, African Americans have a significantly lower prevalence of CAC and average CAC score than whites, despite a higher prevalence of hypertension and diabetes. Nevertheless, the mechanisms accounting for the differential CAC burden in these two populations are poorly understood. These observations suggest that underlying biological processes and genetic predisposition are likely to play a role in the modulation of CAC. Identifying the mechanisms underlying differential manifestations of CAC between these two populations could potentially identify novel targets for prevention of CVD.

Recent studies have demonstrated that vascular calcification is an active biological process similar to bone formation (7). Several signaling pathways, including RANKL and Wnt signaling, have been shown to be crucial in vascular calcification and bone-related diseases. Because these pathways also play an important role in the immune response (14, 25, 35), these findings suggest a potential mechanistic link between atherogenesis, which is promoted by immune and inflammatory response, and vascular calcification. Whole blood consists of several immune cells such as B cells, T cells, and monocytes that are in frequent contact with endothelial cells, transit into the subendothelial vascular wall, and also participate in osteogenesis. Therefore, whole blood may represent a useful surrogate tissue to investigate the pathophysiology of vascular calcification. Studying gene expression patterns of peripheral blood may reveal novel insights into the development of CAC.

In the present study, we performed gene expression profiling of whole blood in a selected sample of women without diabetes or history of clinical cardiovascular disease from the Multi-Ethnic Study of Atherosclerosis (MESA). With its well-documented phenotype data and multiethnic populations, MESA provides a unique opportunity to compare gene expression patterns between African Americans and whites and investigate whether differentially expressed genes between African Americans and whites and genetic pathways identified therein are associated with CAC.

METHODS AND MATERIALS

Participants.

Participants in the present study were selected from the MESA cohort. The MESA was designed to study the prevalence, risk factors, and progression of subclinical cardiovascular disease in a multiethnic cohort. A detailed description of the study design and methods has been published previously (2). In brief, 6,814 participants aged 45–84 yr and free of known clinical CVD were recruited from six US communities (Baltimore, MD; Chicago, IL; Forsyth County, NC; Los Angeles, CA; New York, NY; and St. Paul, MN). The baseline examination was performed between July 2000 and September 2002. Participants included white (38%), African American (28%), Hispanic (22%), and Chinese (12%) individuals. MESA conducted three subsequent examinations of the cohort between 2002 and 2007, and the fifth examination is ongoing from 2010.

Participants in the present study included 67 white and 30 African American women without diabetes or a history of CVD; 26 whites and 13 African Americans had a CAC score ≥ 100 and common carotid intima-media thickness (CC-IMT) ≥ 1.0 mm, while 41 whites and 17 African Americans had a CAC score ≤ 10 and CC-IMT ≤ 0.65 mm. Additionally, 16 Hispanic and 6 Chinese women were included. Three Chinese and six Hispanic women were among the high-CAC and high-IMT group. These 16 women were only included in analyses for gene expression comparison between high-CAC and low-CAC groups. The majority of the 119 women (108) had a low Framingham risk score (i.e., FRS < 10%). Men were not included in the current study because very few MESA male participants had low FRS and high CAC. Carotid IMT was measured for all MESA participants only at the baseline examination. Measurement of CAC was first performed for all MESA participants at the baseline examination. A second CAC measurement was performed during the follow-up: 50% of all participants at the second MESA examination (2002–2003), and the remaining 50% of participants at the third MESA examination (2003–2004). We used the most recent CAC score available for each of these 119 participants. These participants were called back during April-July 2007 to have 5 ml whole blood drawn into two PAXgene tubes for RNA isolation. The Institutional Review Boards at each site approved the study, and all participants gave informed consent.

CAC and IMT assessment.

Scanning centers assessed CAC by chest-computed tomography using either a cardiac-gated electron-beam computed tomography scanner at three field centers or a multidetector computed tomography system at the other three field centers. Each participant was scanned twice, and these scans were read independently at a central reading center, as described previously (5). Carotid IMT was measured using high-resolution B-mode ultrasonography of the right and left near and far walls of the internal carotid and common carotid arteries (29). Given the large number of missing internal carotid IMT data, we elected to use CC-IMT as the selection criterion.

