Correlation of Peripheral-Blood Gene Expression With the Extent of Coronary Artery StenosisCLINICAL PERSPECTIVE
Background— The molecular pathophysiology of coronary artery disease (CAD) includes cytokine release and a localized inflammatory response within the vessel wall. The extent to which CAD and its severity is reflected by gene expression in circulating cells is unknown.
Methods and Results— From an initial coronary catheterization cohort we identified 41 patients, comprising 27 cases with angiographically significant CAD and 14 controls without coronary stenosis. Whole-genome microarray analysis performed on peripheral-blood mononuclear cells yielded 526 genes with >1.3-fold differential expression (P<0.05) between cases and controls. Real-time polymerase chain reaction on 106 genes (the 50 most significant microarray genes and 56 additional literature genes) in an independent subset of 95 patients (63 cases, 32 controls) from the same cohort yielded 14 genes (P<0.05) that independently discriminated CAD state in a multivariable analysis that included clinical and demographic factors. From an independent second catheterization cohort, 215 patients were selected for real-time polymerase chain reaction–based replication. A case:control subset of 107 patients (86 cases, 21 controls) replicated 11 of the 14 multivariably significant genes from the first cohort. An analysis of the 14 genes in the entire set of 215 patients demonstrated that gene expression was proportional to maximal coronary artery stenosis (P<0.001 by ANOVA).
Conclusions— Gene expression in peripheral-blood cells reflects the presence and extent of CAD in patients undergoing angiography.
- gene expression
- coronary disease
- blood, peripheral
- polymerase chain reaction
Received January 31, 2008; accepted June 19, 2008.
Coronary artery disease (CAD) and sequelae of atherosclerotic disease such as stroke and myocardial infarction are the largest source of morbidity and mortality in the developed world. The risk of developing CAD events over time can be estimated with clinical factors and family history, as in the Framingham Risk Score.1 For patients with suspicious clinical histories, extant coronary disease may be diagnosed by indirect methods, including nuclear perfusion imaging and computed tomography angiography, but coronary angiography remains the “gold standard.” These tests have drawbacks, including radiation exposure, contrast agent allergy, nephrotoxicity, and, in the case of coronary angiography, invasiveness of the procedure. Therefore, the development of a blood test that reliably identified patients with CAD would have diagnostic utility.
Editorial p 7
Clinical Perspective p 38
The cellular and molecular basis of atherosclerotic plaque development has a systemic inflammatory component involving primarily monocytes/macrophages and CD4+ T cells.2,3 Oxidized lipids initiate the process with subsequent responses by endothelial, vascular smooth muscle cells, and circulating cells. Peripheral-blood studies have identified gene expression signatures that are correlated with the presence of systemic inflammatory and immune-mediated disorders, as well as cardiovascular diseases,4 suggesting that such an approach might be useful for CAD. However, although investigators have identified profiles for atherosclerosis directly from arterial wall samples in murine models and human atheroma samples,5–9 it is unclear whether such localized processes are detectable or their severity reflected in the peripheral circulation. As a first step, we sought to identify genes for which expression levels distinguished patients with and without significant coronary artery stenosis. We approached this problem by microarray analysis on an angiographically defined patient cohort to identify a set of discriminatory genes. We then replicated these results using real-time polymerase chain reaction (RT-PCR) on 2 additional sets, comprising more than 200 independent patient samples. Finally, we tested whether expression of these genes was quantitatively related to the extent of CAD.
Patients undergoing coronary artery catheterization (N=1090) at the Heart Center, Siegburg, Germany, were enrolled between May and December 2001. The protocol was approved by an institutional review board, and all patients provided written consent. For microarray analysis, we selected 27 cases with CAD defined as ≥70% stenosis in ≥1 major epicardial coronary artery or ≥50% stenosis in ≥2 arteries and 14 controls (0% stenosis) by invasive angiography. All patients presented either for angina (stable or unstable) or suspected CAD, and >50% had preceding positive stress tests. We excluded patients with a history of systemic inflammatory disease.
For RT-PCR replication, the second set of 63 cases and 32 controls was selected using the same case and control definitions. In this group, we excluded patients with a history of inflammatory disease, heart failure, elevated cardiac enzymes (creatine kinase-MB or troponin), elevated white blood cell count (>11×103/μL), or those taking anti-inflammatory medications. Patients with stable angina with any history of myocardial infarction, revascularization, or prior known CAD were also excluded.
