doi:10.1152/physiolgenomics.00015.2007 32:45-57, 2007. First published 18 September 2007;Physiol. Genomics

Ronkainen, Irina Alafuzoff, Helena Kuivaniemi and Gerard TrompShantel Weinsheimer, Guy M. Lenk, Monique van der Voet, Susan Land, Anttianeurysmdata to identify candidate genes in intracranial Integration of expression profiles and genetic mapping

You might find this additional info useful...

for this article can be found at:Supplemental materialhttp://physiolgenomics.physiology.org/content/suppl/2007/10/09/00015.2007.DC1.html

33 articles, 15 of which can be accessed free at:This article cites http://physiolgenomics.physiology.org/content/32/1/45.full.html#ref-list-1

3 other HighWire hosted articlesThis article has been cited by

  [PDF] [Full Text] [Abstract]

, September 9, 2008; 29 (8): 1566-1569.AJNRR. Kadirvel, Y.-H. Ding, D. Dai, D.A. Lewis, S. Raghavakaimal, H.J. Cloft and D.F. KallmesAcid MicroarraysGene Expression Profiling of Experimental Saccular Aneurysms Using Deoxyribonucleic 

[PDF] [Full Text] [Abstract], January 25, 2010; 41 (2): 224-231.Stroke

and Andrzej SzczudlikTomasz Dziedzic, Marcin Piechota, Krzysztof Stachura, Marek Moskala, Ryszard Przewlocki Joanna Pera, Michal Korostynski, Tadeusz Krzyszkowski, Jacek Czopek, Agnieszka Slowik,What Is the Role of Inflammation?Gene Expression Profiles in Human Ruptured and Unruptured Intracranial Aneurysms: 

[PDF] [Full Text] [Abstract], May 24, 2010; 41 (6): 1138-1144.Stroke

Miyamoto and Akio KoizumiNanayakkara, Norio Matsuura, Toshiaki Hitomi, Yasushi Takagi, Nobuo Hashimoto, Susumu Hirokuni Hashikata, Wanyang Liu, Kayoko Inoue, Yohei Mineharu, Shigeki Yamada, ShanikaWith Familial and Sporadic Intracranial Aneurysms in Japanese PatientsConfirmation of an Association of Single-Nucleotide Polymorphism rs1333040 on 9p21

including high resolution figures, can be found at:Updated information and services http://physiolgenomics.physiology.org/content/32/1/45.full.html

can be found at:Physiological Genomicsabout Additional material and information http://www.the-aps.org/publications/pg

This infomation is current as of May 21, 2011. 

the American Physiological Society. ISSN: 1094-8341, ESSN: 1531-2267. Visit our website at http://www.the-aps.org/.July, and October by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 2007 bytechniques linking genes and pathways to physiology, from prokaryotes to eukaryotes. It is published quarterly in January, April,

publishes results of a wide variety of studies from human and from informative model systems withPhysiological Genomics

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

Integration of expression profiles and genetic mapping data to identifycandidate genes in intracranial aneurysm

Shantel Weinsheimer,1 Guy M. Lenk,1 Monique van der Voet,1 Susan Land,1,2,5 Antti Ronkainen,6

Irina Alafuzoff,7 Helena Kuivaniemi,1,3 and Gerard Tromp1,4

1Center for Molecular Medicine and Genetics, Departments of 2Obstetrics and Gynecology, 3Surgery, and 4Neurology,Wayne State University School of Medicine, Detroit; 5Applied Genomics Technology Center, Wayne State University,Detroit, Michigan; and Departments of 6Neurosurgery and 7Clinical Medicine, Neurology, and Pathology, KuopioUniversity, Kuopio, Finland

Submitted 12 January 2007; accepted in final form 12 September 2007

Weinsheimer S, Lenk GM, van der Voet M, Land S, RonkainenA, Alafuzoff I, Kuivaniemi H, Tromp G. Integration of expressionprofiles and genetic mapping data to identify candidate genes in intra-cranial aneurysm. Physiol Genomics 32: 45–57, 2007. First publishedSeptember 18, 2007; doi:10.1152/physiolgenomics.00015.2007.—Intra-cranial aneurysm (IA) is a complex genetic disease for which, to date,10 loci have been identified by linkage. Identification of the risk-conferring genes in the loci has proven difficult, since the regionsoften contain several hundreds of genes. An approach to prioritizepositional candidate genes for further studies is to use gene expressiondata from diseased and nondiseased tissue. Genes that are notexpressed, either in diseased or nondiseased tissue, are ranked asunlikely to contribute to the disease. We demonstrate an approachfor integrating expression and genetic mapping data to identifylikely pathways involved in the pathogenesis of a disease. We usedexpression profiles for IAs and nonaneurysmal intracranial arteries(IVs) together with the 10 reported linkage intervals for IA.Expressed genes were analyzed for membership in Kyoto Ency-clopedia of Genes and Genomes (KEGG) biological pathways. The10 IA loci harbor 1,858 candidate genes, of which 1,561 (84%)were represented on the microarrays. We identified 810 positionalcandidate genes for IA that were expressed in IVs or IAs. Pathwayinformation was available for 294 of these genes and involved 32KEGG biological function pathways represented on at least 2 loci.A likelihood-based score was calculated to rank pathways forinvolvement in the pathogenesis of IA. Adherens junction, MAPK,and Notch signaling pathways ranked high. Integration of geneexpression profiles with genetic mapping data for IA provides anapproach to identify candidate genes that are more likely tofunction in the pathology of IA.

pathway analysis; adherens junction; Notch signaling; MAPK signaling

EACH YEAR, ABOUT 27,000 INDIVIDUALS in the United Statessuffer a rupture of an intracranial aneurysm (IA) (24). IA isa complex disease with both environmental and geneticcomponents, and first-degree relatives of patients with IAhave a higher risk of developing an aneurysm (23, 26).Several approaches have been used to identify factors thatcontribute to a genetic predisposition to IA including DNAlinkage and genetic association studies.

DNA linkage analyses suggest that genetic factors on chro-mosomes 1, 2, 5, 7, 11, 14, 17, 19, and X contribute to IA

(Table 1). Onda et al. (18) found three loci with suggestivelinkage for IA on chromosomes 5q22–q31, 14q22, and 7q11 ina genome-wide linkage study in the Japanese population,including intracranial berry aneurysm-1, known as the ANIB1locus in Online Mendelian Inheritance in Man on chromosome7q11.2. Olson et al. (17) detected two regions of suggestivelinkage: one on chromosome 19q13 and the other on the Xchromosome (17). A follow-up study (28) of the 19q13 locususing additional Finnish samples and covariate affected-rela-tive-pair, as well as model-based, linkage analyses indicatedthat the most likely location for a gene predisposing to IAmapped to a region at chromosome 19q13.3 that spans 6.6 cM,designated the ANIB2 locus. More recently, another genome-wide linkage analysis of 29 Japanese families revealed linkage to3 chromosomal regions including chromosomes 17cen, Xp22, and19q13 (32). A follow-up study of nine of these Japanese IAfamilies confirmed chromosome 19q13.3 as a susceptibility locusfor IA, which overlaps with the Finnish susceptibility locus for IA(15). Farnham et al. (7) studied 13 Utah families with IA andconfirmed linkage to chromosome 7q11 (7).

Four genome-wide linkage studies were performed in singlelarge pedigrees. A Dutch consanguineous family was used tomap a susceptibility locus for IA to chromosome 2p13 (20),although it was subsequently reported that the locus was anartefact because of misclassification of several individuals(22). One North American kindred showed significant linkageto chromosome 1p36.13–p34.3, known as the ANIB3 locus,under a dominant model with high penetrance, suggesting aMendelian form of IA (16). Ozturk et al. (19) performed singlenucleotide polymorphism genotyping for genome-wide linkageanalysis in two kindreds, a Colombian and a North Americanfamily. These analyses provide additional support for linkageto chromosomes 11q and 14q, as previously reported in Japa-nese IA-affected sib pairs. Recently, another susceptibilitylocus, ANIB4, mapping to chromosome 5p15.2–p14.3 wasidentified in a French Canadian family with 12 affected indi-viduals (29).

