Atherosclerosis - Journal of the American College of Cardiology · Atherosclerosis The Path From Genomics to Therapeutics David T. Miller, MD, PHD,*§ Paul M Ridker, MD, MPH, FACC,†‡§

Tmatpdpurm

FCBCtgr(HFDcuhBNG

a

Journal of the American College of Cardiology Vol. 49, No. 15, 2007© 2007 by the American College of Cardiology Foundation ISSN 0735-1097/07/$32.00P

CARDIOVASCULAR GENOMIC MEDICINE

AtherosclerosisThe Path From Genomics to Therapeutics

David T. Miller, MD, PHD,*§ Paul M Ridker, MD, MPH, FACC,†‡§ Peter Libby, MD, FACC,‡§David J. Kwiatkowski, MD, PHD*§

Boston, Massachusetts

Recent rapid advances in genomic tools and techniques hold great promise for transforming the practice of car-diovascular medicine. Resources including the Human Genome Project and the International HapMap project,major technological advances in high-throughput genotyping and methods of statistical analysis, and methodsfor high-throughput gene expression and small molecule profiling allow researchers to confront issues that willfundamentally change the practice of cardiovascular medicine during the 21st century. Genomic, proteomic, andmetabolomic studies of complex cardiovascular diseases such as atherosclerosis will bridge epidemiology andbasic biology, and promise increased understanding of cardiovascular disease processes. Genetic approachesapplied to atherosclerosis will continue to identify genes and pathways involved in the predisposition to andpathophysiology of atherosclerosis. Gene expression profiling refines our understanding of the dynamic na-ture of the atherosclerotic vascular wall and promises discovery and validation of targets for therapeuticintervention. Opportunities to translate genetic, genomic, proteomic, and metabolomic information into car-diovascular clinical practice have never been greater, but their fruition requires validation in large indepen-dent cohorts, achieved only through collaborative effort. Their continued success will depend on ongoingcooperation within the cardiovascular research community. (J Am Coll Cardiol 2007;49:1589–99)© 2007 by the American College of Cardiology Foundation

ublished by Elsevier Inc. doi:10.1016/j.jacc.2006.12.045

ctetmt

donipsplt

rbamgigg

his state-of-the-art review focuses on research that appliesajor technological developments in genomic medicine to

therosclerotic cardiovascular disease. High-throughputechnologies facilitate identification of genetic, genomic,roteomic, and metabolomic markers of coronary arteryisease (CAD) risk that may find a place in clinically usefulrediction algorithms. Such risk models would augment thetility of established clinical tools such as the Framinghamisk score. However, putative genomic, proteomic, andetabolomic markers will require the same rigor in appli-

rom the *Division of Hematology, †Division of Preventive Medicine and Center forardiovascular Disease Prevention, and ‡Division of Cardiovascular Medicine,righam and Women’s Hospital, and the §Donald W. Reynolds Cardiovascularlinical Research Center on Atherosclerosis at Brigham and Women’s Hospital and

he Harvard Medical School, Boston, Massachusetts. This work was supported by arant from the Donald W. Reynolds Foundation (Las Vegas, Nevada). Dr. Millereceived fellowship support from the National Heart, Lung, and Blood InstituteNHLBI 1 F32 HL78274-01). Dr. Ridker is also supported by grants HL43851,

L63293, and HL58755 from the NHLBI, with additional support from the Leducqoundation (Paris, France) and a Distinguished Clinical Scientist Award from theoris Duke Charitable Foundation (New York, New York). Dr. Ridker is listed as a

o-inventor on patents held by the Brigham and Women’s Hospital that relate to these of inflammatory biomarkers in cardiovascular disease and diabetes. Dr. Ridker alsoas received investigator-initiated research support from AstraZeneca, Roche, Dade-ehring, Novartis, and Sanofi-Aventis, and has been a consultant to AstraZeneca,ovartis, and ISIS. Cardiovascular Genomic Medicine series edited by Geoffrey S.insburg, MD, PhD.

nManuscript received June 6, 2006; revised manuscript received October 30, 2006,

ccepted December 4, 2006.

ation as other epidemiologic markers. Ultimately, theranslation of such information to clinical practice shouldnhance “personalized medicine.” The coupling of informa-ion gained through genomic, proteomic, and metabolomicethodologies to traditional tools should sharpen our ability

o assess and modify management of cardiovascular disease.Despite steady progress, atherosclerotic cardiovascular

isease remains a growing public health burden in devel-ped countries, and advances in cardiovascular research willot realize their full impact unless translated to care of

ndividual patients (1). Hope for determining new thera-eutic targets in atherosclerosis increasingly rests on re-earch progress in genetic studies, expression profiling, androteomics (2–7). Combining new markers of cardiovascu-

ar disease with currently available screening tools promiseso promote this translational process.

Before a personalized medicine approach to atheroscle-osis based on genomics, proteomics, and metabolomics canecome reality, researchers must validate novel markerscross different cohorts and in relation to various environ-ental modifiers. The operation of intricate networks of

enes, environmental factors, and gene-by-environmentnteractions further complicates our understanding of theenetic components of atherosclerosis (8,9). Combinedenomic approaches, often called genomic convergence, are
ecessary in atherosclerosis research (7).

ucapgiatstt

GA

GDgtiel

lmu5dmfmvils(

cfm5MlIg(limua

lasriAhrrlgspcOrGmarcamaad

1590 Miller et al. JACC Vol. 49, No. 15, 2007Atherosclerosis: Genomics to Therapeutics April 17, 2007:1589–99

Technological revolutions ingenetics and genomics have facil-itated 2 major approaches to un-derstanding disease pathogenesisand risk. First, the Human Ge-nome Project and InternationalHapMap Project now allow us toobtain with relative speed vastamounts of deoxyribonucleic acid(DNA)-based information appli-cable to research subjects withatherosclerotic cardiovasculardisease. Second, current technol-ogy enables collection of infor-mation on expression for thou-sands of genes in vascular cellsand tissues under different con-ditions. The human DNA se-quence laid the groundwork forstudies of genetic susceptibility todisease, and expression databasesassist definition of disease subtypesand variance related to environ-mental interactions. Althoughcurrently less mature technolo-gies, proteomic and metabolomicstudies promise to complementgenomic approaches. Table 1compares advantages and disad-vantages of methods for markeridentification.