Duke CATHGEN Cohort RT-PCR Set
For the subsequent patient sets, we identified 215 patients among 1311 patients enrolled in the Duke University CATHGEN repository, Durham, NC, between August 2004 and November 2005. For RT-PCR replication, we used a case:control subset of 86 cases and 21 controls who met the Siegburg RT-PCR cohort criteria, except that cases were defined as ≥75% maximum stenosis in 1 vessel or ≥50% in 2 vessels. For the entire group of 215 patients, CAD severity was prospectively divided into 5 angiographically defined categories: none, undetectable (n=39); minimal, stenosis <25% in a major vessel or <50% in a small vessel (n=18); intermediate, stenosis ≥25% and <75% in a major vessel (n=62); significant, 1 major vessel stenosis ≥75% or left main stenosis ≥50% (n=70); and multi-vessel, ≥3 major vessel stenosis ≥75%, with left main stenosis ≥75% counting as 2 vessels (n=26).
All patients enrolled in CATHGEN gave written informed consent for future use of their biological samples, and the current study protocol specifying use of these samples was approved by the Duke University institutional review board.
Blood Collection and RNA Isolation
For the Siegburg cohort, whole blood (16 mL) was collected before angiography and mononuclear cells isolated via density gradient centrifugation through Ficoll-Paque (Vacutainer CPT, Becton-Dickinson, Franklin Lakes, NJ). Peripheral-blood mononuclear cell (PBMC) preps were processed using RLT buffer (Qiagen, Valencia, CA) before freezing at −80°C. RNA was purified using a modification of the RNeasy method (Qiagen). For the Duke cohort, whole blood (2.5 mL) was collected in PAXgene tubes and RNA purified as described (PreAnalytix, Franklin Lakes, NJ).
Microarray Analysis and Gene Selection
Array Processing and Analysis
A 2-color microarray platform was used for gene expression analysis. In brief, 100 ng of sample RNA was labeled with Cy5 (Agilent Technologies, Santa Clara, CA, no. 5184–3523). In parallel, reference PBMC RNA (Clontech, Mountain View, CA; no. 636580) was Cy3 labeled. Both samples were hybridized to 41K Human Whole Genome Arrays (Agilent; no. G4112A) using the Agilent protocol (version 4.0.1) with the following exception: labeled reference samples were pooled in 1.06× Hybridization Buffer and 2× Blocking Buffer, heated (65°C) for 5 minutes, and centrifuged at 13 000 rpm for 10 minutes. Data were extracted on an Agilent Scanner using their Feature Extraction Software (version 188.8.131.52). For the microarrays, 57% of features were present on average.
Data were imported into GeneSpringGx (version 7.3.1), normalized as described (Agilent FE), and subjected to filtering. Retained features had a present or marginal call in at least 1 sample. In addition, Y-linked genes or those involved in X chromosome inactivation were removed.
Samples then were grouped by clinical phenotypes and expression ratios were determined. Genes with a 1.3-fold difference between groups underwent nonparametric analysis (Mann-Whitney). Genes with a P<0.05 were candidates for RT-PCR. For Gene Set Enrichment Analysis,10 version 2.0 was used for microarray analysis, using default settings. The curated set (c2) and the Gene Set Enrichment Analysis software are available from the Broad Institute (http://www.broad.mit.edu/gsea).
Literature Gene Selection
Literature genes were added to complement the relatively limited array patient set (N=41) and to include genes that either (1) have a smaller-fold change only detectable by RT-PCR or (2) did not make the top 50 array genes and yet still could be significant. From published whole genome microarray studies, we identified 261 genes expressed in peripheral blood involved in atherosclerosis or related inflammatory disorders. The number of gene citations was used for ranking, yielding 56 for RT-PCR analysis, with ≥2 independent citations.
Purified RNA was subjected to quantitative (Ribogreen; Molecular Probes, Eugene, OR) and qualitative (Agilent Bioanalyzer) analysis. Genomic DNA contamination was assessed by RT-PCR on RPL28 in the absence of reverse transcriptase. Samples showing genomic contamination underwent DNaseI treatment (Ambion, Austin, TX; no. AM1906) and retesting. RNA was converted to cDNA using Applied Biosystems High Capacity cDNA Archive Kit (ABI, Foster City, CA; no. 4322171). cDNA was stored at −20°C until use.