Identification of multiple susceptibility loci suggests geneticheterogeneity for IA; however, it can reasonably be hypothe-sized that there are a limited number of pathways that contrib-ute to the phenotype and that genetic (locus) heterogeneity cantherefore be utilized to identify the pathway(s). Another hy-pothesis is that genes responsible for diseases that affectparticular tissues are likely to be expressed in that tissue, i.e.,genes not expressed, either in diseased or nondiseased tissue,may be excluded from the candidate list. Microarray technol-ogy offers a high-throughput approach to determine the ex-

Article published online before print. See web site for date of publication(http://physiolgenomics.physiology.org).

Address for reprint requests and other correspondence: G. Tromp, Center forMolecular Medicine and Genetics, Wayne State Univ. School of Medicine,3309 Gordon H. Scott Hall of Basic Medical Sciences, 540 East Canfield Ave.,Detroit, MI 48201 (e-mail: gerard.tromp@sanger.med.wayne.edu).

Physiol Genomics 32: 45–57, 2007.First published September 18, 2007; doi:10.1152/physiolgenomics.00015.2007.

1094-8341/07 $8.00 Copyright © 2007 the American Physiological Society 45

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

pression of thousands of positional candidate genes in diseasedand nondiseased tissues. Integration of gene expression infor-mation and pathway information together with genetic map-ping data will help to prioritize positional candidate genes. Weused microarrays to analyze gene expression in human IAsand contralateral nonaneurysmal intracranial arteries (IVs)as well as intracranial arteries from individuals without IAand integrated this information with published genetic map-ping data to identify positional and functional candidategenes for IA.

EXPERIMENTAL PROCEDURES

Tissue samples, RNA isolation, and microarray analysis. Intracra-nial arteries from a total of 12 Caucasian individuals with IA wereobtained postmortem from autopsies performed within 24 h of deathat the University of Kuopio, Finland. Specimens were collected fromruptured or unruptured IAs and from the contralateral vessel of thesame patients to represent IVs (Table 2). The causes of death for thefour patients with unruptured aneurysms were as follows: accidentaldeath due to trauma, alcoholism, heart disease, and hypertensive heartdisease. The arteries were dissected, placed in RNAlater solution(Ambion, Austin, Texas), and stored frozen at �80°C (Fig. 1).Homogenized IAs or nonaneurysmal IVs from these same patientswere pooled into groups of three to five individuals with similarclinical characteristics including artery type, rupture status (i.e., rup-tured IA or unruptured IA), and sex (Table 2). Additional controlsincluded IVs from non-IA individuals of African American decentthat were obtained postmortem from autopsies performed within 24 hof death (in Detroit, MI) (1). Larger segments were available (n � 5samples from 4 different individuals), and they were analyzed indi-vidually (Table 2).

Total RNA was isolated from blood vessels as previously described(27). IAs and IVs from individuals with IA were analyzed usingAffymetrix microarrays before commercial availability of the Il-lumina platform. Total RNA (300 ng) was amplified one roundusing the MessageAmp II aRNA Kit according to the manual(version 0404; Ambion). Amplified RNA (10 �g) was reversetranscribed into cDNA and then in vitro transcribed and biotinlabeled to produce cRNA, which was subsequently hybridized to

Affymetrix HG_U133 Plus 2.0 microarrays according to the Af-fymetrix GeneChip Expression Analysis Technical Manual (Af-fymetrix, Santa Clara, CA). The arrays were scanned using theGeneArray scanner (Affymetrix). Image analysis was performedwith MicroArray Suite 5.0 software. Affymetrix HG_U133 Plus2.0 microarrays contain 54,675 probes that represent 19,221 dis-tinct and nonpseudo- genes. Affymetrix data were analyzed withthe use of GeneSpring version 7.2 (Silicon Genetics), Bioconduc-tor, and R (version 2.3.0), an open source language and environ-ment for statistical computing and graphics (8, 9), and the datawere preprocessed by applying a sequence-specific expressionmodel known as gcrma (robust multiarray analysis with adjustmentfor GC content of probes), which is a perfect-match probe method(31). Adjusted microarray data were then subjected to an expres-sion filter based on adjusted probe intensity levels to determine theset of genes expressed in either IAs or IVs. A gcrma expressionvalue �3.91 was chosen as the threshold, resulting in the samenumber of genes that were identified in the Illumina gene expres-sion experiments when the detection level was set to �0.99 (seebelow).

On availability of the Sentrix Human-6 Whole Genome Expres-sion BeadChips (Sentrix Human WG-6; Illumina, San Diego, CA),we evaluated gene expression in control arteries (IVs) from indi-viduals without IA as follows. Total RNA (150 ng) extracted fromarteries was reverse transcribed and cRNA prepared, and 1.5 �g ofbiotinylated cRNA were hybridized to Sentrix Human WG-6arrays, washed, and scanned on the BeadArray Reader, as de-scribed by Illumina at http://www.illumina.com. Illumina Bead-Chips contain 47,293 gene targets, representing 18,025 distinctRefSeq genes that are not pseudogenes. Gene expression data wereanalyzed using BeadStudio version 1.5.0.34 software (Illumina).Array data were fit to a cubic spline in BeadStudio, and theexpressed gene targets were identified using a detection level�0.99. We used two different microarray platforms for geneexpression analysis for two main reasons: 1) the Illumina Human-6Whole Genome Expression BeadChips became available only afterwe had performed the initial Affymetrix microarray analysis forthe samples from individuals with IA, and 2) the Illumina platformrequires less RNA and offers an independent gene expressionmeasurement to validate the results of the Affymetrix analyses. We

Table 1. Distribution of expressed genes across IA loci

Genetic Locus Study Population Flanking Marker

Genes, no.

Total Probedb (%) Expressedc (%)d

1p34.3–p36.13 North American (16) D1S199–D1S496 249 221 (88.8) 124 (56.1; 24.9)2p13e Dutch (20) D2S2206–D2S2977 79 65 (82.3) 44 (67.7; 33.8)5p15.2–p14.3 French Canadian (29) D5S2095–D5S2031 57 28 (49.1) 14 (50.0; 28.6)5q22–q31 Japanese (18) D5S471–D5S2010 225 193 (85.8) 102 (52.8; 22.8)7q11 Japanese (18) D7S2415–D7S657 248 164 (66.1) 89 (54.3; 21.3)7q11 North American (7) D7S2415–D7S2421 195 117 (60.0) 62 (53.0; 21.4)11q24–q25 Colombian (19) rs618176–rs1940033 35 32 (91.4) 14 (43.8; 9.4)14q22, q23–31 Japanese (18) D14S258–D14S74 93 87 (93.5) 47 (54.0; 25.3)14q22, q23–31 North American (19) rs2359991–rs2373098 44 39 (88.6) 20 (51.3; 28.2)17cen Japanese (32) D17S921–D17S1800 192 151 (78.6) 83 (55.0; 17.9)19q13 Finnish (17, 28) D19S587–D19S601 572 525 (91.8) 249 (47.4; 13.1)19q13 Japanese (15) D19S198–D19S596 216 200 (92.6) 95 (47.5; 15.5)19q13 Japanese (15) D19S198–D19S596 216 200 (92.6) 95 (47.5; 15.5)Xp22 Finnish (17) DXS1091–DXS9770a 92 83 (90.2) 38 (45.8; 21.7)Xp22 Japanese (32) DXS987–DXS7593 50 47 (94.0) 19 (40.4; 21.3)

aFlanking markers for 10-Mbp region centered at DXS987. bProbed for on either Affymetrix or Illumina microarray platform; nos. in parentheses represent%total genes in interval. cThe %probed genes in interval. dSecond percentage listed indicates %expression of class 5 genes. eLocus recently shown to be anartefact (22). There are 1,858 candidate genes for intracranial aneurysm (IA) that map to 10 distinct linkage regions across the genome, as 4 regions have beenconfirmed in separate studies. Our microarray experiments probed for 1,561 of the 1,858 IA candidate genes. Gene lists from the candidate intervals were obtainedfrom the National Center for Biotechnology Information using genome build 36.2. A total of 810 (52% of 1,561 probed genes) candidate genes were expressedin intracranial arteries.