A patient-specific risk profilefor cardiovascular disease gener-ated by knowledge of the genetic

nderpinnings of human disease risk would likely assist thelinician in providing targeted interventions, e.g., drugs thatct on only a subset of the population or avoiding aarticular environmental exposure known to interact with aenetic variant. Both examples illustrate personalized med-cine. Although current technology enables researchers tomass huge amounts of genomic data in a relatively shortime frame, translation to the clinic will require demon-trated effects on outcomes. Indeed, the bottleneck forranslation of genomic medicine to everyday practice lies athat intersection of knowledge.

enetic Studies oftherosclerotic Cardiovascular Disease

enetic linkage studies. Using families and anonymousNA markers, genetic linkage studies correlate inherited

enomic regions with inherited familial characteristics. Al-hough the many genes and environmental factors influenc-ng atherosclerosis make direct linkage difficult, numerousxamples link phenotypes with atherosclerotic cardiovascu-

Abbreviationsand Acronyms

ALOX5AP � 5-lipoxygenaseactivating protein gene

CAD � coronary arterydisease

CRP � C-reactive protein

FLAP � 5-lipoxygenaseactivating protein

HMG-CoA � 5-hydroxy-3-methylglutaryl-coenzyme A

HSP27 � heat shockprotein-27

LDL � low-densitylipoprotein

LTA4H � leukotriene A4hydrolase gene

LTB4 � leukotriene B4

MI � myocardial infarction

MMP � matrixmetalloproteinase

OR � odds ratio

Osm � oncostatin M

Osmr � oncostatin Mreceptor

PCSK9 � proproteinconvertase subtilisin/kexintype 9 serine protease

SNP � single nucleotidepolymorphism

TNFSF4 � tumor necrosisfactor (ligand) superfamily,member 4

ar disease. The work of Goldstein and Brown (10) on a

ow-density lipoprotein (LDL)-receptor mutations in fa-ilial hypercholesterolemia furnished a foundation for

nderstanding the mechanism of LDL lowering by-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) re-uctase inhibitors. However, single-gene defects in lipidetabolic pathways account only for a very small portion of

amilial and sporadic atherosclerosis (11). In most people,ultiple genes and environmental factors determine ele-

ated lipid levels and atherosclerotic events. Among familiesn which a single gene plays a stronger role, evidence ofinkage frequently identifies a chromosomal locus but notingle genes for atherosclerotic cardiovascular disease12–16).

Notably, linkage analysis has successfully identified spe-ific genes that may contribute to cardiovascular event risk,or example, myocyte enhancer factor-2 (MEF2A) withyocardial infarction (MI) risk (3) and arachidonate

-lipoxygenase activating protein gene (ALOX5AP) withI and stroke risk (17). Helgadottir et al. (17) showed

inkage between the ALOX5AP gene region and MI in 296celandic families, including 713 subjects. The ALOX5APene encodes the enzyme 5-lipoxygenase activating proteinFLAP), which participates in leukotriene synthesis. Sucheukotrienes, especially leukotriene B4 (LTB4), mediatenflammation within the vasculature and participate in

urine atherosclerosis (18). Helgadottir et al. (17) followedp their suspicions about ALOX5AP with a genetic associ-tion study approach (see the following text).

Human families with CAD are not the only resource forinkage analysis. Almost 2 decades ago, linkage analysismong inbred mouse strains identified an atherosclerosisusceptibility region on mouse chromosome 1 (Ath1). Thisegion contains the tumor necrosis factor (ligand) superfam-ly member 4 (Tnfsf4) gene, also termed Ox40l (19).therogenic in mice, OX40L stimulates T cells. Otheruman studies recently linked the TNFSF4 region and theegion containing the gene for OX40 (i.e., the OX40Leceptor) to CAD (16,20), but did not identify the under-ying gene in either case. Comparison of mouse and humanenetics ultimately led to TNFSF4, and further studieshowed an association between a TNFSF4 single nucleotideolymorphism (SNP) and MI risk in women from 2 humanohorts (19). Most recently, another study associated anX40 SNP with MI risk (21), suggesting that this ligand-

eceptor interaction may provide a therapeutic target.enetic association studies. Many genetic and environ-ental factors likely influence complex diseases such as

therosclerosis. Viewed separately, each factor exerts aelatively small effect. Combined, they might explain in-reased disease risk. Compared with linkage studies, geneticssociation studies provide greater statistical power andore detailed knowledge of a genetic region, enhancing our

ttempt to understand the genetics of complex diseases suchs atherosclerosis. In general, these studies seek to identifyifferences in the inheritance of particular SNP alleles
mong subjects with a differing clinical phenotype, such as

Mb(CmstgFctnbscCac(

nCwtCSlgd

fiaced

catfterwv(11rhf1egrd

d

C

S

1591JACC Vol. 49, No. 15, 2007 Miller et al.April 17, 2007:1589–99 Atherosclerosis: Genomics to Therapeutics

I (cases vs. control subjects) or differing levels of aiomarker such as C-reactive protein (CRP) or cholesterolhigh vs. low levels).

andidate gene association studies. Association studiesay take a candidate gene or genome-wide approach. Such

tudies often base selection of candidate genes on assump-ions about biologically relevant genes. Thus, candidateene studies are biased against identification of novel genes.or example, several groups including our own publishedandidate gene association studies on the relationship be-ween SNPs in the CRP gene and CRP levels based onumerous prospective studies showing the influence ofaseline CRP levels on cardiovascular risk (22,23). Threetudies in particular provide comprehensive coverage of allommon CRP SNPs (24–26). Figure 1 shows 7 commonRP SNPs (Fig. 1A), illustrating the sequence difference atn SNP within exon 2 (Fig. 1B) and the generation of 6ommon haplotypes that result when SNP alleles combineFig. 1C).

The SNPs associated with an intermediate phenotype areot always associated with the disease state itself. The thirdRP SNP in Figure 1, rs3091244, consistently associatesith higher CRP levels in multiple studies (24–27). Despite

hese consistent associations, the proportion of variance inRP levels explained by CRP genotypes is only �2%. TheseNPs thus have only a modest influence on plasma CRP

evels. Family studies indicate that there must be additionalenetic factors contributing to CRP levels. To date, candi-

omparison of Genetic, Genomic, Proteomic, and Metabolomic Res

Table 1 Comparison of Genetic, Genomic, Proteomic, and Meta

Method Advantages

Genetic linkage study Identifies single genes with largSurveys the entire genome

Genetic association study More statistical power for compcompared with linkage analy

Could be family-based or unrelaFiner resolution than linkage an

Candidate gene (100–1,000 SNPs) Lower cost than genome-wide sCustomized choice of SNPs

Whole genome (100,000–500,000 SNPs) Surveys entire genome, nothingHuge amount of information peIdentifies variants with relatively

Gene expression study Information at the tissue levelReal-time information about pa

changesCustomized chips for panels of

a process (e.g., inflammation

Proteomic/metabolomic study Information at the tissue levelInformation about protein levels

genotype or mRNA levels

NP � single nucleotide polymorphism.

ate gene SNP selection has achieved limited success in r

nding genetic variants of CRP level, an example of theforementioned bias against novel gene identification. Case-ontrol studies to date have not been adequately powered tovaluate whether or not these SNPs impact on incidentisease (25,26,28).Testing greater numbers of SNPs can increase the

hances of finding a disease-causing variant. Technologicaldvances that reduce genotyping costs now allow researcherso “supersize” their candidate gene studies. A recent studyocused on 11,053 SNPs located within genes, reasoninghat those SNPs more likely would affect gene function orxpression (29). Using a case-control design and 3 rounds ofeplication, Shiffman et al. (29) identified 4 SNPs associatedith risk of MI (all odds ratios [ORs] indicate carriers of 2s. 0 risk alleles): the cytoskeletal protein paladinKIAA0992 [OR 1.40]), a tyrosine kinase (ROS1 [OR.75]), and 2 G-protein coupled receptors (TAS2R50 [OR.58] and OR13G1 [OR 1.40]). Although these SNPsepresent risks on par with accepted risk markers such asypertension and CRP (23), other attempts at replicationor this panel of SNPs yielded mixed results with as few as

in 5 SNPs having a similar magnitude and direction offfect (30). The determination of the generalizability ofenetic associations observed in a given study populationequires replication in independent samples, as discussed inetail elsewhere (31).Combining different genetic methodologies facilitates

iscovery. For example, genetic association studies help

h Methods

mic Research Methods

Disadvantages

t Limited power for complex diseasesLow resolution; must be followed by fine mapping and/or

candidate gene studiesMust be family-based; family may be affected by a rare gene

that has little influence among the general population

eases

bjects

Requires additional samples/cohorts for replicationNeed large number of subjects (�2,000 cases � 2,000 controls)

to detect genetic markers with small effect

Only find associations in the genes chosen for analysis

d

l effect

Expensive because of SNP volumeBioinformatics challengeMay miss a region of interest because of poor coverageRequires multiple rounds of replication

siological

related to

Source tissue (e.g., biopsy) often poor quality or difficult toobtain, resulting in small sample sizes