RT-PCR assays used minor-groove binding moiety (MGB)-containing TaqMan probes.11 Target sequences were masked for single-nucleotide polymorphisms via BLAST against dbSNP before design with PrimerExpress (version 3.0) or FileBuilder (version 3.0; ABI). Amplicons optimally spanned the transcript 3′ splice-site; otherwise, they were designed as close to the 3′ end as possible. Primer pairs were aligned to genomic DNA to ensure they mapped to single loci. Amplification efficiency was evaluated using a PBMC cDNA standard curve, and amplicon identity (size) and specificity by gel electrophoresis (4% agarose, Invitrogen, Carlsbad, CA).
Assays (10-μL final volume in 384-well plates; Eppendorf, no. 951020702)) contained 8 μL of assay mix (250 nM probe, 900 nM each primer) plus Master Mix (ABI; no. 4324020) and 2 ng of cDNA in 2 μL. For each sample, target genes were assayed once per plate, and 2 normalization genes with the lowest SDs across all samples and no discrimination between cases and controls were assayed in triplicate. Plates containing assay mix were stored at −20°C. Complete assay plates were sealed, centrifuged (2 minutes at 2000 rpm), and subjected to PCR (ABI 7900HT), using ABI-suggested cycling parameters (50°C for 2 minutes, 95°C for 10 minutes, then 40 cycles of 95°C for 15 seconds/60°C for 1 minute). Data were exported using a 0.2 threshold, with 3 to 15 cycles as baseline.
PCR Statistical Analysis Methods
Data quality was assessed by the average pairwise correlation metric across all samples. Samples with <92% average correlation were excluded from analysis. Cycle threshold (Ct) values were normalized by the geometric means of RPL28 and PRO1853 for the Siegburg cohort and RPL28 and 18s for the CATHGEN cohort. Normalized Ct values were analyzed using robust linear regression to assess association between disease status and gene expression.12 To estimate the fraction of false positives, false discovery rates were calculated for each patient set as described.13
The regression analysis fits a model of the form: y=a + b*x, where x is a 0/1 indicator of disease status, y is a gene Ct value, a is the intercept, and b is the slope (because Ct values are on the log scale, b corresponds to the log fold change). Probability values from this analysis were derived from the estimate of b divided by the standard error of b as in a standard regression model. This regression framework was also used to adjust for clinical covariates (sex, age, cholesterol, aspirin, statins, and diuretics). The adjusted model had the form: y=a+b*x+b2*z, where z are 1 or more clinical covariates and b2 are their coefficients in the regression model (b is then an adjusted log fold change).
A gene expression score was defined as the sum of the 14 normalized Ct values for the genes that survived multivariable analysis in the Siegburg RT-PCR set. Scores were calculated for all patients in the Duke PCR set of 215 patients. To test for significance of average score differences between categories, 2-sided t tests and an ANOVA with linear trend model were used. All analyses were performed using the R statistical package.14
The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.
To identify genes expressed in peripheral blood that may be sensitive to the presence of current CAD, we used a multi-step approach, starting with gene discovery from microarrays and the literature followed by RT-PCR replication in two cohorts. Figure 1 provides an overview of the overall study design and results.
Siegburg Cohort Gene Discovery
For gene discovery on microarrays, CAD (n=27) and control (n=14) patients were selected from the Siegburg cohort based on the results of cardiac catheterization and clinical presentation of either stable or unstable angina. Demographics of these 41 patients are shown in Table 1. Patients with CAD more often had diabetes, hypertension, and elevated cholesterol and were more often treated with lipid-lowering agents; aortic stenosis was more frequent among controls (n=14). From this case:control microarray comparison, a total of 526 genes showed significant differences (P<0.05 and fold change >1.3), with a large proportion of these genes (n=431) upregulated in CAD. This gene list is shown in Table S1 and the original microarray data are available at GEO (Accession No. GSE10195). In addition to the primary analysis, Gene Set Enrichment Analysis was performed to identify gene expression changes in predefined functionally distinct gene sets.10 At an false discovery rate of <5%, 343 curated gene sets were significant, with many of the most significant genes overlapping the primary analysis gene set (data not shown).