46 GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

analyzed 5 samples or sample pools on each platform for a total of10 independent array experiments. In this way, the use of twodifferent microarray platforms may complicate the analysis, but itserves as a validation of the Affymetrix experiments.

To compare gene expression in arteries from IA individuals(aneurysmal arteries and contralateral control arteries) with arter-ies from non-IA individuals (IVs), we combined the Affymetrixand Illumina data by identifying the genes represented on bothplatforms for use as a common reference gene set. The most recentannotations for both platforms were used. The annotations for theHG_U133 Plus 2.0 (version na23; annotated on July 11, 2007)were obtained from Affymetrix (http://www.affymetric.com/). Togenerate the present annotation for the Illumina chip, we comparedthe probe sequences with the National Center for BiotechnologyInformation RefSeq (Release 24, dated July 10, 2007) sequencesusing stand-alone basic local alignment search tool (BLAST;http://www.ncbi.nlm.nih.gov; ftp://ftp.ncbi.nih.gov/blast) with pa-rameters ensuring near-perfect matches. The matches were evalu-ated to ensure that the best match was unique, i.e., no secondarymatches with one or two mismatches or gaps. There were 16,538distinct genes with unambiguous gene IDs that were not pseudo-genes in common between the Affymetrix (n � 19,221) andIllumina (n � 18,025) platforms. For IA expression data to becombined from both platform analyses, the 20,708 distinct genesthat were not pseudogenes probed for on either platform weregrouped into distinct classes corresponding to no expression,expression on one platform only, and expression on both platformsand were assigned expression confidence scores as shown inTable 3. The raw microarray data sets (series no. GSE6551) can beaccessed at the Gene Expression Omnibus website (http://www.ncbi.nlm.nih.gov/geo).

The Human Investigation Committee of Wayne State University andthe Ethics Committee of the University of Kuopio approved this study.

Real-time quantitative RT-PCR. We performed real-time quanti-tative RT-PCR (qRT-PCR) using validated TaqMan assaysavailable from Applied Biosystems for six IA candidate genes:calpain, small subunit-1 (CAPNS1); serine protease inhibitor,Kunitz type, 2 (SPINT2); nuclear factor of �-light polypeptide geneenhancer in B-cells inhibitor, beta (NFKBIB); p21 (CDKN1A)-activated kinase-4 (PAK4); zinc finger protein-36, C3H type,(ZFP36); and latent transforming growth factor-� binding protein-4(LTBP4). Eight pools of intracranial artery specimens of differentparts of the intracranial vessels originating from 11 individuals wereevaluated as described in Supplemental Table S1 (supplemental dataare available at the online version of this article).

The qRT-PCR assays were performed using 50 ng of arteryRNA at the Applied Genomics Technology Center of Wayne StateUniversity. cDNA synthesis and quantification of gene expressionwere carried out as previously described (27). The messengerRNAs for the six genes were quantified using the delta crossoverthreshold (�Ct) method, on an average of three replicates, whichprovides the difference between the Ct values of the target geneand that of the reference gene. Samples were standardized withrespect to 18S ribosomal RNA TaqMan probes. We calculated thecorrelation between the Illumina gene expression data and theRT-PCR data using the Pearson correlation statistic.

Functional classification of genes. We identified genes expressedin aneurysmal or control intracranial arteries that map to all 10reported IA susceptibility loci (Fig. 1 and Table 1). These ex-pressed genes were analyzed for membership in specific KyotoEncyclopedia of Genes and Genomes (KEGG) biological path-ways, using the WebGestalt resource (web-based gene set analysis

Table 2. Tissue samples used for microarray analysis

Control Aneurysm Ethnicity Sex Age, yrDiagnosis/Cause

of Death§ Artery Aneurysm Size, mm

Affymetrix array (n � 5)

Affy-1-Ac* Affy-2-Aa C F 50 SAH MCA 10Affy-1-Bc Affy-2-Ba C F 66 SAH MCA 5Affy-1-Cc Affy-2-Ca C F 53 SAH MCA 10Affy-1-Dc Affy-2-Da C F 54 SAH MCA 9Affy-1-Ec Affy-2-Ea C F 73 SAH MCA 5

Mean 59.2† Mean 7.8‡Affy-3-Fc Affy-4-Fa C M 45 SAH aCommA �10Affy-3-Gc Affy-4-Ga C M 64 SAH aCommA 5Affy-3-Hc Affy-4-Ha C M 52 SAH aCommA 5

Mean 53.7† Mean 5.9‡Affy-5-Ic — C M 75 UIA aCommA 3Affy-5-Jc — C M 61 UIA aCommA 5Affy-5-Kc — C M 53 UIA aCommA 2Affy-5-Lc — C M 62 UIA aCommA 2

Mean 62.8† Mean 3.0‡

Illumina array (n � 5)

Illumina-1-Mc* — AA M 52 Acute heroinintoxication

M1 MCA ICA —

Illumina-2-Nc — AA M 47 GSW M1 MCA ICA —Illumina-3-Oc — AA M 34 GSW M1 MCA ICA —Illumina-4-Pc — AA M 23 GSW M1 MCA ICA —Illumina-5-Pc — AA M 23 GSW M2 M3 MCA —

*Identifier � platform type-array no.-individual no.artery affected status; Affymetrix analysis included RNA from 12 Finnish Caucasian (C) individuals, andIllumina analysis included RNA from 4 African-American (AA) individuals. †Mean age for pooled samples. ‡Mean aneurysm size for pooled samples. §In thiscolumn, “diagnosis” applies to Affymetrix array (top), and “cause of death” applies to Illumina array (bottom). Affymetrix HG_U133 Plus 2.0 microarray samplesincluded 3 control pools with RNA from 5, 3, and 4 Finnish individuals, respectively, and 2 aneurysm pools with RNA from 5 and 3 Finnish individuals,respectively. Illumina Human WG-6 microarray samples included 5 control samples from 4 distinct individuals, all African-Americans. F, female; M, male; UIA,unruptured IA; SAH, ruptured IA as subarachnoid hemorrhage; GSW, gun shot wound; aCommA, anterior communicating artery; MCA, middle cerebral artery;M1, primary branch MCA; M2, secondary branch MCA; M3, tertiary branch MCA; ICA, internal carotid artery.

47GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

toolkit, available at http://bioinfo.vanderbilt.edu/webgestalt) (33).Taking into consideration the gene expression data, we devel-oped a likelihood-based score to rank pathways for involvement inthe pathogenesis of IA, weighted exponentially by the number of loci

Li,data � j

�Pi, j � Pei, j�, (1)

Pi, j �gwi, j

gwi

�gpj

i

gpj

(2)

Pei, j � 1 ��gei, j � Cj � 0.5�

j

�gpi, j � 1�(3)

(See supplemental data.) The computation allows for locus-specificweights to reflect the a priori confidence in a specific locus as well asthe use of weights for gene expression. We weighted our geneexpression based on our expression confidence scores (Table 3).Ranking of the pathway scores in ascending order gives an indicationof which pathway is most likely to be involved in IA relative to other

pathways. The individual genes in that pathway, which are located inthe candidate intervals, then become functional positional candidates.These analyses were carried out with genes belonging to the func-tional category hsa04nnn that correspond to biological function path-ways and include cell signaling and cellular processes.