Variability between assays/samples because of experimentaland pathophysiological variation

Possible bias in types of genes discovered based on which chipis selected

mRNA levels and protein levels not perfectly correlated

dless ofPost-translational processing leads to variabilityNeed source tissue (e.g., vascular biopsy)Variability between assays/samples because of experimental

and pathophysiological variation (e.g., variability in sampleprocessing)

Not as amenable (yet) to high-throughput format

earc

bolo

e effec

lex dissisted sualysis

tudy

misser assaysmal

thophy

genes)

regar

efine regions identified by linkage studies. After identifying

lHaapadp�h20iwApMapiTos

cslsDssaSci0

viteiAcscsutpsglp(Cp

as(wositrcGosegtnp

c


inkage between MI and the region containing ALOX5AP,elgadottir et al. (17) focused more closely on ALOX5AP,

s a candidate gene, sequencing the entire gene in 93ffected individuals and 93 control subjects to identify aanel of 48 common SNPs in ALOX5AP. A case-controlssociation study genotyped these 48 ALOX5AP SNPs toetermine differences in the inheritance of a particularattern of SNPs, or haplotype, between subjects with MI (n

779) and control subjects (n � 624). An inherited 4-SNPaplotype, HapA, conferred an increased risk (nearly-fold) for MI and stroke (relative risk 1.8, adjusted p �.005). The LTB4 production in stimulated neutrophilsncreased substantially in individuals with MI comparedith control subjects, especially in male carriers ofLOX5AP HapA, who showed significantly greater LTB4roduction compared with control subjects (p � 0.0042).ale carriers of the at-risk haplotype had the strongest

ssociation with disease and also had significantly greaterroduction of leukotriene-B4 (LTB4). The HapA alsoncreased risk of stroke (relative risk 1.67, p � 0.000095).o date, an association between ALOX5AP and stroke wasbserved in more than 1 cohort (32,33), but replication

Figure 1 Genetic Information: SNPs and Haplotypes

The C-reactive protein (CRP ) gene structure, single nucleotide polymorphisms(SNPs), and haplotypes; (A) CRP consists of 2 exons comprising codingregions (darkest shading) and noncoding regions (medium shading) and has 7common SNPs indicated by rs numbers and relative position. (B) Sequencingtraces of a G/C SNP in exon 2 shows the G/G version (black peak, top) andG/C version (black/blue peak, bottom). (C) Different combinations of allelesat each of the 7 SNP positions create 6 common haplotypes for this set ofCRP SNPs. Illustration by Rob Flewell.

tudies for ALOX5AP have provided mixed results (34). p

Most recently, Helgadottir et al. (35) used additionalandidate gene association studies to expand the ALOX5APtory (35). Using a pathway approach, they studied theeukotriene A4 hydrolase (LTA4H) gene, the enzyme re-ponsible for the production of LTB4. They performedNA sequencing of the LTA4H gene region (42 kb) in 93

ubjects with MI to identify 8 novel SNPs. Considered asingle markers, none of those 8 SNPs was significantlyssociated with MI in further studies. Considering groups ofNPs as haplotype blocks revealed a particular haplotype,alled HapK, significantly associated with MI. Each inher-ted copy of HapK conferred a relative risk of 1.45 (p �.035 after adjusting for the number of haplotypes tested).Replication of genetic associations among subjects of

arious ethnicities addresses variability in SNP allele inher-tance patterns among ethnic groups that could influencehe way a particular variant affects disease risk. As onexample, Helgadottir et al. (35) studied LTA4H in 3ndependent MI cohorts of European-Americans andfrican Americans from the U.S. The HapK was signifi-

antly associated with MI among the European-Americanubset of this cohort (relative risk 1.19; p � 0.006), and alsoonferred higher relative risk among the African-Americanubset (relative risk 3.57; p � 0.000022). Presumably,nderlying genetic and environmental differences betweenhe European-American and African-American subsets ex-lain the difference in LTA4H-associated MI risk, under-coring the need to include different ethnic populations inenetic association studies (36). Rare variants in one popu-ation may be relatively common in another, for example,roprotein convertase subtilisin/kexin type 9 serine proteasePCSK9), where 2 nonsense mutations not present inaucasians confer lower levels of LDL cholesterol androtection from CAD to African Americans (37).Other efforts to determine genetic associations with

therosclerosis have achieved limited success in explaining aignificant proportion of the population-attributable risk38,39), including several associations summarized else-here (3,40). This inherent problem, not limited to the studyf cardiovascular disease, occurs in all genetic associationtudies of complex disease in which many genetic markersndividually explain only a fraction of the genetic contributiono disease. Moreover, results from a genetic association studyequire replication through independent studies, as dis-ussed in detail elsewhere (41).

enome-wide association studies. As the per-SNP costsf high-throughput genotyping decline, cardiovascular re-earchers are turning to genetic association studies on anven larger scale. In large prospective cohorts, so-calledenome-wide association studies examine hundreds ofhousands of SNPs throughout the genome. Such studies,ot hypothesis-driven, are suited ideally to discoveringreviously unimagined pathways for particular diseases.The most significant disadvantage of genome-wide asso-

iation studies involves the statistical conundrum of multi-
le comparisons inherent in simultaneously performing

a5gnagda

Gi

Vrcsenrt

bppLb(i

nasfgsIatio

sgdtmetmidg

ptsegwbdph

r(csacwo(cgp


ssociation tests on thousands of markers. A study of00,000 SNPs with a false-positive rate of 0.1% wouldenerate 500 false-positive results, a very large number thatecessitates multiple rounds of replication to confirm anssociation. Thus, thoroughly replicated results fromenome-wide association studies applied to cardiovascularisease and related phenotypes have just now begun toppear (42).

ene Expression Profilingn Cardiovascular Disease

alidating new discoveries related to atherosclerosis willequire multiple complementary approaches (Fig. 2), in-luding gene expression profiling. The SNP associationtudies cannot determine which SNPs may affect genexpression or function. Gene expression microarray tech-ology, reviewed elsewhere (43), provides a way to scanapidly specific tissues for patterns of expression among

Figure 2 Levels of Genomic and Proteomic Information

Capturing all genomic and proteomic information requires a combinedapproach that targets DNA, RNA, and proteins. The single nucleotide polymor-phisms (SNPs) may influence gene expression (e.g., by altering a transcrip-tional activator binding site in the promoter), or gene function (e.g., by causingan amino acid change). Transcriptional activators (TA), repressors (TR), andmRNA turnover influence mRNA levels independent of SNP genotypes; thesemechanisms would only be detectable by expression profiling. Although mRNAlevels may vary within and among cells, regulation of protein trafficking, turn-over, or structure could affect a protein level or function independently fromgenotype or mRNA regulation. Illustration by Rob Flewell.

housands of genes across the genome. Several studies in H

oth human and animal models have applied expressionrofiling technology to atherosclerosis. The general ap-roach is exemplified by studies such as that by Ma andiew (44), which used transcriptional profiling of peripherallood leukocytes among subjects with coronary heart disease80% to 90% stenosis) versus healthy control subjects todentify 108 differentially expressed genes.