RNA extraction and microarray experiment.

Five milliliters of whole blood from each participant were drawn into two PAXgene tubes and incubated at room temperature for 3 h before being frozen at −70°C. RNAs were extracted using the PAXgene blood RNA extraction kit according to the manufacturer's protocol. The concentration of the extracted total RNA and their quality were measured by NanoDrop (Thermofisher) and Bioanalzyer 2100 (Agilent), respectively. Total RNA preparations with 260/280 ratio between 1.98–2.22 and RNA integrity number > 7.4 with sufficient quantity were used for the mRNA expression analysis. Globin reduction was performed using the Ambion GLOBINclear kit. The quality of the globin-reduced RNA samples was assessed by the Bioanalyzer 2100. High-quality samples were used to make first- and second-strand DNA followed by an IVT reaction. The size distribution of the resulting biotin-labeled cRNA, and the yield was checked by Agilent 2100 and NanoDrop, respectively. A normalized amount of labeled cRNA was hybridized to the Human Ref-8 bead chips for 18 h at 55°C. After being washed and stained with Cy3, the chips were scanned on the Illumina iScan. The Ref-8 BeadChip allows genome-wide expression profiling of >22,000 gene transcripts and known alternative splice variants from the RefSeq database.

The microarray analysis was performed at the Northwestern Genomic Core Facility at the Center for Genetic Medicine. All the 119 RNA samples were run on 5 different days within 1 wk. Samples collected from six MESA field centers were randomly assigned to different chips and days. All samples with low and high CAC in each MESA field center were also randomly assigned to different chips and days. Correlations of gene expression between two replicates run in different days were >0.99. Cluster analysis of normalized gene expression data revealed no significant chip-to-chip or day-to-day variation.

Quantitative RT-PCR.

To validate the expression levels in microarrays, total RNA samples from the low-CAC white and low-CAC African American women were analyzed for mRNA expression by quantitative real-time RT-PCR (QRT-PCR) for GAB2, NOTCH1, MMP9, QPCT, TNFSF14, TLR8, and IL1RN genes. Reverse transcription was performed by the SuperScript III VILO cDNA synthesis kit (Invitrogen). Real-time PCR was performed by the Taqman gene expression assay system (Applied Biosystems, Emeryville, CA). The primer and probe mixes were purchased from Applied Biosystems referenced to the probe location for each gene applied in the microarray assay. GAPDH was used as the endogenous control. The relative gene expression was determined by the ΔΔ-C_T method. All experiments were performed in duplicate.

Data analysis.

BeadStudio software was used to translate the scanned images into expression data, which were further log-transformed and normalized by the quantile normalization procedure using the Bioconductor package affy. All data analyses were based on 2,057 probes representing 1,967 genes that met two criteria: 1) coefficient of variation > 0.03 across all samples, and 2) P value for expression detection call < 0.01 for 60 samples or more. Significance analysis of microarrays (40) (SAM) was used to identify differentially expressed genes for a two-group comparison. Standardized fold changes were calculated as mean expression differences divided by pooled standard deviation in the log scale. To account for potential confounding factors, body mass index (BMI), triglyceride, and insulin resistance measure using the homeostatic model assessment (HOMA) were included in the regression models, and FDR for differentially expressed genes (16) were calculated. To account for European admixture in African Americans, we further adjusted for the first principle component derived from the MESA genome-wide association data among African Americans that are available from the National Center for Biotechnology Information (NCBI) dbGaP database (26). The MESA SHARe project (study accession phs000209, http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000209.v5.p1) includes 8,227 individuals (2,686 Caucasians, 777 Chinese, 2,590 non-Hispanic African-Americans, and 2,174 Hispanics). All of them were genotyped using the Affymetrix Human SNP Array 6.0. We excluded monomorphic single nucleotide polymorphisms (SNPs); SNPs with missing rate > 5% or observed heterozygosity > 53%; SNPs that were not in Hardy-Weinberg equilibrium (P < 10⁻⁵); as well as SNP pairs in high linkage disequilibrium (LD) (r² ≥ 0.8). We further removed 6,849 SNPs in genomic regions that have been shown to harbor long-range LD. This left a final sample of 718,707 autosomal SNPs from which we performed Principal Component Analysis as implemented in the program SMARTPCA (30, 32) from the software package EIGENSTRAT to compute principal components of ancestry. Ingenuity Pathway Analysis (IPA, http://www.ingenuity.com) and PANTHER ontology analysis (39) were used to identify enriched pathways or biological functions among differentially expressed genes. All microarray data were submitted to the NCBI Gene Expression Omnibus data repository under accession number GSE20129.