RT-PCR Replication in Siegburg Cohort
A total of 106 genes were selected for replication by RT-PCR, including the top 50 from the microarrays and 56 from the literature of genes implicated in atherosclerotic and inflammatory diseases (Figure 1, Table S2). The baseline characteristics of the 95 distinct Siegburg cohort patients used in this first replication step are shown in Table 1. Patients with CAD were more likely to be men and older than controls and had higher cholesterol. In addition, they were significantly more likely to be treated with aspirin, lipid-lowering agents, and diuretics. Patients with stable angina with previously identified CAD, myocardial infarction, or revascularization were excluded. RT-PCR analysis of the 106 selected genes on PBMC RNAs from this patient set resulted in 35 genes with P<0.05, as shown in Table 2. Permutation analysis estimated the false discovery rate as being 7% to 15%.13 Nineteen of these genes were derived from microarray analysis and 16 from the literature. Fourteen genes retained significance in a multivariable model that adjusted for sex, age, cholesterol, aspirin, statins, and diuretics (Table 1).
RT-PCR Replication in Duke CATHGEN Cohort
The baseline characteristics of the 215 patients from the Duke CATHGEN study cohort are shown in Table S3. We conducted a second RT-PCR experiment of the 106 genes on this entire patient set. The entire gene set was evaluated, as the Duke samples are whole blood rather than PBMCs, containing additional cell types, especially granulocytes, which may be biologically important in CAD. We first analyzed a patient subset comprising 86 cases and 21 controls, which met the same inclusion criteria as the Siegburg PCR set, to test all the genes identified from that cohort. A graphical representation of the significance of genes in both RT-PCR case:control replication sets is shown in Figure 2. Twenty-seven of the 35 genes with P<0.05 in univariate analysis from the Siegburg RT-PCR set were confirmed in the Duke set, including 11 of the 14 that survived multivariable analysis. All the genes in the Duke RT-PCR replication set with P<0.05 are listed in Table S4 (estimated false discovery rate=5% to 11%).
After stratification of the entire 215 patients from the Duke cohort into 5 maximum stenosis groups (none, minimal, intermediate, significant, and multi-vessel disease), we as sessed the relationship of disease severity with a gene expression score constructed from the Ct values of the 14 genes that survived the multivariable case:control analysis in the Siegburg cohort. These results (Figure 3) suggest that peripheral-blood gene expression measurements are sensitive to the extent of maximum coronary stenosis, either by paired t tests between groups or ANOVA (P<0.001).
This study examined whether patients with and without angiographically defined significant CAD could be distinguished by gene expression analysis of peripheral-blood cells. Gene discovery and replication by RT-PCR were used to identify 14 genes that were independently significant in multivariable analysis that included clinical and demographic factors. Eleven of the 14 genes were confirmed in an independent replication cohort from a second institution. Furthermore, a gene expression score created by summing the expression levels of the 14 genes was proportional to the degree of maximum stenosis.
Previous studies have examined the relationships between existing CAD and the expression levels of specific genes or proteins in peripheral blood theoretically involved in CAD development.15–17 A predominant role for monocytic and T cells in atherosclerotic plaque formation has been described, although other cell types, including mast cells, have been implicated.18 Also, studies on gene expression in diseased artery samples from both mice and humans have been performed.5–7 The current study confirmed genes described by other investigators and identified new genes not previously described as involved in atherosclerosis.
Biological Plausibility of Identified Genes
For the 14 genes that survived the combined microarray and RT-PCR based analyses of the Siegburg cohort, 11 were confirmed in the Duke RT-PCR set. These 11 genes were all upregulated in patients with significant CAD. This may reflect their largely monocytic origin, as increased monocyte counts have been associated with CAD.19 Analysis of biological pathways for these genes shows that pro- and antioxidant molecules, extracellular matrix, cell motility proteins, and signaling receptors and transcription factors were represented.
The role of oxidized low-density lipoprotein in CAD pathogenesis has been well documented; we observe upregulation of 5-lipoxygenase, a generator of leukotrienes, which may play a key role in atherosclerosis development.20,21 Conversely, the upregulation of glutathione-S-transferase, a member of the antioxidant detoxification pathways, may represent response to oxidative insult.