Although SNPs represent static information (the ge-ome), expression profiles are dynamic (the transcriptome)nd may show physiological fluctuations (Fig. 2). Expres-ion profiling of atheromata in human aortic tissue collectedrom 63 heart donors indicated significant differences inene expression patterns for 212 genes, based on diseaseeverity and lesion location within the thoracic aorta (45).mportantly, the investigators verified transcriptome reli-bility and accuracy by comparing results obtained by moreraditional histological methods. Indeed, expression profil-ng correctly classified disease severity and location in 93%f samples tested.Extending those findings, Karra et al. (46) recently reported

tudies of gene expression during atherosclerotic lesion pro-ression. These studies compared a histologically determinedisease state with patterns of gene expression in mouse aorticissue from wild-type and apolipoprotein E�/� C57BL6/Jice exposed to different diets and at different ages. Gene

xpression patterns varied according to disease stage: no diseaseo early disease (197 genes, including many involved in lipidetabolism), early to intermediate disease (146 genes, includ-

ng many involved in inflammation), intermediate to moderateisease (110 genes), and moderate to severe disease (650enes).

Importantly, Karra et al. (46) and Tabibiazar et al. (47)rovide additional validation to these gene expression pat-erns by showing consistent cross-species pattern compari-ons. Referring the results of their mouse experiments toarlier results in humans (45), Karra et al. (46) identified 40enes among 650 (p � 0.0001) that significantly correlatedith disease progression in both mice and humans. Analysesased on the Gene Ontology database, a bioinformatics toolesigned to facilitate clustering of genes based on functionalathways (48), bolstered the concordance between mice andumans in both studies.Overlapping genes identified in the 2 studies may provide a

eliable molecular signature for atherosclerotic disease statesTable 2). Many overlapping genes—including chemokines,hemokine receptors, and cytokine-related genes, which con-istently increased in both studies and at various stages oftherosclerotic lesion progression, and major histocompatibilityomplex molecules such as H2-Eb1 and H2-Ab1—associateith inflammation. Both studies identified increased levels ofncostatin M receptor (Osmr) in atherosclerosis. Oncostatin MOsm) belongs to a cytokine family that regulates endothelialell production of other cytokines, including interleukin-6,ranulocyte colony-stimulating factor, and granulocyte macro-hage colony-stimulating factor. Also Osm induces Abca1 in
epG2 cells (49) and Mmp3 and Timp3 gene expression via

JsOs(g

eoepfiVoediwsgp

gupitseraf

nmYitntor

cfceaasdthplitWin(n

D

G


anus kinase/signal transducers and activators of transcriptionignaling (50). Both studies observed that increased levels ofsmr, Jak3, and Abca1 associate with atherosclerosis progres-

ion. Both studies also detected increased levels of osteopontinSpp1), a component of cell-mediated immunity (51), and aene intensively studied in relation to atherosclerosis.

Such unbiased transcriptional profiling approaches discov-red genes such as Runx2, an Spp1-related transcript, previ-usly not implicated in atherosclerosis. In addition, Tabibiazart al. (47) identified several other Spp1-related transcripts notreviously associated with atherosclerosis. Both studies identi-ed genes previously associated with atherosclerosis such ascam1 and Timp1. Additionally, each study identified severalther novel genes, and some appeared in both studies. Forxample, Tabibiazar et al. (47) and Karra et al. (46) bothetermined that the differential regulation of Xin, a genenvolved in cardiac and skeletal muscle development, associatesith atherosclerosis and arterial repair, respectively. Both

tudies also determined that differential regulation of Sgcg, aene strongly expressed in skeletal and heart muscle as well asroliferating myoblasts, associates with atherosclerosis.Transcriptional profiling is especially advantageous in

enomic studies designed to detect an environmental stim-lus. This extension of the concept to cell culture hasermitted investigation of the influence of vascular dynam-cs on transcripts related to atherosclerosis. Studying cul-ured human vascular endothelial cells exposed to varyinghear stress, oxygen gradient, and oxidized LDL, Warabit al. (52) showed differential expression of several Nrf2-elated genes in response to laminar flow, underscoring thedvantage of precise environmental regulation in cell culture

ifferences in Gene Expression Associated With Atherosclerosis

Table 2 Differences in Gene Expression Associated With Ather

Class Symbol Gene

Inflammation Abca1 ATP-binding cassette subfamily A

Ccl9 Chemokine (C-C motif) ligand 9

CCr2 Chemokine (C-C motif) receptor

CCr5 Chemokine (C-C motif) receptor

Cklfsf7 Chemokine-like factor superfam

Cxcl1 Chemokine (C-X-C motif) ligand 1

Cxcr4 Chemokine (C-X-C motif) recepto

IL1rn Interleukin 1 receptor antagonist

IL7r Interleukin 7 receptor

H2-Eb1 Histocompatibility 2, class II anti

H2-Ab1 Response to metastatic cancers

Jak3 Janus kinase 3

Osmr Oncostatin M receptor

Runx2 Runt-related transcription factor

Adhesion Spp1 Secreted phosphoprotein 1

Timp1 Tissue inhibitor of metalloprotein

Vcam1 Vascular cell adhesion molecule

Development Sgcg Sarcoglycan, gamma

Xin/Cmya1 Cardiomyopathy associated 1

enes identified in 2 independent expression studies (46,47).ATP � adenosine triphosphate.

or transcriptional profiling. g

Detection of transcripts in circulating cells offers conve-ient clinical application of transcriptional profiling, butay not reflect gene expression in atherosclerotic lesions.et one such strategy may nonetheless offer novel insight

nto the mechanisms of coronary thrombosis. Transcrip-ional profiling of platelets from patients with acute coro-ary syndromes offers the ability to examine gene transcrip-ion that occurred in megakaryocytes some weeks before thenset of symptoms. This approach has identified myeloid-elated protein-14 as a novel marker of coronary risk (53).

How do transcriptional profiling experiments impactardiovascular therapeutics? Validated markers that emergerom gene expression profiling might improve risk stratifi-ation and personalization of treatment decisions. Genexpression profiling may aid identification of drug targetsnd validation of candidate drugs for the management oftherosclerosis. As proof of principle, Tuomisto et al. (54)howed that HMG-CoA reductase (the target of statinrugs) increased in atherosclerotic plaque. Using laser cap-ure microdissection to isolate macrophage-rich tissue fromuman atherosclerotic lesions, they compared expressionrofiles to normal intimal tissue and THP-1 macrophage-

ike cells and identified augmented expression of 72 genesncluding HMG-CoA reductase. Studying the transcrip-ional effect of statin exposure on peripheral monocytes,

aehre et al. (55) provided evidence showing that statinsnhibit expression of inflammatory cytokine interleukin-1�,ormally present at high levels among subjects with CAD55). Such findings show the potential of genomic tech-iques to identify additional drug targets and provide

rosis

Function

ber 1 Cell growth and/or maintenance, lipid metabolism

Signal transduction, immune response

Signal transduction

Signal transduction

Response to external stimulus

Response to external stimulus, cell growth and/or maintenance

Signal transduction

Signal transduction, immune response, lipid metabolism

Signal transduction

beta Response to biotic stimulus, immune response

Response to biotic stimulus, immune response

Signal transduction, organogenesis

Signal transduction

Signal transduction, cell development

Cell adhesion, organogenesis

Cell adhesion

Cell adhesion

Cell growth and/or maintenance

Cell growth and/or maintenance

oscle

mem

2

5

ily 7

r 4

gen E

2

2

ase 1

1

reater understanding of their effects.