RESULTS

Characteristics of the study sample.

Characteristics of the 97 white and African American participants in the study sample are shown in Table 1 stratified by CAC status and race. Among all CVD risk factors, triglyceride levels were significantly different between white and African American women in both high-CAC and low-CAC groups, while BMI was significantly different between white and African American women only in the low-CAC group. Among the low-CAC group, lipid-lowering medication use was substantially higher among whites, while antihypertension medication use was higher among African Americans, although neither of these differences was statistically significant. No significant differences were observed for blood pressure, total, LDL, or HDL cholesterol, smoking status, fasting glucose, insulin, and insulin resistance measure of HOMA between these two ethnic groups in both high-CAC and low-CAC groups.

Table 1.

Characteristics of 97 MESA nondiabetic women by race and by coronary artery calcification status

	CAC <10		CAC >100
Variable	African American	White	African American	White
n	17	41	13	26
Age, yr	64.1 ± 7.7	62.9 ± 6.8	68.3 ± 8.5	69.0 ± 6.5
Body mass index, kg/m2†	32.0 ± 6.4	25.5 ± 4.2	30.9 ± 7.1	27.4 ± 4.6
Systolic BP, mmHg	119.1 ± 17.3	114.4 ± 18.7	130.6 ± 18.2	130.0 ± 22.6
Diastolic BP, mmHg	68.6 ± 7.8	65.9 ± 11.1	71.4 ± 9.5	67.1 ± 8.7
Total cholesterol, mg/dl	191.8 ± 32.4	197.5 ± 31.0	199.7 ± 29.7	204.0 ± 37.0
LDL cholesterol, mg/dl	113.5 ± 27.7	110.2 ± 27.5	122.6 ± 30.4	119.0 ± 39.1
HDL cholesterol, mg/dl	61.6 ± 13.1	65.0 ± 15.6	61.3 ± 13.5	59.1 ± 12.6
Triglycerides, mg/dl†§	83.7 ± 32.8	111.3 ± 67.5	78.6 ± 29.8	129.2 ± 72.3
Fasting glucose, mg/dl	91.5 ± 8.0	89.7 ± 8.5	90.6 ± 7.5	91.0 ± 8.7
HOMA insulin resistance	0.92 ± 0.45	0.74 ± 0.39	1.04 ± 0.55	1.06 ± 0.70
Insulin, uIU/ml	4.4 ± 2.0	3.6 ± 1.8	4.8 ± 2.4	4.9 ± 3.1
Current smoker, %	23.5	37.5	53.8	48.0
Lipid-lowering medication use, %	5.8	15.3	38.4	52.0
Antihypertension medication use, %	47.0	25.6	53.8	48.0

MESA, Multi-Ethnic Study of Atherosclerosis.

^†P < 0.05 comparing African American with white among women with coronary artery calcification (CAC) <10.

^§P < 0.05 comparing African American with white among women with CAC >100. Note: P value by t-test or χ² test as appropriate.

Identifying differentially expressed genes between African Americans and whites.