We also investigated the interrelatedness of the pathways, as theyare not entirely separate from one another. The KEGG pathwaydatabase includes multiple references to other KEGG pathways, asillustrated by a rectangle with rounded corners in the pathway image(see Fig. 3). We manually extracted this interconnectedness informa-tion from the KEGG pathways and illustrated the interactions using aconnecting line between two related pathways. CytoScape software(http://www.cytoscape.org) was used to generate the framework foran interaction map of the KEGG biological pathways (25). Finalimages were created in Adobe Illustrator (version 12.0.1, AdobeSystems).

RESULTS

Genes expressed in intracranial arteries. To identify genesthat may be involved in the molecular pathogenesis of IA, weperformed gene expression profiling to find genes expressed inintracranial arteries or aneurysmal arteries. We combined datafrom two gene expression platforms: Affymetrix HG_U133Plus 2.0 GeneChips (n � 5) and Illumina WG-6 BeadChips(n � 5). Genes that were represented on either platform and notpseudogenes (20,708) were scored according to the platform inwhich they were identified as being expressed. Twenty-onepercent (4,250) of the genes were identified as expressed onboth platforms (class 5 genes; Table 3), 3% (635) were probedand detected on one platform only (class 4 or 3), 25% (5,281)were expressed on one platform (class 2 or 1), and 52%(10,989) of the genes were not expressed on either platform(class 0). The 10,989 genes with a expression class of 0 cantherefore be excluded as candidate genes for IA, as they are notexpressed in nonaneurysmal or aneurysmal intracranial arter-ies. Expression class 5 genes are regarded as having the highestconfidence in expression, as they were expressed on bothmicroarray platforms.

An independent set of eight pools of RNA samples (seeSupplemental Table S1) was used for validation of geneexpression with RT-PCR. The Pearson correlation between theIllumina microarray expression data and the RT-PCR expres-sion data for the six genes analyzed was 0.963, indicating thatthe two gene expression measures are in good agreement.Specifically, this positive correlation coefficient suggests thatthere is a linear relationship between the two sets of geneexpression data. The RT-PCR results for these six genesvalidated their intracranial artery expression, which was de-tected by both Affymetrix and Illumina microarray platforms.

The difficulty of obtaining IA samples, the need to amplifythem, and the fact that the data were generated on two plat-forms precluded effective differential expression analysis.

Fig. 1. Outline of the study. We integrated global gene expression profiles ofintracranial arteries with genetic mapping data for intracranial aneurysm (IA)to prioritize candidate genes.

Table 3. Gene expression classes and scores

Expression Class Gene Probed on Platform Class Confidence Expression Score

Not expressed Affymetrix or Illumina only, or both 0 0Expressed on Affymetrix only Both 1 Low 0.5Expressed on Illumina only Both 2 Low 0.65Expressed on Affymetrix only Affymetrix only 3 Medium 0.75Expressed on Illumina only Illumina only 4 Medium 0.9Expressed on both Both 5 High 1.0

48 GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

Characteristics of IA positional candidate genes. The ge-nome-wide mRNA expression data provide means for priori-tizing candidate genes for IA from all reported linkage regions(i.e., the most plausible candidates for IA). So far, 10 loci havebeen linked to IA, containing a total of 1,858 distinct andnonpseudo- genes (Table 1, Supplemental Table S2). Thecandidate intervals for chromosomes 7q11 and 19q13 havebeen identified in two or more studies. In addition, the candi-date regions on chromosome 14q overlap by 56 genes, andthose of chromosome Xp22 overlap by 44 genes. The 1,858genes for IA include 335 LOC (hypothetical protein in whichorthologs have not yet been determined), 49 FLJ (Japanesedatabase full-length human cDNA clone), 18 MGC (mamma-lian gene collection clone), 9 KIAA [human novel large (�4kb) cDNA identified in the Human Unidentified Gene-Encoded(HUGE) protein database (http://www.kazusa.or.jp/huge)], 66open reading frame (ORF), and 7 DFKZP (cDNA produced byDeutsches Krebsforschungszentrum) genes as well as 36 non-coding RNAs, leaving 1,338 genes with known function (nothypothetical) as candidates for IA (Table 4). The Affymetrixand Illumina microarray platforms probed for 1,561 (84%) ofthe 1,858 positional IA candidate genes, including 1,298 (97%)of the 1,338 genes with known function. Altogether, 810 (52%)of the 1,561 probed genes were expressed in aneurysmal orcontrol intracranial arteries including 727 (56%) of the 1,298IA candidate genes with known function (see SupplementalTable S2). Table 1 describes the distribution of the expressedgenes across all IA-linked regions, in which linked loci on thesame chromosomal region were combined. The number ofpotential candidate genes per linkage interval ranged from 35on the chromosome 11q24–q25 interval to 572 on the chro-mosome 19q13 interval. The coverage on the microarray plat-forms was best for the chromosome 19q13 interval, with 91.8%of the genes being represented on the arrays, and worst for thechromosome 5p15.2–p14.3 interval, with only 49.1% cover-age. The coverage for known genes was good to excellent,ranging from 89% for chromosome 5q22–q31 to 100% forchromosomes 2p13, 5p15.2–p14.3, and 11q24–q25, with thecoverage for all other loci being �97%. The average propor-tion of genes in all intervals that were expressed with highconfidence (class 5) was 17% (9–28%) for all genes and 19%(9–34%) for probed genes, increasing to 23% (12–41%) for

known genes. The lowest proportion of expressed genes was inthe chromosome 11q24–q25 interval and the highest in thechromosome 2p13 interval (see Table 1).

Integration of genetic mapping and expression data to iden-tify functional pathways in IA. Only �22% (177 genes) of the810 expressed IA genes have KEGG pathway annotations.These include 93 genes involved in 45 biological functionpathways (hsa04nnn), where pathways were represented on atleast 1 locus, and 92 genes involved in 32 biological functionpathways represented on at least 2 loci. The pathway member-ship of the 177 expressed and annotated linked IA genes (Table4) was determined for each pathway and for each of the 15chromosomal regions (10 distinct), where pathways were rep-resented on at least 2 loci (n � 32 KEGG pathways; Table 5,Fig. 2). We grouped the pathways into five categories: neuro-logical, signaling, immunological, cell-cell adhesion, and basalcell processes. Figure 2 and Supplemental Table S3 show thatthe pathways are highly interrelated, with many genes havingmembership in more than one biological function pathway.The locus on chromosome 2p13 recently was shown to be anartefact; this was due largely to misclassification (22), and itwas eliminated from further analyses.

All pathways were ranked according to the likelihood-basedscore that accounts for the number of genes expressed andnumber of linked loci with genes mapping to a particularpathway, and in which the number of linked loci contributessubstantially to the score. We considered two ways to rank thepathways: 1) pathways weighted by expressed gene count,where all genes with nominal expression are ranked equally,and 2) pathways weighted by a confidence score in theexpression of the gene based on whether the gene wasexpressed on one or both microarray platforms (Table 5).With the use of this approach, the adherens junction andMAPK signaling pathway had the highest likelihood-basedscores (Table 5). There were altogether 83 expressed posi-tional candidate genes in the top 20 biological pathways(Table 6); 29 (35%) of them belonged to more than onepathway.