a(spsrcvTtvct

PP

EstcpmtgiiLert

cecpcbidslctcaplcplg

re

iac(tpts

yhppDaahtto

spmdaficrcraim

spdimcitpavcmmga

S

G


In another study based on expression profiling, CRPugmented the expression of matrix metalloproteinaseMMP)-1 and MMP-10 in human endothelial cell tran-criptional profiling experiments (56). The MMPs arearticipants in plaque disruption and thrombosis. In anothertudy, purified CRP caused increased expression of geneselated to programmed cell death such as GADD153 inultured vascular smooth muscle cells (57). Apoptosis ofascular smooth muscle cells occurs in atheromata (58).hus, expression profiling described a molecular link be-

ween inflammation and atherosclerotic lesions and pro-ided a means for testing the effect of statins or otherompounds on apoptosis-associated genes, pointing to novelherapeutic approaches to atherosclerosis.

roteomic and Metabolomicrofiling in Atherosclerosis

ven if replicated, association or transcriptional profilingtudies cannot assess potentially important post-transcrip-ional variables including alternative splicing of mRNA,ontrol subjects on protein translation, and post-translationalrocessing of proteins (Fig. 2). Protein markers may provideore accurate real-time information about pathophysiology

han stable germ-line markers such as SNPs. As withenomics, potential benefits to the clinical communitynclude better tools for diagnosis (cardiac biomarkers) anddentification of therapeutic targets. The National Heart,ung, and Blood Institute has supported this effort bystablishing several Proteomics Centers (59), and severalecent reports highlight progress and further challenges inhis area of cardiovascular research.

Challenges in the application of proteomics to cardiovas-ular disease begin with the selection of tissue samples. Forxample, one study that sampled endarterectomy sectionsontaining atherosclerotic plaque determined decreased ex-ression of heat shock protein-27 (HSP27) in plaqueompared with healthy tissue, and confirmed these resultsy showing a similar trend for the amount of soluble HSP27n plasma of subjects with atherosclerotic cardiovascularisease (60). Additionally, recent evidence suggests a pos-ible connection between HSP27 and atherosclerosis at theevel of estrogen signaling (61). Others have pointed toonfounding in studies of atherosclerotic plaque because ofhe heterogeneity of cell types (62). Yet breaking down theomponents of a plaque without inducing experimentalrtifact also presents challenges. Although circulatinglasma may not reflect all aspects of an atherosclerotic

esion, it offers accessibility and reproducibility of sampleollection. Given the overwhelming abundance of certainroteins such as albumin in blood, ferreting out changes in

evels of much less abundant proteins such as cytokines androwth factors presents a formidable technical challenge.

As in the case of genomic markers, proteomic studiesequire rigorous validation of technology platforms and
xperimental results. Providing a foundation for study f
nterpretation, the Human Proteome Organization initiatedPlasma Proteome Project. The pilot phase identified 345

ardiovascular disease-related proteins in human plasma63,64). These catalogs determine additional novel proteinshat might associate with cardiovascular disease in futureroteomic discovery experiments. Such databases acceleratehe identification of unknown markers present in athero-clerotic cardiovascular disease.

Proteomic studies on isolated plasma lipid fractions haveielded new insight into the composition of LDL andigh-density lipoprotein particles, identifying 3 proteinsreviously not associated with LDL and 2 proteins notreviously associated with high-density lipoprotein (65,66).avidsson et al. (67) described unique patterns of LDL-

ssociated apolipoproteins in subjects with type 2 diabetesnd subclinical peripheral atherosclerosis compared withealthy control subjects, suggesting that particular distribu-ions of LDL-associated apolipoproteins in subjects withype 2 diabetes could contribute to the increased incidencef atherosclerotic cardiovascular disease in that group.Metabolomics seeks to quantify small molecules that

erve as physiological indicators within circulating plasma orarticular cells and tissues. The potential list of smallolecules runs into the thousands and includes carbohy-

rates, peptides, lipids, and metabolic intermediates such asmino acids, organic acids, and drug metabolites. Quanti-cation is generally based on various methods of spectros-opy or chromatography, techniques that could provideapid high-throughput results at relatively low cost inlinical practice. Noninvasive sampling of plasma to identifyeal-time markers of CAD has been reported (68), but clinicalpplications of this technology must first overcome technolog-cal and statistical limitations in simultaneously detecting

yriad compounds across a broad concentration range.As proof of principle, Sabatine et al. (69) used mass

pectrometry-based technology to identify differences inlasma metabolites among 18 subjects with ischemia in-uced by exercise stress testing compared with nonischemicndividuals who also exercised; specifically, changes in 6

etabolites, including citric acid, accurately differentiatedases from control subjects (p � 0.0001). Although requir-ng further study, combining small molecule profiling withhe arsenal of genomic and proteomic technologies shouldrovide additional diagnostic information and improve ourbility to identify new therapeutic targets. Identifying rele-ant proteomic and metabolomic markers will benefit fromomparison with genetic and genomic data. Future experi-ents should compare fluctuations of cardiovascular bio-arkers with other epidemiologic factors such as age,

ender, ethnicity, and a variety of environmental exposureslready recognized to influence disease risk and outcome.

cientific Community Genomic Resources

enetic and genomic studies of atherosclerosis are moving
orward rapidly thanks to sharing of resources such as the

H(sshciggstpgtmPep

G

Igsaitspittm

fcgdncparb

CG

GCsgmertg

cwmtctrtp

wmgrfaCfQti


uman Genome Project and The International HapMapi.e., haplotype map) Project. HapMap recently completedystematic genotyping of 3 million SNPs in each of 269ubjects representing 4 different ethnic groups, offeringigh-resolution information about inheritance patterns ofommon SNP alleles throughout the genome (70). Suchnformation facilitates development of high-throughputenotyping technologies allowing researchers to performenome-wide association studies at ever lower costs perample. Detailed genotype information in HapMap facili-ates fine mapping of areas identified in linkage studies androvides a valuable community resource for patterns ofenetic variation in diverse ethnic groups, thus simplifyinghe statistical analysis of genetic association studies. Com-unity efforts such as the MicroArray Quality Controlroject aim to improve reliability and validity of genexpression data (71), and the same will be necessary forroteomics and metabolomics.

enomic Data Integration

ntegrating various forms of genomic data, so-calledenomic convergence, would yield a whole greater than theum of its parts. Overlapping signals from linkage analysis,ssociation studies, and expression profiling would facilitatedentification of relevant markers. Identifying markershrough multiple experimental approaches reduces bias andtrengthens validity. Such nonbiased approaches are ofarticular interest in the discovery of genomic markers thatmpact complex traits, such as atherosclerosis. More impor-antly to clinical translation, cross-referencing with pro-eomic and metabolomic profiles could identify markersore easily measured in a clinical setting.Comparisons of such large data sets are biased toward

alse-positive results. Statistical methods for meeting thishallenge are manifold and reviewed elsewhere (72,73). Ineneral terms, approaches to large genomic data sets includeata normalization, filtering, and correction algorithms for theumerous significant p values generated by thousands ofomparisons. At best, these measures will help refine the list ofrime candidates for further replication. Identifying clinicallypplicable genomic, proteomic, and metabolomic markersequires consistent replication across populations, a challengeest addressed through collaborative research efforts.