The gene expression pattern was distinct between 41 whites and 17 African Americans in the low-CAC group. SAM identified 431 probes representing 409 genes [false discovery rate (FDR) < 5%] differentially expressed between these two groups (Supplemental Fig. S1A and Supplemental Table S1).¹ In contrast, only 30 differentially expressed genes were identified at the FDR < 5% level between the 26 whites and 13 African Americans in the high-CAC group (Supplemental Fig. S1B). Because BMI and triglyceride levels are significantly different between African American and white women in our data, we fitted a linear regression model adjusted for BMI and triglyceride for each of the 2,057 probes from the low-CAC group comparison. Only 86 probes were associated with BMI and 37 probes associated with triglyceride with nominal P values < 0.05 (FDR 51 and 99%, respectively). Race was associated with 484 probes (nominal P value < 0.05, FDR 14%), with a majority of these being (409) race-associated genes (333, 81.4%). Regression analysis with additional adjustment for insulin resistance or adjusted only for BMI showed a similar result. When the same regression analysis was performed with additional adjustment for African ancestry in African American women, 28 probes were associated with BMI and 191 probes associated with triglyceride with nominal P values < 0.05 (FDR 99 and 36%, respectively). Race was associated with 524 probes (nominal P value <0.05, FDR 11%), among which the majority (342, 84%) were race-associated genes. These results suggest that the association between expression levels of race-associated genes and race are not explained by BMI and triglyceride.

We further compared our 409 race-associated genes with those reported by Schisler et al. (34), who examined the expression profiles of whole blood from 17 African American and 30 Caucasian men and women without coronary artery disease. Schisler et al. reported 151 geo-ancestral genes differentially expressed between these two populations. These geo-ancestral genes were defined as containing at least one SNP within 10 kb of the untranslated regions that has significant difference in allele frequency between the two HapMap populations, the Yoruba of Ibadan, Nigeria (YRI), and the CEPH population (Utah residents with ancestry in northern and western Europe). Only 56 of Schisler's geo-ancestral genes were found among the 1,967 genes in our study because the microarray platform, gene filtering processes, and criteria for selecting differentially expressed genes used in Schisler's study differed from those in our study. Nevertheless, a striking overlap (36 genes, P < 10⁻¹⁶) was observed between the 56 genes and our 409 race-associated genes. Specifically, 4 of 11 genes expressed at a higher level in African Americans and 32 of the 45 genes expressed at a lower level in African Americans in Schisler's list appeared in our gene list (see Table 2). Of note, all of the 36 genes had concordant expression directions between Schisler's and our study. More importantly, among the 32 genes expressed at a lower level in American Americans, 9 of the 10 genes with the largest fold changes in Schisler's list (S100P, MMP9, C20orf3, QPCT, KRT23, USP10, NOV, HK2, and PGD) appeared in our race-associated gene list. The significant overlap between our and Schisler's gene lists suggests a high true positive rate of the identified race-associated genes. Several studies have linked genetic variations to gene expression among various HapMap populations and other Caucasian populations. Comparing the compilation of those studies available in the University of Chicago expression quantitative trait locus (eQTL) database (http://eqtl.uchicago.edu/cgi-bin/gbrowse/eqtl/) identified 18 genes (50%) with at least one eQTL (or exon-QTL) (see Table 2), suggesting that genetic variation is attributable to the expression differences between African Americans and whites.

Table 2.