Versican, a chondroitin sulfate proteoglycan, has been previously implicated in atherosclerosis.22 Extracellular matrix regulation is correlated with plaque erosion and rupture and with myocardial infarction.23,24 The gelsolin-like capping protein, CAPG, is involved in cell motility; in vivo, it has a role in host pathogen defense and thus in innate immunity.25 CAPG overexpression in atherosclerotic aortic samples has been reported.6
A number of signaling receptors were also upregulated in patients with significant CAD. VSIG4, a member of the B7 receptor family, is expressed on macrophages in carotid atherosclerosis26 and negatively regulates T cell activation.27 The receptors for granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor (CSF3R and CSF2RA, respectively) both were upregulated in this study; a role for these colony-stimulating factor pathways in atherosclerosis has been suggested by mouse studies, although the data are somewhat contradictory.28,29 Interestingly, granulocyte-macrophage colony-stimulating factor null mice show decreased interleukin-1 receptor antagonist expression, another gene upregulated in this study.28 Expression levels of this gene affect development of atherosclerosis in mouse models,30 and its protein product is found in human atherosclerotic arteries.31
Hexokinase 3 is upregulated in atherosclerotic disease in human aortic samples.6 CREB5 and NS5ATP13T are members of transcription factor families for which further work is needed to understand their relationship to atherosclerosis.
This study has several limitations. First, this was a diagnostic study with maximum stenosis determined by invasive coronary angiography as an end point and does not examine the ability of gene expression to predict future clinical events. Second, we used a series of small, independent retrospective case:control studies and multivariable analysis to evaluate the relative roles of gene expression measurements and clinical and demographic factors. One particularly significant imbalance in our cohorts is due to sex, a known risk factor for CAD. A larger, truly prospective study might be less influenced by unidentified confounders and would possibly identify genes important in specific demographic subsets, such as women. Third, it is not possible to assign causation to the gene expression changes we observed; rather we have demonstrated relationships that could be causal or responsive to the existence of CAD, but also could have no mechanistic relationship with CAD or its extent.
Experimentally, the small size of the microarray cohort and use of PBMC RNAs in the initial discovery experiment likely limited the number of significant genes identified.32 We attempted to mitigate this by adding literature genes associated with CAD, but a larger discovery study using whole-blood RNA might find additional significant genes. Despite these challenges, the majority of the univariable (27 of 35) and multivariable (11 of 14) significant genes from the Siegburg study were replicated in the Duke whole-blood set. This is perhaps not surprising, as mononuclear cells are a subset of whole blood, and leukocyte and monocytes are often cited as key cell types in atherosclerosis development.15
Although we used the highly reproducible method of RT-PCR for our replication studies, the fold changes observed are quite modest, although similar to those seen in other studies of peripheral-blood gene expression.33 Our small sample sizes and the large number of genes investigated also renders our study vulnerable to false discovery. Despite our attempts to account for this by using multiple independent cohorts, our results should be considered hypothesis generating while awaiting validation in larger cohorts.
Additionally, we only examined gene expression in peripheral blood, rather than in atherosclerotic plaque. It is known that a complex interplay between circulating cells and the endothelium leads to development of atherosclerosis, so circulating peripheral-blood cells may directly reflect only one aspect of a complex pathophysiology. Other methods including proteomic or metabolomic approaches may reflect additional components of the disease process, such as endothelium and liver, and should be viewed as potentially complementary to our approach. Finally, because we were measuring overall gene expression levels in peripheral blood, it is likely that specific cell population shifts and gene expression changes within specific lineages may have contributed to the observed results.
We have used gene expression analysis to identify and confirm genes that distinguish patients with and without angiographically defined CAD. We also observed that gene expression scores based on the 14 genes independently associated with the presence or absence of CAD were proportional to the extent of disease burden, as defined by maximum stenosis. It is possible that these measurements solely reflect the presence of a disease process. Alternatively, gene expression might reflect the extent of overall atheromatous burden and the rate of disease progression. Our current results are more consistent with this latter hypothesis, although further work is required with serial sampling of patients to determine a correlation with disease progression. Overall, this study suggests that it may be possible to identify and monitor the severity and progression of CAD from gene expression analysis of peripheral-blood samples.