linical Applications ofenomics to Cardiovascular Medicine

enomic-based cardiovascular risk prediction models.ardiovascular genomic, proteomic, and metabolomic re-

earch will continue to identify molecular markers—enotypes, RNA expression levels, proteomic andetabolomic markers—associated with cardiovascular dis-

ase, thus revealing new therapeutic targets. In atheroscle-osis, many factors influence disease predisposition andherapeutic response, providing an appropriate testing
round for a personalized medicine approach based on o
onvergence of information. Future risk assessment modelsill likely incorporate a patient’s genomic, proteomic,etabolomic, and environmental information, using statis-

ical models to identify marker-disease associations andorrect for confounders such as gene-environment interac-ions (i.e., patients with a particular marker will have highisk only if exposed to a particular stimulus). Avoidance ofhat stimulus offers a more tractable intervention for bothatient and physician.Data networks that combine various forms of information

ill greatly assist this process by identifying, for example,etabolomic markers that vary in response to a patient’s

enetic makeup and environmental exposures. Decision algo-ithms will harness information generated by medical andamily histories, clinical criteria such as the Framingham Score,nd multiple genetic, genomic, and biomarker tests (Fig. 3).linicians might wonder why we are not yet testing patients

or genetic markers such as ALOX5AP or TNFSF4 SNPs.uite simply, our current knowledge of the risk attributable to

hese variants is applicable only to a population, not tondividuals. Providing patient-specific risk information based

Figure 3 Coordination of Informationfor Clinical Decision-Making

Informed clinical decision-making based on genomic information will requiresynthesis of diverse pieces of data. The relative significance of genomic, pro-teomic, and metabolomic markers could vary among patients. Such informationmust be considered in the context of personal medical history, family history,other clinical and laboratory data, and environmental variables. ECG � electro-cardiography. Illustration by Rob Flewell.

n genomic markers awaits a more comprehensive understand-

ib

G

Nnt2mcftep1tv

dmAlPpetMAc

S

Rtdcmpcacp

sCwattripra

a

cSwmdibtfpiedpd

tbCbotpmwa

RWcPPr

R

1


ng of how these variants interact with each patient’s geneticackground and other risk factors.

enomics for Identifying Therapeutic Targets

ewer genome-wide approaches may uncover additionalovel markers of disease risk that represent new drugargets. The human genome likely contains more than2,000 genes (see the preceding text), of which currentedications target only �500 (74). Genomic approaches

an identify novel drug targets. Unfortunately, the pathrom potential target discovery to development of usefulherapeutics presents considerable challenges. Microarrayxperiments might identify multiple potential targets, butrioritization of that list presents a challenge. Given a list of00 or 1,000 candidate genes, researchers likely will focus onhe ones they recognize; however, this bias could overlookiable targets.

However, progress in CAD gene identification drivesevelopment of potential therapeutics. For example, treat-ent of 191 subjects who carried at-risk variants in theLOX5AP or LTA4H genes with a FLAP inhibitor reduced

evels of LTB4, a biomarker associated with MI risk (75),CSK9 inhibitors seem a ripe target for adjunct therapy inatients in whom LDL targets are not achieved on statinsither as monotherapy or combined with cholesterol absorp-ion inhibitors, or who cannot tolerate high-dose statins.

oreover, studies of LTA4H and PCSK9 among Africanmericans may increase our understanding of ethnicity’s

ontribution to CAD.

ummary for Clinicians

apid advances in genomic, proteomic, and metabolomicechnology provide researchers with valuable tools for un-erstanding the genetic predisposition to atheroscleroticardiovascular disease. Genetic, proteomic, and metabolo-ic markers will add diagnostic accuracy and perhaps

redict the onset and severity of atherosclerotic cardiovas-ular disease. Early results from genetic association studiesnd genomic profiling of gene expression provide proof ofoncept for consistent and reproducibly associated predis-osition of genomic markers to cardiovascular disease.Clinical translation will require validation in prospective

tudies that encompass large, ethnically diverse cohorts.ombining classical epidemiology with modern genomicsill yield unprecedented insight into mechanisms of disease,

nd also will generate ever more risk markers. Performingests on patients is relatively easy. Despite test sophistica-ion, high-throughput automation will allow clinical labo-atories to offer testing at costs similar to most diagnosticmaging studies. Interpreting such test results will trulyrovide a challenge. Indeed, prospective validation of newisk markers likely will limit the rate of progress more thanny technical or experimental factors.

Genetic, genomic, proteomic, and metabolomic tests will
dd to rather than replace the clinical utility of traditional
1

ardiovascular risk markers. Elements of the Framinghamcore will remain relevant. However, newer informationill temper and refine the relative importance of traditionalarkers. Framingham Scores do not include family history

ata, leaving room to incorporate genetic and genomicnformation into the screening process. Because DNA-ased genetic markers remain static throughout life, predic-ive genetic tests could be performed at very young ages toacilitate early intervention. On the other hand, gene ex-ression profiling and serum biomarkers provide real-timenformation that integrates current environmental influ-nces, and likely aid diagnostic categorization and the earlyetection of disease. Again, a coordinated screening ap-roach that maximizes the benefits of each screening mo-ality will provide the most clinical utility.Finally, genomic approaches have already identified po-

ential drug targets. Inhibitors of FLAP, OX40L-OX40inding, MMP, and PCSK9 are all attractive candidates forAD risk modifiers. Although inherent challenges ofringing any new drug to the clinic will influence the pacef introduction of therapeutic changes, genomic approacheso cardiology promise to bring lasting improvements toatient care. Clinical translation of genomic, proteomic, andetabolomic information will require collaborative effortsithin the cardiovascular disease community at both bench

nd bedside.

eprint requests and correspondence: Dr. Peter Libby, Donald. Reynolds Cardiovascular Clinical Research Center, Cardiovas-

ular Medicine, Brigham and Women’s Hospital, Mallinckrodtrofessor of Medicine, Harvard Medical School, 77 Avenue Louisasteur, NRB 7, Boston, Massachusetts 02115. E-mail: plibby@

ics.bwh.harvard.edu.

EFERENCES

1. Rosamond WD, Chambless LE, Folsom AR, et al. Trends in theincidence of myocardial infarction and in mortality due to coronaryheart disease, 1987 to 1994. N Engl J Med 1998;339:861–7.

2. Wang Q. Advances in the genetic basis of coronary artery disease. CurrAtheroscler Rep 2005;7:235–41.

3. Topol EJ. The genomic basis of myocardial infarction. J Am CollCardiol 2005;46:1456–65.

4. Watkins H, Farrall M. Genetic susceptibility to coronary arterydisease: from promise to progress. Nat Rev Genet 2006;7:163–73.

5. Tuomisto TT, Yla-Herttuala S. What have we learnt about microarrayanalyses of atherogenesis? Curr Opin Lipidol 2005;16:201–5.