36 common race-associated genes found in both MESA and Schisler's studies

Gene Symbol	Gene Name	Entrez Gene ID	Fold Change	Fold Change (Schisler's study)	With eQTL or exon-QTL
Genes expressed lower in African American
S100P	S100 calcium binding protein P	6286	2.8	2.79	Y
MMP9	matrix metallopeptidase 9	4318	2.18	1.98
C20orf3	chromosome 20 open reading frame 3	57136	1.21	1.87	Y
QPCT	glutaminyl-peptide cyclotransferase	25797	1.3	1.68	Y
KRT23	keratin 23	25984	1.19	1.51	Y
USP10	ubiquitin specific peptidase 10	9100	1.25	1.5
NOV	nephroblastoma overexpressed gene	4856	1.43	1.5
HK2	hexokinase 2	3099	1.17	1.49
PGD	phosphogluconate dehydrogenase	5226	1.28	1.49	Y
SULF2	sulfatase 2	55959	1.27	1.48	Y
LAMP2	lysosomal-associated membrane protein 2	3920	1.22	1.46
GPR97	G protein-coupled receptor 97	222487	1.19	1.42
KIAA0319L	KIAA0319-like	79932	1.2	1.42	Y
NADK	NAD kinase	65220	1.25	1.41
MME	membrane metallo-endopeptidase	4311	1.34	1.41
HEBP2	heme binding protein 2	23593	1.37	1.39	Y
TMEM45B	transmembrane protein 45B	120224	1.21	1.39
PLAUR	plasminogen activator, urokinase receptor	5329	1.31	1.36	Y
ACOX1	acyl-Coenzyme A oxidase 1, palmitoyl	51	1.25	1.36	Y
STX3	syntaxin 3	6809	1.21	1.35
RNF135	ring finger protein 135	84282	1.26	1.35
SRPK1	SFRS protein kinase 1	6732	1.16	1.35
UBN1	ubinuclein 1	29855	1.3	1.35	Y
FLOT1	flotillin 1	10211	1.2	1.34	Y
TMEM55A	transmembrane protein 55A	55529	1.24	1.34
REPS2	RALBP1 associated Eps domain containing 2	9185	1.24	1.33
ANPEP	alanyl (membrane) aminopeptidase	290	1.31	1.33	Y
PYGL	phosphorylase, glycogen, liver	5836	1.25	1.33
GAB2	GRB2-associated binding protein 2	9846	1.34	1.33	Y
ST6GALNAC2	ST6 (alpha-N-acetyl-neuraminyl-2,3-beta-galactosyl-1,3)-N-acetylgalactosaminide alpha-2,6-sialyltransferase 2	10610	1.31	1.32	Y
CDA	cytidine deaminase	978	1.5	1.31	Y
AGPAT9	1-acylglycerol-3-phosphate O-acyltransferase 9	84803	1.36	1.31	Y
Genes expressed higher in African American
CRIP1	cysteine-rich protein 1 (intestinal)	1396	1.27	1.3
NKG7	natural killer cell group 7 sequence	4818	1.33	1.36	Y
GZMH	granzyme H (cathepsin G-like 2, protein h-CCPX)	2999	1.67	1.38
IGJ	immunoglobulin J polypeptide, linker protein for immunoglobulin alpha and mu polypeptides	3512	1.99	1.46

eQTL, expression quantitative trait locus.

Comparison with the CAC-associated gene expression pattern.

A high correlation of global expression changes was observed between the CAC and race-associated gene expression patterns. We calculated the standardized fold changes of all 2,057 probes between 71 low-CAC and 48 high-CAC participants. Interestingly, a high correlation (Pearson correlation 0.57, P = 0.009, permutation test) of fold changes was observed between the two low-CAC race group (i.e., low-CAC white vs. low-CAC African American) comparison and the two CAC group (i.e., all low-CAC vs. all high-CAC) comparison (Fig. 1). Among the 409 race-associated genes, the correlation was 0.69 (P = 0.002, permutation test) between these two comparisons and 88% of these genes had concordant fold changes. This correlation (0.39) was attenuated when the two high-CAC race groups were used for comparison. SAM identified 337 differentially expressed probes representing 317 genes (FDR < 22%) between the high- and low-CAC groups. A highly significant overlap of these two gene lists was observed (148 genes, P < 10⁻⁶). More than half of the race-associated genes (145/284 genes) expressed at a lower level in African American women were found among the CAC associated genes (P < 10⁻¹⁵). The result was similar when 16 Hispanic and 6 Chinese women were excluded from the analysis. Furthermore, we fit a linear regression model with CAC, race, and the CAC-race interaction term for each of the 2,057 probes from the 97 African American and white women. A total of 110 of the 409 race-associated genes were associated with CAC (nominal P value < 0.05, FDR < 4%). In contrast, only 98 of the remaining 1,626 probes were associated with CAC. Comparing the regression coefficients of race and CAC, a higher correlation (0.79) and a higher concordance (94%) of coefficient signs were observed for the 409 race-associated genes. Taken together, these results suggest that some race-associated genes may contribute to the development of high CAC (P = 0.0057). The results also suggest that genes expressed at a lower (or higher) level in African American women tend to be expressed at a lower (or higher) level in those with low CAC.

Fig. 1.

Correlation of standardized fold changes (log-scale) of 1,967 genes between the low-coronary artery calcium (CAC) race comparison (x-axis) and high-/low-CAC comparison (y-axis). A high correlation (0.57, P = 0.009) of standardized fold changes was observed between these 2 group comparisons. Among the 409 race-associated genes (×), the correlation was 0.69 (P = 0.002).