We thank the patients and teams at both The Siegburg Heart Center and Duke University Medical Center for all their efforts in providing samples and clinical information; Rachel Nuttall, Michael Doctolero, Lori Littleford, Nicholas Paoni, Dick Lawn, and other colleagues at CardioDx for their work; and Eric Topol, Fred E. Cohen, Lewis T. Williams, and Joffre Baker for their review of this manuscript.
Source of Funding
This work was supported by CardioDx, Inc.
Drs Wingrove, Daniels, Sehnert, Tingley, Elashoff, and Rosenberg are employed by and own stock in CardioDx, Inc. Dr Rosenberg has also served as a consultant to or on the advisory board of XDx, Inc. Dr Newby has received research support from BG Medicine and Biosite and has served as a consultant to or on the advisory boards of Roche Diagnostics and CV Therapeutics. Dr Ginsburg has received a research grant from and owns stock in CardioDx, Inc. The other authors report no potential conflicts.
Wilson PW, D'Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998; 97: 1837–1847.
Libby P, Theroux P. Pathophysiology of coronary artery disease. Circulation. 2005; 111: 3481–3488.
King JY, Ferrara R, Tabibiazar R, Spin JM, Chen MM, Kuchinsky A, Vailaya A, Kincaid R, Tsalenko A, Deng DX, Connolly A, Zhang P, Yang E, Watt C, Yakhini Z, Ben-Dor A, Adler A, Bruhn L, Tsao P, Quertermous T, Ashley EA. Pathway analysis of coronary atherosclerosis. Physiol Genomics. 2005; 23: 103–118.
Seo D, Wang T, Dressman H, Herderick EE, Iversen ES, Dong C, Vata K, Milano CA, Rigat F, Pittman J, Nevins JR, West M, Goldschmidt-Clermont PJ. Gene expression phenotypes of atherosclerosis. Arterioscler Thromb Vasc Biol. 2004; 24: 1922–1927.
Tabibiazar R, Wagner RA, Ashley EA, King JY, Ferrara R, Spin JM, Sanan DA, Narasimhan B, Tibshirani R, Tsao PS, Efron B, Quertermous T. Signature patterns of gene expression in mouse atherosclerosis and their correlation to human coronary disease. Physiol Genomics. 2005; 22: 213–226.
Donahue MP, Rose K, Hochstrasser D, Vonderscher J, Grass P, Chibout SD, Nelson CL, Sinnaeve P, Goldschmidt-Clermont PJ, Granger CB. Discovery of proteins related to coronary artery disease using industrial-scale proteomics analysis of pooled plasma. Am Heart J. 2006; 152: 478–485.
Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005; 102: 15545–15550.
Kutyavin IV, Afonina IA, Mills A, Gorn VV, Lukhtanov EA, Belousov ES, Singer MJ, Walburger DK, Lokhov SG, Gall AA, Dempcy R, Reed MW, Meyer RB, Hedgpeth J. 3′-minor groove binder-DNA probes increase sequence specificity at PCR extension temperatures. Nucleic Acids Res. 2000; 28: 655–661.
Huber PJ. Robust Statistics. New York: Wiley; 1981.
Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A. 2003; 100: 9440–9445.
Team RDC. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2007.
Vanderlaan PA, Reardon CA. Thematic review series: the immune system and atherogenesis: the unusual suspects: an overview of the minor leukocyte populations in atherosclerosis. J Lipid Res. 2005; 46: 829–838.
Patino WD, Mian OY, Kang JG, Matoba S, Bartlett LD, Holbrook B, Trout HH, III, Kozloff L, Hwang PM. Circulating transcriptome reveals markers of atherosclerosis. Proc Natl Acad Sci U S A. 2005; 102: 3423–3428.
Tabibiazar R, Wagner RA, Deng A, Tsao PS, Quertermous T. Proteomic profiles of serum inflammatory markers accurately predict atherosclerosis in mice. Physiol Genomics. 2006; 25: 194–202.
Qiu H, Gabrielsen A, Agardh HE, Wan M, Wetterholm A, Wong CH, Hedin U, Swedenborg J, Hansson GK, Samuelsson B, Paulsson-Berne G, Haeggstrom JZ. Expression of 5-lipoxygenase and leukotriene A4 hydrolase in human atherosclerotic lesions correlates with symptoms of plaque instability. Proc Natl Acad Sci U S A. 2006; 103: 8161–8166.