6. Bijnens AP, Lutgens E, Ayoubi T, Kuiper J, Horrevoets AJ, DaemenMJ. Genome-wide expression studies of atherosclerosis: critical issuesin methodology, analysis, interpretation of transcriptomics data.Arterioscler Thromb Vasc Biol 2006;26:1226–35.

7. Tuomisto TT, Binder BR, Yla-Herttuala S. Genetics, genomics andproteomics in atherosclerosis research. Ann Med 2005;37:323–32.

8. Boerwinkle E, Ellsworth DL, Hallman DM, Biddinger A. Geneticanalysis of atherosclerosis: a research paradigm for the commonchronic diseases. Hum Mol Genet 1996;5:1405–10.

9. Hunter DJ. Gene-environment interactions in human diseases. NatRev Genet 2005;6:287–98.

0. Goldstein JL, Brown MS. The LDL receptor defect in familialhypercholesterolemia. Implications for pathogenesis and therapy. MedClin North Am 1982;66:335–62.

1. Breslow JL. Genetic differences in endothelial cells may determineatherosclerosis susceptibility. Circulation 2000;102:5–6.

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

5

5

5

5


2. Pajukanta P, Cargill M, Viitanen L, et al. Two loci on chromosomes2 and X for premature coronary heart disease identified in early- andlate-settlement populations of Finland. Am J Hum Genet 2000;67:1481–93.

3. Francke S, Manraj M, Lacquemant C, et al. A genome-wide scan forcoronary heart disease suggests in Indo-Mauritians a susceptibilitylocus on chromosome 16p13 and replicates linkage with the metabolicsyndrome on 3q27. Hum Mol Genet 2001;10:2751–65.

4. Harrap SB, Zammit KS, Wong ZY, et al. Genome-wide linkageanalysis of the acute coronary syndrome suggests a locus on chromo-some 2. Arterioscler Thromb Vasc Biol 2002;22:874–8.

5. Broeckel U, Hengstenberg C, Mayer B, et al. A comprehensive linkageanalysis for myocardial infarction and its related risk factors. NatGenet 2002;30:210–4.

6. Wang Q, Rao S, Shen GQ, et al. Premature myocardial infarctionnovel susceptibility locus on chromosome 1P34-36 identified bygenomewide linkage analysis. Am J Hum Genet 2004;74:262–71.

7. Helgadottir A, Manolescu A, Thorleifsson G, et al. The geneencoding 5-lipoxygenase activating protein confers risk of myocardialinfarction and stroke. Nat Genet 2004;36:233–9.

8. Mehrabian M, Allayee H, Wong J, et al. Identification of5-lipoxygenase as a major gene contributing to atherosclerosis suscep-tibility in mice. Circ Res 2002;91:120–6.

9. Wang X, Ria M, Kelmenson PM, et al. Positional identification ofTNFSF4, encoding OX40 ligand, as a gene that influences atheroscle-rosis susceptibility. Nat Genet 2005;37:365–72.

0. Hauser ER, Crossman DC, Granger CB, et al. A genomewide scan forearly-onset coronary artery disease in 438 families: the GENECARDStudy. Am J Hum Genet 2004;75:436–47.

1. Ria M, Eriksson P, Boquist S, Ericsson CG, Hamsten A, LagercrantzJ. Human genetic evidence that OX40 is implicated in myocardialinfarction. Biochem Biophys Res Commun 2006;339:1001–6.

2. Ridker PM, Rifai N, Rose L, Buring JE, Cook NR. Comparison ofC-reactive protein and low-density lipoprotein cholesterol levels in theprediction of first cardiovascular events. N Engl J Med 2002;347:1557–65.

3. Danesh J, Wheeler JG, Hirschfield GM, et al. C-reactive protein andother circulating markers of inflammation in the prediction of coronaryheart disease. N Engl J Med 2004;350:1387–97.

4. Carlson CS, Aldred SF, Lee PK, et al. Polymorphisms within theC-reactive protein (CRP) promoter region are associated with plasmaCRP levels. Am J Hum Genet 2005;77:64–77.

5. Miller DT, Zee RYL, Suk Danik HJ, et al. Association of commonCRP gene variants with CRP levels and cardiovascular events. AnnHum Genet 2005;69:623–38.

6. Kathiresan S, Larson MG, Vasan RS, et al. Contribution of clinicalcorrelates and 13 C-reactive protein gene polymorphisms to interin-dividual variability in serum C-reactive protein level. Circulation2006;113:1415–23.

7. Kovacs A, Green F, Hansson LO, et al. A novel common singlenucleotide polymorphism in the promoter region of the C-reactiveprotein gene associated with the plasma concentration of C-reactiveprotein. Atherosclerosis 2005;178:193–8.

8. Casas JP, Shah T, Cooper J, et al. Insight into the nature of theCRP-coronary event association using Mendelian randomization. IntJ Epidemiol 2006;35:922–31.

9. Shiffman D, Ellis SG, Rowland CM, et al. Identification of four genevariants associated with myocardial infarction. Am J Hum Genet2005;77:596–605.

0. Zee RYL, Michaud S, Hegener HH, Diehl KA, Ridker PM. Aprospective replication-study of five gene variants previously associatedwith risk of myocardial infarction. J Thromb Hemost 2006;4:2093–5.

1. Hirschhorn JN, Lohmueller K, Byrne E, Hirschhorn K. A compre-hensive review of genetic association studies. Genet Med 2002;4:45– 61.

2. Helgadottir A, Gretarsdottir S, St. Clair D, et al. Association betweenthe gene encoding 5-lipoxygenase-activating protein and stroke repli-cated in a Scottish population. Am J Hum Genet 2005;76:505–9.

3. Lohmussaar E, Gschwendtner A, Mueller JC, et al. ALOX5AP geneand the PDE4D gene in a Central European population of strokepatients. Stroke 2005;36:731–6.

4. Zee RY, Cheng S, Hegener HH, Erlich HA, Ridker PM. Genetic
variants of arachidonate 5-lipoxygenase-activating protein, and risk of
incident myocardial infarction and ischemic stroke: a nested case-control approach. Stroke 2006;37:2007–11.

5. Helgadottir A, Manolescu A, Helgason A, et al. A variant of the geneencoding leukotriene A4 hydrolase confers ethnicity-specific risk ofmyocardial infarction. Nat Genet 2006;38:68–74.

6. Tang H. Confronting ethnicity-specific disease risk. Nat Genet2006;38:13–5.

7. Cohen JC, Boerwinkle E, Mosley TH Jr., Hobbs HH. Sequencevariations in PCSK9, low LDL, and protection against coronary heartdisease. N Engl J Med 2006;354:1264–72.

8. Ozaki K, Ohnishi Y, Iida A, et al. Functional SNPs in thelymphotoxin-alpha gene that are associated with susceptibility tomyocardial infarction. Nat Genet 2002;32:650–4.

9. McCarthy JJ, Parker A, Salem R, et al. Large scale association analysisfor identification of genes underlying premature coronary heart disease:cumulative perspective from analysis of 111 candidate genes. J MedGenet 2004;41:334–41.

0. Gibbons GH, Liew CC, Goodarzi MO, et al. Genetic markers:progress and potential for cardiovascular disease. Circulation 2004;109:IV47–58.

1. Casas JP, Cooper J, Miller GJ, Hingorani AD, Humphries SE.Investigating the genetic determinants of cardiovascular disease usingcandidate genes and meta-analysis of association studies. Ann HumGenet 2006;70:145–69.