The adherens junction pathway includes 77 genes, of which59 were identified as expressed in aneurysm or control arteries,37 of which were at high confidence (Fig. 3 and SupplementalTable S4). Eight of the 77 adherens junction pathway genes

Table 4. Summary of positional IA candidate genes

Total* Known DKFZP FLJ KIAA LOC MGC ORF ncRNA

All Intervals 1,858 1,338 7 49 9 335 18 66 36Probed 1,561 1,298 7 46 9 121 15 65

KEGGall �1 loci 294 292 2KEGGall �2 loci 284 282 2KEGGbf �1 loci 176 176KEGGbf �2 loci 173 174

Expressed 810 727 2 12 6 31 4 28KEGGall �1 loci 177 173 1KEGGall �2 loci 169 163 1KEGGbf �1 loci 93 93KEGGbf �2 loci 92 92

*Does not include the ncRNA that are known. ncRNA, noncoding RNA, e.g., snoRNA, miRNA, snRNA; KEGGall, all 161 Kyoto Encyclopedia of Genes andGenomes (KEGG) pathways based on the July 3, 2007 revisions of KEGG; KEGGbf, 45 KEGG pathways related to biological function (hsa04nnn). LOC,hypothetical protein in which orthologs have not yet been determined; KIAA, human novel large (�4 kb) cDNA identified in the Human UnidentifiedGene-Encoded (HUGE) protein database (http://www.kazusa.or.jp/huge); FLJ, Japanese database full-length human cDNA clone; MGC, mammalian genecollection clone; DKFZP, cDNA produced by Deutsches Krebsforschungszentrum.

49GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

map to 5 distinct IA linkage intervals. Six of these genes wereexpressed at high confidence: ACTN4, CDC42, EGFR, NLK,PVRL2, and TCF7. Two were expressed at lower confidence:CTNNA1 and WASF2.

The MAPK pathway includes 256 genes, 165 of which wereexpressed, with 68 at high confidence (Supplemental Table S5).Thirty genes map to 13 IA linkage intervals, of which 8 aredistinct. Eighteen of the 30 genes were expressed, 7 at highconfidence: CDC42, EGFR, NLK, RPS6KA3, STMN1, TAOK1,and TGFB3. Eleven were expressed at low confidence: CD14,FOS, MAP2K3, MAP3K6, MAP3K10, MAP4K1, MAPK7, NF1,PPP3R1, PPP5C, and RRAS.

The Notch pathway was among the top biologically relevantpathways and includes 47 genes, of which 40 were identified asexpressed in aneurysm or control arteries, 19 of which at highconfidence (Fig. 4 and Supplemental Table S6). Eight of the 47Notch pathway genes are located in IA loci and can thereforebe considered IA positional candidate genes, including 6 that

were identified as expressed in IA or control arteries, DTX2,SNW1, HDAC1, PSENEN, NUMB, and NUMBL, and two thatwere not detected as expressed, DLL3 and PSEN1. Three geneswere expressed at high confidence: HDAC1, NUMB, andSNW1.

DISCUSSION

The global gene expression data together with previousgenetic mapping data for IA gave us the tools to identify thepositional candidate genes that are expressed in intracranialarteries. A gene involved in disease may influence the concen-tration of protein isoforms or protein functionality or efficiencyor alter the protein responsiveness to environmental factors thatsignal for gene expression (30). The genes identified as ex-pressed are involved in several biological pathways includingadherens junction, MAPK signaling, focal adhesion, regulationof actin cytoskeleton, calcium signaling, and Notch signaling.

Fig. 2. Interaction map of Kyoto Encyclopedia of Genes and Genomes (KEGG) biological pathways for IA. Pathway annotations (nodes) were available for 177of the 810 expressed positional candidate genes. Presence or absence of pathway genes in linked loci is indicated by dark pie chart sections in nodes for pathwayswith genes on at least 2 loci (n � 32 KEGG pathways; n � 92 genes). Node diameters are proportional to pathway size (in genes), node circumference colorindicates pathway category, and colored edges for nodes in the center of the interaction map differentiate pathways with connections to multiple pathwaycategories. Pathway abbreviations are shown in Table 5. See EXPERIMENTAL PROCEDURES for details on how the interactions were identified.

Table 5. Most significant KEGG biological pathways in IA

KEGG ID Pathway Name Abbrev.a Ngb Ngp

b Ngeb Ngle

b Nlocb Fractionc Scored Rsum

e Rcnte

hsa04520 Adherens junction AJ 77 8 8 11 8 0.30 1.80E-12 1 2hsa04010 MAPK signaling pathway MAPK 256 30 18 23 13 1 2.52E-12 2 1hsa04510 Focal adhesion FA 195 12 9 14 8 0.76 3.13E-11 3 4hsa04810 Regulation of actin cytoskeleton RAC 208 17 15 21 9 0.81 6.78E-11 4 3hsa04912 GnRH signaling pathway GnRH 97 12 6 9 8 0.38 2.66E-10 5 5hsa04020 Calcium signaling pathway Ca2 175 16 7 13 10 0.68 8.59E-10 6 6hsa04330 Notch signaling pathway Notch 47 8 6 8 6 0.18 1.29E-09 7 7hsa04910 Insulin signaling pathway INS 135 9 6 11 6 0.53 2.62E-09 8 9hsa04110 Cell cycle CC 114 11 8 10 6 0.45 4.74E-09 9 11hsa04670 Leukocyte transendothelial migration LTEM 109 8 6 10 6 0.43 4.94E-09 10 10hsa04530 Tight junction TJ 119 14 10 13 8 0.46 6.02E-09 11 8hsa04720 Long-term potentiation LTP 69 5 3 6 6 0.27 6.20E-09 12 12hsa04150 mTOR signaling pathway mTOR 47 6 3 4 4 0.18 4.66E-08 13 13hsa04350 TGF-beta signaling pathway TGF-� 84 6 5 6 4 0.33 8.15E-08 14 14hsa04916 Melanogenesis MG 102 7 4 6 5 0.40 9.64E-08 15 16hsa04320 Dorso-ventral axis formation DVAF 28 2 2 3 3 0.11 1.02E-07 16 15hsa04610 Complement and coagulation cascades CCC 69 5 4 8 4 0.27 3.44E-07 17 17hsa04514 Cell adhesion molecules, CAMs CAM 132 8 3 7 4 0.52 4.83E-07 18 19hsa04360 Axon guidance AG 128 12 8 9 6 0.50 5.68E-07 19 18hsa04540 Gap junction GJ 92 3 3 4 3 0.36 1.01E-06 20 20hsa04060 Cytokine-cytokine receptor interaction CCRI 256 20 5 7 6 1 2.07E-06 21 22hsa04620 Toll-like receptor signaling pathway TLR 90 7 4 4 4 0.35 1.32e-05 22 21hsa04310 Wnt signaling pathway Wnt 149 11 7 7 5 0.58 2.49e-05 23 23hsa04012 ErbB signaling pathway ErbB 87 7 3 4 4 0.34 2.42e-05 24 24hsa04370 VEGF signaling pathway VEGF 70 9 3 5 5 0.27 2.09e-05 25 25hsa04130 SNARE interactions in vesicular transport SNARE 36 2 2 2 2 0.14 0.00031 26 26hsa04660 T-cell receptor signaling pathway T cell 93 11 5 5 4 0.36 0.0006 27 27hsa04730 Long-term depression LTD 76 9 2 2 2 0.30 0.0020 28 28hsa04662 B-cell receptor signaling pathway B cell 63 6 3 3 3 0.25 0.0034 29 29hsa04080 Neuroactive ligand-receptor interaction NLRI 254 20 5 9 4 0.99 0.018 30 30hsa04640 Hematopoietic cell lineage HP 88 10 2 2 2 0.34 0.031 31 31hsa04650 Natural killer cell-mediated cytotoxicity NK 131 5 3 3 2 0.51 0.12 32 32

aPathway name abbreviation. bCounts of genes or loci: Ng, count of genes in pathway; Ngp, count of genes in pathway that were in the intervals and were alsoprobed; Nge, count of genes in pathway that were in the intervals, were probed, and were also expressed; Ngle, count of genes in pathway that were expressed,including multiples from overlapping intervals (this is the count used for scoring); Nloc, count of loci that had one or more genes in the pathway that wereexpressed. cFraction is the proportion with respect to the largest pathway (MAPK, n � 256 genes). dPathway ranking score. eRank order of pathways: Rsum, basedon using the sum of the gene expression confidence scores; Rcnt, based on count of genes expressed. Boldface nos. indicate where rank by count differs fromrank by sum of gene expression scores.