Wight TN, Merrilees MJ. Proteoglycans in atherosclerosis and restenosis: key roles for versican. Circ Res. 2004; 94: 1158–1167.
Gabrielsen A, Lawler PR, Yongzhong W, Steinbruchel D, Blagoja D, Paulsson-Berne G, Kastrup J, Hansson GK. Gene expression signals involved in ischemic injury, extracellular matrix composition and fibrosis defined by global mRNA profiling of the human left ventricular myocardium. J Mol Cell Cardiol. 2007; 42: 870–883.
Parikh SS, Litherland SA, Clare-Salzler MJ, Li W, Gulig PA, Southwick FS. CapG(-/-) mice have specific host defense defects that render them more susceptible than CapG(+/+) mice to Listeria monocytogenes infection but not to Salmonella enterica serovar Typhimurium infection. Infect Immun. 2003; 71: 6582–6590.
Lee MY, Kim WJ, Kang YJ, Jung YM, Kang YM, Suk K, Park JE, Choi EM, Choi BK, Kwon BS, Lee WH. Z39Ig is expressed on macrophages and may mediate inflammatory reactions in arthritis and atherosclerosis. J Leukoc Biol. 2006; 80: 922–928.
Ditiatkovski M, Toh BH, Bobik A. GM-CSF deficiency reduces macrophage PPAR-gamma expression and aggravates atherosclerosis in ApoE-deficient mice. Arterioscler Thromb Vasc Biol. 2006; 26: 2337–2344.
Haghighat A, Weiss D, Whalin MK, Cowan DP, Taylor WR. Granulocyte colony-stimulating factor and granulocyte macrophage colony-stimulating factor exacerbate atherosclerosis in apolipoprotein E-deficient mice. Circulation. 2007; 115: 2049–2054.
Dewberry R, Holden H, Crossman D, Francis S. Interleukin-1 receptor antagonist expression in human endothelial cells and atherosclerosis. Arterioscler Thromb Vasc Biol. 2000; 20: 2394–2400.
Simon R. Roadmap for developing and validating therapeutically relevant genomic classifiers. J Clin Oncol. 2005; 23: 7332–7341.
Deng MC, Eisen HJ, Mehra MR, Billingham M, Marboe CC, Berry G, Kobashigawa J, Johnson FL, Starling RC, Murali S, Pauly DF, Baron H, Wohlgemuth JG, Woodward RN, Klingler TM, Walther D, Lal PG, Rosenberg S, Hunt S. Noninvasive discrimination of rejection in cardiac allograft recipients using gene expression profiling. Am J Transplant. 2006; 6: 150–160.
This work examined the hypothesis that peripheral-blood gene expression reflected the extent of coronary artery stenosis, as defined by invasive angiography. For patients with suspicious clinical histories or chest pain of uncertain etiology, diagnosis of extant coronary artery disease is often accomplished by a variety of methods, including nuclear perfusion imaging and computed tomography angiography, but coronary angiography remains the “gold standard.” These tests have associated risks, including radiation exposure, contrast agent allergy, nephrotoxicity, and, in the case of coronary angiography, invasiveness of the procedure. Therefore, the development of a blood test that reliably identifies patients with significant coronary artery disease who might most benefit from further testing with its significant burden and risks would have clinical utility. Gene discovery by gene expression microarrays and the literature were used to identify a set of candidate genes. These results were then replicated using real-time polymerase chain reaction in two additional sample sets, comprising more than 200 patients from two independent patient cohorts. Finally, the candidate genes that were significant in multivariable analysis in the first cohort demonstrated a quantitative relationship with the extent of coronary artery disease in the second cohort. These results suggest that measurement of gene expression in peripheral-blood cells could be of clinical utility in the diagnosis of coronary artery disease with the derivation of an explicit classifier and permit, after validation in additional studies, the development of a blood-based diagnostic test to identify patients with clinically significant coronary artery disease.
Correspondence to CardioDx, Inc, 2500 Faber Place, Palo Alto, CA 94303. E-mail email@example.com
The online Data Supplement can be found with this article at http://circgenetics.ahajournals.org/cgi/content/full/1/1/31/DC1.