Ontology enrichment analysis of race-associated genes between two low-CAC groups.

Among the 409 race-associated genes, IPA identified several enriched functional categories related to immune and inflammatory response and apoptosis. Simultaneous survey and evaluation of these functional categories enabled us to identify molecular mechanisms important to the process of vascular calcification. Examples of the diversity in these mechanisms include 1) increased expression (in whites) of members of Toll/IL-1R signaling (TLR1, TLR8, IL1RN, IL1R2), a key pathway critical for the activation of innate immunity and development of adaptive immunity; 2) increased expression of ATP-binding cassette transporter family genes (ABCG1, ABCC5) that control cholesterol efflux; 3) increased expression of matrix metalloproteinases and their endogenous inhibitors (MMP9, MMP25, and TIMP2), key regulators of bone formation and extracellular matrix degradation; 4) increased expression of FRAT1 and LRP10, critical for regulation of the Wnt/β-catenin signaling pathway; and 5) increased expression of cytosolic regulatory components (NCF1, NCF4) of NADPH oxidase pathway, a major source of generating reactive oxygen species in the artery. Of note, mobilization of calcium was one of the most enriched functional categories (30 genes, P = 8·10⁻¹⁰). Systemic examination of these 30 genes revealed an involvement of the majority (25/30) of these genes in immune and inflammatory responses (see Table 3 and Supplemental Fig. S2). For example, IPA identified an inflammatory and immune response network that included seven calcium-related genes (P = 10⁻⁷, Supplemental Fig. S3). A literature search for the functions of these 409 genes further identified several additional genes crucial to bone metabolism or bone-related diseases, including TNFSF14 (12), NOTCH1 (15), MMP9 (41), and QPCT (20). Consistent with the IPA results, PANTHER ontology analysis also revealed significant enrichment of B-cell and T-cell-mediated immunity and defense, inflammation mediated by chemokine and cytokine signaling pathway, and apoptosis. Taken together, these observations suggest that many of the race-associated genes involved in multiple biological processes synergistically contribute to the development of CAC.

Table 3.

30 race-associated genes involved in mobilization of calcium

Gene Symbol	Gene Name	Entrez Gene ID	Fold Change*	Involved in Immune Response	Involved in Inflammatory Response
ANXA1	annexin A1	301	−1.39	X	X
C5AR1	complement component 5a receptor 1	728	1.26	X	X
CAMP	cathelicidin antimicrobial peptide	820	1.5	X	X
CCL5	chemokine (C-C motif) ligand 5	6352	−1.3	X	X
CCR3	chemokine (C-C motif) receptor 3	1232	1.36		X
CD2	CD2 molecule	914	−1.25	X
CX3CR1	chemokine (C-X3-C motif) receptor 1	1524	−1.2	X	X
CXCL1	chemokine (C-X-C motif) ligand 1 (melanoma growth stimulating activity, alpha)	2919	1.26	X	X
F2R	coagulation factor II (thrombin) receptor	2149	−1.21	X	X
F2RL1	coagulation factor II (thrombin) receptor-like 1	2150	1.33	X	X
FCRL3	Fc receptor-like 3	115352	−1.31
FPR2	formyl peptide receptor 2	2358	1.38	X	X
GAB2	GRB2-associated binding protein 2	9846	1.34
GNAQ	guanine nucleotide binding protein (G protein), q polypeptide	2776	1.18
HEBP1	heme binding protein 1	50865	1.24		X
IL8	interleukin 8	3576	1.36	X	X
CXCR1	chemokine (C-X-C motif) receptor 1	3577	1.21		X
CXCR2	chemokine (C-X-C motif) receptor 2	3579	1.21	X	X
ITGB1	integrin, beta 1 (fibronectin receptor, beta polypeptide, antigen CD29 includes MDF2, MSK12)	3688	−1.32		X
ITK	IL2-inducible T-cell kinase	3702	−1.24	X
KLRD1	killer cell lectin-like receptor subfamily D, member 1	3824	−1.52	X
KLRF1	killer cell lectin-like receptor subfamily F, member 1	51348	−1.27	X
KLRK1	killer cell lectin-like receptor subfamily K, member 1	22914	−1.35	X
LAT2	linker for activation of T cells family, member 2	7462	1.24	X
LPAR2	lysophosphatidic acid receptor 2	9170	1.2
LTB4R	leukotriene B4 receptor	1241	1.21	X	X
PILRA	paired immunoglobin-like type 2 receptor alpha	29992	1.29
PLAUR	plasminogen activator, urokinase receptor	5329	1.31		X
PTAFR	platelet-activating factor receptor	5724	1.2	X	X
PTGDR	prostaglandin D2 receptor (DP)	5729	−1.22		X