50 GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

51GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

Table 6. Expressed positional candidate genes with membership in the 20 most significant KEGG biological pathways in IA

Symbol Gene ID Locus Score

Rank

2

(hsa04010)

6

(hsa04020)

9

(hsa04110)

13

(hsa04150)

16

(hsa04320)

7

(hsa04330)

14

(hsa04350)

19

(hsa04360)

ACTN4 81 19q13 5ARHGEF1 9138 19q13 1C1QA 712 1p34.3–p36.13 2C1QC 714 1p34.3–p36.13 2C5AR1 728 19q13 5CALM3 808 19q13 5 XCD14 929 5q22–q31 2 XCDC23 8697 5q22–q31 5 XCDC42 998 1p34.3–p36.13 5 X XCDK6 1021 7q11 1 XCTNNA1 1495 5q22–q31 3DBF4 10926 7q11 2 XDIAPH1 1729 5q22–q31 2DTX2 113878 7q11 2 XEGFR 1956 7q11 5 X X XEPB41 2035 1p34.3–p36.13 3ETS1 2113 11q24–q25 1 XFLOT2 2319 17cen 2FOS 2353 14q22 2 XFOS 2353 14q23-31 2 XFZD1 8321 7q11 2GIT1 28964 17cen 1GNAI1 2770 7q11 2 XGRLF1 2909 19q13 5GYS1 2997 19q13 2HBEGF 1839 5q22–q31 2HDAC1 3065 1p34.3–p36.13 5 X XHGF 3082 7q11 1HTR6 3362 1p34.3–p36.13 1 XID3 3399 1p34.3–p36.13 2 XLIMK1 3984 7q11 2 XLIPE 3991 19q13 1LLGL1 3996 17cen 1MAGI2 9863 7q11 5MAP2K3 5606 17cen 1 XMAP3K10 4294 19q13 1 XMAP3K6 9064 1p34.3–p36.13 2 XMAP4K1 11184 19q13 2 XMAPK7 5598 17cen 2 XNF1 4763 17cen 1 XNLK 51701 17cen 5 XNRXN3 9369 14q22 5NRXN3 9369 14q23-31 5NUMB 8650 14q22 5 XNUMB 8650 14q23-31 5 XNUMBL 9253 19q13 1 XPAK4 10298 19q13 5 XPGF 5228 14q22 5 XPGF 5228 14q23-31 5 XPHKA2 5256 Xp22 2 XPLAUR 5329 19q13 2PPP2CA 5515 5q22–q31 2 XPPP3R1 5534 2p13 2 X X XPPP5C 5536 19q13 2 XPSENEN 55851 19q13 2 XPVR 5817 19q13 5PVRL2 5819 19q13 5RPS6KA3 6197 Xp22 5 X XRRAS 6237 19q13 2 XSEMA3C 10512 7q11 5 XSEMA3E 9723 7q11 2 XSEMA5A 9037 5p15.2–p14.3 3 XSFN 2810 1p34.3–p36.13 1 XSKP1A 6500 5q22–q31 5 X XSLC9A1 6548 1p34.3–p36.13 5SMAD5 4090 5q22–q31 5 XSNW1 22938 14q22 5 XSNW1 22938 14q23-31 5 XSPHK2 56848 19q13 2 XSREBF1 6720 17cen 2SSH2 85464 17cen 1STMN1 3925 1p34.3–p36.13 5 XSYMPK 8189 19q13 1TAOK1 57551 17cen 5 XTCF7 6932 5q22–q31 5TGFB3 7043 14q22 5 X X XTGFB3 7043 14q23-31 5 X X XTMSB4X 7114 Xp22 4ULK2 9706 17cen 5 XVASP 7408 19q13 2VDAC1 7416 5q22–q31 5 XWASF2 10163 1p34.3–p36.13 1YWHAG 7532 7q11 5 X

Continued

52 GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

Table 6.—Continued

Rank

3

(hsa04510)

18

(hsa04514)

1

(hsa04520)

11

(hsa04530)

20

(hsa04540)

17

(hsa04610)

10

(hsa04670)

12

(hsa04720)

4

(hsa04810)

8

(hsa04910)

5

(hsa04912)

15

(hsa04916)

X X X X XX

XXX

X X X XX

X X X X X X

X X X

X X

X X X X XX

X

XX

X X X XX X X

XX

X

XX

XX

X

X X

XXX

X XXX

XX

XX

XX X

XX X

X

XX

X

X X

X

X X

X X

53GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

Such pathways regulate important processes such as cellgrowth, tissue remodeling, inflammation, and wound healing,processes that likely contribute to the pathophysiology of IA.Here we discuss in some detail three signaling pathways,adherens junction, MAPK, and Notch signaling, that are func-tionally relevant to the pathogenesis of IA.

Adherens junction. The most significant biological pathwayidentified for IA was the adherens junction pathway. Adherensjunctions are complex dynamic structures that include bothadhesive and signaling molecules that maintain contact inhibi-tion of endothelial cell growth and vascular permeability toinflammatory cells and solutes and are required for the orga-nization of new vessels in angiogenesis (3). In vascular endo-thelial cells, adherens junctions are important structures thatmaintain the integrity of the vessel wall and are also involvedin endothelium remodeling in various physiological and patho-

logical processes (14). Adherens junction proteins can transferintracellular signals to �-catenin, which has the ability totranslocate to the nucleus and associate with transcriptionfactors to regulate the expression of genes involved in endo-thelial cell growth, differentiation, and apoptosis (3).

The adherens junction pathway is connected to severalbiological pathways including MAPK, Wnt, and transforminggrowth factor-� (TGF-�) signaling, cytokine-cytokine receptorinteraction, and the cell adhesion molecule pathway. Theinteraction of key integral membrane and intracellular proteinsto form appropriate endothelial cell-cell junctions is importantto maintain vascular homeostasis. Our gene expression analy-sis indicated that the majority of genes in the adherens junctionpathway were expressed at moderate-to-high confidence levels.It is noteworthy that CTNND2, mapping to the IA locus onchromosome 5p, encodes catenin delta-2, a protein that may be

Fig. 3. Adherens junction pathway (hsa04520) modified from KEGG pathway. Protein symbols were replaced by gene symbols to reflect gene-centric data. Fordetails on the scoring system, see EXPERIMENTAL PROCEDURES and Table 3. Supplemental Table S4 provides details on the expression of each gene in this pathway.

54 GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

involved in the regulation of adhesion molecules in the brain(29). Positional IA candidate genes in the adherens junctionpathway, such as the CTNNA1gene encoding catenin alpha-1,may therefore have a specific role in maintaining the integrityof intracranial arteries.

MAPK signaling. MAPK is one of the most interconnectedpathways (hsa04010; Fig. 2, Supplemental Table S3). Thethree major branches of the MAPK pathway include the clas-sical MAPK pathway, c-Jun NH2-terminal kinase (JNK) andp38 MAPK pathway, and the ERK5 pathway, which involvessignaling systems that regulate cell proliferation, differentia-tion, inflammation, and apoptosis. MAPKs are expressed inmultiple vascular cell types including cardiomyocytes, vascu-lar endothelial cells, and vascular smooth muscle cells(VSMCs) and function to regulate cardiovascular signal trans-duction pathways (12). The MAPKs mediate signals triggeredby cytokines, growth factors, and environmental stress includ-ing ischemia, shear stress, and vasoactive agents. Studies ofMAPK signaling in the cardiovascular system indicate thatthese cascades regulate vascular endothelial cell permeability,production of cytokines, modulation of vasomotor function,and mediation of reperfusion injury (12).