2. Herbert A, Gerry NP, McQueen MB, et al. A common geneticvariant is associated with adult and childhood obesity. Science2006;312:279 – 83.

3. Cook SA, Rosenzweig A. DNA microarrays: implications for cardio-vascular medicine. Circ Res 2002;91:559–64.

4. Ma J, Liew CC. Gene profiling identifies secreted protein transcriptsfrom peripheral blood cells in coronary artery disease. J Mol CellCardiol 2003;35:993–8.

5. Seo D, Wang T, Dressman H, et al. Gene expression phenotypes ofatherosclerosis. Arterioscler Thromb Vasc Biol 2004;24:1922–7.

6. Karra R, Vemullapalli S, Dong C, et al. Molecular evidence forarterial repair in atherosclerosis. Proc Natl Acad Sci U S A 2005;102:16789 –94.

7. Tabibiazar R, Wagner RA, Ashley EA, et al. Signature patterns ofgene expression in mouse atherosclerosis and their correlation tohuman coronary disease. Physiol Genomics 2005;22:213–26.

8. The Gene Ontology Consortium. Gene ontology: tool for the unifi-cation of biology. Nat Genet 2000;25:25–9.

9. Langmann T, Porsch-Ozcurumez M, Heimerl S, et al. Identificationof sterol-independent regulatory elements in the human ATP-bindingcassette transporter A1 promoter. Role of Sp1/3, E-box bindingfactors, and an oncostatin M-responsive element. J Biol Chem2002;277:14443–50.

0. Li WQ, Dehnade F, Zafarullah M. Oncostatin M-induced matrixmetalloproteinase and tissue inhibitor of metalloproteinase-3 genesexpression in chondrocytes requires Janus kinase/STAT signalingpathway. J Immunol 2001;166:3491–8.

1. Ashkar S, Weber GF, Panoutsakopoulou V, et al. Eta-1 (osteopontin):an early component of type-1 (cell-mediated) immunity. Science2000;287:860–4.

2. Warabi E, Wada Y, Kajiwara H, et al. Effect on endothelial cell geneexpression of shear stress, oxygen concentration, and low-densitylipoprotein as studied by a novel flow cell culture system. Free RadicBiol Med 2004;37:682–94.

3. Healy AM, Pickard MD, Pradhan AD, et al. Platelet expressionprofiling and clinical validation of myeloid-resistant protein-14 as anovel determinant of cardiovascular events. Circulation 2006;113:2278–84.

4. Tuomisto TT, Korkeela A, Rutanen J, et al. Gene expression inmacrophage-rich inflammatory cell infiltrates in human atheroscleroticlesions as studied by laser microdissection and DNA array: overexpres-sion of HMG-CoA reductase, colony stimulating factor receptors,CD11A/CD18 integrins, and interleukin receptors. ArteriosclerThromb Vasc Biol 2003;23:2235–40.

5. Waehre T, Yndestad A, Smith C, et al. Increased expression ofinterleukin-1 in coronary artery disease with downregulatory effects ofHMG-CoA reductase inhibitors. Circulation 2004;109:1966–72.

6. Montero I, Orbe J, Varo N, et al. C-reactive protein induces matrix
metalloproteinase-1 and -10 in human endothelial cells: implications

5

5

5

6

6

6

6

6

6

6

6

6

6

7

7

7

7

7

7


for clinical and subclinical atherosclerosis. J Am Coll Cardiol 2006;47:1369–78.

7. Blaschke F, Bruemmer D, Yin F, et al. C-reactive protein inducesapoptosis in human coronary vascular smooth muscle cells. Circulation2004;110:579–87.

8. Geng YJ, Libby P. Progression of atheroma: a struggle between deathand procreation. Arterioscler Thromb Vasc Biol 2002;22:1370–80.

9. Granger CB, Van Eyk JE, Mockrin SC, Anderson NL, on behalf ofthe Working Group Members. National Heart, Lung, and BloodInstitute Clinical Proteomics Working Group Report. Circulation2004;109:1697–703.

0. Martin-Ventura JL, Duran MC, Blanco-Colio LM, et al. Identifica-tion by a differential proteomic approach of heat shock protein 27 as apotential marker of atherosclerosis. Circulation 2004;110:2216–9.

1. Miller H, Poon S, Hibbert B, Rayner K, Chen Y-X, O’Brien ER.Modulation of estrogen signaling by the novel interaction of heatshock protein 27, a biomarker for atherosclerosis, and estrogenreceptor �: mechanistic insight into the vascular effects of estrogens.Arterioscler Thromb Vasc Biol 2005;25:e10–4.

2. Duran MC, Mas S, Martin-Ventura JL, et al. Proteomic analysis ofhuman vessels: application to atherosclerotic plaques. Proteomics2003;3:973–8.

3. Berhane BT, Zong C, Liem DA, et al. Cardiovascular-related proteinsidentified in human plasma by the HUPO Plasma Proteome Projectpilot phase. Proteomics 2005;5:3520–30.

4. Ping P, Vondriska TM, Creighton CJ, et al. A functional annotationof subproteomes in human plasma. Proteomics 2005;5:3506–19.

5. Karlsson H, Leanderson P, Tagesson C, Lindahl M. LipoproteomicsI: mapping of proteins in low-density lipoprotein using two-dimensional gel electrophoresis and mass spectrometry. Proteomics
2005;5:551–65.
6. Karlsson H, Leanderson P, Tagesson C, Lindahl M. LipoproteomicsII: mapping of proteins in high-density lipoprotein using two-dimensional gel electrophoresis and mass spectrometry. Proteomics2005;5:1431–45.

7. Davidsson P, Hulthe J, Fagerberg B, et al. A proteomic study of theapolipoproteins in LDL subclasses in patients with the metabolicsyndrome and type 2 diabetes. J Lipid Res 2005;46:1999–2006.

8. Brindle JT, Antti H, Holmes E, et al. Rapid and noninvasive diagnosisof the presence and severity of coronary heart disease using 1H-NMR-based metabonomics. Nat Med 2002;8:1439–44.

9. Sabatine MS, Liu E, Morrow DA, et al. Metabolomic identificationof novel biomarkers of myocardial ischemia. Circulation 2005;112:3868 –75.

0. Altshuler D, Brooks LD, Chakravarti A, Collins FS, Daly MJ,Donnelly P. A haplotype map of the human genome. Nature 2005;437:1299–320.

1. Draghici S, Khatri P, Eklund AC, Szallasi Z. Reliability and repro-ducibility issues in DNA microarray measurements. Trends Genet2006;22:101–9.

2. Ehm MG, Nelson MR, Spurr NK. Guidelines for conducting andreporting whole genome/large-scale association studies. Hum MolGenet 2005;14:2485–8.

3. Allison DB, Cui X, Page GP, Sabripour M. Microarray data analysis:from disarray to consolidation and consensus. Nat Rev Genet 2006;7:55–65.

4. Drews J. Drug discovery: a historical perspective. Science 2000;287:1960–4.

5. Hakonarson H, Thorvaldsson S, Helgadottir A, et al. Effects of a5-lipoxygenase-activating protein inhibitor on biomarkers associatedwith risk of myocardial infarction: a randomized trial. JAMA
2005;293:2245–56.

Documents

Atherosclerosis - Journal of the American College of Cardiology · Atherosclerosis The Path From Genomics to Therapeutics David T. Miller, MD, PHD,*§ Paul M Ridker, MD, MPH, FACC,†‡§