^*Compare white with African American.

Validation of gene expression data.

Quantitative real-time RT-PCR was performed on the low-CAC African Americans and whites to confirm the microarray expression levels of seven selected genes with potential roles in RANK pathway, lipid metabolism, immune and inflammatory response, and vascular calcification (GAB2, TNFSF14, TLR8, IL1RN, MMP9, NOTCH1, QPCT). We confirmed the expression results for all these seven genes except NOTCH1 (P ≤ 0.05) (Fig. 2).

Fig. 2.

QRT-PCR validation of 7 race-associated genes identified from the microarray experiment. Bars represent fold changes comparing expression between white and African American women in the low-CAC group. A fold change >1 represents upregulation of gene expression in white women. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001.

DISCUSSION

African Americans are less likely to develop CAC and have lower CAC scores on average than whites. However, the molecular mechanism responsible for this difference is largely unknown. In the present study, we compared gene expression profiles of whole blood from African American and white women. We found a large number of differentially expressed genes between these two ethnic groups among those with low CAC scores. Interestingly, the expression differences of these genes significantly correlated with the expression differences between groups with negligible or no CAC and high CAC. Investigation of the functional modules of the race-associated genes revealed enrichment and substantial overlap of two major functional axes: calcium mobilization and immune response. The finding of these molecular modules is consistent with that by Storey et al. (36), who identified between the HapMap CEPH and YRI samples a large number of differentially expressed genes that were strongly enriched in inflammatory and immune-related pathways. Studies have shown that RANKL and Wnt signaling pathways are crucial in vascular calcification and bone-related diseases and play an important role in the immune response (14, 25, 35). Thus, these pathways provide a potential mechanistic link between immune and inflammatory response and vascular calcification. Our result not only supports the emerging concept of cross talk between immunity and osteogenic pathways but also suggests that a large proportion of the race-associated genes identified in the present study may be associated with the development of CAC. Although other pathways could possibly explain the difference in CAC prevalence and severity between African Americans and whites, we found no major evidence for this in our gene expression data. Taken together, these results provide a potential and compelling mechanistic explanation for the greater CAC prevalence and severity among whites compared with African Americans, seen even after adjustment for traditional CVD risk factor levels. Our study benefited from using a sample of participants from a well-phenotyped cohort study specifically designed to explore differences in subclinical CVD between race/ethnic groups, thus minimizing the effect of potential confounding factors on gene expression patterns associated with CAC.

Vascular calcification, once regarded as a passive degenerative disease, involves a complex mineralization process similar to bone formation. Remodeling of bone involves synthesis of bone matrix by osteoblasts and coordinates bone resorption by osteoclasts. Recent efforts to improve our knowledge of this disease have led to several findings and provided novel research directions.

An emerging area is the study of key regulators of the immune and bone systems including receptor activator of NF-κB (RANK), its ligand (RANKL), and osteoprotegerin (OPG). However, the molecular mechanism and coupling molecules that control RANKL signaling have not been fully characterized. Our study has identified several novel genes that might act as coupling molecules between the immune and bone systems. For example, we found a novel gene, GAB2, that was expressed at a lower level in African Americans and those with low CAC scores. The gene, GAB2, was recently shown to be a key regulatory scaffold molecule that controls select RANK signaling pathways and to have a crucial role in the differentiation of human progenitor cells into osteoclasts (42). We also identified TNFSF14, a member of the TNF superfamily that promotes a RANKL-mediated osteoclastogenesis and can induce osteoclast formation independent of RANKL (12, 21).