There are 18 expressed IA candidate genes that play a rolein MAPK signaling, including MAPK7, TGFB3, CDC42, NLK,STMN1, PPP3R1, and RPS6KA3 (see Table 6). Additionally,genes encoding heat shock proteins such as HSPA9B andHSPB1 are expressed in IA. Other genes that influence orinteract with the MAPK signaling pathway and are expressedin intracranial arteries include EGFR, TGFBR1, TNFRSF1A,CASP2, IL1B, IL1R1, JUN, MAPK8 (JNK), HSPA5, andDUSP10. HSP5A is a positional candidate gene for IA thatencodes a heat shock protein involved in the negative regu-lation of JNK. DUSP10 encodes a dual-specificity MAPKphosphatase that has a principal function in both innate andadaptive immune responses, as it inhibits JNK, which regulatesthe transcription factor activator protein-1 (AP1), implicated in

the controlled expression of many genes involved in the im-mune response (4, 34).

Recently, TNFRSF13B, located on chromosome 17, wasidentified as a susceptibility gene for IA (10). It encodestransmembrane activator and calcium modulator ligand inter-actor (TACI), involved in immunity by mediation of isotypeswitching in B cells. TNFRSF13B is not known to be one ofthe TNF receptors in the MAPK signaling pathway; however,a related TNF receptor superfamily member, TNFRSF1A, isinvolved in the MAPK pathway and is expressed in IA.

Notch signaling. Several linked genes for IA are part of theNotch signaling pathway (hsa04330; see Table 6) (5), which isinvolved in the development of cardiovascular structures inmammals and is important for endothelial and smooth musclecells to form arteries and veins. Its relevance in the adultvascular system is demonstrated by mutations in NOTCH3 thatcause cerebral autosomal dominant arteriopathy with subcor-tical infarcts and leukoencephalopathy (CADASIL), a vasculardegenerative disease that is the most common form of hered-itary stroke disorder (11). The underlying pathology involvesprogressive degeneration of the VSMCs in arteries. Mutationsin NOTCH3 cause abnormal accumulation of the Notch3 re-ceptor at the cytoplasmic membrane of VSMCs in intracranialvessels (5). The disruption of the normal anchorage of VSMCsto the extracellular matrix and other nearby cells, in addition tocytoskeletal alterations, seems to trigger the VSMC degenera-tion observed in CADASIL patients (21). The loss of VSMCshas been shown to lead to endothelial cell abnormalities in tightjunctions and gap junctions (6). These cell-cell adhesion path-ways are connected to other pathways that function to maintaincell junctions, including adherens junction, cell adhesion mol-ecules, regulation of the actin cytoskeleton, and MAPK andCa2 signaling pathways (Fig. 2).

Study limitations. Although gene expression analysis is avaluable experimental tool, KEGG annotation is �75% incom-plete for expressed IA positional candidate genes, a significant

Fig. 4. Notch signaling pathway (hsa04330) modified from KEGG pathway. Protein symbols were replaced by gene symbols to reflect gene-centric data. Fordetails on the scoring system, see EXPERIMENTAL PROCEDURES and Table 3. Supplemental Table S6 provides details on the expression of each gene in this pathway.

55GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

limitation. The experiments described here evaluated geneexpression as it appears in end-stage IA rather than differentialgene expression in the development of aneurysmal arteriescompared with control arteries. To determine differential geneexpression in the development and rupture of IA, a largersample set needs to be evaluated, preferably in tissues excisedduring aneurysm repair surgery. Also, to understand whyaneurysms are prone to form at specific locations along thecircle of Willis, there is a need for a comprehensive analysis ofgene expression of the different intracranial arteries. Addition-ally, the intracranial arteries evaluated in this study come fromtwo different ethnic groups, Finnish Caucasians and AfricanAmericans. It is possible that vascular gene expression mightdiffer between populations. Extensive analysis of a largersample set of arteries from these two populations would berequired to quantify such differences.

The intracranial artery tissue samples were obtained atautopsy; thus we cannot completely exclude the possibility ofpostmortem changes in the analyzed samples. All autopsieswere performed within 24 h of death. In addition, the brainsused for IV tissue here were also used by Albertson et al. (1)to obtain good-quality RNA from the accumbens for geneexpression experiments. Also, our recent microarray expres-sion study on abdominal aortic aneurysms, described in Lenket al. (13), showed that high-quality RNA can be obtained fromautopsies. Our approach may have missed genes involved in IApathogenesis, because expression of causative genes may occuronly at certain times in development and not in adult tissues,although this is unlikely, since most IAs develop later in life.Technical problems (e.g., so-called nonperforming probes)with the microarray probes representing some genes have beennoted previously (2). This is unlikely to influence these results,since two different microarray platforms were used, and theyagree extensively (see also Lenk et al., Ref. 13). There couldhave been a loss of transcripts in the RNA isolation or ampli-fication process. Nevertheless, the microarray data provideuseful qualitative gene expression information that assists inthe identification of functional candidate genes involved in theformation or rupture of IA, or both. These data will also beuseful for studies of other cerebrovascular diseases.

In summary, the 10 IA susceptibility loci contain 810 ex-pressed positional candidate genes. These genes function inbiological pathways such as adherens junction, MAPK signal-ing, and Notch signaling that may be involved in the patho-genesis of IA. This work demonstrates a method for integratingglobal gene expression profiles of nonaneurysmal and aneu-rysmal intracranial arteries with genetic mapping data to pri-oritize positional candidate genes for further study.

ACKNOWLEDGMENTS

We thank Daniel Lott and Sara McNorton (Applied Genomics TechnologyCenter at Wayne State University) for work on microarray experiments,Kristine Barker and Dr. Michael Bannon (Dept. of Pharmacology at WayneState University) for the contribution of collecting intracranial artery speci-mens, and Katariina Helin (Kuopio University) for assistance in the collectionof IA samples.

Present addresses: S. Weinsheimer, Genomics Division, Lawrence Berke-ley National Laboratory, Berkeley, CA; and M. van der Voet, Dept. ofDevelopmental Biology, Utrecht University, Utrecht, The Netherlands.

GRANTS

This project was funded in part by a grant from the National Institute ofNeurological Disorders and Stroke (NS-034395 to G. Tromp). S. Weinsheimer

(0410051Z) and G. M. Lenk (0510063Z and 0710099Z) were recipients ofpredoctoral fellowships from the American Heart Association, Greater Mid-west Affiliate.

REFERENCES

1. Albertson DN, Schmidt CJ, Kapatos G, Bannon MJ. Distinctive pro-files of gene expression in the human nucleus accumbens associated withcocaine and heroin abuse. Neuropsychopharmacology 31: 2304–2312,2006.

2. Barnes M, Freudenberg J, Thompson S, Aronow B, Pavlidis P.Experimental comparison and cross-validation of the Affymetrix andIllumina gene expression analysis platforms. Nucleic Acids Res 33: 5914–5923, 2005.

3. Bazzoni G, Dejana E. Endothelial cell-to-cell junctions: molecular orga-nization and role in vascular homeostasis. Physiol Rev 84: 869–901, 2004.

4. Bennett BL, Satoh Y, Lewis AJ. JNK: a new therapeutic target fordiabetes. Curr Opin Pharmacol 3: 420–425, 2003.

5. Bianchi S, Dotti MT, Federico A. Physiology and pathology of notchsignalling system. J Cell Physiol 207: 300–308, 2006.

6. Brulin P, Godfraind C, Leteurtre E, Ruchoux MM. Morphometricanalysis of ultrastructural vascular changes in CADASIL: analysis of 50skin biopsy specimens and pathogenic implications. Acta Neuropathol(Berl) 104: 241–248, 2002.

7. Farnham JM, Camp NJ, Neuhausen SL, Tsuruda J, Parker D,MacDonald J, Cannon-Albright LA. Confirmation of chromosome 7q11locus for predisposition to intracranial aneurysm. Hum Genet 114: 250–255, 2004.

8. Gautier L, Cope L, Bolstad BM, Irizarry RA. affy–analysis of Af-fymetrix GeneChip data at the probe level. Bioinformatics 20: 307–315,2004.

9. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, DudoitS, Ellis B, Gautier L, Ge Y, Gentry J, Hornik K, Hothorn T, Huber W,Iacus S, Irizarry R, Leisch F, Li C, Maechler M, Rossini AJ, SawitzkiG, Smith C, Smyth G, Tierney L, Yang JY, Zhang J. Bioconductor:open software development for computational biology and bioinformatics.Genome Biol 5: R80, 2004.

10. Inoue K, Mineharu Y, Inoue S, Yamada S, Matsuda F, Nozaki K,Takenaka K, Hashimoto N, Koizumi A. Search on chromosome 17centromere reveals TNFRSF13B as a susceptibility gene for intracranialaneurysm: a preliminary study. Circulation 113: 2002–2010, 2006.

11. Joutel A, Andreux F, Gaulis S, Domenga V, Cecillon M, Battail N,Piga N, Chapon F, Godfrain C, Tournier-Lasserve E. The ectodomainof the Notch3 receptor accumulates within the cerebrovasculature ofCADASIL patients. J Clin Invest 105: 597–605, 2000.

12. Khan TA, Bianchi C, Ruel M, Voisine P, Sellke FW. Mitogen-activatedprotein kinase pathways and cardiac surgery. J Thorac Cardiovasc Surg127: 806–811, 2004.

13. Lenk GM, Tromp G, Weinsheimer S, Gatalica Z, Berguer R, Kuiva-niemi H. Whole genome expression profiling reveals a significant role forimmune function in human abdominal aortic aneurysms. BMC Genomics8: 237, 2007.

14. Liebner S, Cavallaro U, Dejana E. The multiple languages of endothelialcell-to-cell communication. Arterioscler Thromb Vasc Biol 26: 1431–1438, 2006.

15. Mineharu Y, Inoue K, Inoue S, Yamada S, Nozaki K, Hashimoto N,Koizumi A. Model-based linkage analyses confirm chromosome 19q13.3as a susceptibility locus for intracranial aneurysm. Stroke 38: 1174–1178,2007.

16. Nahed BV, Seker A, Guclu B, Ozturk AK, Finberg K, Hawkins AA,DiLuna ML, State M, Lifton RP, Gunel M. Mapping a Mendelian formof intracranial aneurysm to 1p34.3–p36.13. Am J Hum Genet 76: 172–179,2005.

17. Olson JM, Vongpunsawad S, Kuivaniemi H, Ronkainen A, Hernes-niemi J, Ryynanen M, Kim LL, Tromp G. Search for intracranialaneurysm susceptibility gene(s) using Finnish families. BMC Med Genet3: 7, 2002.

18. Onda H, Kasuya H, Yoneyama T, Takakura K, Hori T, Takeda J,Nakajima T, Inoue I. Genomewide-linkage and haplotype-associationstudies map intracranial aneurysm to chromosome 7q11. Am J Hum Genet69: 804–819, 2001.

19. Ozturk AK, Nahed BV, Bydon M, Bilguvar K, Goksu E, Bademci G,Guclu B, Johnson MH, Amar A, Lifton RP, Gunel M. Molecular geneticanalysis of two large kindreds with intracranial aneurysms demonstrateslinkage to 11q24–25 and 14q23–31. Stroke 37: 1021–1027, 2006.

56 GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from

20. Roos YB, Pals G, Struycken PM, Rinkel GJ, Limburg M, Pronk JC, vanden Berg JS, Luijten JA, Pearson PL, Vermeulen M, Westerveld A.Genome-wide linkage in a large Dutch consanguineous family maps a locusfor intracranial aneurysms to chromosome 2p13. Stroke 35: 2276–2281, 2004.

21. Ruchoux MM, Domenga V, Brulin P, Maciazek J, Limol S, Tournier-Lasserve E, Joutel A. Transgenic mice expressing mutant Notch3 de-velop vascular alterations characteristic of cerebral autosomal dominantarteriopathy with subcortical infarcts and leukoencephalopathy. Am JPathol 162: 329–342, 2003.

22. Ruigrok YM. From intracranial aneurysm to subarachnoid hemorrhage:unraveling the genetics (PhD Dissertation). Utrecht, The Netherlands:Utrecht University, 2006.

23. Ruigrok YM, Rinkel GJ, Wijmenga C. Genetics of intracranial aneu-rysms. Lancet Neurol 4: 179–189, 2005.

24. Schievink WI. Genetics of intracranial aneurysms. Neurosurgery 40:651–662, 1997.

25. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D,Amin N, Schwikowski B, Ideker T. Cytoscape: a software environmentfor integrated models of biomolecular interaction networks. Genome Res13: 2498–2504, 2003.

26. Teasdale GM, Wardlaw JM, White PM, Murray G, Teasdale EM,Easton V. The familial risk of subarachnoid haemorrhage. Brain 128:1677–1685, 2005.

27. Tromp G, Gatalica Z, Skunca M, Berguer R, Siegel T, Kline RA,Kuivaniemi H. Elevated expression of matrix metalloproteinase-13 inabdominal aortic aneurysms. Ann Vasc Surg 18: 414–420, 2004.

28. van der Voet M, Olson JM, Kuivaniemi H, Dudek DM, Skunca M,Ronkainen A, Niemela M, Jaaskelainen J, Hernesniemi J, Helin K,Leinonen E, Biswas M, Tromp G. Intracranial aneurysms in Finnishfamilies: confirmation of linkage and refinement of the interval to chro-mosome 19q13.3. Am J Hum Genet 74: 564–571, 2004.

29. Verlaan DJ, Dube MP, St-Onge J, Noreau A, Roussel J, Satge N,Wallace MC, Rouleau GA. A new locus for autosomal dominant intra-cranial aneurysm, ANIB4, maps to chromosome 5p15.2–14.3. J MedGenet 43: e31, 2006.

30. Vink JM, Boomsma DI. Gene finding strategies. Biol Psychol 61: 53–71,2002.

31. Wu Z, Irizarray R, Gentleman R, Martinez-Murillo F, Spencer F. Amodel-based background adjustment for oligonucleotide expression ar-rays. J Am Stat Assoc 99: 909–917, 2004.

32. Yamada S, Utsunomiya M, Inoue K, Nozaki K, Inoue S, Takenaka K,Hashimoto N, Koizumi A. Genome-wide scan for Japanese familialintracranial aneurysms: linkage to several chromosomal regions. Circula-tion 110: 3727–3733, 2004.

33. Zhang B, Kirov S, Snoddy J. WebGestalt: an integrated system forexploring gene sets in various biological contexts. Nucleic Acids Res 33:W741–W748, 2005.

34. Zhang Y, Blattman JN, Kennedy NJ, Duong J, Nguyen T, Wang Y,Davis RJ, Greenberg PD, Flavell RA, Dong C. Regulation of innate andadaptive immune responses by MAP kinase phosphatase 5. Nature 430:793–797, 2004.

57GENE EXPRESSION IN INTRACRANIAL ANEURYSM

Physiol Genomics • VOL 32 • www.physiolgenomics.org

on May 21, 2011

physiolgenomics.physiology.org

Dow

nloaded from