Another area of active research area is the Wnt3a/β-catenin and LDLR-related protein (LRP5)-dependent activation of the canonical Wnt signaling cascade in calcifying human aortic valves (4). We also observed FRAT1 and LRP10, two molecules essential for the regulation of Wnt3a-induced Wnt signaling and canonical Wnt/β-catenin signaling pathways (18, 22). Further investigation of these genes may deepen our understanding of the complex mechanism of vascular calcification regulated by these signaling pathways.

Gene expression is influenced by genetic variation (6, 37, 38). Storey et al. (36) analyzed gene expression profiles of B lymphoblastoid cells from the CEPH and YRI samples used in the International HapMap project. Among the genes differentially expressed between these two populations, they identified molecular functions and signaling pathways including inflammation mediated by chemokine and cytokine, T-cell and B-cell activation, VEGF signaling, and Toll-like receptor signaling. Their gene ontology enrichment analysis results are similar to ours even though they analyzed different cell types and lifestyle differences exist between African Americans and YRI. Additionally, the geo-ancestral genes identified by Schisler et al. (34) contained at least one SNP that has significantly different allele frequency between the HapMap YRI and CEPH populations and significantly overlapped with our gene list. Therefore, underlying genetic variation between these two populations is likely to influence gene expression to a large extent. Future studies combining genetic and expression data may identify genetic variants that influence the initiation and progression of vascular calcification.

Although genetic variation influences an individual's susceptibility to diseases, environmental factors, and lifestyle can contribute substantially to changes in gene expression (9, 11, 23, 43). We observed that differences in gene expression profiles between asymptomatic African American and white women were attenuated considerably when the comparison was made between two high-CAC race groups. This suggests that reductions in the number of differentially expressed genes and the fold change magnitudes are due to the constellation of CVD risk factors influencing gene expression levels. This finding may have an important public health implication because individuals at higher risk for vascular calcification may be able to reduce their susceptibility to the diseases by changing their lifestyle.

The present study included only women for gene expression profile comparisons. It is unknown whether our race-associated genes are differentially expressed between men in these two populations. The significant overlap between our and Schisler's gene lists, however, suggests a lack of sex-related bias. In addition, the race-associated genes may correlate with IMT because participants in the high-CAC group also had higher IMT. Nevertheless, epidemiological data have shown that African Americans tend to have thicker CC-IMT than whites (19, 27). Therefore, the observed positive correlation between race-associated genes and CAC-associated genes is more likely to reflect the difference in CAC rather than IMT. Lastly, the present study has a small sample size in the low-CAC group that allowed us to adjust only a limited number of confounding factors. Therefore, potential residual confounding may exist. Future studies with larger samples and a different design are needed to address the specific confounding issues.

Since CAC is a strong indicator of atherosclerosis, some race-associated genes may also be involved in atherogenesis. For example, our gene list contains several crucial atherogenic genes including PLAUR, MMP9, and several chemokines and cytokines (IL8, CXCL1, CCL5, IL6R). This result is not surprising because several molecular mechanisms are shared between CAC and atherosclerosis (1). To our knowledge, CAC has not been used as a sole phenotype in any gene expression profile studies. Further studies are needed to verify our finding.

In summary, we have reported the first correlation of CAC with gene expression profiles among African Americans and whites. Our data provide a potential explanation for the lower prevalence of CAC among African Americans compared with whites. Functional enrichment analysis of our race-associated genes supports an important role of the immune system in vascular calcification. Joint epidemiological and laboratory investigation of several novel genes discovered in the present study may elucidate the determinants of CAC and other bone-related diseases such as valvular calcification and osteoporosis. Several race-associated genes like cytokines and chemokines could be used, in combination with conventional CVD risk factors, to identify individuals at high risk for developing vascular calcification.

GRANTS

This research was supported by National Institutes of Health (NIH) Grant R01 HL-086678 (C.-C. Huang). The MESA was supported by NIH contracts N01-HC-95159 through N01-HC-95166 and N01-HC-95169.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).